^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1) // SPDX-License-Identifier: GPL-2.0-only
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3) * Kernel-based Virtual Machine driver for Linux
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5) * derived from drivers/kvm/kvm_main.c
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7) * Copyright (C) 2006 Qumranet, Inc.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8) * Copyright (C) 2008 Qumranet, Inc.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9) * Copyright IBM Corporation, 2008
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10) * Copyright 2010 Red Hat, Inc. and/or its affiliates.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 12) * Authors:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 13) * Avi Kivity <avi@qumranet.com>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 14) * Yaniv Kamay <yaniv@qumranet.com>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 15) * Amit Shah <amit.shah@qumranet.com>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 16) * Ben-Ami Yassour <benami@il.ibm.com>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 17) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 18)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 19) #include <linux/kvm_host.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 20) #include "irq.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 21) #include "ioapic.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 22) #include "mmu.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 23) #include "i8254.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 24) #include "tss.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 25) #include "kvm_cache_regs.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 26) #include "kvm_emulate.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 27) #include "x86.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 28) #include "cpuid.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 29) #include "pmu.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 30) #include "hyperv.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 31) #include "lapic.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 32)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 33) #include <linux/clocksource.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 34) #include <linux/interrupt.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 35) #include <linux/kvm.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 36) #include <linux/fs.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 37) #include <linux/vmalloc.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 38) #include <linux/export.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 39) #include <linux/moduleparam.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 40) #include <linux/mman.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 41) #include <linux/highmem.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 42) #include <linux/iommu.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 43) #include <linux/intel-iommu.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 44) #include <linux/cpufreq.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 45) #include <linux/user-return-notifier.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 46) #include <linux/srcu.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 47) #include <linux/slab.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 48) #include <linux/perf_event.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 49) #include <linux/uaccess.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 50) #include <linux/hash.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 51) #include <linux/pci.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 52) #include <linux/timekeeper_internal.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 53) #include <linux/pvclock_gtod.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 54) #include <linux/kvm_irqfd.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 55) #include <linux/irqbypass.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 56) #include <linux/sched/stat.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 57) #include <linux/sched/isolation.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 58) #include <linux/mem_encrypt.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 59) #include <linux/entry-kvm.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 60)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 61) #include <trace/events/kvm.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 62)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 63) #include <asm/debugreg.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 64) #include <asm/msr.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 65) #include <asm/desc.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 66) #include <asm/mce.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 67) #include <linux/kernel_stat.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 68) #include <asm/fpu/internal.h> /* Ugh! */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 69) #include <asm/pvclock.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 70) #include <asm/div64.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 71) #include <asm/irq_remapping.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 72) #include <asm/mshyperv.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 73) #include <asm/hypervisor.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 74) #include <asm/tlbflush.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 75) #include <asm/intel_pt.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 76) #include <asm/emulate_prefix.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 77) #include <clocksource/hyperv_timer.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 78)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 79) #define CREATE_TRACE_POINTS
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 80) #include "trace.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 81)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 82) #define MAX_IO_MSRS 256
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 83) #define KVM_MAX_MCE_BANKS 32
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 84) u64 __read_mostly kvm_mce_cap_supported = MCG_CTL_P | MCG_SER_P;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 85) EXPORT_SYMBOL_GPL(kvm_mce_cap_supported);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 86)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 87) #define emul_to_vcpu(ctxt) \
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 88) ((struct kvm_vcpu *)(ctxt)->vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 89)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 90) /* EFER defaults:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 91) * - enable syscall per default because its emulated by KVM
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 92) * - enable LME and LMA per default on 64 bit KVM
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 93) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 94) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 95) static
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 96) u64 __read_mostly efer_reserved_bits = ~((u64)(EFER_SCE | EFER_LME | EFER_LMA));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 97) #else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 98) static u64 __read_mostly efer_reserved_bits = ~((u64)EFER_SCE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 99) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 100)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 101) static u64 __read_mostly cr4_reserved_bits = CR4_RESERVED_BITS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 102)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 103) #define KVM_X2APIC_API_VALID_FLAGS (KVM_X2APIC_API_USE_32BIT_IDS | \
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 104) KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 105)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 106) static void update_cr8_intercept(struct kvm_vcpu *vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 107) static void process_nmi(struct kvm_vcpu *vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 108) static void process_smi(struct kvm_vcpu *vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 109) static void enter_smm(struct kvm_vcpu *vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 110) static void __kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 111) static void store_regs(struct kvm_vcpu *vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 112) static int sync_regs(struct kvm_vcpu *vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 113)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 114) struct kvm_x86_ops kvm_x86_ops __read_mostly;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 115) EXPORT_SYMBOL_GPL(kvm_x86_ops);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 116)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 117) static bool __read_mostly ignore_msrs = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 118) module_param(ignore_msrs, bool, S_IRUGO | S_IWUSR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 119)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 120) static bool __read_mostly report_ignored_msrs = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 121) module_param(report_ignored_msrs, bool, S_IRUGO | S_IWUSR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 122)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 123) unsigned int min_timer_period_us = 200;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 124) module_param(min_timer_period_us, uint, S_IRUGO | S_IWUSR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 125)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 126) static bool __read_mostly kvmclock_periodic_sync = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 127) module_param(kvmclock_periodic_sync, bool, S_IRUGO);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 128)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 129) bool __read_mostly kvm_has_tsc_control;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 130) EXPORT_SYMBOL_GPL(kvm_has_tsc_control);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 131) u32 __read_mostly kvm_max_guest_tsc_khz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 132) EXPORT_SYMBOL_GPL(kvm_max_guest_tsc_khz);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 133) u8 __read_mostly kvm_tsc_scaling_ratio_frac_bits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 134) EXPORT_SYMBOL_GPL(kvm_tsc_scaling_ratio_frac_bits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 135) u64 __read_mostly kvm_max_tsc_scaling_ratio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 136) EXPORT_SYMBOL_GPL(kvm_max_tsc_scaling_ratio);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 137) u64 __read_mostly kvm_default_tsc_scaling_ratio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 138) EXPORT_SYMBOL_GPL(kvm_default_tsc_scaling_ratio);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 139)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 140) /* tsc tolerance in parts per million - default to 1/2 of the NTP threshold */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 141) static u32 __read_mostly tsc_tolerance_ppm = 250;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 142) module_param(tsc_tolerance_ppm, uint, S_IRUGO | S_IWUSR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 143)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 144) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 145) * lapic timer advance (tscdeadline mode only) in nanoseconds. '-1' enables
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 146) * adaptive tuning starting from default advancment of 1000ns. '0' disables
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 147) * advancement entirely. Any other value is used as-is and disables adaptive
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 148) * tuning, i.e. allows priveleged userspace to set an exact advancement time.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 149) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 150) static int __read_mostly lapic_timer_advance_ns = -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 151) module_param(lapic_timer_advance_ns, int, S_IRUGO | S_IWUSR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 152)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 153) static bool __read_mostly vector_hashing = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 154) module_param(vector_hashing, bool, S_IRUGO);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 155)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 156) bool __read_mostly enable_vmware_backdoor = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 157) module_param(enable_vmware_backdoor, bool, S_IRUGO);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 158) EXPORT_SYMBOL_GPL(enable_vmware_backdoor);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 159)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 160) static bool __read_mostly force_emulation_prefix = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 161) module_param(force_emulation_prefix, bool, S_IRUGO);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 162)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 163) int __read_mostly pi_inject_timer = -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 164) module_param(pi_inject_timer, bint, S_IRUGO | S_IWUSR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 165)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 166) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 167) * Restoring the host value for MSRs that are only consumed when running in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 168) * usermode, e.g. SYSCALL MSRs and TSC_AUX, can be deferred until the CPU
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 169) * returns to userspace, i.e. the kernel can run with the guest's value.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 170) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 171) #define KVM_MAX_NR_USER_RETURN_MSRS 16
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 172)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 173) struct kvm_user_return_msrs_global {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 174) int nr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 175) u32 msrs[KVM_MAX_NR_USER_RETURN_MSRS];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 176) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 177)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 178) struct kvm_user_return_msrs {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 179) struct user_return_notifier urn;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 180) bool registered;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 181) struct kvm_user_return_msr_values {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 182) u64 host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 183) u64 curr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 184) } values[KVM_MAX_NR_USER_RETURN_MSRS];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 185) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 186)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 187) static struct kvm_user_return_msrs_global __read_mostly user_return_msrs_global;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 188) static struct kvm_user_return_msrs __percpu *user_return_msrs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 189)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 190) #define KVM_SUPPORTED_XCR0 (XFEATURE_MASK_FP | XFEATURE_MASK_SSE \
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 191) | XFEATURE_MASK_YMM | XFEATURE_MASK_BNDREGS \
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 192) | XFEATURE_MASK_BNDCSR | XFEATURE_MASK_AVX512 \
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 193) | XFEATURE_MASK_PKRU)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 194)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 195) u64 __read_mostly host_efer;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 196) EXPORT_SYMBOL_GPL(host_efer);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 197)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 198) bool __read_mostly allow_smaller_maxphyaddr = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 199) EXPORT_SYMBOL_GPL(allow_smaller_maxphyaddr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 200)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 201) static u64 __read_mostly host_xss;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 202) u64 __read_mostly supported_xss;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 203) EXPORT_SYMBOL_GPL(supported_xss);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 204)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 205) struct kvm_stats_debugfs_item debugfs_entries[] = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 206) VCPU_STAT("pf_fixed", pf_fixed),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 207) VCPU_STAT("pf_guest", pf_guest),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 208) VCPU_STAT("tlb_flush", tlb_flush),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 209) VCPU_STAT("invlpg", invlpg),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 210) VCPU_STAT("exits", exits),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 211) VCPU_STAT("io_exits", io_exits),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 212) VCPU_STAT("mmio_exits", mmio_exits),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 213) VCPU_STAT("signal_exits", signal_exits),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 214) VCPU_STAT("irq_window", irq_window_exits),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 215) VCPU_STAT("nmi_window", nmi_window_exits),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 216) VCPU_STAT("halt_exits", halt_exits),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 217) VCPU_STAT("halt_successful_poll", halt_successful_poll),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 218) VCPU_STAT("halt_attempted_poll", halt_attempted_poll),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 219) VCPU_STAT("halt_poll_invalid", halt_poll_invalid),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 220) VCPU_STAT("halt_wakeup", halt_wakeup),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 221) VCPU_STAT("hypercalls", hypercalls),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 222) VCPU_STAT("request_irq", request_irq_exits),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 223) VCPU_STAT("irq_exits", irq_exits),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 224) VCPU_STAT("host_state_reload", host_state_reload),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 225) VCPU_STAT("fpu_reload", fpu_reload),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 226) VCPU_STAT("insn_emulation", insn_emulation),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 227) VCPU_STAT("insn_emulation_fail", insn_emulation_fail),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 228) VCPU_STAT("irq_injections", irq_injections),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 229) VCPU_STAT("nmi_injections", nmi_injections),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 230) VCPU_STAT("req_event", req_event),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 231) VCPU_STAT("l1d_flush", l1d_flush),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 232) VCPU_STAT("halt_poll_success_ns", halt_poll_success_ns),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 233) VCPU_STAT("halt_poll_fail_ns", halt_poll_fail_ns),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 234) VM_STAT("mmu_shadow_zapped", mmu_shadow_zapped),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 235) VM_STAT("mmu_pte_write", mmu_pte_write),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 236) VM_STAT("mmu_pde_zapped", mmu_pde_zapped),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 237) VM_STAT("mmu_flooded", mmu_flooded),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 238) VM_STAT("mmu_recycled", mmu_recycled),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 239) VM_STAT("mmu_cache_miss", mmu_cache_miss),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 240) VM_STAT("mmu_unsync", mmu_unsync),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 241) VM_STAT("remote_tlb_flush", remote_tlb_flush),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 242) VM_STAT("largepages", lpages, .mode = 0444),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 243) VM_STAT("nx_largepages_splitted", nx_lpage_splits, .mode = 0444),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 244) VM_STAT("max_mmu_page_hash_collisions", max_mmu_page_hash_collisions),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 245) { NULL }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 246) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 247)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 248) u64 __read_mostly host_xcr0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 249) u64 __read_mostly supported_xcr0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 250) EXPORT_SYMBOL_GPL(supported_xcr0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 251)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 252) static struct kmem_cache *x86_fpu_cache;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 253)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 254) static struct kmem_cache *x86_emulator_cache;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 255)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 256) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 257) * When called, it means the previous get/set msr reached an invalid msr.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 258) * Return true if we want to ignore/silent this failed msr access.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 259) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 260) static bool kvm_msr_ignored_check(struct kvm_vcpu *vcpu, u32 msr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 261) u64 data, bool write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 262) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 263) const char *op = write ? "wrmsr" : "rdmsr";
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 264)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 265) if (ignore_msrs) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 266) if (report_ignored_msrs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 267) kvm_pr_unimpl("ignored %s: 0x%x data 0x%llx\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 268) op, msr, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 269) /* Mask the error */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 270) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 271) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 272) kvm_debug_ratelimited("unhandled %s: 0x%x data 0x%llx\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 273) op, msr, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 274) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 275) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 276) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 277)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 278) static struct kmem_cache *kvm_alloc_emulator_cache(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 279) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 280) unsigned int useroffset = offsetof(struct x86_emulate_ctxt, src);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 281) unsigned int size = sizeof(struct x86_emulate_ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 282)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 283) return kmem_cache_create_usercopy("x86_emulator", size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 284) __alignof__(struct x86_emulate_ctxt),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 285) SLAB_ACCOUNT, useroffset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 286) size - useroffset, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 287) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 288)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 289) static int emulator_fix_hypercall(struct x86_emulate_ctxt *ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 290)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 291) static inline void kvm_async_pf_hash_reset(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 292) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 293) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 294) for (i = 0; i < ASYNC_PF_PER_VCPU; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 295) vcpu->arch.apf.gfns[i] = ~0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 296) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 297)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 298) static void kvm_on_user_return(struct user_return_notifier *urn)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 299) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 300) unsigned slot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 301) struct kvm_user_return_msrs *msrs
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 302) = container_of(urn, struct kvm_user_return_msrs, urn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 303) struct kvm_user_return_msr_values *values;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 304) unsigned long flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 305)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 306) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 307) * Disabling irqs at this point since the following code could be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 308) * interrupted and executed through kvm_arch_hardware_disable()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 309) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 310) local_irq_save(flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 311) if (msrs->registered) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 312) msrs->registered = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 313) user_return_notifier_unregister(urn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 314) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 315) local_irq_restore(flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 316) for (slot = 0; slot < user_return_msrs_global.nr; ++slot) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 317) values = &msrs->values[slot];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 318) if (values->host != values->curr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 319) wrmsrl(user_return_msrs_global.msrs[slot], values->host);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 320) values->curr = values->host;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 321) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 322) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 323) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 324)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 325) int kvm_probe_user_return_msr(u32 msr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 326) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 327) u64 val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 328) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 329)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 330) preempt_disable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 331) ret = rdmsrl_safe(msr, &val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 332) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 333) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 334) ret = wrmsrl_safe(msr, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 335) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 336) preempt_enable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 337) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 338) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 339) EXPORT_SYMBOL_GPL(kvm_probe_user_return_msr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 340)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 341) void kvm_define_user_return_msr(unsigned slot, u32 msr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 342) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 343) BUG_ON(slot >= KVM_MAX_NR_USER_RETURN_MSRS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 344) user_return_msrs_global.msrs[slot] = msr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 345) if (slot >= user_return_msrs_global.nr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 346) user_return_msrs_global.nr = slot + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 347) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 348) EXPORT_SYMBOL_GPL(kvm_define_user_return_msr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 349)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 350) static void kvm_user_return_msr_cpu_online(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 351) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 352) unsigned int cpu = smp_processor_id();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 353) struct kvm_user_return_msrs *msrs = per_cpu_ptr(user_return_msrs, cpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 354) u64 value;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 355) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 356)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 357) for (i = 0; i < user_return_msrs_global.nr; ++i) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 358) rdmsrl_safe(user_return_msrs_global.msrs[i], &value);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 359) msrs->values[i].host = value;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 360) msrs->values[i].curr = value;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 361) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 362) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 363)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 364) int kvm_set_user_return_msr(unsigned slot, u64 value, u64 mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 365) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 366) unsigned int cpu = smp_processor_id();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 367) struct kvm_user_return_msrs *msrs = per_cpu_ptr(user_return_msrs, cpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 368) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 369)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 370) value = (value & mask) | (msrs->values[slot].host & ~mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 371) if (value == msrs->values[slot].curr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 372) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 373) err = wrmsrl_safe(user_return_msrs_global.msrs[slot], value);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 374) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 375) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 376)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 377) msrs->values[slot].curr = value;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 378) if (!msrs->registered) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 379) msrs->urn.on_user_return = kvm_on_user_return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 380) user_return_notifier_register(&msrs->urn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 381) msrs->registered = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 382) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 383) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 384) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 385) EXPORT_SYMBOL_GPL(kvm_set_user_return_msr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 386)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 387) static void drop_user_return_notifiers(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 388) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 389) unsigned int cpu = smp_processor_id();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 390) struct kvm_user_return_msrs *msrs = per_cpu_ptr(user_return_msrs, cpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 391)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 392) if (msrs->registered)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 393) kvm_on_user_return(&msrs->urn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 394) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 395)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 396) u64 kvm_get_apic_base(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 397) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 398) return vcpu->arch.apic_base;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 399) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 400) EXPORT_SYMBOL_GPL(kvm_get_apic_base);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 401)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 402) enum lapic_mode kvm_get_apic_mode(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 403) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 404) return kvm_apic_mode(kvm_get_apic_base(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 405) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 406) EXPORT_SYMBOL_GPL(kvm_get_apic_mode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 407)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 408) int kvm_set_apic_base(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 409) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 410) enum lapic_mode old_mode = kvm_get_apic_mode(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 411) enum lapic_mode new_mode = kvm_apic_mode(msr_info->data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 412) u64 reserved_bits = ((~0ULL) << cpuid_maxphyaddr(vcpu)) | 0x2ff |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 413) (guest_cpuid_has(vcpu, X86_FEATURE_X2APIC) ? 0 : X2APIC_ENABLE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 414)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 415) if ((msr_info->data & reserved_bits) != 0 || new_mode == LAPIC_MODE_INVALID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 416) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 417) if (!msr_info->host_initiated) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 418) if (old_mode == LAPIC_MODE_X2APIC && new_mode == LAPIC_MODE_XAPIC)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 419) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 420) if (old_mode == LAPIC_MODE_DISABLED && new_mode == LAPIC_MODE_X2APIC)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 421) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 422) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 423)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 424) kvm_lapic_set_base(vcpu, msr_info->data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 425) kvm_recalculate_apic_map(vcpu->kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 426) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 427) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 428) EXPORT_SYMBOL_GPL(kvm_set_apic_base);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 429)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 430) asmlinkage __visible noinstr void kvm_spurious_fault(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 431) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 432) /* Fault while not rebooting. We want the trace. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 433) BUG_ON(!kvm_rebooting);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 434) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 435) EXPORT_SYMBOL_GPL(kvm_spurious_fault);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 436)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 437) #define EXCPT_BENIGN 0
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 438) #define EXCPT_CONTRIBUTORY 1
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 439) #define EXCPT_PF 2
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 440)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 441) static int exception_class(int vector)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 442) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 443) switch (vector) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 444) case PF_VECTOR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 445) return EXCPT_PF;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 446) case DE_VECTOR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 447) case TS_VECTOR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 448) case NP_VECTOR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 449) case SS_VECTOR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 450) case GP_VECTOR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 451) return EXCPT_CONTRIBUTORY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 452) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 453) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 454) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 455) return EXCPT_BENIGN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 456) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 457)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 458) #define EXCPT_FAULT 0
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 459) #define EXCPT_TRAP 1
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 460) #define EXCPT_ABORT 2
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 461) #define EXCPT_INTERRUPT 3
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 462)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 463) static int exception_type(int vector)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 464) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 465) unsigned int mask;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 466)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 467) if (WARN_ON(vector > 31 || vector == NMI_VECTOR))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 468) return EXCPT_INTERRUPT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 469)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 470) mask = 1 << vector;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 471)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 472) /* #DB is trap, as instruction watchpoints are handled elsewhere */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 473) if (mask & ((1 << DB_VECTOR) | (1 << BP_VECTOR) | (1 << OF_VECTOR)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 474) return EXCPT_TRAP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 475)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 476) if (mask & ((1 << DF_VECTOR) | (1 << MC_VECTOR)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 477) return EXCPT_ABORT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 478)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 479) /* Reserved exceptions will result in fault */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 480) return EXCPT_FAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 481) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 482)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 483) void kvm_deliver_exception_payload(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 484) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 485) unsigned nr = vcpu->arch.exception.nr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 486) bool has_payload = vcpu->arch.exception.has_payload;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 487) unsigned long payload = vcpu->arch.exception.payload;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 488)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 489) if (!has_payload)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 490) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 491)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 492) switch (nr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 493) case DB_VECTOR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 494) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 495) * "Certain debug exceptions may clear bit 0-3. The
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 496) * remaining contents of the DR6 register are never
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 497) * cleared by the processor".
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 498) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 499) vcpu->arch.dr6 &= ~DR_TRAP_BITS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 500) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 501) * DR6.RTM is set by all #DB exceptions that don't clear it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 502) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 503) vcpu->arch.dr6 |= DR6_RTM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 504) vcpu->arch.dr6 |= payload;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 505) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 506) * Bit 16 should be set in the payload whenever the #DB
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 507) * exception should clear DR6.RTM. This makes the payload
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 508) * compatible with the pending debug exceptions under VMX.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 509) * Though not currently documented in the SDM, this also
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 510) * makes the payload compatible with the exit qualification
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 511) * for #DB exceptions under VMX.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 512) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 513) vcpu->arch.dr6 ^= payload & DR6_RTM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 514)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 515) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 516) * The #DB payload is defined as compatible with the 'pending
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 517) * debug exceptions' field under VMX, not DR6. While bit 12 is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 518) * defined in the 'pending debug exceptions' field (enabled
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 519) * breakpoint), it is reserved and must be zero in DR6.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 520) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 521) vcpu->arch.dr6 &= ~BIT(12);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 522) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 523) case PF_VECTOR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 524) vcpu->arch.cr2 = payload;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 525) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 526) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 527)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 528) vcpu->arch.exception.has_payload = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 529) vcpu->arch.exception.payload = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 530) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 531) EXPORT_SYMBOL_GPL(kvm_deliver_exception_payload);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 532)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 533) static void kvm_multiple_exception(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 534) unsigned nr, bool has_error, u32 error_code,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 535) bool has_payload, unsigned long payload, bool reinject)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 536) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 537) u32 prev_nr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 538) int class1, class2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 539)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 540) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 541)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 542) if (!vcpu->arch.exception.pending && !vcpu->arch.exception.injected) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 543) queue:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 544) if (reinject) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 545) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 546) * On vmentry, vcpu->arch.exception.pending is only
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 547) * true if an event injection was blocked by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 548) * nested_run_pending. In that case, however,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 549) * vcpu_enter_guest requests an immediate exit,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 550) * and the guest shouldn't proceed far enough to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 551) * need reinjection.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 552) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 553) WARN_ON_ONCE(vcpu->arch.exception.pending);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 554) vcpu->arch.exception.injected = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 555) if (WARN_ON_ONCE(has_payload)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 556) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 557) * A reinjected event has already
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 558) * delivered its payload.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 559) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 560) has_payload = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 561) payload = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 562) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 563) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 564) vcpu->arch.exception.pending = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 565) vcpu->arch.exception.injected = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 566) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 567) vcpu->arch.exception.has_error_code = has_error;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 568) vcpu->arch.exception.nr = nr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 569) vcpu->arch.exception.error_code = error_code;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 570) vcpu->arch.exception.has_payload = has_payload;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 571) vcpu->arch.exception.payload = payload;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 572) if (!is_guest_mode(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 573) kvm_deliver_exception_payload(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 574) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 575) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 576)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 577) /* to check exception */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 578) prev_nr = vcpu->arch.exception.nr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 579) if (prev_nr == DF_VECTOR) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 580) /* triple fault -> shutdown */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 581) kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 582) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 583) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 584) class1 = exception_class(prev_nr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 585) class2 = exception_class(nr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 586) if ((class1 == EXCPT_CONTRIBUTORY && class2 == EXCPT_CONTRIBUTORY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 587) || (class1 == EXCPT_PF && class2 != EXCPT_BENIGN)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 588) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 589) * Generate double fault per SDM Table 5-5. Set
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 590) * exception.pending = true so that the double fault
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 591) * can trigger a nested vmexit.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 592) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 593) vcpu->arch.exception.pending = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 594) vcpu->arch.exception.injected = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 595) vcpu->arch.exception.has_error_code = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 596) vcpu->arch.exception.nr = DF_VECTOR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 597) vcpu->arch.exception.error_code = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 598) vcpu->arch.exception.has_payload = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 599) vcpu->arch.exception.payload = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 600) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 601) /* replace previous exception with a new one in a hope
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 602) that instruction re-execution will regenerate lost
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 603) exception */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 604) goto queue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 605) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 606)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 607) void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 608) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 609) kvm_multiple_exception(vcpu, nr, false, 0, false, 0, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 610) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 611) EXPORT_SYMBOL_GPL(kvm_queue_exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 612)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 613) void kvm_requeue_exception(struct kvm_vcpu *vcpu, unsigned nr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 614) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 615) kvm_multiple_exception(vcpu, nr, false, 0, false, 0, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 616) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 617) EXPORT_SYMBOL_GPL(kvm_requeue_exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 618)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 619) void kvm_queue_exception_p(struct kvm_vcpu *vcpu, unsigned nr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 620) unsigned long payload)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 621) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 622) kvm_multiple_exception(vcpu, nr, false, 0, true, payload, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 623) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 624) EXPORT_SYMBOL_GPL(kvm_queue_exception_p);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 625)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 626) static void kvm_queue_exception_e_p(struct kvm_vcpu *vcpu, unsigned nr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 627) u32 error_code, unsigned long payload)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 628) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 629) kvm_multiple_exception(vcpu, nr, true, error_code,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 630) true, payload, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 631) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 632)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 633) int kvm_complete_insn_gp(struct kvm_vcpu *vcpu, int err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 634) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 635) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 636) kvm_inject_gp(vcpu, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 637) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 638) return kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 639)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 640) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 641) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 642) EXPORT_SYMBOL_GPL(kvm_complete_insn_gp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 643)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 644) void kvm_inject_page_fault(struct kvm_vcpu *vcpu, struct x86_exception *fault)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 645) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 646) ++vcpu->stat.pf_guest;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 647) vcpu->arch.exception.nested_apf =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 648) is_guest_mode(vcpu) && fault->async_page_fault;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 649) if (vcpu->arch.exception.nested_apf) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 650) vcpu->arch.apf.nested_apf_token = fault->address;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 651) kvm_queue_exception_e(vcpu, PF_VECTOR, fault->error_code);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 652) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 653) kvm_queue_exception_e_p(vcpu, PF_VECTOR, fault->error_code,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 654) fault->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 655) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 656) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 657) EXPORT_SYMBOL_GPL(kvm_inject_page_fault);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 658)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 659) bool kvm_inject_emulated_page_fault(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 660) struct x86_exception *fault)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 661) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 662) struct kvm_mmu *fault_mmu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 663) WARN_ON_ONCE(fault->vector != PF_VECTOR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 664)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 665) fault_mmu = fault->nested_page_fault ? vcpu->arch.mmu :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 666) vcpu->arch.walk_mmu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 667)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 668) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 669) * Invalidate the TLB entry for the faulting address, if it exists,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 670) * else the access will fault indefinitely (and to emulate hardware).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 671) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 672) if ((fault->error_code & PFERR_PRESENT_MASK) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 673) !(fault->error_code & PFERR_RSVD_MASK))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 674) kvm_mmu_invalidate_gva(vcpu, fault_mmu, fault->address,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 675) fault_mmu->root_hpa);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 676)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 677) fault_mmu->inject_page_fault(vcpu, fault);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 678) return fault->nested_page_fault;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 679) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 680) EXPORT_SYMBOL_GPL(kvm_inject_emulated_page_fault);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 681)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 682) void kvm_inject_nmi(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 683) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 684) atomic_inc(&vcpu->arch.nmi_queued);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 685) kvm_make_request(KVM_REQ_NMI, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 686) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 687) EXPORT_SYMBOL_GPL(kvm_inject_nmi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 688)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 689) void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 690) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 691) kvm_multiple_exception(vcpu, nr, true, error_code, false, 0, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 692) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 693) EXPORT_SYMBOL_GPL(kvm_queue_exception_e);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 694)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 695) void kvm_requeue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 696) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 697) kvm_multiple_exception(vcpu, nr, true, error_code, false, 0, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 698) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 699) EXPORT_SYMBOL_GPL(kvm_requeue_exception_e);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 700)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 701) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 702) * Checks if cpl <= required_cpl; if true, return true. Otherwise queue
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 703) * a #GP and return false.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 704) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 705) bool kvm_require_cpl(struct kvm_vcpu *vcpu, int required_cpl)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 706) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 707) if (kvm_x86_ops.get_cpl(vcpu) <= required_cpl)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 708) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 709) kvm_queue_exception_e(vcpu, GP_VECTOR, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 710) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 711) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 712) EXPORT_SYMBOL_GPL(kvm_require_cpl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 713)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 714) bool kvm_require_dr(struct kvm_vcpu *vcpu, int dr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 715) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 716) if ((dr != 4 && dr != 5) || !kvm_read_cr4_bits(vcpu, X86_CR4_DE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 717) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 718)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 719) kvm_queue_exception(vcpu, UD_VECTOR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 720) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 721) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 722) EXPORT_SYMBOL_GPL(kvm_require_dr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 723)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 724) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 725) * This function will be used to read from the physical memory of the currently
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 726) * running guest. The difference to kvm_vcpu_read_guest_page is that this function
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 727) * can read from guest physical or from the guest's guest physical memory.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 728) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 729) int kvm_read_guest_page_mmu(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 730) gfn_t ngfn, void *data, int offset, int len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 731) u32 access)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 732) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 733) struct x86_exception exception;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 734) gfn_t real_gfn;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 735) gpa_t ngpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 736)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 737) ngpa = gfn_to_gpa(ngfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 738) real_gfn = mmu->translate_gpa(vcpu, ngpa, access, &exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 739) if (real_gfn == UNMAPPED_GVA)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 740) return -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 741)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 742) real_gfn = gpa_to_gfn(real_gfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 743)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 744) return kvm_vcpu_read_guest_page(vcpu, real_gfn, data, offset, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 745) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 746) EXPORT_SYMBOL_GPL(kvm_read_guest_page_mmu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 747)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 748) static int kvm_read_nested_guest_page(struct kvm_vcpu *vcpu, gfn_t gfn,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 749) void *data, int offset, int len, u32 access)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 750) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 751) return kvm_read_guest_page_mmu(vcpu, vcpu->arch.walk_mmu, gfn,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 752) data, offset, len, access);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 753) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 754)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 755) static inline u64 pdptr_rsvd_bits(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 756) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 757) return rsvd_bits(cpuid_maxphyaddr(vcpu), 63) | rsvd_bits(5, 8) |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 758) rsvd_bits(1, 2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 759) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 760)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 761) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 762) * Load the pae pdptrs. Return 1 if they are all valid, 0 otherwise.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 763) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 764) int load_pdptrs(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, unsigned long cr3)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 765) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 766) gfn_t pdpt_gfn = cr3 >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 767) unsigned offset = ((cr3 & (PAGE_SIZE-1)) >> 5) << 2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 768) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 769) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 770) u64 pdpte[ARRAY_SIZE(mmu->pdptrs)];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 771)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 772) ret = kvm_read_guest_page_mmu(vcpu, mmu, pdpt_gfn, pdpte,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 773) offset * sizeof(u64), sizeof(pdpte),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 774) PFERR_USER_MASK|PFERR_WRITE_MASK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 775) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 776) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 777) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 778) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 779) for (i = 0; i < ARRAY_SIZE(pdpte); ++i) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 780) if ((pdpte[i] & PT_PRESENT_MASK) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 781) (pdpte[i] & pdptr_rsvd_bits(vcpu))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 782) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 783) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 784) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 785) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 786) ret = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 787)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 788) memcpy(mmu->pdptrs, pdpte, sizeof(mmu->pdptrs));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 789) kvm_register_mark_dirty(vcpu, VCPU_EXREG_PDPTR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 790)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 791) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 792)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 793) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 794) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 795) EXPORT_SYMBOL_GPL(load_pdptrs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 796)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 797) bool pdptrs_changed(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 798) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 799) u64 pdpte[ARRAY_SIZE(vcpu->arch.walk_mmu->pdptrs)];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 800) int offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 801) gfn_t gfn;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 802) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 803)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 804) if (!is_pae_paging(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 805) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 806)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 807) if (!kvm_register_is_available(vcpu, VCPU_EXREG_PDPTR))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 808) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 809)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 810) gfn = (kvm_read_cr3(vcpu) & 0xffffffe0ul) >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 811) offset = (kvm_read_cr3(vcpu) & 0xffffffe0ul) & (PAGE_SIZE - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 812) r = kvm_read_nested_guest_page(vcpu, gfn, pdpte, offset, sizeof(pdpte),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 813) PFERR_USER_MASK | PFERR_WRITE_MASK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 814) if (r < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 815) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 816)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 817) return memcmp(pdpte, vcpu->arch.walk_mmu->pdptrs, sizeof(pdpte)) != 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 818) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 819) EXPORT_SYMBOL_GPL(pdptrs_changed);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 820)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 821) int kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 822) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 823) unsigned long old_cr0 = kvm_read_cr0(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 824) unsigned long pdptr_bits = X86_CR0_CD | X86_CR0_NW | X86_CR0_PG;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 825) unsigned long update_bits = X86_CR0_PG | X86_CR0_WP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 826)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 827) cr0 |= X86_CR0_ET;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 828)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 829) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 830) if (cr0 & 0xffffffff00000000UL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 831) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 832) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 833)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 834) cr0 &= ~CR0_RESERVED_BITS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 835)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 836) if ((cr0 & X86_CR0_NW) && !(cr0 & X86_CR0_CD))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 837) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 838)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 839) if ((cr0 & X86_CR0_PG) && !(cr0 & X86_CR0_PE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 840) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 841)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 842) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 843) if ((vcpu->arch.efer & EFER_LME) && !is_paging(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 844) (cr0 & X86_CR0_PG)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 845) int cs_db, cs_l;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 846)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 847) if (!is_pae(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 848) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 849) kvm_x86_ops.get_cs_db_l_bits(vcpu, &cs_db, &cs_l);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 850) if (cs_l)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 851) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 852) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 853) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 854) if (!(vcpu->arch.efer & EFER_LME) && (cr0 & X86_CR0_PG) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 855) is_pae(vcpu) && ((cr0 ^ old_cr0) & pdptr_bits) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 856) !load_pdptrs(vcpu, vcpu->arch.walk_mmu, kvm_read_cr3(vcpu)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 857) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 858)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 859) if (!(cr0 & X86_CR0_PG) && kvm_read_cr4_bits(vcpu, X86_CR4_PCIDE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 860) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 861)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 862) kvm_x86_ops.set_cr0(vcpu, cr0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 863)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 864) if ((cr0 ^ old_cr0) & X86_CR0_PG) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 865) kvm_clear_async_pf_completion_queue(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 866) kvm_async_pf_hash_reset(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 867) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 868)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 869) if ((cr0 ^ old_cr0) & update_bits)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 870) kvm_mmu_reset_context(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 871)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 872) if (((cr0 ^ old_cr0) & X86_CR0_CD) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 873) kvm_arch_has_noncoherent_dma(vcpu->kvm) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 874) !kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_CD_NW_CLEARED))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 875) kvm_zap_gfn_range(vcpu->kvm, 0, ~0ULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 876)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 877) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 878) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 879) EXPORT_SYMBOL_GPL(kvm_set_cr0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 880)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 881) void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 882) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 883) (void)kvm_set_cr0(vcpu, kvm_read_cr0_bits(vcpu, ~0x0eul) | (msw & 0x0f));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 884) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 885) EXPORT_SYMBOL_GPL(kvm_lmsw);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 886)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 887) void kvm_load_guest_xsave_state(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 888) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 889) if (kvm_read_cr4_bits(vcpu, X86_CR4_OSXSAVE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 890)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 891) if (vcpu->arch.xcr0 != host_xcr0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 892) xsetbv(XCR_XFEATURE_ENABLED_MASK, vcpu->arch.xcr0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 893)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 894) if (vcpu->arch.xsaves_enabled &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 895) vcpu->arch.ia32_xss != host_xss)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 896) wrmsrl(MSR_IA32_XSS, vcpu->arch.ia32_xss);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 897) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 898)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 899) if (static_cpu_has(X86_FEATURE_PKU) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 900) (kvm_read_cr4_bits(vcpu, X86_CR4_PKE) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 901) (vcpu->arch.xcr0 & XFEATURE_MASK_PKRU)) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 902) vcpu->arch.pkru != vcpu->arch.host_pkru)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 903) __write_pkru(vcpu->arch.pkru);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 904) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 905) EXPORT_SYMBOL_GPL(kvm_load_guest_xsave_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 906)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 907) void kvm_load_host_xsave_state(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 908) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 909) if (static_cpu_has(X86_FEATURE_PKU) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 910) (kvm_read_cr4_bits(vcpu, X86_CR4_PKE) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 911) (vcpu->arch.xcr0 & XFEATURE_MASK_PKRU))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 912) vcpu->arch.pkru = rdpkru();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 913) if (vcpu->arch.pkru != vcpu->arch.host_pkru)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 914) __write_pkru(vcpu->arch.host_pkru);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 915) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 916)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 917) if (kvm_read_cr4_bits(vcpu, X86_CR4_OSXSAVE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 918)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 919) if (vcpu->arch.xcr0 != host_xcr0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 920) xsetbv(XCR_XFEATURE_ENABLED_MASK, host_xcr0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 921)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 922) if (vcpu->arch.xsaves_enabled &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 923) vcpu->arch.ia32_xss != host_xss)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 924) wrmsrl(MSR_IA32_XSS, host_xss);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 925) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 926)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 927) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 928) EXPORT_SYMBOL_GPL(kvm_load_host_xsave_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 929)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 930) static int __kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 931) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 932) u64 xcr0 = xcr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 933) u64 old_xcr0 = vcpu->arch.xcr0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 934) u64 valid_bits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 935)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 936) /* Only support XCR_XFEATURE_ENABLED_MASK(xcr0) now */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 937) if (index != XCR_XFEATURE_ENABLED_MASK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 938) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 939) if (!(xcr0 & XFEATURE_MASK_FP))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 940) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 941) if ((xcr0 & XFEATURE_MASK_YMM) && !(xcr0 & XFEATURE_MASK_SSE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 942) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 943)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 944) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 945) * Do not allow the guest to set bits that we do not support
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 946) * saving. However, xcr0 bit 0 is always set, even if the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 947) * emulated CPU does not support XSAVE (see fx_init).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 948) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 949) valid_bits = vcpu->arch.guest_supported_xcr0 | XFEATURE_MASK_FP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 950) if (xcr0 & ~valid_bits)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 951) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 952)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 953) if ((!(xcr0 & XFEATURE_MASK_BNDREGS)) !=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 954) (!(xcr0 & XFEATURE_MASK_BNDCSR)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 955) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 956)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 957) if (xcr0 & XFEATURE_MASK_AVX512) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 958) if (!(xcr0 & XFEATURE_MASK_YMM))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 959) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 960) if ((xcr0 & XFEATURE_MASK_AVX512) != XFEATURE_MASK_AVX512)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 961) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 962) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 963) vcpu->arch.xcr0 = xcr0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 964)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 965) if ((xcr0 ^ old_xcr0) & XFEATURE_MASK_EXTEND)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 966) kvm_update_cpuid_runtime(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 967) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 968) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 969)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 970) int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 971) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 972) if (kvm_x86_ops.get_cpl(vcpu) != 0 ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 973) __kvm_set_xcr(vcpu, index, xcr)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 974) kvm_inject_gp(vcpu, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 975) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 976) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 977) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 978) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 979) EXPORT_SYMBOL_GPL(kvm_set_xcr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 980)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 981) int kvm_valid_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 982) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 983) if (cr4 & cr4_reserved_bits)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 984) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 985)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 986) if (cr4 & vcpu->arch.cr4_guest_rsvd_bits)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 987) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 988)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 989) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 990) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 991) EXPORT_SYMBOL_GPL(kvm_valid_cr4);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 992)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 993) int kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 994) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 995) unsigned long old_cr4 = kvm_read_cr4(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 996) unsigned long pdptr_bits = X86_CR4_PGE | X86_CR4_PSE | X86_CR4_PAE |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 997) X86_CR4_SMEP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 998) unsigned long mmu_role_bits = pdptr_bits | X86_CR4_SMAP | X86_CR4_PKE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 999)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1000) if (kvm_valid_cr4(vcpu, cr4))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1001) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1002)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1003) if (is_long_mode(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1004) if (!(cr4 & X86_CR4_PAE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1005) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1006) if ((cr4 ^ old_cr4) & X86_CR4_LA57)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1007) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1008) } else if (is_paging(vcpu) && (cr4 & X86_CR4_PAE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1009) && ((cr4 ^ old_cr4) & pdptr_bits)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1010) && !load_pdptrs(vcpu, vcpu->arch.walk_mmu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1011) kvm_read_cr3(vcpu)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1012) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1013)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1014) if ((cr4 & X86_CR4_PCIDE) && !(old_cr4 & X86_CR4_PCIDE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1015) if (!guest_cpuid_has(vcpu, X86_FEATURE_PCID))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1016) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1017)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1018) /* PCID can not be enabled when cr3[11:0]!=000H or EFER.LMA=0 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1019) if ((kvm_read_cr3(vcpu) & X86_CR3_PCID_MASK) || !is_long_mode(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1020) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1021) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1022)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1023) if (kvm_x86_ops.set_cr4(vcpu, cr4))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1024) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1025)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1026) if (((cr4 ^ old_cr4) & mmu_role_bits) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1027) (!(cr4 & X86_CR4_PCIDE) && (old_cr4 & X86_CR4_PCIDE)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1028) kvm_mmu_reset_context(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1029)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1030) if ((cr4 ^ old_cr4) & (X86_CR4_OSXSAVE | X86_CR4_PKE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1031) kvm_update_cpuid_runtime(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1032)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1033) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1034) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1035) EXPORT_SYMBOL_GPL(kvm_set_cr4);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1036)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1037) int kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1038) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1039) bool skip_tlb_flush = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1040) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1041) bool pcid_enabled = kvm_read_cr4_bits(vcpu, X86_CR4_PCIDE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1042)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1043) if (pcid_enabled) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1044) skip_tlb_flush = cr3 & X86_CR3_PCID_NOFLUSH;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1045) cr3 &= ~X86_CR3_PCID_NOFLUSH;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1046) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1047) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1048)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1049) if (cr3 == kvm_read_cr3(vcpu) && !pdptrs_changed(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1050) if (!skip_tlb_flush) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1051) kvm_mmu_sync_roots(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1052) kvm_make_request(KVM_REQ_TLB_FLUSH_CURRENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1053) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1054) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1055) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1056)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1057) if (is_long_mode(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1058) (cr3 & vcpu->arch.cr3_lm_rsvd_bits))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1059) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1060) else if (is_pae_paging(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1061) !load_pdptrs(vcpu, vcpu->arch.walk_mmu, cr3))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1062) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1063)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1064) kvm_mmu_new_pgd(vcpu, cr3, skip_tlb_flush, skip_tlb_flush);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1065) vcpu->arch.cr3 = cr3;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1066) kvm_register_mark_available(vcpu, VCPU_EXREG_CR3);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1067)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1068) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1069) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1070) EXPORT_SYMBOL_GPL(kvm_set_cr3);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1071)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1072) int kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1073) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1074) if (cr8 & CR8_RESERVED_BITS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1075) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1076) if (lapic_in_kernel(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1077) kvm_lapic_set_tpr(vcpu, cr8);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1078) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1079) vcpu->arch.cr8 = cr8;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1080) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1081) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1082) EXPORT_SYMBOL_GPL(kvm_set_cr8);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1083)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1084) unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1085) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1086) if (lapic_in_kernel(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1087) return kvm_lapic_get_cr8(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1088) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1089) return vcpu->arch.cr8;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1090) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1091) EXPORT_SYMBOL_GPL(kvm_get_cr8);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1092)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1093) static void kvm_update_dr0123(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1094) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1095) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1096)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1097) if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1098) for (i = 0; i < KVM_NR_DB_REGS; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1099) vcpu->arch.eff_db[i] = vcpu->arch.db[i];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1100) vcpu->arch.switch_db_regs |= KVM_DEBUGREG_RELOAD;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1101) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1102) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1103)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1104) void kvm_update_dr7(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1105) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1106) unsigned long dr7;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1107)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1108) if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1109) dr7 = vcpu->arch.guest_debug_dr7;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1110) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1111) dr7 = vcpu->arch.dr7;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1112) kvm_x86_ops.set_dr7(vcpu, dr7);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1113) vcpu->arch.switch_db_regs &= ~KVM_DEBUGREG_BP_ENABLED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1114) if (dr7 & DR7_BP_EN_MASK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1115) vcpu->arch.switch_db_regs |= KVM_DEBUGREG_BP_ENABLED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1116) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1117) EXPORT_SYMBOL_GPL(kvm_update_dr7);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1118)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1119) static u64 kvm_dr6_fixed(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1120) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1121) u64 fixed = DR6_FIXED_1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1122)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1123) if (!guest_cpuid_has(vcpu, X86_FEATURE_RTM))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1124) fixed |= DR6_RTM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1125) return fixed;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1126) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1127)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1128) static int __kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1129) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1130) size_t size = ARRAY_SIZE(vcpu->arch.db);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1131)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1132) switch (dr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1133) case 0 ... 3:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1134) vcpu->arch.db[array_index_nospec(dr, size)] = val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1135) if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1136) vcpu->arch.eff_db[dr] = val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1137) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1138) case 4:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1139) case 6:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1140) if (!kvm_dr6_valid(val))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1141) return -1; /* #GP */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1142) vcpu->arch.dr6 = (val & DR6_VOLATILE) | kvm_dr6_fixed(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1143) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1144) case 5:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1145) default: /* 7 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1146) if (!kvm_dr7_valid(val))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1147) return -1; /* #GP */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1148) vcpu->arch.dr7 = (val & DR7_VOLATILE) | DR7_FIXED_1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1149) kvm_update_dr7(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1150) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1151) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1152)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1153) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1154) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1155)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1156) int kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1157) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1158) if (__kvm_set_dr(vcpu, dr, val)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1159) kvm_inject_gp(vcpu, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1160) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1161) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1162) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1163) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1164) EXPORT_SYMBOL_GPL(kvm_set_dr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1165)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1166) int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1167) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1168) size_t size = ARRAY_SIZE(vcpu->arch.db);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1169)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1170) switch (dr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1171) case 0 ... 3:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1172) *val = vcpu->arch.db[array_index_nospec(dr, size)];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1173) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1174) case 4:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1175) case 6:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1176) *val = vcpu->arch.dr6;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1177) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1178) case 5:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1179) default: /* 7 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1180) *val = vcpu->arch.dr7;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1181) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1182) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1183) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1184) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1185) EXPORT_SYMBOL_GPL(kvm_get_dr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1186)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1187) bool kvm_rdpmc(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1188) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1189) u32 ecx = kvm_rcx_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1190) u64 data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1191) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1192)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1193) err = kvm_pmu_rdpmc(vcpu, ecx, &data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1194) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1195) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1196) kvm_rax_write(vcpu, (u32)data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1197) kvm_rdx_write(vcpu, data >> 32);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1198) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1199) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1200) EXPORT_SYMBOL_GPL(kvm_rdpmc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1201)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1202) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1203) * List of msr numbers which we expose to userspace through KVM_GET_MSRS
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1204) * and KVM_SET_MSRS, and KVM_GET_MSR_INDEX_LIST.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1205) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1206) * The three MSR lists(msrs_to_save, emulated_msrs, msr_based_features)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1207) * extract the supported MSRs from the related const lists.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1208) * msrs_to_save is selected from the msrs_to_save_all to reflect the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1209) * capabilities of the host cpu. This capabilities test skips MSRs that are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1210) * kvm-specific. Those are put in emulated_msrs_all; filtering of emulated_msrs
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1211) * may depend on host virtualization features rather than host cpu features.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1212) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1213)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1214) static const u32 msrs_to_save_all[] = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1215) MSR_IA32_SYSENTER_CS, MSR_IA32_SYSENTER_ESP, MSR_IA32_SYSENTER_EIP,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1216) MSR_STAR,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1217) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1218) MSR_CSTAR, MSR_KERNEL_GS_BASE, MSR_SYSCALL_MASK, MSR_LSTAR,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1219) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1220) MSR_IA32_TSC, MSR_IA32_CR_PAT, MSR_VM_HSAVE_PA,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1221) MSR_IA32_FEAT_CTL, MSR_IA32_BNDCFGS, MSR_TSC_AUX,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1222) MSR_IA32_SPEC_CTRL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1223) MSR_IA32_RTIT_CTL, MSR_IA32_RTIT_STATUS, MSR_IA32_RTIT_CR3_MATCH,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1224) MSR_IA32_RTIT_OUTPUT_BASE, MSR_IA32_RTIT_OUTPUT_MASK,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1225) MSR_IA32_RTIT_ADDR0_A, MSR_IA32_RTIT_ADDR0_B,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1226) MSR_IA32_RTIT_ADDR1_A, MSR_IA32_RTIT_ADDR1_B,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1227) MSR_IA32_RTIT_ADDR2_A, MSR_IA32_RTIT_ADDR2_B,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1228) MSR_IA32_RTIT_ADDR3_A, MSR_IA32_RTIT_ADDR3_B,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1229) MSR_IA32_UMWAIT_CONTROL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1230)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1231) MSR_ARCH_PERFMON_FIXED_CTR0, MSR_ARCH_PERFMON_FIXED_CTR1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1232) MSR_ARCH_PERFMON_FIXED_CTR0 + 2,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1233) MSR_CORE_PERF_FIXED_CTR_CTRL, MSR_CORE_PERF_GLOBAL_STATUS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1234) MSR_CORE_PERF_GLOBAL_CTRL, MSR_CORE_PERF_GLOBAL_OVF_CTRL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1235) MSR_ARCH_PERFMON_PERFCTR0, MSR_ARCH_PERFMON_PERFCTR1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1236) MSR_ARCH_PERFMON_PERFCTR0 + 2, MSR_ARCH_PERFMON_PERFCTR0 + 3,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1237) MSR_ARCH_PERFMON_PERFCTR0 + 4, MSR_ARCH_PERFMON_PERFCTR0 + 5,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1238) MSR_ARCH_PERFMON_PERFCTR0 + 6, MSR_ARCH_PERFMON_PERFCTR0 + 7,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1239) MSR_ARCH_PERFMON_PERFCTR0 + 8, MSR_ARCH_PERFMON_PERFCTR0 + 9,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1240) MSR_ARCH_PERFMON_PERFCTR0 + 10, MSR_ARCH_PERFMON_PERFCTR0 + 11,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1241) MSR_ARCH_PERFMON_PERFCTR0 + 12, MSR_ARCH_PERFMON_PERFCTR0 + 13,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1242) MSR_ARCH_PERFMON_PERFCTR0 + 14, MSR_ARCH_PERFMON_PERFCTR0 + 15,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1243) MSR_ARCH_PERFMON_PERFCTR0 + 16, MSR_ARCH_PERFMON_PERFCTR0 + 17,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1244) MSR_ARCH_PERFMON_EVENTSEL0, MSR_ARCH_PERFMON_EVENTSEL1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1245) MSR_ARCH_PERFMON_EVENTSEL0 + 2, MSR_ARCH_PERFMON_EVENTSEL0 + 3,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1246) MSR_ARCH_PERFMON_EVENTSEL0 + 4, MSR_ARCH_PERFMON_EVENTSEL0 + 5,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1247) MSR_ARCH_PERFMON_EVENTSEL0 + 6, MSR_ARCH_PERFMON_EVENTSEL0 + 7,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1248) MSR_ARCH_PERFMON_EVENTSEL0 + 8, MSR_ARCH_PERFMON_EVENTSEL0 + 9,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1249) MSR_ARCH_PERFMON_EVENTSEL0 + 10, MSR_ARCH_PERFMON_EVENTSEL0 + 11,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1250) MSR_ARCH_PERFMON_EVENTSEL0 + 12, MSR_ARCH_PERFMON_EVENTSEL0 + 13,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1251) MSR_ARCH_PERFMON_EVENTSEL0 + 14, MSR_ARCH_PERFMON_EVENTSEL0 + 15,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1252) MSR_ARCH_PERFMON_EVENTSEL0 + 16, MSR_ARCH_PERFMON_EVENTSEL0 + 17,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1253)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1254) MSR_K7_EVNTSEL0, MSR_K7_EVNTSEL1, MSR_K7_EVNTSEL2, MSR_K7_EVNTSEL3,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1255) MSR_K7_PERFCTR0, MSR_K7_PERFCTR1, MSR_K7_PERFCTR2, MSR_K7_PERFCTR3,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1256) MSR_F15H_PERF_CTL0, MSR_F15H_PERF_CTL1, MSR_F15H_PERF_CTL2,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1257) MSR_F15H_PERF_CTL3, MSR_F15H_PERF_CTL4, MSR_F15H_PERF_CTL5,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1258) MSR_F15H_PERF_CTR0, MSR_F15H_PERF_CTR1, MSR_F15H_PERF_CTR2,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1259) MSR_F15H_PERF_CTR3, MSR_F15H_PERF_CTR4, MSR_F15H_PERF_CTR5,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1260) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1261)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1262) static u32 msrs_to_save[ARRAY_SIZE(msrs_to_save_all)];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1263) static unsigned num_msrs_to_save;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1264)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1265) static const u32 emulated_msrs_all[] = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1266) MSR_KVM_SYSTEM_TIME, MSR_KVM_WALL_CLOCK,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1267) MSR_KVM_SYSTEM_TIME_NEW, MSR_KVM_WALL_CLOCK_NEW,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1268) HV_X64_MSR_GUEST_OS_ID, HV_X64_MSR_HYPERCALL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1269) HV_X64_MSR_TIME_REF_COUNT, HV_X64_MSR_REFERENCE_TSC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1270) HV_X64_MSR_TSC_FREQUENCY, HV_X64_MSR_APIC_FREQUENCY,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1271) HV_X64_MSR_CRASH_P0, HV_X64_MSR_CRASH_P1, HV_X64_MSR_CRASH_P2,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1272) HV_X64_MSR_CRASH_P3, HV_X64_MSR_CRASH_P4, HV_X64_MSR_CRASH_CTL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1273) HV_X64_MSR_RESET,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1274) HV_X64_MSR_VP_INDEX,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1275) HV_X64_MSR_VP_RUNTIME,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1276) HV_X64_MSR_SCONTROL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1277) HV_X64_MSR_STIMER0_CONFIG,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1278) HV_X64_MSR_VP_ASSIST_PAGE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1279) HV_X64_MSR_REENLIGHTENMENT_CONTROL, HV_X64_MSR_TSC_EMULATION_CONTROL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1280) HV_X64_MSR_TSC_EMULATION_STATUS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1281) HV_X64_MSR_SYNDBG_OPTIONS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1282) HV_X64_MSR_SYNDBG_CONTROL, HV_X64_MSR_SYNDBG_STATUS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1283) HV_X64_MSR_SYNDBG_SEND_BUFFER, HV_X64_MSR_SYNDBG_RECV_BUFFER,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1284) HV_X64_MSR_SYNDBG_PENDING_BUFFER,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1285)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1286) MSR_KVM_ASYNC_PF_EN, MSR_KVM_STEAL_TIME,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1287) MSR_KVM_PV_EOI_EN, MSR_KVM_ASYNC_PF_INT, MSR_KVM_ASYNC_PF_ACK,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1288)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1289) MSR_IA32_TSC_ADJUST,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1290) MSR_IA32_TSCDEADLINE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1291) MSR_IA32_ARCH_CAPABILITIES,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1292) MSR_IA32_PERF_CAPABILITIES,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1293) MSR_IA32_MISC_ENABLE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1294) MSR_IA32_MCG_STATUS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1295) MSR_IA32_MCG_CTL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1296) MSR_IA32_MCG_EXT_CTL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1297) MSR_IA32_SMBASE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1298) MSR_SMI_COUNT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1299) MSR_PLATFORM_INFO,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1300) MSR_MISC_FEATURES_ENABLES,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1301) MSR_AMD64_VIRT_SPEC_CTRL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1302) MSR_IA32_POWER_CTL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1303) MSR_IA32_UCODE_REV,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1304)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1305) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1306) * The following list leaves out MSRs whose values are determined
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1307) * by arch/x86/kvm/vmx/nested.c based on CPUID or other MSRs.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1308) * We always support the "true" VMX control MSRs, even if the host
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1309) * processor does not, so I am putting these registers here rather
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1310) * than in msrs_to_save_all.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1311) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1312) MSR_IA32_VMX_BASIC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1313) MSR_IA32_VMX_TRUE_PINBASED_CTLS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1314) MSR_IA32_VMX_TRUE_PROCBASED_CTLS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1315) MSR_IA32_VMX_TRUE_EXIT_CTLS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1316) MSR_IA32_VMX_TRUE_ENTRY_CTLS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1317) MSR_IA32_VMX_MISC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1318) MSR_IA32_VMX_CR0_FIXED0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1319) MSR_IA32_VMX_CR4_FIXED0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1320) MSR_IA32_VMX_VMCS_ENUM,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1321) MSR_IA32_VMX_PROCBASED_CTLS2,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1322) MSR_IA32_VMX_EPT_VPID_CAP,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1323) MSR_IA32_VMX_VMFUNC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1324)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1325) MSR_K7_HWCR,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1326) MSR_KVM_POLL_CONTROL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1327) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1328)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1329) static u32 emulated_msrs[ARRAY_SIZE(emulated_msrs_all)];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1330) static unsigned num_emulated_msrs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1331)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1332) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1333) * List of msr numbers which are used to expose MSR-based features that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1334) * can be used by a hypervisor to validate requested CPU features.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1335) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1336) static const u32 msr_based_features_all[] = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1337) MSR_IA32_VMX_BASIC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1338) MSR_IA32_VMX_TRUE_PINBASED_CTLS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1339) MSR_IA32_VMX_PINBASED_CTLS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1340) MSR_IA32_VMX_TRUE_PROCBASED_CTLS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1341) MSR_IA32_VMX_PROCBASED_CTLS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1342) MSR_IA32_VMX_TRUE_EXIT_CTLS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1343) MSR_IA32_VMX_EXIT_CTLS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1344) MSR_IA32_VMX_TRUE_ENTRY_CTLS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1345) MSR_IA32_VMX_ENTRY_CTLS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1346) MSR_IA32_VMX_MISC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1347) MSR_IA32_VMX_CR0_FIXED0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1348) MSR_IA32_VMX_CR0_FIXED1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1349) MSR_IA32_VMX_CR4_FIXED0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1350) MSR_IA32_VMX_CR4_FIXED1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1351) MSR_IA32_VMX_VMCS_ENUM,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1352) MSR_IA32_VMX_PROCBASED_CTLS2,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1353) MSR_IA32_VMX_EPT_VPID_CAP,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1354) MSR_IA32_VMX_VMFUNC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1355)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1356) MSR_F10H_DECFG,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1357) MSR_IA32_UCODE_REV,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1358) MSR_IA32_ARCH_CAPABILITIES,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1359) MSR_IA32_PERF_CAPABILITIES,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1360) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1361)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1362) static u32 msr_based_features[ARRAY_SIZE(msr_based_features_all)];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1363) static unsigned int num_msr_based_features;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1364)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1365) static u64 kvm_get_arch_capabilities(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1366) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1367) u64 data = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1368)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1369) if (boot_cpu_has(X86_FEATURE_ARCH_CAPABILITIES))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1370) rdmsrl(MSR_IA32_ARCH_CAPABILITIES, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1371)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1372) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1373) * If nx_huge_pages is enabled, KVM's shadow paging will ensure that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1374) * the nested hypervisor runs with NX huge pages. If it is not,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1375) * L1 is anyway vulnerable to ITLB_MULTIHIT explots from other
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1376) * L1 guests, so it need not worry about its own (L2) guests.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1377) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1378) data |= ARCH_CAP_PSCHANGE_MC_NO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1379)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1380) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1381) * If we're doing cache flushes (either "always" or "cond")
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1382) * we will do one whenever the guest does a vmlaunch/vmresume.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1383) * If an outer hypervisor is doing the cache flush for us
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1384) * (VMENTER_L1D_FLUSH_NESTED_VM), we can safely pass that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1385) * capability to the guest too, and if EPT is disabled we're not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1386) * vulnerable. Overall, only VMENTER_L1D_FLUSH_NEVER will
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1387) * require a nested hypervisor to do a flush of its own.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1388) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1389) if (l1tf_vmx_mitigation != VMENTER_L1D_FLUSH_NEVER)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1390) data |= ARCH_CAP_SKIP_VMENTRY_L1DFLUSH;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1391)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1392) if (!boot_cpu_has_bug(X86_BUG_CPU_MELTDOWN))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1393) data |= ARCH_CAP_RDCL_NO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1394) if (!boot_cpu_has_bug(X86_BUG_SPEC_STORE_BYPASS))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1395) data |= ARCH_CAP_SSB_NO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1396) if (!boot_cpu_has_bug(X86_BUG_MDS))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1397) data |= ARCH_CAP_MDS_NO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1398)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1399) if (!boot_cpu_has(X86_FEATURE_RTM)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1400) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1401) * If RTM=0 because the kernel has disabled TSX, the host might
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1402) * have TAA_NO or TSX_CTRL. Clear TAA_NO (the guest sees RTM=0
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1403) * and therefore knows that there cannot be TAA) but keep
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1404) * TSX_CTRL: some buggy userspaces leave it set on tsx=on hosts,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1405) * and we want to allow migrating those guests to tsx=off hosts.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1406) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1407) data &= ~ARCH_CAP_TAA_NO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1408) } else if (!boot_cpu_has_bug(X86_BUG_TAA)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1409) data |= ARCH_CAP_TAA_NO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1410) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1411) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1412) * Nothing to do here; we emulate TSX_CTRL if present on the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1413) * host so the guest can choose between disabling TSX or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1414) * using VERW to clear CPU buffers.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1415) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1416) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1417)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1418) return data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1419) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1420)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1421) static int kvm_get_msr_feature(struct kvm_msr_entry *msr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1422) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1423) switch (msr->index) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1424) case MSR_IA32_ARCH_CAPABILITIES:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1425) msr->data = kvm_get_arch_capabilities();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1426) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1427) case MSR_IA32_UCODE_REV:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1428) rdmsrl_safe(msr->index, &msr->data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1429) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1430) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1431) return kvm_x86_ops.get_msr_feature(msr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1432) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1433) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1434) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1435)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1436) static int do_get_msr_feature(struct kvm_vcpu *vcpu, unsigned index, u64 *data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1437) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1438) struct kvm_msr_entry msr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1439) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1440)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1441) msr.index = index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1442) r = kvm_get_msr_feature(&msr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1443)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1444) if (r == KVM_MSR_RET_INVALID) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1445) /* Unconditionally clear the output for simplicity */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1446) *data = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1447) if (kvm_msr_ignored_check(vcpu, index, 0, false))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1448) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1449) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1450)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1451) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1452) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1453)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1454) *data = msr.data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1455)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1456) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1457) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1458)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1459) static bool __kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1460) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1461) if (efer & EFER_FFXSR && !guest_cpuid_has(vcpu, X86_FEATURE_FXSR_OPT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1462) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1463)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1464) if (efer & EFER_SVME && !guest_cpuid_has(vcpu, X86_FEATURE_SVM))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1465) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1466)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1467) if (efer & (EFER_LME | EFER_LMA) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1468) !guest_cpuid_has(vcpu, X86_FEATURE_LM))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1469) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1470)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1471) if (efer & EFER_NX && !guest_cpuid_has(vcpu, X86_FEATURE_NX))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1472) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1473)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1474) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1475)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1476) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1477) bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1478) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1479) if (efer & efer_reserved_bits)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1480) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1481)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1482) return __kvm_valid_efer(vcpu, efer);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1483) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1484) EXPORT_SYMBOL_GPL(kvm_valid_efer);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1485)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1486) static int set_efer(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1487) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1488) u64 old_efer = vcpu->arch.efer;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1489) u64 efer = msr_info->data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1490) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1491)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1492) if (efer & efer_reserved_bits)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1493) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1494)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1495) if (!msr_info->host_initiated) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1496) if (!__kvm_valid_efer(vcpu, efer))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1497) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1498)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1499) if (is_paging(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1500) (vcpu->arch.efer & EFER_LME) != (efer & EFER_LME))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1501) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1502) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1503)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1504) efer &= ~EFER_LMA;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1505) efer |= vcpu->arch.efer & EFER_LMA;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1506)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1507) r = kvm_x86_ops.set_efer(vcpu, efer);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1508) if (r) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1509) WARN_ON(r > 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1510) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1511) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1512)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1513) /* Update reserved bits */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1514) if ((efer ^ old_efer) & EFER_NX)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1515) kvm_mmu_reset_context(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1516)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1517) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1518) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1519)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1520) void kvm_enable_efer_bits(u64 mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1521) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1522) efer_reserved_bits &= ~mask;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1523) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1524) EXPORT_SYMBOL_GPL(kvm_enable_efer_bits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1525)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1526) bool kvm_msr_allowed(struct kvm_vcpu *vcpu, u32 index, u32 type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1527) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1528) struct kvm_x86_msr_filter *msr_filter;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1529) struct msr_bitmap_range *ranges;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1530) struct kvm *kvm = vcpu->kvm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1531) bool allowed;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1532) int idx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1533) u32 i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1534)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1535) /* x2APIC MSRs do not support filtering. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1536) if (index >= 0x800 && index <= 0x8ff)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1537) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1538)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1539) idx = srcu_read_lock(&kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1540)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1541) msr_filter = srcu_dereference(kvm->arch.msr_filter, &kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1542) if (!msr_filter) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1543) allowed = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1544) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1545) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1546)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1547) allowed = msr_filter->default_allow;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1548) ranges = msr_filter->ranges;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1549)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1550) for (i = 0; i < msr_filter->count; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1551) u32 start = ranges[i].base;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1552) u32 end = start + ranges[i].nmsrs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1553) u32 flags = ranges[i].flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1554) unsigned long *bitmap = ranges[i].bitmap;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1555)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1556) if ((index >= start) && (index < end) && (flags & type)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1557) allowed = !!test_bit(index - start, bitmap);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1558) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1559) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1560) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1561)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1562) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1563) srcu_read_unlock(&kvm->srcu, idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1564)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1565) return allowed;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1566) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1567) EXPORT_SYMBOL_GPL(kvm_msr_allowed);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1568)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1569) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1570) * Write @data into the MSR specified by @index. Select MSR specific fault
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1571) * checks are bypassed if @host_initiated is %true.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1572) * Returns 0 on success, non-0 otherwise.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1573) * Assumes vcpu_load() was already called.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1574) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1575) static int __kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1576) bool host_initiated)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1577) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1578) struct msr_data msr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1579)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1580) if (!host_initiated && !kvm_msr_allowed(vcpu, index, KVM_MSR_FILTER_WRITE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1581) return KVM_MSR_RET_FILTERED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1582)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1583) switch (index) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1584) case MSR_FS_BASE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1585) case MSR_GS_BASE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1586) case MSR_KERNEL_GS_BASE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1587) case MSR_CSTAR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1588) case MSR_LSTAR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1589) if (is_noncanonical_address(data, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1590) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1591) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1592) case MSR_IA32_SYSENTER_EIP:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1593) case MSR_IA32_SYSENTER_ESP:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1594) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1595) * IA32_SYSENTER_ESP and IA32_SYSENTER_EIP cause #GP if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1596) * non-canonical address is written on Intel but not on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1597) * AMD (which ignores the top 32-bits, because it does
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1598) * not implement 64-bit SYSENTER).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1599) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1600) * 64-bit code should hence be able to write a non-canonical
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1601) * value on AMD. Making the address canonical ensures that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1602) * vmentry does not fail on Intel after writing a non-canonical
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1603) * value, and that something deterministic happens if the guest
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1604) * invokes 64-bit SYSENTER.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1605) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1606) data = get_canonical(data, vcpu_virt_addr_bits(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1607) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1608)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1609) msr.data = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1610) msr.index = index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1611) msr.host_initiated = host_initiated;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1612)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1613) return kvm_x86_ops.set_msr(vcpu, &msr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1614) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1615)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1616) static int kvm_set_msr_ignored_check(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1617) u32 index, u64 data, bool host_initiated)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1618) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1619) int ret = __kvm_set_msr(vcpu, index, data, host_initiated);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1620)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1621) if (ret == KVM_MSR_RET_INVALID)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1622) if (kvm_msr_ignored_check(vcpu, index, data, true))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1623) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1624)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1625) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1626) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1627)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1628) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1629) * Read the MSR specified by @index into @data. Select MSR specific fault
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1630) * checks are bypassed if @host_initiated is %true.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1631) * Returns 0 on success, non-0 otherwise.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1632) * Assumes vcpu_load() was already called.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1633) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1634) int __kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1635) bool host_initiated)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1636) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1637) struct msr_data msr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1638) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1639)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1640) if (!host_initiated && !kvm_msr_allowed(vcpu, index, KVM_MSR_FILTER_READ))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1641) return KVM_MSR_RET_FILTERED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1642)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1643) msr.index = index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1644) msr.host_initiated = host_initiated;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1645)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1646) ret = kvm_x86_ops.get_msr(vcpu, &msr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1647) if (!ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1648) *data = msr.data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1649) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1650) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1651)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1652) static int kvm_get_msr_ignored_check(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1653) u32 index, u64 *data, bool host_initiated)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1654) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1655) int ret = __kvm_get_msr(vcpu, index, data, host_initiated);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1656)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1657) if (ret == KVM_MSR_RET_INVALID) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1658) /* Unconditionally clear *data for simplicity */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1659) *data = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1660) if (kvm_msr_ignored_check(vcpu, index, 0, false))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1661) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1662) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1663)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1664) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1665) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1666)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1667) int kvm_get_msr(struct kvm_vcpu *vcpu, u32 index, u64 *data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1668) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1669) return kvm_get_msr_ignored_check(vcpu, index, data, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1670) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1671) EXPORT_SYMBOL_GPL(kvm_get_msr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1672)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1673) int kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1674) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1675) return kvm_set_msr_ignored_check(vcpu, index, data, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1676) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1677) EXPORT_SYMBOL_GPL(kvm_set_msr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1678)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1679) static int complete_emulated_msr(struct kvm_vcpu *vcpu, bool is_read)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1680) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1681) if (vcpu->run->msr.error) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1682) kvm_inject_gp(vcpu, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1683) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1684) } else if (is_read) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1685) kvm_rax_write(vcpu, (u32)vcpu->run->msr.data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1686) kvm_rdx_write(vcpu, vcpu->run->msr.data >> 32);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1687) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1688)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1689) return kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1690) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1691)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1692) static int complete_emulated_rdmsr(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1693) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1694) return complete_emulated_msr(vcpu, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1695) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1696)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1697) static int complete_emulated_wrmsr(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1698) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1699) return complete_emulated_msr(vcpu, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1700) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1701)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1702) static u64 kvm_msr_reason(int r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1703) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1704) switch (r) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1705) case KVM_MSR_RET_INVALID:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1706) return KVM_MSR_EXIT_REASON_UNKNOWN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1707) case KVM_MSR_RET_FILTERED:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1708) return KVM_MSR_EXIT_REASON_FILTER;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1709) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1710) return KVM_MSR_EXIT_REASON_INVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1711) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1712) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1713)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1714) static int kvm_msr_user_space(struct kvm_vcpu *vcpu, u32 index,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1715) u32 exit_reason, u64 data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1716) int (*completion)(struct kvm_vcpu *vcpu),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1717) int r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1718) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1719) u64 msr_reason = kvm_msr_reason(r);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1720)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1721) /* Check if the user wanted to know about this MSR fault */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1722) if (!(vcpu->kvm->arch.user_space_msr_mask & msr_reason))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1723) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1724)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1725) vcpu->run->exit_reason = exit_reason;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1726) vcpu->run->msr.error = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1727) memset(vcpu->run->msr.pad, 0, sizeof(vcpu->run->msr.pad));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1728) vcpu->run->msr.reason = msr_reason;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1729) vcpu->run->msr.index = index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1730) vcpu->run->msr.data = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1731) vcpu->arch.complete_userspace_io = completion;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1732)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1733) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1734) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1735)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1736) static int kvm_get_msr_user_space(struct kvm_vcpu *vcpu, u32 index, int r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1737) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1738) return kvm_msr_user_space(vcpu, index, KVM_EXIT_X86_RDMSR, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1739) complete_emulated_rdmsr, r);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1740) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1741)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1742) static int kvm_set_msr_user_space(struct kvm_vcpu *vcpu, u32 index, u64 data, int r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1743) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1744) return kvm_msr_user_space(vcpu, index, KVM_EXIT_X86_WRMSR, data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1745) complete_emulated_wrmsr, r);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1746) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1747)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1748) int kvm_emulate_rdmsr(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1749) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1750) u32 ecx = kvm_rcx_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1751) u64 data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1752) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1753)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1754) r = kvm_get_msr(vcpu, ecx, &data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1755)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1756) /* MSR read failed? See if we should ask user space */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1757) if (r && kvm_get_msr_user_space(vcpu, ecx, r)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1758) /* Bounce to user space */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1759) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1760) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1761)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1762) /* MSR read failed? Inject a #GP */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1763) if (r) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1764) trace_kvm_msr_read_ex(ecx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1765) kvm_inject_gp(vcpu, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1766) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1767) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1768)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1769) trace_kvm_msr_read(ecx, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1770)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1771) kvm_rax_write(vcpu, data & -1u);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1772) kvm_rdx_write(vcpu, (data >> 32) & -1u);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1773) return kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1774) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1775) EXPORT_SYMBOL_GPL(kvm_emulate_rdmsr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1776)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1777) int kvm_emulate_wrmsr(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1778) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1779) u32 ecx = kvm_rcx_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1780) u64 data = kvm_read_edx_eax(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1781) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1782)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1783) r = kvm_set_msr(vcpu, ecx, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1784)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1785) /* MSR write failed? See if we should ask user space */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1786) if (r && kvm_set_msr_user_space(vcpu, ecx, data, r))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1787) /* Bounce to user space */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1788) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1789)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1790) /* Signal all other negative errors to userspace */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1791) if (r < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1792) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1793)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1794) /* MSR write failed? Inject a #GP */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1795) if (r > 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1796) trace_kvm_msr_write_ex(ecx, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1797) kvm_inject_gp(vcpu, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1798) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1799) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1800)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1801) trace_kvm_msr_write(ecx, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1802) return kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1803) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1804) EXPORT_SYMBOL_GPL(kvm_emulate_wrmsr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1805)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1806) bool kvm_vcpu_exit_request(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1807) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1808) return vcpu->mode == EXITING_GUEST_MODE || kvm_request_pending(vcpu) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1809) xfer_to_guest_mode_work_pending();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1810) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1811) EXPORT_SYMBOL_GPL(kvm_vcpu_exit_request);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1812)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1813) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1814) * The fast path for frequent and performance sensitive wrmsr emulation,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1815) * i.e. the sending of IPI, sending IPI early in the VM-Exit flow reduces
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1816) * the latency of virtual IPI by avoiding the expensive bits of transitioning
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1817) * from guest to host, e.g. reacquiring KVM's SRCU lock. In contrast to the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1818) * other cases which must be called after interrupts are enabled on the host.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1819) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1820) static int handle_fastpath_set_x2apic_icr_irqoff(struct kvm_vcpu *vcpu, u64 data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1821) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1822) if (!lapic_in_kernel(vcpu) || !apic_x2apic_mode(vcpu->arch.apic))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1823) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1824)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1825) if (((data & APIC_SHORT_MASK) == APIC_DEST_NOSHORT) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1826) ((data & APIC_DEST_MASK) == APIC_DEST_PHYSICAL) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1827) ((data & APIC_MODE_MASK) == APIC_DM_FIXED) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1828) ((u32)(data >> 32) != X2APIC_BROADCAST)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1829)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1830) data &= ~(1 << 12);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1831) kvm_apic_send_ipi(vcpu->arch.apic, (u32)data, (u32)(data >> 32));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1832) kvm_lapic_set_reg(vcpu->arch.apic, APIC_ICR2, (u32)(data >> 32));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1833) kvm_lapic_set_reg(vcpu->arch.apic, APIC_ICR, (u32)data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1834) trace_kvm_apic_write(APIC_ICR, (u32)data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1835) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1836) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1837)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1838) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1839) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1840)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1841) static int handle_fastpath_set_tscdeadline(struct kvm_vcpu *vcpu, u64 data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1842) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1843) if (!kvm_can_use_hv_timer(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1844) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1845)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1846) kvm_set_lapic_tscdeadline_msr(vcpu, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1847) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1848) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1849)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1850) fastpath_t handle_fastpath_set_msr_irqoff(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1851) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1852) u32 msr = kvm_rcx_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1853) u64 data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1854) fastpath_t ret = EXIT_FASTPATH_NONE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1855)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1856) switch (msr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1857) case APIC_BASE_MSR + (APIC_ICR >> 4):
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1858) data = kvm_read_edx_eax(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1859) if (!handle_fastpath_set_x2apic_icr_irqoff(vcpu, data)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1860) kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1861) ret = EXIT_FASTPATH_EXIT_HANDLED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1862) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1863) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1864) case MSR_IA32_TSCDEADLINE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1865) data = kvm_read_edx_eax(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1866) if (!handle_fastpath_set_tscdeadline(vcpu, data)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1867) kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1868) ret = EXIT_FASTPATH_REENTER_GUEST;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1869) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1870) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1871) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1872) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1873) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1874)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1875) if (ret != EXIT_FASTPATH_NONE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1876) trace_kvm_msr_write(msr, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1877)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1878) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1879) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1880) EXPORT_SYMBOL_GPL(handle_fastpath_set_msr_irqoff);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1881)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1882) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1883) * Adapt set_msr() to msr_io()'s calling convention
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1884) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1885) static int do_get_msr(struct kvm_vcpu *vcpu, unsigned index, u64 *data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1886) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1887) return kvm_get_msr_ignored_check(vcpu, index, data, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1888) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1889)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1890) static int do_set_msr(struct kvm_vcpu *vcpu, unsigned index, u64 *data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1891) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1892) return kvm_set_msr_ignored_check(vcpu, index, *data, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1893) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1894)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1895) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1896) struct pvclock_clock {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1897) int vclock_mode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1898) u64 cycle_last;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1899) u64 mask;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1900) u32 mult;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1901) u32 shift;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1902) u64 base_cycles;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1903) u64 offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1904) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1905)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1906) struct pvclock_gtod_data {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1907) seqcount_t seq;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1908)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1909) struct pvclock_clock clock; /* extract of a clocksource struct */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1910) struct pvclock_clock raw_clock; /* extract of a clocksource struct */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1911)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1912) ktime_t offs_boot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1913) u64 wall_time_sec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1914) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1915)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1916) static struct pvclock_gtod_data pvclock_gtod_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1917)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1918) static void update_pvclock_gtod(struct timekeeper *tk)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1919) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1920) struct pvclock_gtod_data *vdata = &pvclock_gtod_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1921)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1922) write_seqcount_begin(&vdata->seq);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1923)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1924) /* copy pvclock gtod data */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1925) vdata->clock.vclock_mode = tk->tkr_mono.clock->vdso_clock_mode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1926) vdata->clock.cycle_last = tk->tkr_mono.cycle_last;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1927) vdata->clock.mask = tk->tkr_mono.mask;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1928) vdata->clock.mult = tk->tkr_mono.mult;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1929) vdata->clock.shift = tk->tkr_mono.shift;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1930) vdata->clock.base_cycles = tk->tkr_mono.xtime_nsec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1931) vdata->clock.offset = tk->tkr_mono.base;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1932)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1933) vdata->raw_clock.vclock_mode = tk->tkr_raw.clock->vdso_clock_mode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1934) vdata->raw_clock.cycle_last = tk->tkr_raw.cycle_last;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1935) vdata->raw_clock.mask = tk->tkr_raw.mask;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1936) vdata->raw_clock.mult = tk->tkr_raw.mult;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1937) vdata->raw_clock.shift = tk->tkr_raw.shift;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1938) vdata->raw_clock.base_cycles = tk->tkr_raw.xtime_nsec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1939) vdata->raw_clock.offset = tk->tkr_raw.base;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1940)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1941) vdata->wall_time_sec = tk->xtime_sec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1942)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1943) vdata->offs_boot = tk->offs_boot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1944)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1945) write_seqcount_end(&vdata->seq);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1946) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1947)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1948) static s64 get_kvmclock_base_ns(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1949) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1950) /* Count up from boot time, but with the frequency of the raw clock. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1951) return ktime_to_ns(ktime_add(ktime_get_raw(), pvclock_gtod_data.offs_boot));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1952) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1953) #else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1954) static s64 get_kvmclock_base_ns(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1955) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1956) /* Master clock not used, so we can just use CLOCK_BOOTTIME. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1957) return ktime_get_boottime_ns();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1958) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1959) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1960)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1961) static void kvm_write_wall_clock(struct kvm *kvm, gpa_t wall_clock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1962) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1963) int version;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1964) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1965) struct pvclock_wall_clock wc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1966) u64 wall_nsec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1967)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1968) kvm->arch.wall_clock = wall_clock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1969)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1970) if (!wall_clock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1971) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1972)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1973) r = kvm_read_guest(kvm, wall_clock, &version, sizeof(version));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1974) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1975) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1976)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1977) if (version & 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1978) ++version; /* first time write, random junk */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1979)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1980) ++version;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1981)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1982) if (kvm_write_guest(kvm, wall_clock, &version, sizeof(version)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1983) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1984)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1985) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1986) * The guest calculates current wall clock time by adding
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1987) * system time (updated by kvm_guest_time_update below) to the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1988) * wall clock specified here. We do the reverse here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1989) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1990) wall_nsec = ktime_get_real_ns() - get_kvmclock_ns(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1991)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1992) wc.nsec = do_div(wall_nsec, 1000000000);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1993) wc.sec = (u32)wall_nsec; /* overflow in 2106 guest time */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1994) wc.version = version;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1995)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1996) kvm_write_guest(kvm, wall_clock, &wc, sizeof(wc));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1997)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1998) version++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1999) kvm_write_guest(kvm, wall_clock, &version, sizeof(version));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2000) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2001)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2002) static void kvm_write_system_time(struct kvm_vcpu *vcpu, gpa_t system_time,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2003) bool old_msr, bool host_initiated)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2004) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2005) struct kvm_arch *ka = &vcpu->kvm->arch;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2006)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2007) if (vcpu->vcpu_id == 0 && !host_initiated) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2008) if (ka->boot_vcpu_runs_old_kvmclock != old_msr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2009) kvm_make_request(KVM_REQ_MASTERCLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2010)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2011) ka->boot_vcpu_runs_old_kvmclock = old_msr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2012) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2013)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2014) vcpu->arch.time = system_time;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2015) kvm_make_request(KVM_REQ_GLOBAL_CLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2016)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2017) /* we verify if the enable bit is set... */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2018) vcpu->arch.pv_time_enabled = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2019) if (!(system_time & 1))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2020) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2021)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2022) if (!kvm_gfn_to_hva_cache_init(vcpu->kvm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2023) &vcpu->arch.pv_time, system_time & ~1ULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2024) sizeof(struct pvclock_vcpu_time_info)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2025) vcpu->arch.pv_time_enabled = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2026)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2027) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2028) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2029)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2030) static uint32_t div_frac(uint32_t dividend, uint32_t divisor)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2031) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2032) do_shl32_div32(dividend, divisor);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2033) return dividend;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2034) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2035)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2036) static void kvm_get_time_scale(uint64_t scaled_hz, uint64_t base_hz,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2037) s8 *pshift, u32 *pmultiplier)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2038) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2039) uint64_t scaled64;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2040) int32_t shift = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2041) uint64_t tps64;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2042) uint32_t tps32;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2043)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2044) tps64 = base_hz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2045) scaled64 = scaled_hz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2046) while (tps64 > scaled64*2 || tps64 & 0xffffffff00000000ULL) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2047) tps64 >>= 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2048) shift--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2049) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2050)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2051) tps32 = (uint32_t)tps64;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2052) while (tps32 <= scaled64 || scaled64 & 0xffffffff00000000ULL) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2053) if (scaled64 & 0xffffffff00000000ULL || tps32 & 0x80000000)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2054) scaled64 >>= 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2055) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2056) tps32 <<= 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2057) shift++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2058) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2059)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2060) *pshift = shift;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2061) *pmultiplier = div_frac(scaled64, tps32);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2062) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2063)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2064) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2065) static atomic_t kvm_guest_has_master_clock = ATOMIC_INIT(0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2066) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2067)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2068) static DEFINE_PER_CPU(unsigned long, cpu_tsc_khz);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2069) static unsigned long max_tsc_khz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2070)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2071) static u32 adjust_tsc_khz(u32 khz, s32 ppm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2072) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2073) u64 v = (u64)khz * (1000000 + ppm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2074) do_div(v, 1000000);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2075) return v;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2076) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2077)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2078) static int set_tsc_khz(struct kvm_vcpu *vcpu, u32 user_tsc_khz, bool scale)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2079) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2080) u64 ratio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2081)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2082) /* Guest TSC same frequency as host TSC? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2083) if (!scale) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2084) vcpu->arch.tsc_scaling_ratio = kvm_default_tsc_scaling_ratio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2085) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2086) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2087)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2088) /* TSC scaling supported? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2089) if (!kvm_has_tsc_control) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2090) if (user_tsc_khz > tsc_khz) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2091) vcpu->arch.tsc_catchup = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2092) vcpu->arch.tsc_always_catchup = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2093) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2094) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2095) pr_warn_ratelimited("user requested TSC rate below hardware speed\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2096) return -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2097) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2098) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2099)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2100) /* TSC scaling required - calculate ratio */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2101) ratio = mul_u64_u32_div(1ULL << kvm_tsc_scaling_ratio_frac_bits,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2102) user_tsc_khz, tsc_khz);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2103)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2104) if (ratio == 0 || ratio >= kvm_max_tsc_scaling_ratio) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2105) pr_warn_ratelimited("Invalid TSC scaling ratio - virtual-tsc-khz=%u\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2106) user_tsc_khz);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2107) return -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2108) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2109)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2110) vcpu->arch.tsc_scaling_ratio = ratio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2111) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2112) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2113)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2114) static int kvm_set_tsc_khz(struct kvm_vcpu *vcpu, u32 user_tsc_khz)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2115) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2116) u32 thresh_lo, thresh_hi;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2117) int use_scaling = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2118)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2119) /* tsc_khz can be zero if TSC calibration fails */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2120) if (user_tsc_khz == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2121) /* set tsc_scaling_ratio to a safe value */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2122) vcpu->arch.tsc_scaling_ratio = kvm_default_tsc_scaling_ratio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2123) return -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2124) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2125)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2126) /* Compute a scale to convert nanoseconds in TSC cycles */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2127) kvm_get_time_scale(user_tsc_khz * 1000LL, NSEC_PER_SEC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2128) &vcpu->arch.virtual_tsc_shift,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2129) &vcpu->arch.virtual_tsc_mult);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2130) vcpu->arch.virtual_tsc_khz = user_tsc_khz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2131)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2132) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2133) * Compute the variation in TSC rate which is acceptable
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2134) * within the range of tolerance and decide if the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2135) * rate being applied is within that bounds of the hardware
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2136) * rate. If so, no scaling or compensation need be done.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2137) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2138) thresh_lo = adjust_tsc_khz(tsc_khz, -tsc_tolerance_ppm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2139) thresh_hi = adjust_tsc_khz(tsc_khz, tsc_tolerance_ppm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2140) if (user_tsc_khz < thresh_lo || user_tsc_khz > thresh_hi) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2141) pr_debug("kvm: requested TSC rate %u falls outside tolerance [%u,%u]\n", user_tsc_khz, thresh_lo, thresh_hi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2142) use_scaling = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2143) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2144) return set_tsc_khz(vcpu, user_tsc_khz, use_scaling);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2145) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2146)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2147) static u64 compute_guest_tsc(struct kvm_vcpu *vcpu, s64 kernel_ns)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2148) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2149) u64 tsc = pvclock_scale_delta(kernel_ns-vcpu->arch.this_tsc_nsec,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2150) vcpu->arch.virtual_tsc_mult,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2151) vcpu->arch.virtual_tsc_shift);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2152) tsc += vcpu->arch.this_tsc_write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2153) return tsc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2154) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2155)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2156) static inline int gtod_is_based_on_tsc(int mode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2157) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2158) return mode == VDSO_CLOCKMODE_TSC || mode == VDSO_CLOCKMODE_HVCLOCK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2159) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2160)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2161) static void kvm_track_tsc_matching(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2162) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2163) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2164) bool vcpus_matched;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2165) struct kvm_arch *ka = &vcpu->kvm->arch;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2166) struct pvclock_gtod_data *gtod = &pvclock_gtod_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2167)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2168) vcpus_matched = (ka->nr_vcpus_matched_tsc + 1 ==
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2169) atomic_read(&vcpu->kvm->online_vcpus));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2170)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2171) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2172) * Once the masterclock is enabled, always perform request in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2173) * order to update it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2174) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2175) * In order to enable masterclock, the host clocksource must be TSC
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2176) * and the vcpus need to have matched TSCs. When that happens,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2177) * perform request to enable masterclock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2178) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2179) if (ka->use_master_clock ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2180) (gtod_is_based_on_tsc(gtod->clock.vclock_mode) && vcpus_matched))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2181) kvm_make_request(KVM_REQ_MASTERCLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2182)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2183) trace_kvm_track_tsc(vcpu->vcpu_id, ka->nr_vcpus_matched_tsc,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2184) atomic_read(&vcpu->kvm->online_vcpus),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2185) ka->use_master_clock, gtod->clock.vclock_mode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2186) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2187) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2188)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2189) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2190) * Multiply tsc by a fixed point number represented by ratio.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2191) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2192) * The most significant 64-N bits (mult) of ratio represent the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2193) * integral part of the fixed point number; the remaining N bits
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2194) * (frac) represent the fractional part, ie. ratio represents a fixed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2195) * point number (mult + frac * 2^(-N)).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2196) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2197) * N equals to kvm_tsc_scaling_ratio_frac_bits.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2198) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2199) static inline u64 __scale_tsc(u64 ratio, u64 tsc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2200) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2201) return mul_u64_u64_shr(tsc, ratio, kvm_tsc_scaling_ratio_frac_bits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2202) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2203)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2204) u64 kvm_scale_tsc(struct kvm_vcpu *vcpu, u64 tsc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2205) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2206) u64 _tsc = tsc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2207) u64 ratio = vcpu->arch.tsc_scaling_ratio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2208)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2209) if (ratio != kvm_default_tsc_scaling_ratio)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2210) _tsc = __scale_tsc(ratio, tsc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2211)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2212) return _tsc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2213) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2214) EXPORT_SYMBOL_GPL(kvm_scale_tsc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2215)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2216) static u64 kvm_compute_tsc_offset(struct kvm_vcpu *vcpu, u64 target_tsc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2217) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2218) u64 tsc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2219)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2220) tsc = kvm_scale_tsc(vcpu, rdtsc());
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2221)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2222) return target_tsc - tsc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2223) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2224)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2225) u64 kvm_read_l1_tsc(struct kvm_vcpu *vcpu, u64 host_tsc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2226) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2227) return vcpu->arch.l1_tsc_offset + kvm_scale_tsc(vcpu, host_tsc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2228) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2229) EXPORT_SYMBOL_GPL(kvm_read_l1_tsc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2230)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2231) static void kvm_vcpu_write_tsc_offset(struct kvm_vcpu *vcpu, u64 offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2232) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2233) vcpu->arch.l1_tsc_offset = offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2234) vcpu->arch.tsc_offset = kvm_x86_ops.write_l1_tsc_offset(vcpu, offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2235) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2236)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2237) static inline bool kvm_check_tsc_unstable(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2238) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2239) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2240) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2241) * TSC is marked unstable when we're running on Hyper-V,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2242) * 'TSC page' clocksource is good.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2243) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2244) if (pvclock_gtod_data.clock.vclock_mode == VDSO_CLOCKMODE_HVCLOCK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2245) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2246) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2247) return check_tsc_unstable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2248) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2249)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2250) static void kvm_synchronize_tsc(struct kvm_vcpu *vcpu, u64 data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2251) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2252) struct kvm *kvm = vcpu->kvm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2253) u64 offset, ns, elapsed;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2254) unsigned long flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2255) bool matched;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2256) bool already_matched;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2257) bool synchronizing = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2258)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2259) raw_spin_lock_irqsave(&kvm->arch.tsc_write_lock, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2260) offset = kvm_compute_tsc_offset(vcpu, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2261) ns = get_kvmclock_base_ns();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2262) elapsed = ns - kvm->arch.last_tsc_nsec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2263)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2264) if (vcpu->arch.virtual_tsc_khz) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2265) if (data == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2266) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2267) * detection of vcpu initialization -- need to sync
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2268) * with other vCPUs. This particularly helps to keep
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2269) * kvm_clock stable after CPU hotplug
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2270) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2271) synchronizing = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2272) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2273) u64 tsc_exp = kvm->arch.last_tsc_write +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2274) nsec_to_cycles(vcpu, elapsed);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2275) u64 tsc_hz = vcpu->arch.virtual_tsc_khz * 1000LL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2276) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2277) * Special case: TSC write with a small delta (1 second)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2278) * of virtual cycle time against real time is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2279) * interpreted as an attempt to synchronize the CPU.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2280) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2281) synchronizing = data < tsc_exp + tsc_hz &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2282) data + tsc_hz > tsc_exp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2283) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2284) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2285)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2286) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2287) * For a reliable TSC, we can match TSC offsets, and for an unstable
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2288) * TSC, we add elapsed time in this computation. We could let the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2289) * compensation code attempt to catch up if we fall behind, but
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2290) * it's better to try to match offsets from the beginning.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2291) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2292) if (synchronizing &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2293) vcpu->arch.virtual_tsc_khz == kvm->arch.last_tsc_khz) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2294) if (!kvm_check_tsc_unstable()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2295) offset = kvm->arch.cur_tsc_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2296) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2297) u64 delta = nsec_to_cycles(vcpu, elapsed);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2298) data += delta;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2299) offset = kvm_compute_tsc_offset(vcpu, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2300) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2301) matched = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2302) already_matched = (vcpu->arch.this_tsc_generation == kvm->arch.cur_tsc_generation);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2303) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2304) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2305) * We split periods of matched TSC writes into generations.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2306) * For each generation, we track the original measured
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2307) * nanosecond time, offset, and write, so if TSCs are in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2308) * sync, we can match exact offset, and if not, we can match
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2309) * exact software computation in compute_guest_tsc()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2310) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2311) * These values are tracked in kvm->arch.cur_xxx variables.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2312) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2313) kvm->arch.cur_tsc_generation++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2314) kvm->arch.cur_tsc_nsec = ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2315) kvm->arch.cur_tsc_write = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2316) kvm->arch.cur_tsc_offset = offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2317) matched = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2318) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2319)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2320) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2321) * We also track th most recent recorded KHZ, write and time to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2322) * allow the matching interval to be extended at each write.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2323) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2324) kvm->arch.last_tsc_nsec = ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2325) kvm->arch.last_tsc_write = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2326) kvm->arch.last_tsc_khz = vcpu->arch.virtual_tsc_khz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2327)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2328) vcpu->arch.last_guest_tsc = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2329)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2330) /* Keep track of which generation this VCPU has synchronized to */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2331) vcpu->arch.this_tsc_generation = kvm->arch.cur_tsc_generation;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2332) vcpu->arch.this_tsc_nsec = kvm->arch.cur_tsc_nsec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2333) vcpu->arch.this_tsc_write = kvm->arch.cur_tsc_write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2334)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2335) kvm_vcpu_write_tsc_offset(vcpu, offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2336) raw_spin_unlock_irqrestore(&kvm->arch.tsc_write_lock, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2337)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2338) spin_lock(&kvm->arch.pvclock_gtod_sync_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2339) if (!matched) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2340) kvm->arch.nr_vcpus_matched_tsc = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2341) } else if (!already_matched) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2342) kvm->arch.nr_vcpus_matched_tsc++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2343) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2344)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2345) kvm_track_tsc_matching(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2346) spin_unlock(&kvm->arch.pvclock_gtod_sync_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2347) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2348)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2349) static inline void adjust_tsc_offset_guest(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2350) s64 adjustment)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2351) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2352) u64 tsc_offset = vcpu->arch.l1_tsc_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2353) kvm_vcpu_write_tsc_offset(vcpu, tsc_offset + adjustment);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2354) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2355)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2356) static inline void adjust_tsc_offset_host(struct kvm_vcpu *vcpu, s64 adjustment)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2357) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2358) if (vcpu->arch.tsc_scaling_ratio != kvm_default_tsc_scaling_ratio)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2359) WARN_ON(adjustment < 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2360) adjustment = kvm_scale_tsc(vcpu, (u64) adjustment);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2361) adjust_tsc_offset_guest(vcpu, adjustment);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2362) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2363)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2364) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2365)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2366) static u64 read_tsc(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2367) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2368) u64 ret = (u64)rdtsc_ordered();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2369) u64 last = pvclock_gtod_data.clock.cycle_last;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2370)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2371) if (likely(ret >= last))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2372) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2373)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2374) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2375) * GCC likes to generate cmov here, but this branch is extremely
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2376) * predictable (it's just a function of time and the likely is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2377) * very likely) and there's a data dependence, so force GCC
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2378) * to generate a branch instead. I don't barrier() because
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2379) * we don't actually need a barrier, and if this function
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2380) * ever gets inlined it will generate worse code.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2381) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2382) asm volatile ("");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2383) return last;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2384) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2385)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2386) static inline u64 vgettsc(struct pvclock_clock *clock, u64 *tsc_timestamp,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2387) int *mode)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2388) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2389) long v;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2390) u64 tsc_pg_val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2391)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2392) switch (clock->vclock_mode) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2393) case VDSO_CLOCKMODE_HVCLOCK:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2394) tsc_pg_val = hv_read_tsc_page_tsc(hv_get_tsc_page(),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2395) tsc_timestamp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2396) if (tsc_pg_val != U64_MAX) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2397) /* TSC page valid */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2398) *mode = VDSO_CLOCKMODE_HVCLOCK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2399) v = (tsc_pg_val - clock->cycle_last) &
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2400) clock->mask;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2401) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2402) /* TSC page invalid */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2403) *mode = VDSO_CLOCKMODE_NONE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2404) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2405) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2406) case VDSO_CLOCKMODE_TSC:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2407) *mode = VDSO_CLOCKMODE_TSC;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2408) *tsc_timestamp = read_tsc();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2409) v = (*tsc_timestamp - clock->cycle_last) &
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2410) clock->mask;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2411) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2412) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2413) *mode = VDSO_CLOCKMODE_NONE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2414) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2415)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2416) if (*mode == VDSO_CLOCKMODE_NONE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2417) *tsc_timestamp = v = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2418)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2419) return v * clock->mult;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2420) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2421)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2422) static int do_monotonic_raw(s64 *t, u64 *tsc_timestamp)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2423) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2424) struct pvclock_gtod_data *gtod = &pvclock_gtod_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2425) unsigned long seq;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2426) int mode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2427) u64 ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2428)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2429) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2430) seq = read_seqcount_begin(>od->seq);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2431) ns = gtod->raw_clock.base_cycles;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2432) ns += vgettsc(>od->raw_clock, tsc_timestamp, &mode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2433) ns >>= gtod->raw_clock.shift;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2434) ns += ktime_to_ns(ktime_add(gtod->raw_clock.offset, gtod->offs_boot));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2435) } while (unlikely(read_seqcount_retry(>od->seq, seq)));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2436) *t = ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2437)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2438) return mode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2439) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2440)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2441) static int do_realtime(struct timespec64 *ts, u64 *tsc_timestamp)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2442) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2443) struct pvclock_gtod_data *gtod = &pvclock_gtod_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2444) unsigned long seq;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2445) int mode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2446) u64 ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2447)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2448) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2449) seq = read_seqcount_begin(>od->seq);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2450) ts->tv_sec = gtod->wall_time_sec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2451) ns = gtod->clock.base_cycles;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2452) ns += vgettsc(>od->clock, tsc_timestamp, &mode);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2453) ns >>= gtod->clock.shift;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2454) } while (unlikely(read_seqcount_retry(>od->seq, seq)));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2455)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2456) ts->tv_sec += __iter_div_u64_rem(ns, NSEC_PER_SEC, &ns);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2457) ts->tv_nsec = ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2458)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2459) return mode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2460) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2461)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2462) /* returns true if host is using TSC based clocksource */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2463) static bool kvm_get_time_and_clockread(s64 *kernel_ns, u64 *tsc_timestamp)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2464) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2465) /* checked again under seqlock below */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2466) if (!gtod_is_based_on_tsc(pvclock_gtod_data.clock.vclock_mode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2467) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2468)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2469) return gtod_is_based_on_tsc(do_monotonic_raw(kernel_ns,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2470) tsc_timestamp));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2471) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2472)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2473) /* returns true if host is using TSC based clocksource */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2474) static bool kvm_get_walltime_and_clockread(struct timespec64 *ts,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2475) u64 *tsc_timestamp)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2476) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2477) /* checked again under seqlock below */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2478) if (!gtod_is_based_on_tsc(pvclock_gtod_data.clock.vclock_mode))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2479) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2480)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2481) return gtod_is_based_on_tsc(do_realtime(ts, tsc_timestamp));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2482) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2483) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2484)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2485) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2486) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2487) * Assuming a stable TSC across physical CPUS, and a stable TSC
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2488) * across virtual CPUs, the following condition is possible.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2489) * Each numbered line represents an event visible to both
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2490) * CPUs at the next numbered event.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2491) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2492) * "timespecX" represents host monotonic time. "tscX" represents
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2493) * RDTSC value.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2494) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2495) * VCPU0 on CPU0 | VCPU1 on CPU1
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2496) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2497) * 1. read timespec0,tsc0
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2498) * 2. | timespec1 = timespec0 + N
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2499) * | tsc1 = tsc0 + M
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2500) * 3. transition to guest | transition to guest
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2501) * 4. ret0 = timespec0 + (rdtsc - tsc0) |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2502) * 5. | ret1 = timespec1 + (rdtsc - tsc1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2503) * | ret1 = timespec0 + N + (rdtsc - (tsc0 + M))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2504) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2505) * Since ret0 update is visible to VCPU1 at time 5, to obey monotonicity:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2506) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2507) * - ret0 < ret1
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2508) * - timespec0 + (rdtsc - tsc0) < timespec0 + N + (rdtsc - (tsc0 + M))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2509) * ...
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2510) * - 0 < N - M => M < N
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2511) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2512) * That is, when timespec0 != timespec1, M < N. Unfortunately that is not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2513) * always the case (the difference between two distinct xtime instances
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2514) * might be smaller then the difference between corresponding TSC reads,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2515) * when updating guest vcpus pvclock areas).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2516) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2517) * To avoid that problem, do not allow visibility of distinct
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2518) * system_timestamp/tsc_timestamp values simultaneously: use a master
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2519) * copy of host monotonic time values. Update that master copy
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2520) * in lockstep.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2521) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2522) * Rely on synchronization of host TSCs and guest TSCs for monotonicity.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2523) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2524) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2525)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2526) static void pvclock_update_vm_gtod_copy(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2527) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2528) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2529) struct kvm_arch *ka = &kvm->arch;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2530) int vclock_mode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2531) bool host_tsc_clocksource, vcpus_matched;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2532)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2533) vcpus_matched = (ka->nr_vcpus_matched_tsc + 1 ==
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2534) atomic_read(&kvm->online_vcpus));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2535)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2536) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2537) * If the host uses TSC clock, then passthrough TSC as stable
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2538) * to the guest.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2539) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2540) host_tsc_clocksource = kvm_get_time_and_clockread(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2541) &ka->master_kernel_ns,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2542) &ka->master_cycle_now);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2543)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2544) ka->use_master_clock = host_tsc_clocksource && vcpus_matched
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2545) && !ka->backwards_tsc_observed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2546) && !ka->boot_vcpu_runs_old_kvmclock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2547)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2548) if (ka->use_master_clock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2549) atomic_set(&kvm_guest_has_master_clock, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2550)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2551) vclock_mode = pvclock_gtod_data.clock.vclock_mode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2552) trace_kvm_update_master_clock(ka->use_master_clock, vclock_mode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2553) vcpus_matched);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2554) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2555) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2556)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2557) void kvm_make_mclock_inprogress_request(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2558) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2559) kvm_make_all_cpus_request(kvm, KVM_REQ_MCLOCK_INPROGRESS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2560) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2561)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2562) static void kvm_gen_update_masterclock(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2563) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2564) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2565) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2566) struct kvm_vcpu *vcpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2567) struct kvm_arch *ka = &kvm->arch;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2568)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2569) spin_lock(&ka->pvclock_gtod_sync_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2570) kvm_make_mclock_inprogress_request(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2571) /* no guest entries from this point */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2572) pvclock_update_vm_gtod_copy(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2573)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2574) kvm_for_each_vcpu(i, vcpu, kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2575) kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2576)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2577) /* guest entries allowed */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2578) kvm_for_each_vcpu(i, vcpu, kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2579) kvm_clear_request(KVM_REQ_MCLOCK_INPROGRESS, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2580)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2581) spin_unlock(&ka->pvclock_gtod_sync_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2582) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2583) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2584)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2585) u64 get_kvmclock_ns(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2586) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2587) struct kvm_arch *ka = &kvm->arch;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2588) struct pvclock_vcpu_time_info hv_clock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2589) u64 ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2590)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2591) spin_lock(&ka->pvclock_gtod_sync_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2592) if (!ka->use_master_clock) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2593) spin_unlock(&ka->pvclock_gtod_sync_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2594) return get_kvmclock_base_ns() + ka->kvmclock_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2595) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2596)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2597) hv_clock.tsc_timestamp = ka->master_cycle_now;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2598) hv_clock.system_time = ka->master_kernel_ns + ka->kvmclock_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2599) spin_unlock(&ka->pvclock_gtod_sync_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2600)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2601) /* both __this_cpu_read() and rdtsc() should be on the same cpu */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2602) get_cpu();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2603)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2604) if (__this_cpu_read(cpu_tsc_khz)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2605) kvm_get_time_scale(NSEC_PER_SEC, __this_cpu_read(cpu_tsc_khz) * 1000LL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2606) &hv_clock.tsc_shift,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2607) &hv_clock.tsc_to_system_mul);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2608) ret = __pvclock_read_cycles(&hv_clock, rdtsc());
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2609) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2610) ret = get_kvmclock_base_ns() + ka->kvmclock_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2611)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2612) put_cpu();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2613)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2614) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2615) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2616)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2617) static void kvm_setup_pvclock_page(struct kvm_vcpu *v)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2618) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2619) struct kvm_vcpu_arch *vcpu = &v->arch;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2620) struct pvclock_vcpu_time_info guest_hv_clock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2621)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2622) if (unlikely(kvm_read_guest_cached(v->kvm, &vcpu->pv_time,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2623) &guest_hv_clock, sizeof(guest_hv_clock))))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2624) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2625)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2626) /* This VCPU is paused, but it's legal for a guest to read another
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2627) * VCPU's kvmclock, so we really have to follow the specification where
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2628) * it says that version is odd if data is being modified, and even after
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2629) * it is consistent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2630) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2631) * Version field updates must be kept separate. This is because
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2632) * kvm_write_guest_cached might use a "rep movs" instruction, and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2633) * writes within a string instruction are weakly ordered. So there
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2634) * are three writes overall.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2635) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2636) * As a small optimization, only write the version field in the first
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2637) * and third write. The vcpu->pv_time cache is still valid, because the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2638) * version field is the first in the struct.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2639) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2640) BUILD_BUG_ON(offsetof(struct pvclock_vcpu_time_info, version) != 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2641)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2642) if (guest_hv_clock.version & 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2643) ++guest_hv_clock.version; /* first time write, random junk */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2644)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2645) vcpu->hv_clock.version = guest_hv_clock.version + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2646) kvm_write_guest_cached(v->kvm, &vcpu->pv_time,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2647) &vcpu->hv_clock,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2648) sizeof(vcpu->hv_clock.version));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2649)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2650) smp_wmb();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2651)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2652) /* retain PVCLOCK_GUEST_STOPPED if set in guest copy */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2653) vcpu->hv_clock.flags |= (guest_hv_clock.flags & PVCLOCK_GUEST_STOPPED);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2654)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2655) if (vcpu->pvclock_set_guest_stopped_request) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2656) vcpu->hv_clock.flags |= PVCLOCK_GUEST_STOPPED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2657) vcpu->pvclock_set_guest_stopped_request = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2658) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2659)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2660) trace_kvm_pvclock_update(v->vcpu_id, &vcpu->hv_clock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2661)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2662) kvm_write_guest_cached(v->kvm, &vcpu->pv_time,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2663) &vcpu->hv_clock,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2664) sizeof(vcpu->hv_clock));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2665)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2666) smp_wmb();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2667)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2668) vcpu->hv_clock.version++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2669) kvm_write_guest_cached(v->kvm, &vcpu->pv_time,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2670) &vcpu->hv_clock,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2671) sizeof(vcpu->hv_clock.version));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2672) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2673)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2674) static int kvm_guest_time_update(struct kvm_vcpu *v)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2675) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2676) unsigned long flags, tgt_tsc_khz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2677) struct kvm_vcpu_arch *vcpu = &v->arch;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2678) struct kvm_arch *ka = &v->kvm->arch;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2679) s64 kernel_ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2680) u64 tsc_timestamp, host_tsc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2681) u8 pvclock_flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2682) bool use_master_clock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2683)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2684) kernel_ns = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2685) host_tsc = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2686)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2687) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2688) * If the host uses TSC clock, then passthrough TSC as stable
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2689) * to the guest.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2690) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2691) spin_lock(&ka->pvclock_gtod_sync_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2692) use_master_clock = ka->use_master_clock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2693) if (use_master_clock) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2694) host_tsc = ka->master_cycle_now;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2695) kernel_ns = ka->master_kernel_ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2696) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2697) spin_unlock(&ka->pvclock_gtod_sync_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2698)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2699) /* Keep irq disabled to prevent changes to the clock */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2700) local_irq_save(flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2701) tgt_tsc_khz = __this_cpu_read(cpu_tsc_khz);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2702) if (unlikely(tgt_tsc_khz == 0)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2703) local_irq_restore(flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2704) kvm_make_request(KVM_REQ_CLOCK_UPDATE, v);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2705) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2706) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2707) if (!use_master_clock) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2708) host_tsc = rdtsc();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2709) kernel_ns = get_kvmclock_base_ns();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2710) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2711)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2712) tsc_timestamp = kvm_read_l1_tsc(v, host_tsc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2713)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2714) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2715) * We may have to catch up the TSC to match elapsed wall clock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2716) * time for two reasons, even if kvmclock is used.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2717) * 1) CPU could have been running below the maximum TSC rate
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2718) * 2) Broken TSC compensation resets the base at each VCPU
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2719) * entry to avoid unknown leaps of TSC even when running
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2720) * again on the same CPU. This may cause apparent elapsed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2721) * time to disappear, and the guest to stand still or run
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2722) * very slowly.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2723) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2724) if (vcpu->tsc_catchup) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2725) u64 tsc = compute_guest_tsc(v, kernel_ns);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2726) if (tsc > tsc_timestamp) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2727) adjust_tsc_offset_guest(v, tsc - tsc_timestamp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2728) tsc_timestamp = tsc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2729) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2730) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2731)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2732) local_irq_restore(flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2733)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2734) /* With all the info we got, fill in the values */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2735)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2736) if (kvm_has_tsc_control)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2737) tgt_tsc_khz = kvm_scale_tsc(v, tgt_tsc_khz);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2738)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2739) if (unlikely(vcpu->hw_tsc_khz != tgt_tsc_khz)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2740) kvm_get_time_scale(NSEC_PER_SEC, tgt_tsc_khz * 1000LL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2741) &vcpu->hv_clock.tsc_shift,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2742) &vcpu->hv_clock.tsc_to_system_mul);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2743) vcpu->hw_tsc_khz = tgt_tsc_khz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2744) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2745)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2746) vcpu->hv_clock.tsc_timestamp = tsc_timestamp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2747) vcpu->hv_clock.system_time = kernel_ns + v->kvm->arch.kvmclock_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2748) vcpu->last_guest_tsc = tsc_timestamp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2749)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2750) /* If the host uses TSC clocksource, then it is stable */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2751) pvclock_flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2752) if (use_master_clock)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2753) pvclock_flags |= PVCLOCK_TSC_STABLE_BIT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2754)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2755) vcpu->hv_clock.flags = pvclock_flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2756)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2757) if (vcpu->pv_time_enabled)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2758) kvm_setup_pvclock_page(v);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2759) if (v == kvm_get_vcpu(v->kvm, 0))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2760) kvm_hv_setup_tsc_page(v->kvm, &vcpu->hv_clock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2761) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2762) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2763)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2764) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2765) * kvmclock updates which are isolated to a given vcpu, such as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2766) * vcpu->cpu migration, should not allow system_timestamp from
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2767) * the rest of the vcpus to remain static. Otherwise ntp frequency
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2768) * correction applies to one vcpu's system_timestamp but not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2769) * the others.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2770) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2771) * So in those cases, request a kvmclock update for all vcpus.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2772) * We need to rate-limit these requests though, as they can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2773) * considerably slow guests that have a large number of vcpus.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2774) * The time for a remote vcpu to update its kvmclock is bound
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2775) * by the delay we use to rate-limit the updates.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2776) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2777)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2778) #define KVMCLOCK_UPDATE_DELAY msecs_to_jiffies(100)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2779)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2780) static void kvmclock_update_fn(struct work_struct *work)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2781) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2782) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2783) struct delayed_work *dwork = to_delayed_work(work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2784) struct kvm_arch *ka = container_of(dwork, struct kvm_arch,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2785) kvmclock_update_work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2786) struct kvm *kvm = container_of(ka, struct kvm, arch);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2787) struct kvm_vcpu *vcpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2788)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2789) kvm_for_each_vcpu(i, vcpu, kvm) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2790) kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2791) kvm_vcpu_kick(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2792) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2793) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2794)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2795) static void kvm_gen_kvmclock_update(struct kvm_vcpu *v)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2796) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2797) struct kvm *kvm = v->kvm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2798)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2799) kvm_make_request(KVM_REQ_CLOCK_UPDATE, v);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2800) schedule_delayed_work(&kvm->arch.kvmclock_update_work,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2801) KVMCLOCK_UPDATE_DELAY);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2802) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2803)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2804) #define KVMCLOCK_SYNC_PERIOD (300 * HZ)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2805)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2806) static void kvmclock_sync_fn(struct work_struct *work)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2807) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2808) struct delayed_work *dwork = to_delayed_work(work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2809) struct kvm_arch *ka = container_of(dwork, struct kvm_arch,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2810) kvmclock_sync_work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2811) struct kvm *kvm = container_of(ka, struct kvm, arch);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2812)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2813) if (!kvmclock_periodic_sync)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2814) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2815)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2816) schedule_delayed_work(&kvm->arch.kvmclock_update_work, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2817) schedule_delayed_work(&kvm->arch.kvmclock_sync_work,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2818) KVMCLOCK_SYNC_PERIOD);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2819) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2820)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2821) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2822) * On AMD, HWCR[McStatusWrEn] controls whether setting MCi_STATUS results in #GP.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2823) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2824) static bool can_set_mci_status(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2825) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2826) /* McStatusWrEn enabled? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2827) if (guest_cpuid_is_amd_or_hygon(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2828) return !!(vcpu->arch.msr_hwcr & BIT_ULL(18));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2829)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2830) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2831) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2832)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2833) static int set_msr_mce(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2834) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2835) u64 mcg_cap = vcpu->arch.mcg_cap;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2836) unsigned bank_num = mcg_cap & 0xff;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2837) u32 msr = msr_info->index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2838) u64 data = msr_info->data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2839)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2840) switch (msr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2841) case MSR_IA32_MCG_STATUS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2842) vcpu->arch.mcg_status = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2843) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2844) case MSR_IA32_MCG_CTL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2845) if (!(mcg_cap & MCG_CTL_P) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2846) (data || !msr_info->host_initiated))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2847) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2848) if (data != 0 && data != ~(u64)0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2849) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2850) vcpu->arch.mcg_ctl = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2851) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2852) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2853) if (msr >= MSR_IA32_MC0_CTL &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2854) msr < MSR_IA32_MCx_CTL(bank_num)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2855) u32 offset = array_index_nospec(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2856) msr - MSR_IA32_MC0_CTL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2857) MSR_IA32_MCx_CTL(bank_num) - MSR_IA32_MC0_CTL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2858)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2859) /* only 0 or all 1s can be written to IA32_MCi_CTL
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2860) * some Linux kernels though clear bit 10 in bank 4 to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2861) * workaround a BIOS/GART TBL issue on AMD K8s, ignore
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2862) * this to avoid an uncatched #GP in the guest
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2863) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2864) if ((offset & 0x3) == 0 &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2865) data != 0 && (data | (1 << 10)) != ~(u64)0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2866) return -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2867)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2868) /* MCi_STATUS */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2869) if (!msr_info->host_initiated &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2870) (offset & 0x3) == 1 && data != 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2871) if (!can_set_mci_status(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2872) return -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2873) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2874)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2875) vcpu->arch.mce_banks[offset] = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2876) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2877) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2878) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2879) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2880) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2881) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2882)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2883) static int xen_hvm_config(struct kvm_vcpu *vcpu, u64 data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2884) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2885) struct kvm *kvm = vcpu->kvm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2886) int lm = is_long_mode(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2887) u8 *blob_addr = lm ? (u8 *)(long)kvm->arch.xen_hvm_config.blob_addr_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2888) : (u8 *)(long)kvm->arch.xen_hvm_config.blob_addr_32;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2889) u8 blob_size = lm ? kvm->arch.xen_hvm_config.blob_size_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2890) : kvm->arch.xen_hvm_config.blob_size_32;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2891) u32 page_num = data & ~PAGE_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2892) u64 page_addr = data & PAGE_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2893) u8 *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2894)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2895) if (page_num >= blob_size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2896) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2897)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2898) page = memdup_user(blob_addr + (page_num * PAGE_SIZE), PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2899) if (IS_ERR(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2900) return PTR_ERR(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2901)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2902) if (kvm_vcpu_write_guest(vcpu, page_addr, page, PAGE_SIZE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2903) kfree(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2904) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2905) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2906) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2907) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2908)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2909) static inline bool kvm_pv_async_pf_enabled(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2910) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2911) u64 mask = KVM_ASYNC_PF_ENABLED | KVM_ASYNC_PF_DELIVERY_AS_INT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2912)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2913) return (vcpu->arch.apf.msr_en_val & mask) == mask;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2914) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2915)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2916) static int kvm_pv_enable_async_pf(struct kvm_vcpu *vcpu, u64 data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2917) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2918) gpa_t gpa = data & ~0x3f;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2919)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2920) /* Bits 4:5 are reserved, Should be zero */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2921) if (data & 0x30)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2922) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2923)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2924) if (!guest_pv_has(vcpu, KVM_FEATURE_ASYNC_PF_VMEXIT) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2925) (data & KVM_ASYNC_PF_DELIVERY_AS_PF_VMEXIT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2926) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2927)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2928) if (!guest_pv_has(vcpu, KVM_FEATURE_ASYNC_PF_INT) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2929) (data & KVM_ASYNC_PF_DELIVERY_AS_INT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2930) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2931)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2932) if (!lapic_in_kernel(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2933) return data ? 1 : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2934)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2935) vcpu->arch.apf.msr_en_val = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2936)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2937) if (!kvm_pv_async_pf_enabled(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2938) kvm_clear_async_pf_completion_queue(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2939) kvm_async_pf_hash_reset(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2940) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2941) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2942)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2943) if (kvm_gfn_to_hva_cache_init(vcpu->kvm, &vcpu->arch.apf.data, gpa,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2944) sizeof(u64)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2945) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2946)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2947) vcpu->arch.apf.send_user_only = !(data & KVM_ASYNC_PF_SEND_ALWAYS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2948) vcpu->arch.apf.delivery_as_pf_vmexit = data & KVM_ASYNC_PF_DELIVERY_AS_PF_VMEXIT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2949)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2950) kvm_async_pf_wakeup_all(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2951)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2952) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2953) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2954)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2955) static int kvm_pv_enable_async_pf_int(struct kvm_vcpu *vcpu, u64 data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2956) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2957) /* Bits 8-63 are reserved */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2958) if (data >> 8)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2959) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2960)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2961) if (!lapic_in_kernel(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2962) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2963)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2964) vcpu->arch.apf.msr_int_val = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2965)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2966) vcpu->arch.apf.vec = data & KVM_ASYNC_PF_VEC_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2967)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2968) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2969) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2970)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2971) static void kvmclock_reset(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2972) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2973) vcpu->arch.pv_time_enabled = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2974) vcpu->arch.time = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2975) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2976)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2977) static void kvm_vcpu_flush_tlb_all(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2978) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2979) ++vcpu->stat.tlb_flush;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2980) kvm_x86_ops.tlb_flush_all(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2981) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2982)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2983) static void kvm_vcpu_flush_tlb_guest(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2984) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2985) ++vcpu->stat.tlb_flush;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2986) kvm_x86_ops.tlb_flush_guest(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2987) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2988)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2989) static void record_steal_time(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2990) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2991) struct kvm_host_map map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2992) struct kvm_steal_time *st;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2993)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2994) if (!(vcpu->arch.st.msr_val & KVM_MSR_ENABLED))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2995) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2996)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2997) /* -EAGAIN is returned in atomic context so we can just return. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2998) if (kvm_map_gfn(vcpu, vcpu->arch.st.msr_val >> PAGE_SHIFT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2999) &map, &vcpu->arch.st.cache, false))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3000) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3001)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3002) st = map.hva +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3003) offset_in_page(vcpu->arch.st.msr_val & KVM_STEAL_VALID_BITS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3004)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3005) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3006) * Doing a TLB flush here, on the guest's behalf, can avoid
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3007) * expensive IPIs.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3008) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3009) if (guest_pv_has(vcpu, KVM_FEATURE_PV_TLB_FLUSH)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3010) trace_kvm_pv_tlb_flush(vcpu->vcpu_id,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3011) st->preempted & KVM_VCPU_FLUSH_TLB);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3012) if (xchg(&st->preempted, 0) & KVM_VCPU_FLUSH_TLB)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3013) kvm_vcpu_flush_tlb_guest(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3014) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3015) st->preempted = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3016) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3017)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3018) vcpu->arch.st.preempted = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3019)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3020) if (st->version & 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3021) st->version += 1; /* first time write, random junk */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3022)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3023) st->version += 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3024)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3025) smp_wmb();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3026)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3027) st->steal += current->sched_info.run_delay -
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3028) vcpu->arch.st.last_steal;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3029) vcpu->arch.st.last_steal = current->sched_info.run_delay;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3030)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3031) smp_wmb();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3032)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3033) st->version += 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3034)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3035) kvm_unmap_gfn(vcpu, &map, &vcpu->arch.st.cache, true, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3036) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3037)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3038) int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3039) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3040) bool pr = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3041) u32 msr = msr_info->index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3042) u64 data = msr_info->data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3043)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3044) switch (msr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3045) case MSR_AMD64_NB_CFG:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3046) case MSR_IA32_UCODE_WRITE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3047) case MSR_VM_HSAVE_PA:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3048) case MSR_AMD64_PATCH_LOADER:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3049) case MSR_AMD64_BU_CFG2:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3050) case MSR_AMD64_DC_CFG:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3051) case MSR_F15H_EX_CFG:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3052) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3053)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3054) case MSR_IA32_UCODE_REV:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3055) if (msr_info->host_initiated)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3056) vcpu->arch.microcode_version = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3057) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3058) case MSR_IA32_ARCH_CAPABILITIES:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3059) if (!msr_info->host_initiated)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3060) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3061) vcpu->arch.arch_capabilities = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3062) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3063) case MSR_IA32_PERF_CAPABILITIES: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3064) struct kvm_msr_entry msr_ent = {.index = msr, .data = 0};
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3065)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3066) if (!msr_info->host_initiated)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3067) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3068) if (kvm_get_msr_feature(&msr_ent))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3069) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3070) if (data & ~msr_ent.data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3071) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3072)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3073) vcpu->arch.perf_capabilities = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3074)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3075) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3076) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3077) case MSR_EFER:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3078) return set_efer(vcpu, msr_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3079) case MSR_K7_HWCR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3080) data &= ~(u64)0x40; /* ignore flush filter disable */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3081) data &= ~(u64)0x100; /* ignore ignne emulation enable */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3082) data &= ~(u64)0x8; /* ignore TLB cache disable */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3083)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3084) /* Handle McStatusWrEn */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3085) if (data == BIT_ULL(18)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3086) vcpu->arch.msr_hwcr = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3087) } else if (data != 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3088) vcpu_unimpl(vcpu, "unimplemented HWCR wrmsr: 0x%llx\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3089) data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3090) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3091) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3092) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3093) case MSR_FAM10H_MMIO_CONF_BASE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3094) if (data != 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3095) vcpu_unimpl(vcpu, "unimplemented MMIO_CONF_BASE wrmsr: "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3096) "0x%llx\n", data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3097) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3098) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3099) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3100) case MSR_IA32_DEBUGCTLMSR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3101) if (!data) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3102) /* We support the non-activated case already */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3103) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3104) } else if (data & ~(DEBUGCTLMSR_LBR | DEBUGCTLMSR_BTF)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3105) /* Values other than LBR and BTF are vendor-specific,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3106) thus reserved and should throw a #GP */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3107) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3108) } else if (report_ignored_msrs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3109) vcpu_unimpl(vcpu, "%s: MSR_IA32_DEBUGCTLMSR 0x%llx, nop\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3110) __func__, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3111) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3112) case 0x200 ... 0x2ff:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3113) return kvm_mtrr_set_msr(vcpu, msr, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3114) case MSR_IA32_APICBASE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3115) return kvm_set_apic_base(vcpu, msr_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3116) case APIC_BASE_MSR ... APIC_BASE_MSR + 0xff:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3117) return kvm_x2apic_msr_write(vcpu, msr, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3118) case MSR_IA32_TSCDEADLINE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3119) kvm_set_lapic_tscdeadline_msr(vcpu, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3120) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3121) case MSR_IA32_TSC_ADJUST:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3122) if (guest_cpuid_has(vcpu, X86_FEATURE_TSC_ADJUST)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3123) if (!msr_info->host_initiated) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3124) s64 adj = data - vcpu->arch.ia32_tsc_adjust_msr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3125) adjust_tsc_offset_guest(vcpu, adj);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3126) /* Before back to guest, tsc_timestamp must be adjusted
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3127) * as well, otherwise guest's percpu pvclock time could jump.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3128) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3129) kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3130) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3131) vcpu->arch.ia32_tsc_adjust_msr = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3132) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3133) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3134) case MSR_IA32_MISC_ENABLE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3135) if (!kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_MISC_ENABLE_NO_MWAIT) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3136) ((vcpu->arch.ia32_misc_enable_msr ^ data) & MSR_IA32_MISC_ENABLE_MWAIT)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3137) if (!guest_cpuid_has(vcpu, X86_FEATURE_XMM3))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3138) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3139) vcpu->arch.ia32_misc_enable_msr = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3140) kvm_update_cpuid_runtime(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3141) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3142) vcpu->arch.ia32_misc_enable_msr = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3143) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3144) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3145) case MSR_IA32_SMBASE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3146) if (!msr_info->host_initiated)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3147) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3148) vcpu->arch.smbase = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3149) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3150) case MSR_IA32_POWER_CTL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3151) vcpu->arch.msr_ia32_power_ctl = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3152) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3153) case MSR_IA32_TSC:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3154) if (msr_info->host_initiated) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3155) kvm_synchronize_tsc(vcpu, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3156) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3157) u64 adj = kvm_compute_tsc_offset(vcpu, data) - vcpu->arch.l1_tsc_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3158) adjust_tsc_offset_guest(vcpu, adj);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3159) vcpu->arch.ia32_tsc_adjust_msr += adj;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3160) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3161) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3162) case MSR_IA32_XSS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3163) if (!msr_info->host_initiated &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3164) !guest_cpuid_has(vcpu, X86_FEATURE_XSAVES))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3165) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3166) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3167) * KVM supports exposing PT to the guest, but does not support
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3168) * IA32_XSS[bit 8]. Guests have to use RDMSR/WRMSR rather than
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3169) * XSAVES/XRSTORS to save/restore PT MSRs.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3170) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3171) if (data & ~supported_xss)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3172) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3173) vcpu->arch.ia32_xss = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3174) kvm_update_cpuid_runtime(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3175) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3176) case MSR_SMI_COUNT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3177) if (!msr_info->host_initiated)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3178) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3179) vcpu->arch.smi_count = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3180) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3181) case MSR_KVM_WALL_CLOCK_NEW:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3182) if (!guest_pv_has(vcpu, KVM_FEATURE_CLOCKSOURCE2))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3183) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3184)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3185) kvm_write_wall_clock(vcpu->kvm, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3186) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3187) case MSR_KVM_WALL_CLOCK:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3188) if (!guest_pv_has(vcpu, KVM_FEATURE_CLOCKSOURCE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3189) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3190)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3191) kvm_write_wall_clock(vcpu->kvm, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3192) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3193) case MSR_KVM_SYSTEM_TIME_NEW:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3194) if (!guest_pv_has(vcpu, KVM_FEATURE_CLOCKSOURCE2))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3195) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3196)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3197) kvm_write_system_time(vcpu, data, false, msr_info->host_initiated);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3198) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3199) case MSR_KVM_SYSTEM_TIME:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3200) if (!guest_pv_has(vcpu, KVM_FEATURE_CLOCKSOURCE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3201) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3202)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3203) kvm_write_system_time(vcpu, data, true, msr_info->host_initiated);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3204) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3205) case MSR_KVM_ASYNC_PF_EN:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3206) if (!guest_pv_has(vcpu, KVM_FEATURE_ASYNC_PF))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3207) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3208)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3209) if (kvm_pv_enable_async_pf(vcpu, data))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3210) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3211) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3212) case MSR_KVM_ASYNC_PF_INT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3213) if (!guest_pv_has(vcpu, KVM_FEATURE_ASYNC_PF_INT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3214) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3215)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3216) if (kvm_pv_enable_async_pf_int(vcpu, data))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3217) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3218) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3219) case MSR_KVM_ASYNC_PF_ACK:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3220) if (!guest_pv_has(vcpu, KVM_FEATURE_ASYNC_PF_INT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3221) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3222) if (data & 0x1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3223) vcpu->arch.apf.pageready_pending = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3224) kvm_check_async_pf_completion(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3225) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3226) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3227) case MSR_KVM_STEAL_TIME:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3228) if (!guest_pv_has(vcpu, KVM_FEATURE_STEAL_TIME))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3229) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3230)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3231) if (unlikely(!sched_info_on()))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3232) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3233)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3234) if (data & KVM_STEAL_RESERVED_MASK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3235) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3236)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3237) vcpu->arch.st.msr_val = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3238)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3239) if (!(data & KVM_MSR_ENABLED))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3240) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3241)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3242) kvm_make_request(KVM_REQ_STEAL_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3243)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3244) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3245) case MSR_KVM_PV_EOI_EN:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3246) if (!guest_pv_has(vcpu, KVM_FEATURE_PV_EOI))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3247) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3248)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3249) if (kvm_lapic_enable_pv_eoi(vcpu, data, sizeof(u8)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3250) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3251) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3252)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3253) case MSR_KVM_POLL_CONTROL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3254) if (!guest_pv_has(vcpu, KVM_FEATURE_POLL_CONTROL))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3255) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3256)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3257) /* only enable bit supported */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3258) if (data & (-1ULL << 1))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3259) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3260)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3261) vcpu->arch.msr_kvm_poll_control = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3262) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3263)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3264) case MSR_IA32_MCG_CTL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3265) case MSR_IA32_MCG_STATUS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3266) case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3267) return set_msr_mce(vcpu, msr_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3268)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3269) case MSR_K7_PERFCTR0 ... MSR_K7_PERFCTR3:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3270) case MSR_P6_PERFCTR0 ... MSR_P6_PERFCTR1:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3271) pr = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3272) fallthrough;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3273) case MSR_K7_EVNTSEL0 ... MSR_K7_EVNTSEL3:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3274) case MSR_P6_EVNTSEL0 ... MSR_P6_EVNTSEL1:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3275) if (kvm_pmu_is_valid_msr(vcpu, msr))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3276) return kvm_pmu_set_msr(vcpu, msr_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3277)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3278) if (pr || data != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3279) vcpu_unimpl(vcpu, "disabled perfctr wrmsr: "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3280) "0x%x data 0x%llx\n", msr, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3281) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3282) case MSR_K7_CLK_CTL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3283) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3284) * Ignore all writes to this no longer documented MSR.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3285) * Writes are only relevant for old K7 processors,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3286) * all pre-dating SVM, but a recommended workaround from
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3287) * AMD for these chips. It is possible to specify the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3288) * affected processor models on the command line, hence
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3289) * the need to ignore the workaround.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3290) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3291) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3292) case HV_X64_MSR_GUEST_OS_ID ... HV_X64_MSR_SINT15:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3293) case HV_X64_MSR_SYNDBG_CONTROL ... HV_X64_MSR_SYNDBG_PENDING_BUFFER:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3294) case HV_X64_MSR_SYNDBG_OPTIONS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3295) case HV_X64_MSR_CRASH_P0 ... HV_X64_MSR_CRASH_P4:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3296) case HV_X64_MSR_CRASH_CTL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3297) case HV_X64_MSR_STIMER0_CONFIG ... HV_X64_MSR_STIMER3_COUNT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3298) case HV_X64_MSR_REENLIGHTENMENT_CONTROL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3299) case HV_X64_MSR_TSC_EMULATION_CONTROL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3300) case HV_X64_MSR_TSC_EMULATION_STATUS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3301) return kvm_hv_set_msr_common(vcpu, msr, data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3302) msr_info->host_initiated);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3303) case MSR_IA32_BBL_CR_CTL3:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3304) /* Drop writes to this legacy MSR -- see rdmsr
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3305) * counterpart for further detail.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3306) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3307) if (report_ignored_msrs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3308) vcpu_unimpl(vcpu, "ignored wrmsr: 0x%x data 0x%llx\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3309) msr, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3310) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3311) case MSR_AMD64_OSVW_ID_LENGTH:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3312) if (!guest_cpuid_has(vcpu, X86_FEATURE_OSVW))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3313) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3314) vcpu->arch.osvw.length = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3315) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3316) case MSR_AMD64_OSVW_STATUS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3317) if (!guest_cpuid_has(vcpu, X86_FEATURE_OSVW))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3318) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3319) vcpu->arch.osvw.status = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3320) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3321) case MSR_PLATFORM_INFO:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3322) if (!msr_info->host_initiated ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3323) (!(data & MSR_PLATFORM_INFO_CPUID_FAULT) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3324) cpuid_fault_enabled(vcpu)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3325) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3326) vcpu->arch.msr_platform_info = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3327) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3328) case MSR_MISC_FEATURES_ENABLES:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3329) if (data & ~MSR_MISC_FEATURES_ENABLES_CPUID_FAULT ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3330) (data & MSR_MISC_FEATURES_ENABLES_CPUID_FAULT &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3331) !supports_cpuid_fault(vcpu)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3332) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3333) vcpu->arch.msr_misc_features_enables = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3334) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3335) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3336) if (msr && (msr == vcpu->kvm->arch.xen_hvm_config.msr))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3337) return xen_hvm_config(vcpu, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3338) if (kvm_pmu_is_valid_msr(vcpu, msr))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3339) return kvm_pmu_set_msr(vcpu, msr_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3340) return KVM_MSR_RET_INVALID;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3341) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3342) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3343) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3344) EXPORT_SYMBOL_GPL(kvm_set_msr_common);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3345)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3346) static int get_msr_mce(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata, bool host)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3347) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3348) u64 data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3349) u64 mcg_cap = vcpu->arch.mcg_cap;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3350) unsigned bank_num = mcg_cap & 0xff;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3351)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3352) switch (msr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3353) case MSR_IA32_P5_MC_ADDR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3354) case MSR_IA32_P5_MC_TYPE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3355) data = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3356) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3357) case MSR_IA32_MCG_CAP:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3358) data = vcpu->arch.mcg_cap;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3359) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3360) case MSR_IA32_MCG_CTL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3361) if (!(mcg_cap & MCG_CTL_P) && !host)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3362) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3363) data = vcpu->arch.mcg_ctl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3364) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3365) case MSR_IA32_MCG_STATUS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3366) data = vcpu->arch.mcg_status;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3367) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3368) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3369) if (msr >= MSR_IA32_MC0_CTL &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3370) msr < MSR_IA32_MCx_CTL(bank_num)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3371) u32 offset = array_index_nospec(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3372) msr - MSR_IA32_MC0_CTL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3373) MSR_IA32_MCx_CTL(bank_num) - MSR_IA32_MC0_CTL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3374)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3375) data = vcpu->arch.mce_banks[offset];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3376) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3377) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3378) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3379) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3380) *pdata = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3381) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3382) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3383)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3384) int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3385) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3386) switch (msr_info->index) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3387) case MSR_IA32_PLATFORM_ID:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3388) case MSR_IA32_EBL_CR_POWERON:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3389) case MSR_IA32_DEBUGCTLMSR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3390) case MSR_IA32_LASTBRANCHFROMIP:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3391) case MSR_IA32_LASTBRANCHTOIP:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3392) case MSR_IA32_LASTINTFROMIP:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3393) case MSR_IA32_LASTINTTOIP:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3394) case MSR_K8_SYSCFG:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3395) case MSR_K8_TSEG_ADDR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3396) case MSR_K8_TSEG_MASK:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3397) case MSR_VM_HSAVE_PA:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3398) case MSR_K8_INT_PENDING_MSG:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3399) case MSR_AMD64_NB_CFG:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3400) case MSR_FAM10H_MMIO_CONF_BASE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3401) case MSR_AMD64_BU_CFG2:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3402) case MSR_IA32_PERF_CTL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3403) case MSR_AMD64_DC_CFG:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3404) case MSR_F15H_EX_CFG:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3405) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3406) * Intel Sandy Bridge CPUs must support the RAPL (running average power
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3407) * limit) MSRs. Just return 0, as we do not want to expose the host
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3408) * data here. Do not conditionalize this on CPUID, as KVM does not do
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3409) * so for existing CPU-specific MSRs.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3410) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3411) case MSR_RAPL_POWER_UNIT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3412) case MSR_PP0_ENERGY_STATUS: /* Power plane 0 (core) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3413) case MSR_PP1_ENERGY_STATUS: /* Power plane 1 (graphics uncore) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3414) case MSR_PKG_ENERGY_STATUS: /* Total package */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3415) case MSR_DRAM_ENERGY_STATUS: /* DRAM controller */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3416) msr_info->data = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3417) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3418) case MSR_F15H_PERF_CTL0 ... MSR_F15H_PERF_CTR5:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3419) case MSR_K7_EVNTSEL0 ... MSR_K7_EVNTSEL3:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3420) case MSR_K7_PERFCTR0 ... MSR_K7_PERFCTR3:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3421) case MSR_P6_PERFCTR0 ... MSR_P6_PERFCTR1:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3422) case MSR_P6_EVNTSEL0 ... MSR_P6_EVNTSEL1:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3423) if (kvm_pmu_is_valid_msr(vcpu, msr_info->index))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3424) return kvm_pmu_get_msr(vcpu, msr_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3425) msr_info->data = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3426) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3427) case MSR_IA32_UCODE_REV:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3428) msr_info->data = vcpu->arch.microcode_version;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3429) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3430) case MSR_IA32_ARCH_CAPABILITIES:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3431) if (!msr_info->host_initiated &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3432) !guest_cpuid_has(vcpu, X86_FEATURE_ARCH_CAPABILITIES))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3433) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3434) msr_info->data = vcpu->arch.arch_capabilities;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3435) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3436) case MSR_IA32_PERF_CAPABILITIES:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3437) if (!msr_info->host_initiated &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3438) !guest_cpuid_has(vcpu, X86_FEATURE_PDCM))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3439) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3440) msr_info->data = vcpu->arch.perf_capabilities;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3441) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3442) case MSR_IA32_POWER_CTL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3443) msr_info->data = vcpu->arch.msr_ia32_power_ctl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3444) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3445) case MSR_IA32_TSC: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3446) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3447) * Intel SDM states that MSR_IA32_TSC read adds the TSC offset
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3448) * even when not intercepted. AMD manual doesn't explicitly
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3449) * state this but appears to behave the same.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3450) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3451) * On userspace reads and writes, however, we unconditionally
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3452) * return L1's TSC value to ensure backwards-compatible
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3453) * behavior for migration.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3454) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3455) u64 tsc_offset = msr_info->host_initiated ? vcpu->arch.l1_tsc_offset :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3456) vcpu->arch.tsc_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3457)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3458) msr_info->data = kvm_scale_tsc(vcpu, rdtsc()) + tsc_offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3459) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3460) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3461) case MSR_MTRRcap:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3462) case 0x200 ... 0x2ff:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3463) return kvm_mtrr_get_msr(vcpu, msr_info->index, &msr_info->data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3464) case 0xcd: /* fsb frequency */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3465) msr_info->data = 3;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3466) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3467) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3468) * MSR_EBC_FREQUENCY_ID
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3469) * Conservative value valid for even the basic CPU models.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3470) * Models 0,1: 000 in bits 23:21 indicating a bus speed of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3471) * 100MHz, model 2 000 in bits 18:16 indicating 100MHz,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3472) * and 266MHz for model 3, or 4. Set Core Clock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3473) * Frequency to System Bus Frequency Ratio to 1 (bits
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3474) * 31:24) even though these are only valid for CPU
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3475) * models > 2, however guests may end up dividing or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3476) * multiplying by zero otherwise.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3477) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3478) case MSR_EBC_FREQUENCY_ID:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3479) msr_info->data = 1 << 24;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3480) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3481) case MSR_IA32_APICBASE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3482) msr_info->data = kvm_get_apic_base(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3483) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3484) case APIC_BASE_MSR ... APIC_BASE_MSR + 0xff:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3485) return kvm_x2apic_msr_read(vcpu, msr_info->index, &msr_info->data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3486) case MSR_IA32_TSCDEADLINE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3487) msr_info->data = kvm_get_lapic_tscdeadline_msr(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3488) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3489) case MSR_IA32_TSC_ADJUST:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3490) msr_info->data = (u64)vcpu->arch.ia32_tsc_adjust_msr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3491) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3492) case MSR_IA32_MISC_ENABLE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3493) msr_info->data = vcpu->arch.ia32_misc_enable_msr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3494) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3495) case MSR_IA32_SMBASE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3496) if (!msr_info->host_initiated)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3497) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3498) msr_info->data = vcpu->arch.smbase;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3499) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3500) case MSR_SMI_COUNT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3501) msr_info->data = vcpu->arch.smi_count;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3502) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3503) case MSR_IA32_PERF_STATUS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3504) /* TSC increment by tick */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3505) msr_info->data = 1000ULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3506) /* CPU multiplier */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3507) msr_info->data |= (((uint64_t)4ULL) << 40);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3508) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3509) case MSR_EFER:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3510) msr_info->data = vcpu->arch.efer;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3511) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3512) case MSR_KVM_WALL_CLOCK:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3513) if (!guest_pv_has(vcpu, KVM_FEATURE_CLOCKSOURCE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3514) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3515)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3516) msr_info->data = vcpu->kvm->arch.wall_clock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3517) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3518) case MSR_KVM_WALL_CLOCK_NEW:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3519) if (!guest_pv_has(vcpu, KVM_FEATURE_CLOCKSOURCE2))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3520) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3521)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3522) msr_info->data = vcpu->kvm->arch.wall_clock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3523) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3524) case MSR_KVM_SYSTEM_TIME:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3525) if (!guest_pv_has(vcpu, KVM_FEATURE_CLOCKSOURCE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3526) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3527)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3528) msr_info->data = vcpu->arch.time;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3529) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3530) case MSR_KVM_SYSTEM_TIME_NEW:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3531) if (!guest_pv_has(vcpu, KVM_FEATURE_CLOCKSOURCE2))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3532) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3533)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3534) msr_info->data = vcpu->arch.time;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3535) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3536) case MSR_KVM_ASYNC_PF_EN:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3537) if (!guest_pv_has(vcpu, KVM_FEATURE_ASYNC_PF))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3538) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3539)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3540) msr_info->data = vcpu->arch.apf.msr_en_val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3541) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3542) case MSR_KVM_ASYNC_PF_INT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3543) if (!guest_pv_has(vcpu, KVM_FEATURE_ASYNC_PF_INT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3544) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3545)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3546) msr_info->data = vcpu->arch.apf.msr_int_val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3547) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3548) case MSR_KVM_ASYNC_PF_ACK:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3549) if (!guest_pv_has(vcpu, KVM_FEATURE_ASYNC_PF_INT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3550) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3551)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3552) msr_info->data = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3553) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3554) case MSR_KVM_STEAL_TIME:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3555) if (!guest_pv_has(vcpu, KVM_FEATURE_STEAL_TIME))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3556) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3557)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3558) msr_info->data = vcpu->arch.st.msr_val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3559) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3560) case MSR_KVM_PV_EOI_EN:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3561) if (!guest_pv_has(vcpu, KVM_FEATURE_PV_EOI))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3562) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3563)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3564) msr_info->data = vcpu->arch.pv_eoi.msr_val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3565) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3566) case MSR_KVM_POLL_CONTROL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3567) if (!guest_pv_has(vcpu, KVM_FEATURE_POLL_CONTROL))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3568) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3569)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3570) msr_info->data = vcpu->arch.msr_kvm_poll_control;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3571) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3572) case MSR_IA32_P5_MC_ADDR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3573) case MSR_IA32_P5_MC_TYPE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3574) case MSR_IA32_MCG_CAP:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3575) case MSR_IA32_MCG_CTL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3576) case MSR_IA32_MCG_STATUS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3577) case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3578) return get_msr_mce(vcpu, msr_info->index, &msr_info->data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3579) msr_info->host_initiated);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3580) case MSR_IA32_XSS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3581) if (!msr_info->host_initiated &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3582) !guest_cpuid_has(vcpu, X86_FEATURE_XSAVES))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3583) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3584) msr_info->data = vcpu->arch.ia32_xss;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3585) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3586) case MSR_K7_CLK_CTL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3587) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3588) * Provide expected ramp-up count for K7. All other
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3589) * are set to zero, indicating minimum divisors for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3590) * every field.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3591) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3592) * This prevents guest kernels on AMD host with CPU
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3593) * type 6, model 8 and higher from exploding due to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3594) * the rdmsr failing.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3595) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3596) msr_info->data = 0x20000000;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3597) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3598) case HV_X64_MSR_GUEST_OS_ID ... HV_X64_MSR_SINT15:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3599) case HV_X64_MSR_SYNDBG_CONTROL ... HV_X64_MSR_SYNDBG_PENDING_BUFFER:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3600) case HV_X64_MSR_SYNDBG_OPTIONS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3601) case HV_X64_MSR_CRASH_P0 ... HV_X64_MSR_CRASH_P4:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3602) case HV_X64_MSR_CRASH_CTL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3603) case HV_X64_MSR_STIMER0_CONFIG ... HV_X64_MSR_STIMER3_COUNT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3604) case HV_X64_MSR_REENLIGHTENMENT_CONTROL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3605) case HV_X64_MSR_TSC_EMULATION_CONTROL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3606) case HV_X64_MSR_TSC_EMULATION_STATUS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3607) return kvm_hv_get_msr_common(vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3608) msr_info->index, &msr_info->data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3609) msr_info->host_initiated);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3610) case MSR_IA32_BBL_CR_CTL3:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3611) /* This legacy MSR exists but isn't fully documented in current
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3612) * silicon. It is however accessed by winxp in very narrow
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3613) * scenarios where it sets bit #19, itself documented as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3614) * a "reserved" bit. Best effort attempt to source coherent
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3615) * read data here should the balance of the register be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3616) * interpreted by the guest:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3617) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3618) * L2 cache control register 3: 64GB range, 256KB size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3619) * enabled, latency 0x1, configured
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3620) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3621) msr_info->data = 0xbe702111;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3622) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3623) case MSR_AMD64_OSVW_ID_LENGTH:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3624) if (!guest_cpuid_has(vcpu, X86_FEATURE_OSVW))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3625) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3626) msr_info->data = vcpu->arch.osvw.length;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3627) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3628) case MSR_AMD64_OSVW_STATUS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3629) if (!guest_cpuid_has(vcpu, X86_FEATURE_OSVW))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3630) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3631) msr_info->data = vcpu->arch.osvw.status;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3632) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3633) case MSR_PLATFORM_INFO:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3634) if (!msr_info->host_initiated &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3635) !vcpu->kvm->arch.guest_can_read_msr_platform_info)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3636) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3637) msr_info->data = vcpu->arch.msr_platform_info;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3638) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3639) case MSR_MISC_FEATURES_ENABLES:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3640) msr_info->data = vcpu->arch.msr_misc_features_enables;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3641) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3642) case MSR_K7_HWCR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3643) msr_info->data = vcpu->arch.msr_hwcr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3644) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3645) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3646) if (kvm_pmu_is_valid_msr(vcpu, msr_info->index))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3647) return kvm_pmu_get_msr(vcpu, msr_info);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3648) return KVM_MSR_RET_INVALID;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3649) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3650) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3651) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3652) EXPORT_SYMBOL_GPL(kvm_get_msr_common);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3653)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3654) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3655) * Read or write a bunch of msrs. All parameters are kernel addresses.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3656) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3657) * @return number of msrs set successfully.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3658) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3659) static int __msr_io(struct kvm_vcpu *vcpu, struct kvm_msrs *msrs,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3660) struct kvm_msr_entry *entries,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3661) int (*do_msr)(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3662) unsigned index, u64 *data))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3663) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3664) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3665)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3666) for (i = 0; i < msrs->nmsrs; ++i)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3667) if (do_msr(vcpu, entries[i].index, &entries[i].data))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3668) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3669)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3670) return i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3671) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3672)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3673) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3674) * Read or write a bunch of msrs. Parameters are user addresses.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3675) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3676) * @return number of msrs set successfully.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3677) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3678) static int msr_io(struct kvm_vcpu *vcpu, struct kvm_msrs __user *user_msrs,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3679) int (*do_msr)(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3680) unsigned index, u64 *data),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3681) int writeback)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3682) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3683) struct kvm_msrs msrs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3684) struct kvm_msr_entry *entries;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3685) int r, n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3686) unsigned size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3687)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3688) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3689) if (copy_from_user(&msrs, user_msrs, sizeof(msrs)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3690) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3691)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3692) r = -E2BIG;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3693) if (msrs.nmsrs >= MAX_IO_MSRS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3694) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3695)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3696) size = sizeof(struct kvm_msr_entry) * msrs.nmsrs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3697) entries = memdup_user(user_msrs->entries, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3698) if (IS_ERR(entries)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3699) r = PTR_ERR(entries);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3700) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3701) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3702)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3703) r = n = __msr_io(vcpu, &msrs, entries, do_msr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3704) if (r < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3705) goto out_free;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3706)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3707) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3708) if (writeback && copy_to_user(user_msrs->entries, entries, size))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3709) goto out_free;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3710)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3711) r = n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3712)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3713) out_free:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3714) kfree(entries);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3715) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3716) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3717) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3718)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3719) static inline bool kvm_can_mwait_in_guest(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3720) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3721) return boot_cpu_has(X86_FEATURE_MWAIT) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3722) !boot_cpu_has_bug(X86_BUG_MONITOR) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3723) boot_cpu_has(X86_FEATURE_ARAT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3724) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3725)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3726) int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3727) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3728) int r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3729)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3730) switch (ext) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3731) case KVM_CAP_IRQCHIP:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3732) case KVM_CAP_HLT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3733) case KVM_CAP_MMU_SHADOW_CACHE_CONTROL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3734) case KVM_CAP_SET_TSS_ADDR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3735) case KVM_CAP_EXT_CPUID:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3736) case KVM_CAP_EXT_EMUL_CPUID:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3737) case KVM_CAP_CLOCKSOURCE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3738) case KVM_CAP_PIT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3739) case KVM_CAP_NOP_IO_DELAY:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3740) case KVM_CAP_MP_STATE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3741) case KVM_CAP_SYNC_MMU:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3742) case KVM_CAP_USER_NMI:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3743) case KVM_CAP_REINJECT_CONTROL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3744) case KVM_CAP_IRQ_INJECT_STATUS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3745) case KVM_CAP_IOEVENTFD:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3746) case KVM_CAP_IOEVENTFD_NO_LENGTH:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3747) case KVM_CAP_PIT2:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3748) case KVM_CAP_PIT_STATE2:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3749) case KVM_CAP_SET_IDENTITY_MAP_ADDR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3750) case KVM_CAP_XEN_HVM:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3751) case KVM_CAP_VCPU_EVENTS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3752) case KVM_CAP_HYPERV:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3753) case KVM_CAP_HYPERV_VAPIC:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3754) case KVM_CAP_HYPERV_SPIN:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3755) case KVM_CAP_HYPERV_SYNIC:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3756) case KVM_CAP_HYPERV_SYNIC2:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3757) case KVM_CAP_HYPERV_VP_INDEX:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3758) case KVM_CAP_HYPERV_EVENTFD:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3759) case KVM_CAP_HYPERV_TLBFLUSH:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3760) case KVM_CAP_HYPERV_SEND_IPI:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3761) case KVM_CAP_HYPERV_CPUID:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3762) case KVM_CAP_PCI_SEGMENT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3763) case KVM_CAP_DEBUGREGS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3764) case KVM_CAP_X86_ROBUST_SINGLESTEP:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3765) case KVM_CAP_XSAVE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3766) case KVM_CAP_ASYNC_PF:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3767) case KVM_CAP_ASYNC_PF_INT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3768) case KVM_CAP_GET_TSC_KHZ:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3769) case KVM_CAP_KVMCLOCK_CTRL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3770) case KVM_CAP_READONLY_MEM:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3771) case KVM_CAP_HYPERV_TIME:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3772) case KVM_CAP_IOAPIC_POLARITY_IGNORED:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3773) case KVM_CAP_TSC_DEADLINE_TIMER:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3774) case KVM_CAP_DISABLE_QUIRKS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3775) case KVM_CAP_SET_BOOT_CPU_ID:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3776) case KVM_CAP_SPLIT_IRQCHIP:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3777) case KVM_CAP_IMMEDIATE_EXIT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3778) case KVM_CAP_PMU_EVENT_FILTER:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3779) case KVM_CAP_GET_MSR_FEATURES:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3780) case KVM_CAP_MSR_PLATFORM_INFO:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3781) case KVM_CAP_EXCEPTION_PAYLOAD:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3782) case KVM_CAP_SET_GUEST_DEBUG:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3783) case KVM_CAP_LAST_CPU:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3784) case KVM_CAP_X86_USER_SPACE_MSR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3785) case KVM_CAP_X86_MSR_FILTER:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3786) case KVM_CAP_ENFORCE_PV_FEATURE_CPUID:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3787) r = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3788) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3789) case KVM_CAP_SYNC_REGS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3790) r = KVM_SYNC_X86_VALID_FIELDS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3791) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3792) case KVM_CAP_ADJUST_CLOCK:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3793) r = KVM_CLOCK_TSC_STABLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3794) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3795) case KVM_CAP_X86_DISABLE_EXITS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3796) r |= KVM_X86_DISABLE_EXITS_HLT | KVM_X86_DISABLE_EXITS_PAUSE |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3797) KVM_X86_DISABLE_EXITS_CSTATE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3798) if(kvm_can_mwait_in_guest())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3799) r |= KVM_X86_DISABLE_EXITS_MWAIT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3800) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3801) case KVM_CAP_X86_SMM:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3802) /* SMBASE is usually relocated above 1M on modern chipsets,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3803) * and SMM handlers might indeed rely on 4G segment limits,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3804) * so do not report SMM to be available if real mode is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3805) * emulated via vm86 mode. Still, do not go to great lengths
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3806) * to avoid userspace's usage of the feature, because it is a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3807) * fringe case that is not enabled except via specific settings
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3808) * of the module parameters.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3809) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3810) r = kvm_x86_ops.has_emulated_msr(MSR_IA32_SMBASE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3811) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3812) case KVM_CAP_VAPIC:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3813) r = !kvm_x86_ops.cpu_has_accelerated_tpr();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3814) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3815) case KVM_CAP_NR_VCPUS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3816) r = KVM_SOFT_MAX_VCPUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3817) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3818) case KVM_CAP_MAX_VCPUS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3819) r = KVM_MAX_VCPUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3820) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3821) case KVM_CAP_MAX_VCPU_ID:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3822) r = KVM_MAX_VCPU_ID;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3823) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3824) case KVM_CAP_PV_MMU: /* obsolete */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3825) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3826) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3827) case KVM_CAP_MCE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3828) r = KVM_MAX_MCE_BANKS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3829) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3830) case KVM_CAP_XCRS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3831) r = boot_cpu_has(X86_FEATURE_XSAVE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3832) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3833) case KVM_CAP_TSC_CONTROL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3834) r = kvm_has_tsc_control;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3835) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3836) case KVM_CAP_X2APIC_API:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3837) r = KVM_X2APIC_API_VALID_FLAGS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3838) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3839) case KVM_CAP_NESTED_STATE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3840) r = kvm_x86_ops.nested_ops->get_state ?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3841) kvm_x86_ops.nested_ops->get_state(NULL, NULL, 0) : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3842) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3843) case KVM_CAP_HYPERV_DIRECT_TLBFLUSH:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3844) r = kvm_x86_ops.enable_direct_tlbflush != NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3845) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3846) case KVM_CAP_HYPERV_ENLIGHTENED_VMCS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3847) r = kvm_x86_ops.nested_ops->enable_evmcs != NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3848) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3849) case KVM_CAP_SMALLER_MAXPHYADDR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3850) r = (int) allow_smaller_maxphyaddr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3851) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3852) case KVM_CAP_STEAL_TIME:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3853) r = sched_info_on();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3854) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3855) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3856) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3857) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3858) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3859)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3860) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3861)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3862) long kvm_arch_dev_ioctl(struct file *filp,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3863) unsigned int ioctl, unsigned long arg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3864) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3865) void __user *argp = (void __user *)arg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3866) long r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3867)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3868) switch (ioctl) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3869) case KVM_GET_MSR_INDEX_LIST: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3870) struct kvm_msr_list __user *user_msr_list = argp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3871) struct kvm_msr_list msr_list;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3872) unsigned n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3873)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3874) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3875) if (copy_from_user(&msr_list, user_msr_list, sizeof(msr_list)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3876) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3877) n = msr_list.nmsrs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3878) msr_list.nmsrs = num_msrs_to_save + num_emulated_msrs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3879) if (copy_to_user(user_msr_list, &msr_list, sizeof(msr_list)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3880) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3881) r = -E2BIG;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3882) if (n < msr_list.nmsrs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3883) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3884) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3885) if (copy_to_user(user_msr_list->indices, &msrs_to_save,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3886) num_msrs_to_save * sizeof(u32)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3887) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3888) if (copy_to_user(user_msr_list->indices + num_msrs_to_save,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3889) &emulated_msrs,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3890) num_emulated_msrs * sizeof(u32)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3891) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3892) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3893) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3894) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3895) case KVM_GET_SUPPORTED_CPUID:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3896) case KVM_GET_EMULATED_CPUID: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3897) struct kvm_cpuid2 __user *cpuid_arg = argp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3898) struct kvm_cpuid2 cpuid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3899)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3900) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3901) if (copy_from_user(&cpuid, cpuid_arg, sizeof(cpuid)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3902) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3903)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3904) r = kvm_dev_ioctl_get_cpuid(&cpuid, cpuid_arg->entries,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3905) ioctl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3906) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3907) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3908)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3909) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3910) if (copy_to_user(cpuid_arg, &cpuid, sizeof(cpuid)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3911) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3912) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3913) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3914) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3915) case KVM_X86_GET_MCE_CAP_SUPPORTED:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3916) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3917) if (copy_to_user(argp, &kvm_mce_cap_supported,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3918) sizeof(kvm_mce_cap_supported)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3919) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3920) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3921) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3922) case KVM_GET_MSR_FEATURE_INDEX_LIST: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3923) struct kvm_msr_list __user *user_msr_list = argp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3924) struct kvm_msr_list msr_list;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3925) unsigned int n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3926)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3927) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3928) if (copy_from_user(&msr_list, user_msr_list, sizeof(msr_list)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3929) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3930) n = msr_list.nmsrs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3931) msr_list.nmsrs = num_msr_based_features;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3932) if (copy_to_user(user_msr_list, &msr_list, sizeof(msr_list)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3933) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3934) r = -E2BIG;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3935) if (n < msr_list.nmsrs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3936) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3937) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3938) if (copy_to_user(user_msr_list->indices, &msr_based_features,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3939) num_msr_based_features * sizeof(u32)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3940) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3941) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3942) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3943) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3944) case KVM_GET_MSRS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3945) r = msr_io(NULL, argp, do_get_msr_feature, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3946) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3947) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3948) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3949) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3950) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3951) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3952) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3953) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3954)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3955) static void wbinvd_ipi(void *garbage)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3956) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3957) wbinvd();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3958) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3959)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3960) static bool need_emulate_wbinvd(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3961) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3962) return kvm_arch_has_noncoherent_dma(vcpu->kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3963) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3964)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3965) void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3966) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3967) /* Address WBINVD may be executed by guest */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3968) if (need_emulate_wbinvd(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3969) if (kvm_x86_ops.has_wbinvd_exit())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3970) cpumask_set_cpu(cpu, vcpu->arch.wbinvd_dirty_mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3971) else if (vcpu->cpu != -1 && vcpu->cpu != cpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3972) smp_call_function_single(vcpu->cpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3973) wbinvd_ipi, NULL, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3974) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3975)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3976) kvm_x86_ops.vcpu_load(vcpu, cpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3977)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3978) /* Save host pkru register if supported */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3979) vcpu->arch.host_pkru = read_pkru();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3980)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3981) /* Apply any externally detected TSC adjustments (due to suspend) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3982) if (unlikely(vcpu->arch.tsc_offset_adjustment)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3983) adjust_tsc_offset_host(vcpu, vcpu->arch.tsc_offset_adjustment);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3984) vcpu->arch.tsc_offset_adjustment = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3985) kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3986) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3987)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3988) if (unlikely(vcpu->cpu != cpu) || kvm_check_tsc_unstable()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3989) s64 tsc_delta = !vcpu->arch.last_host_tsc ? 0 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3990) rdtsc() - vcpu->arch.last_host_tsc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3991) if (tsc_delta < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3992) mark_tsc_unstable("KVM discovered backwards TSC");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3993)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3994) if (kvm_check_tsc_unstable()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3995) u64 offset = kvm_compute_tsc_offset(vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3996) vcpu->arch.last_guest_tsc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3997) kvm_vcpu_write_tsc_offset(vcpu, offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3998) vcpu->arch.tsc_catchup = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3999) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4000)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4001) if (kvm_lapic_hv_timer_in_use(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4002) kvm_lapic_restart_hv_timer(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4003)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4004) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4005) * On a host with synchronized TSC, there is no need to update
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4006) * kvmclock on vcpu->cpu migration
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4007) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4008) if (!vcpu->kvm->arch.use_master_clock || vcpu->cpu == -1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4009) kvm_make_request(KVM_REQ_GLOBAL_CLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4010) if (vcpu->cpu != cpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4011) kvm_make_request(KVM_REQ_MIGRATE_TIMER, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4012) vcpu->cpu = cpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4013) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4014)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4015) kvm_make_request(KVM_REQ_STEAL_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4016) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4017)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4018) static void kvm_steal_time_set_preempted(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4019) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4020) struct kvm_host_map map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4021) struct kvm_steal_time *st;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4022)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4023) if (!(vcpu->arch.st.msr_val & KVM_MSR_ENABLED))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4024) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4025)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4026) if (vcpu->arch.st.preempted)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4027) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4028)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4029) if (kvm_map_gfn(vcpu, vcpu->arch.st.msr_val >> PAGE_SHIFT, &map,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4030) &vcpu->arch.st.cache, true))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4031) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4032)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4033) st = map.hva +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4034) offset_in_page(vcpu->arch.st.msr_val & KVM_STEAL_VALID_BITS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4035)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4036) st->preempted = vcpu->arch.st.preempted = KVM_VCPU_PREEMPTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4037)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4038) kvm_unmap_gfn(vcpu, &map, &vcpu->arch.st.cache, true, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4039) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4040)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4041) void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4042) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4043) int idx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4044)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4045) if (vcpu->preempted)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4046) vcpu->arch.preempted_in_kernel = !kvm_x86_ops.get_cpl(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4047)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4048) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4049) * Disable page faults because we're in atomic context here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4050) * kvm_write_guest_offset_cached() would call might_fault()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4051) * that relies on pagefault_disable() to tell if there's a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4052) * bug. NOTE: the write to guest memory may not go through if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4053) * during postcopy live migration or if there's heavy guest
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4054) * paging.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4055) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4056) pagefault_disable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4057) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4058) * kvm_memslots() will be called by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4059) * kvm_write_guest_offset_cached() so take the srcu lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4060) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4061) idx = srcu_read_lock(&vcpu->kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4062) kvm_steal_time_set_preempted(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4063) srcu_read_unlock(&vcpu->kvm->srcu, idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4064) pagefault_enable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4065) kvm_x86_ops.vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4066) vcpu->arch.last_host_tsc = rdtsc();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4067) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4068) * If userspace has set any breakpoints or watchpoints, dr6 is restored
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4069) * on every vmexit, but if not, we might have a stale dr6 from the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4070) * guest. do_debug expects dr6 to be cleared after it runs, do the same.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4071) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4072) set_debugreg(0, 6);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4073) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4074)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4075) static int kvm_vcpu_ioctl_get_lapic(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4076) struct kvm_lapic_state *s)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4077) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4078) if (vcpu->arch.apicv_active)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4079) kvm_x86_ops.sync_pir_to_irr(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4080)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4081) return kvm_apic_get_state(vcpu, s);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4082) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4083)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4084) static int kvm_vcpu_ioctl_set_lapic(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4085) struct kvm_lapic_state *s)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4086) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4087) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4088)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4089) r = kvm_apic_set_state(vcpu, s);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4090) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4091) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4092) update_cr8_intercept(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4093)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4094) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4095) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4096)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4097) static int kvm_cpu_accept_dm_intr(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4098) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4099) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4100) * We can accept userspace's request for interrupt injection
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4101) * as long as we have a place to store the interrupt number.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4102) * The actual injection will happen when the CPU is able to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4103) * deliver the interrupt.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4104) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4105) if (kvm_cpu_has_extint(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4106) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4107)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4108) /* Acknowledging ExtINT does not happen if LINT0 is masked. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4109) return (!lapic_in_kernel(vcpu) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4110) kvm_apic_accept_pic_intr(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4111) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4112)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4113) static int kvm_vcpu_ready_for_interrupt_injection(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4114) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4115) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4116) * Do not cause an interrupt window exit if an exception
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4117) * is pending or an event needs reinjection; userspace
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4118) * might want to inject the interrupt manually using KVM_SET_REGS
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4119) * or KVM_SET_SREGS. For that to work, we must be at an
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4120) * instruction boundary and with no events half-injected.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4121) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4122) return (kvm_arch_interrupt_allowed(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4123) kvm_cpu_accept_dm_intr(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4124) !kvm_event_needs_reinjection(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4125) !vcpu->arch.exception.pending);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4126) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4127)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4128) static int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4129) struct kvm_interrupt *irq)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4130) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4131) if (irq->irq >= KVM_NR_INTERRUPTS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4132) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4133)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4134) if (!irqchip_in_kernel(vcpu->kvm)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4135) kvm_queue_interrupt(vcpu, irq->irq, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4136) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4137) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4138) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4139)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4140) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4141) * With in-kernel LAPIC, we only use this to inject EXTINT, so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4142) * fail for in-kernel 8259.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4143) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4144) if (pic_in_kernel(vcpu->kvm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4145) return -ENXIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4146)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4147) if (vcpu->arch.pending_external_vector != -1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4148) return -EEXIST;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4149)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4150) vcpu->arch.pending_external_vector = irq->irq;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4151) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4152) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4153) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4154)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4155) static int kvm_vcpu_ioctl_nmi(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4156) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4157) kvm_inject_nmi(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4158)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4159) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4160) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4161)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4162) static int kvm_vcpu_ioctl_smi(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4163) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4164) kvm_make_request(KVM_REQ_SMI, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4165)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4166) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4167) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4168)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4169) static int vcpu_ioctl_tpr_access_reporting(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4170) struct kvm_tpr_access_ctl *tac)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4171) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4172) if (tac->flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4173) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4174) vcpu->arch.tpr_access_reporting = !!tac->enabled;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4175) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4176) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4177)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4178) static int kvm_vcpu_ioctl_x86_setup_mce(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4179) u64 mcg_cap)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4180) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4181) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4182) unsigned bank_num = mcg_cap & 0xff, bank;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4183)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4184) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4185) if (!bank_num || bank_num > KVM_MAX_MCE_BANKS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4186) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4187) if (mcg_cap & ~(kvm_mce_cap_supported | 0xff | 0xff0000))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4188) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4189) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4190) vcpu->arch.mcg_cap = mcg_cap;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4191) /* Init IA32_MCG_CTL to all 1s */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4192) if (mcg_cap & MCG_CTL_P)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4193) vcpu->arch.mcg_ctl = ~(u64)0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4194) /* Init IA32_MCi_CTL to all 1s */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4195) for (bank = 0; bank < bank_num; bank++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4196) vcpu->arch.mce_banks[bank*4] = ~(u64)0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4197)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4198) kvm_x86_ops.setup_mce(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4199) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4200) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4201) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4202)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4203) static int kvm_vcpu_ioctl_x86_set_mce(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4204) struct kvm_x86_mce *mce)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4205) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4206) u64 mcg_cap = vcpu->arch.mcg_cap;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4207) unsigned bank_num = mcg_cap & 0xff;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4208) u64 *banks = vcpu->arch.mce_banks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4209)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4210) if (mce->bank >= bank_num || !(mce->status & MCI_STATUS_VAL))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4211) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4212) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4213) * if IA32_MCG_CTL is not all 1s, the uncorrected error
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4214) * reporting is disabled
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4215) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4216) if ((mce->status & MCI_STATUS_UC) && (mcg_cap & MCG_CTL_P) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4217) vcpu->arch.mcg_ctl != ~(u64)0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4218) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4219) banks += 4 * mce->bank;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4220) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4221) * if IA32_MCi_CTL is not all 1s, the uncorrected error
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4222) * reporting is disabled for the bank
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4223) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4224) if ((mce->status & MCI_STATUS_UC) && banks[0] != ~(u64)0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4225) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4226) if (mce->status & MCI_STATUS_UC) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4227) if ((vcpu->arch.mcg_status & MCG_STATUS_MCIP) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4228) !kvm_read_cr4_bits(vcpu, X86_CR4_MCE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4229) kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4230) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4231) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4232) if (banks[1] & MCI_STATUS_VAL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4233) mce->status |= MCI_STATUS_OVER;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4234) banks[2] = mce->addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4235) banks[3] = mce->misc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4236) vcpu->arch.mcg_status = mce->mcg_status;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4237) banks[1] = mce->status;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4238) kvm_queue_exception(vcpu, MC_VECTOR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4239) } else if (!(banks[1] & MCI_STATUS_VAL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4240) || !(banks[1] & MCI_STATUS_UC)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4241) if (banks[1] & MCI_STATUS_VAL)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4242) mce->status |= MCI_STATUS_OVER;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4243) banks[2] = mce->addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4244) banks[3] = mce->misc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4245) banks[1] = mce->status;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4246) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4247) banks[1] |= MCI_STATUS_OVER;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4248) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4249) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4250)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4251) static void kvm_vcpu_ioctl_x86_get_vcpu_events(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4252) struct kvm_vcpu_events *events)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4253) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4254) process_nmi(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4255)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4256) if (kvm_check_request(KVM_REQ_SMI, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4257) process_smi(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4258)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4259) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4260) * In guest mode, payload delivery should be deferred,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4261) * so that the L1 hypervisor can intercept #PF before
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4262) * CR2 is modified (or intercept #DB before DR6 is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4263) * modified under nVMX). Unless the per-VM capability,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4264) * KVM_CAP_EXCEPTION_PAYLOAD, is set, we may not defer the delivery of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4265) * an exception payload and handle after a KVM_GET_VCPU_EVENTS. Since we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4266) * opportunistically defer the exception payload, deliver it if the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4267) * capability hasn't been requested before processing a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4268) * KVM_GET_VCPU_EVENTS.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4269) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4270) if (!vcpu->kvm->arch.exception_payload_enabled &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4271) vcpu->arch.exception.pending && vcpu->arch.exception.has_payload)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4272) kvm_deliver_exception_payload(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4273)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4274) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4275) * The API doesn't provide the instruction length for software
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4276) * exceptions, so don't report them. As long as the guest RIP
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4277) * isn't advanced, we should expect to encounter the exception
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4278) * again.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4279) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4280) if (kvm_exception_is_soft(vcpu->arch.exception.nr)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4281) events->exception.injected = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4282) events->exception.pending = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4283) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4284) events->exception.injected = vcpu->arch.exception.injected;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4285) events->exception.pending = vcpu->arch.exception.pending;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4286) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4287) * For ABI compatibility, deliberately conflate
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4288) * pending and injected exceptions when
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4289) * KVM_CAP_EXCEPTION_PAYLOAD isn't enabled.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4290) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4291) if (!vcpu->kvm->arch.exception_payload_enabled)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4292) events->exception.injected |=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4293) vcpu->arch.exception.pending;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4294) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4295) events->exception.nr = vcpu->arch.exception.nr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4296) events->exception.has_error_code = vcpu->arch.exception.has_error_code;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4297) events->exception.error_code = vcpu->arch.exception.error_code;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4298) events->exception_has_payload = vcpu->arch.exception.has_payload;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4299) events->exception_payload = vcpu->arch.exception.payload;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4300)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4301) events->interrupt.injected =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4302) vcpu->arch.interrupt.injected && !vcpu->arch.interrupt.soft;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4303) events->interrupt.nr = vcpu->arch.interrupt.nr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4304) events->interrupt.soft = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4305) events->interrupt.shadow = kvm_x86_ops.get_interrupt_shadow(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4306)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4307) events->nmi.injected = vcpu->arch.nmi_injected;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4308) events->nmi.pending = vcpu->arch.nmi_pending != 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4309) events->nmi.masked = kvm_x86_ops.get_nmi_mask(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4310) events->nmi.pad = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4311)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4312) events->sipi_vector = 0; /* never valid when reporting to user space */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4313)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4314) events->smi.smm = is_smm(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4315) events->smi.pending = vcpu->arch.smi_pending;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4316) events->smi.smm_inside_nmi =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4317) !!(vcpu->arch.hflags & HF_SMM_INSIDE_NMI_MASK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4318) events->smi.latched_init = kvm_lapic_latched_init(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4319)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4320) events->flags = (KVM_VCPUEVENT_VALID_NMI_PENDING
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4321) | KVM_VCPUEVENT_VALID_SHADOW
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4322) | KVM_VCPUEVENT_VALID_SMM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4323) if (vcpu->kvm->arch.exception_payload_enabled)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4324) events->flags |= KVM_VCPUEVENT_VALID_PAYLOAD;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4325)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4326) memset(&events->reserved, 0, sizeof(events->reserved));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4327) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4328)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4329) static void kvm_smm_changed(struct kvm_vcpu *vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4330)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4331) static int kvm_vcpu_ioctl_x86_set_vcpu_events(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4332) struct kvm_vcpu_events *events)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4333) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4334) if (events->flags & ~(KVM_VCPUEVENT_VALID_NMI_PENDING
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4335) | KVM_VCPUEVENT_VALID_SIPI_VECTOR
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4336) | KVM_VCPUEVENT_VALID_SHADOW
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4337) | KVM_VCPUEVENT_VALID_SMM
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4338) | KVM_VCPUEVENT_VALID_PAYLOAD))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4339) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4340)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4341) if (events->flags & KVM_VCPUEVENT_VALID_PAYLOAD) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4342) if (!vcpu->kvm->arch.exception_payload_enabled)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4343) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4344) if (events->exception.pending)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4345) events->exception.injected = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4346) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4347) events->exception_has_payload = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4348) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4349) events->exception.pending = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4350) events->exception_has_payload = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4351) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4352)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4353) if ((events->exception.injected || events->exception.pending) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4354) (events->exception.nr > 31 || events->exception.nr == NMI_VECTOR))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4355) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4356)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4357) /* INITs are latched while in SMM */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4358) if (events->flags & KVM_VCPUEVENT_VALID_SMM &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4359) (events->smi.smm || events->smi.pending) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4360) vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4361) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4362)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4363) process_nmi(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4364) vcpu->arch.exception.injected = events->exception.injected;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4365) vcpu->arch.exception.pending = events->exception.pending;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4366) vcpu->arch.exception.nr = events->exception.nr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4367) vcpu->arch.exception.has_error_code = events->exception.has_error_code;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4368) vcpu->arch.exception.error_code = events->exception.error_code;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4369) vcpu->arch.exception.has_payload = events->exception_has_payload;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4370) vcpu->arch.exception.payload = events->exception_payload;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4371)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4372) vcpu->arch.interrupt.injected = events->interrupt.injected;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4373) vcpu->arch.interrupt.nr = events->interrupt.nr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4374) vcpu->arch.interrupt.soft = events->interrupt.soft;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4375) if (events->flags & KVM_VCPUEVENT_VALID_SHADOW)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4376) kvm_x86_ops.set_interrupt_shadow(vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4377) events->interrupt.shadow);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4378)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4379) vcpu->arch.nmi_injected = events->nmi.injected;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4380) if (events->flags & KVM_VCPUEVENT_VALID_NMI_PENDING)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4381) vcpu->arch.nmi_pending = events->nmi.pending;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4382) kvm_x86_ops.set_nmi_mask(vcpu, events->nmi.masked);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4383)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4384) if (events->flags & KVM_VCPUEVENT_VALID_SIPI_VECTOR &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4385) lapic_in_kernel(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4386) vcpu->arch.apic->sipi_vector = events->sipi_vector;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4387)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4388) if (events->flags & KVM_VCPUEVENT_VALID_SMM) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4389) if (!!(vcpu->arch.hflags & HF_SMM_MASK) != events->smi.smm) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4390) if (events->smi.smm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4391) vcpu->arch.hflags |= HF_SMM_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4392) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4393) vcpu->arch.hflags &= ~HF_SMM_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4394)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4395) kvm_x86_ops.nested_ops->leave_nested(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4396) kvm_smm_changed(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4397) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4398)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4399) vcpu->arch.smi_pending = events->smi.pending;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4400)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4401) if (events->smi.smm) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4402) if (events->smi.smm_inside_nmi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4403) vcpu->arch.hflags |= HF_SMM_INSIDE_NMI_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4404) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4405) vcpu->arch.hflags &= ~HF_SMM_INSIDE_NMI_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4406) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4407)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4408) if (lapic_in_kernel(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4409) if (events->smi.latched_init)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4410) set_bit(KVM_APIC_INIT, &vcpu->arch.apic->pending_events);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4411) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4412) clear_bit(KVM_APIC_INIT, &vcpu->arch.apic->pending_events);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4413) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4414) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4415)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4416) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4417)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4418) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4419) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4420)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4421) static void kvm_vcpu_ioctl_x86_get_debugregs(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4422) struct kvm_debugregs *dbgregs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4423) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4424) unsigned long val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4425)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4426) memcpy(dbgregs->db, vcpu->arch.db, sizeof(vcpu->arch.db));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4427) kvm_get_dr(vcpu, 6, &val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4428) dbgregs->dr6 = val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4429) dbgregs->dr7 = vcpu->arch.dr7;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4430) dbgregs->flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4431) memset(&dbgregs->reserved, 0, sizeof(dbgregs->reserved));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4432) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4433)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4434) static int kvm_vcpu_ioctl_x86_set_debugregs(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4435) struct kvm_debugregs *dbgregs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4436) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4437) if (dbgregs->flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4438) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4439)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4440) if (dbgregs->dr6 & ~0xffffffffull)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4441) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4442) if (dbgregs->dr7 & ~0xffffffffull)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4443) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4444)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4445) memcpy(vcpu->arch.db, dbgregs->db, sizeof(vcpu->arch.db));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4446) kvm_update_dr0123(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4447) vcpu->arch.dr6 = dbgregs->dr6;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4448) vcpu->arch.dr7 = dbgregs->dr7;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4449) kvm_update_dr7(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4450)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4451) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4452) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4453)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4454) #define XSTATE_COMPACTION_ENABLED (1ULL << 63)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4455)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4456) static void fill_xsave(u8 *dest, struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4457) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4458) struct xregs_state *xsave = &vcpu->arch.guest_fpu->state.xsave;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4459) u64 xstate_bv = xsave->header.xfeatures;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4460) u64 valid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4461)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4462) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4463) * Copy legacy XSAVE area, to avoid complications with CPUID
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4464) * leaves 0 and 1 in the loop below.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4465) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4466) memcpy(dest, xsave, XSAVE_HDR_OFFSET);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4467)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4468) /* Set XSTATE_BV */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4469) xstate_bv &= vcpu->arch.guest_supported_xcr0 | XFEATURE_MASK_FPSSE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4470) *(u64 *)(dest + XSAVE_HDR_OFFSET) = xstate_bv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4471)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4472) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4473) * Copy each region from the possibly compacted offset to the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4474) * non-compacted offset.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4475) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4476) valid = xstate_bv & ~XFEATURE_MASK_FPSSE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4477) while (valid) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4478) u64 xfeature_mask = valid & -valid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4479) int xfeature_nr = fls64(xfeature_mask) - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4480) void *src = get_xsave_addr(xsave, xfeature_nr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4481)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4482) if (src) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4483) u32 size, offset, ecx, edx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4484) cpuid_count(XSTATE_CPUID, xfeature_nr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4485) &size, &offset, &ecx, &edx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4486) if (xfeature_nr == XFEATURE_PKRU)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4487) memcpy(dest + offset, &vcpu->arch.pkru,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4488) sizeof(vcpu->arch.pkru));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4489) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4490) memcpy(dest + offset, src, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4491)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4492) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4493)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4494) valid -= xfeature_mask;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4495) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4496) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4497)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4498) static void load_xsave(struct kvm_vcpu *vcpu, u8 *src)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4499) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4500) struct xregs_state *xsave = &vcpu->arch.guest_fpu->state.xsave;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4501) u64 xstate_bv = *(u64 *)(src + XSAVE_HDR_OFFSET);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4502) u64 valid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4503)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4504) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4505) * Copy legacy XSAVE area, to avoid complications with CPUID
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4506) * leaves 0 and 1 in the loop below.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4507) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4508) memcpy(xsave, src, XSAVE_HDR_OFFSET);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4509)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4510) /* Set XSTATE_BV and possibly XCOMP_BV. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4511) xsave->header.xfeatures = xstate_bv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4512) if (boot_cpu_has(X86_FEATURE_XSAVES))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4513) xsave->header.xcomp_bv = host_xcr0 | XSTATE_COMPACTION_ENABLED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4514)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4515) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4516) * Copy each region from the non-compacted offset to the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4517) * possibly compacted offset.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4518) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4519) valid = xstate_bv & ~XFEATURE_MASK_FPSSE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4520) while (valid) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4521) u64 xfeature_mask = valid & -valid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4522) int xfeature_nr = fls64(xfeature_mask) - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4523) void *dest = get_xsave_addr(xsave, xfeature_nr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4524)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4525) if (dest) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4526) u32 size, offset, ecx, edx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4527) cpuid_count(XSTATE_CPUID, xfeature_nr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4528) &size, &offset, &ecx, &edx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4529) if (xfeature_nr == XFEATURE_PKRU)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4530) memcpy(&vcpu->arch.pkru, src + offset,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4531) sizeof(vcpu->arch.pkru));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4532) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4533) memcpy(dest, src + offset, size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4534) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4535)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4536) valid -= xfeature_mask;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4537) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4538) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4539)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4540) static void kvm_vcpu_ioctl_x86_get_xsave(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4541) struct kvm_xsave *guest_xsave)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4542) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4543) if (boot_cpu_has(X86_FEATURE_XSAVE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4544) memset(guest_xsave, 0, sizeof(struct kvm_xsave));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4545) fill_xsave((u8 *) guest_xsave->region, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4546) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4547) memcpy(guest_xsave->region,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4548) &vcpu->arch.guest_fpu->state.fxsave,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4549) sizeof(struct fxregs_state));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4550) *(u64 *)&guest_xsave->region[XSAVE_HDR_OFFSET / sizeof(u32)] =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4551) XFEATURE_MASK_FPSSE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4552) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4553) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4554)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4555) #define XSAVE_MXCSR_OFFSET 24
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4556)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4557) static int kvm_vcpu_ioctl_x86_set_xsave(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4558) struct kvm_xsave *guest_xsave)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4559) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4560) u64 xstate_bv =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4561) *(u64 *)&guest_xsave->region[XSAVE_HDR_OFFSET / sizeof(u32)];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4562) u32 mxcsr = *(u32 *)&guest_xsave->region[XSAVE_MXCSR_OFFSET / sizeof(u32)];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4563)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4564) if (boot_cpu_has(X86_FEATURE_XSAVE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4565) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4566) * Here we allow setting states that are not present in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4567) * CPUID leaf 0xD, index 0, EDX:EAX. This is for compatibility
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4568) * with old userspace.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4569) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4570) if (xstate_bv & ~supported_xcr0 || mxcsr & ~mxcsr_feature_mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4571) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4572) load_xsave(vcpu, (u8 *)guest_xsave->region);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4573) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4574) if (xstate_bv & ~XFEATURE_MASK_FPSSE ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4575) mxcsr & ~mxcsr_feature_mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4576) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4577) memcpy(&vcpu->arch.guest_fpu->state.fxsave,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4578) guest_xsave->region, sizeof(struct fxregs_state));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4579) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4580) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4581) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4582)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4583) static void kvm_vcpu_ioctl_x86_get_xcrs(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4584) struct kvm_xcrs *guest_xcrs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4585) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4586) if (!boot_cpu_has(X86_FEATURE_XSAVE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4587) guest_xcrs->nr_xcrs = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4588) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4589) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4590)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4591) guest_xcrs->nr_xcrs = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4592) guest_xcrs->flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4593) guest_xcrs->xcrs[0].xcr = XCR_XFEATURE_ENABLED_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4594) guest_xcrs->xcrs[0].value = vcpu->arch.xcr0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4595) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4596)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4597) static int kvm_vcpu_ioctl_x86_set_xcrs(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4598) struct kvm_xcrs *guest_xcrs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4599) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4600) int i, r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4601)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4602) if (!boot_cpu_has(X86_FEATURE_XSAVE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4603) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4604)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4605) if (guest_xcrs->nr_xcrs > KVM_MAX_XCRS || guest_xcrs->flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4606) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4607)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4608) for (i = 0; i < guest_xcrs->nr_xcrs; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4609) /* Only support XCR0 currently */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4610) if (guest_xcrs->xcrs[i].xcr == XCR_XFEATURE_ENABLED_MASK) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4611) r = __kvm_set_xcr(vcpu, XCR_XFEATURE_ENABLED_MASK,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4612) guest_xcrs->xcrs[i].value);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4613) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4614) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4615) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4616) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4617) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4618) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4619)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4620) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4621) * kvm_set_guest_paused() indicates to the guest kernel that it has been
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4622) * stopped by the hypervisor. This function will be called from the host only.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4623) * EINVAL is returned when the host attempts to set the flag for a guest that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4624) * does not support pv clocks.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4625) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4626) static int kvm_set_guest_paused(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4627) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4628) if (!vcpu->arch.pv_time_enabled)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4629) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4630) vcpu->arch.pvclock_set_guest_stopped_request = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4631) kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4632) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4633) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4634)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4635) static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4636) struct kvm_enable_cap *cap)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4637) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4638) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4639) uint16_t vmcs_version;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4640) void __user *user_ptr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4641)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4642) if (cap->flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4643) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4644)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4645) switch (cap->cap) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4646) case KVM_CAP_HYPERV_SYNIC2:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4647) if (cap->args[0])
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4648) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4649) fallthrough;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4650)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4651) case KVM_CAP_HYPERV_SYNIC:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4652) if (!irqchip_in_kernel(vcpu->kvm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4653) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4654) return kvm_hv_activate_synic(vcpu, cap->cap ==
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4655) KVM_CAP_HYPERV_SYNIC2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4656) case KVM_CAP_HYPERV_ENLIGHTENED_VMCS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4657) if (!kvm_x86_ops.nested_ops->enable_evmcs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4658) return -ENOTTY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4659) r = kvm_x86_ops.nested_ops->enable_evmcs(vcpu, &vmcs_version);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4660) if (!r) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4661) user_ptr = (void __user *)(uintptr_t)cap->args[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4662) if (copy_to_user(user_ptr, &vmcs_version,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4663) sizeof(vmcs_version)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4664) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4665) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4666) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4667) case KVM_CAP_HYPERV_DIRECT_TLBFLUSH:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4668) if (!kvm_x86_ops.enable_direct_tlbflush)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4669) return -ENOTTY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4670)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4671) return kvm_x86_ops.enable_direct_tlbflush(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4672)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4673) case KVM_CAP_ENFORCE_PV_FEATURE_CPUID:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4674) vcpu->arch.pv_cpuid.enforce = cap->args[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4675) if (vcpu->arch.pv_cpuid.enforce)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4676) kvm_update_pv_runtime(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4677)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4678) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4679)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4680) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4681) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4682) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4683) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4684)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4685) long kvm_arch_vcpu_ioctl(struct file *filp,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4686) unsigned int ioctl, unsigned long arg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4687) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4688) struct kvm_vcpu *vcpu = filp->private_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4689) void __user *argp = (void __user *)arg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4690) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4691) union {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4692) struct kvm_lapic_state *lapic;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4693) struct kvm_xsave *xsave;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4694) struct kvm_xcrs *xcrs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4695) void *buffer;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4696) } u;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4697)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4698) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4699)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4700) u.buffer = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4701) switch (ioctl) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4702) case KVM_GET_LAPIC: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4703) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4704) if (!lapic_in_kernel(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4705) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4706) u.lapic = kzalloc(sizeof(struct kvm_lapic_state),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4707) GFP_KERNEL_ACCOUNT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4708)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4709) r = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4710) if (!u.lapic)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4711) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4712) r = kvm_vcpu_ioctl_get_lapic(vcpu, u.lapic);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4713) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4714) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4715) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4716) if (copy_to_user(argp, u.lapic, sizeof(struct kvm_lapic_state)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4717) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4718) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4719) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4720) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4721) case KVM_SET_LAPIC: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4722) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4723) if (!lapic_in_kernel(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4724) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4725) u.lapic = memdup_user(argp, sizeof(*u.lapic));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4726) if (IS_ERR(u.lapic)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4727) r = PTR_ERR(u.lapic);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4728) goto out_nofree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4729) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4730)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4731) r = kvm_vcpu_ioctl_set_lapic(vcpu, u.lapic);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4732) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4733) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4734) case KVM_INTERRUPT: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4735) struct kvm_interrupt irq;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4736)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4737) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4738) if (copy_from_user(&irq, argp, sizeof(irq)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4739) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4740) r = kvm_vcpu_ioctl_interrupt(vcpu, &irq);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4741) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4742) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4743) case KVM_NMI: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4744) r = kvm_vcpu_ioctl_nmi(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4745) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4746) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4747) case KVM_SMI: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4748) r = kvm_vcpu_ioctl_smi(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4749) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4750) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4751) case KVM_SET_CPUID: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4752) struct kvm_cpuid __user *cpuid_arg = argp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4753) struct kvm_cpuid cpuid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4754)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4755) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4756) if (copy_from_user(&cpuid, cpuid_arg, sizeof(cpuid)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4757) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4758) r = kvm_vcpu_ioctl_set_cpuid(vcpu, &cpuid, cpuid_arg->entries);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4759) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4760) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4761) case KVM_SET_CPUID2: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4762) struct kvm_cpuid2 __user *cpuid_arg = argp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4763) struct kvm_cpuid2 cpuid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4764)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4765) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4766) if (copy_from_user(&cpuid, cpuid_arg, sizeof(cpuid)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4767) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4768) r = kvm_vcpu_ioctl_set_cpuid2(vcpu, &cpuid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4769) cpuid_arg->entries);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4770) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4771) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4772) case KVM_GET_CPUID2: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4773) struct kvm_cpuid2 __user *cpuid_arg = argp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4774) struct kvm_cpuid2 cpuid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4775)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4776) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4777) if (copy_from_user(&cpuid, cpuid_arg, sizeof(cpuid)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4778) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4779) r = kvm_vcpu_ioctl_get_cpuid2(vcpu, &cpuid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4780) cpuid_arg->entries);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4781) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4782) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4783) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4784) if (copy_to_user(cpuid_arg, &cpuid, sizeof(cpuid)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4785) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4786) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4787) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4788) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4789) case KVM_GET_MSRS: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4790) int idx = srcu_read_lock(&vcpu->kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4791) r = msr_io(vcpu, argp, do_get_msr, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4792) srcu_read_unlock(&vcpu->kvm->srcu, idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4793) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4794) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4795) case KVM_SET_MSRS: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4796) int idx = srcu_read_lock(&vcpu->kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4797) r = msr_io(vcpu, argp, do_set_msr, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4798) srcu_read_unlock(&vcpu->kvm->srcu, idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4799) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4800) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4801) case KVM_TPR_ACCESS_REPORTING: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4802) struct kvm_tpr_access_ctl tac;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4803)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4804) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4805) if (copy_from_user(&tac, argp, sizeof(tac)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4806) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4807) r = vcpu_ioctl_tpr_access_reporting(vcpu, &tac);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4808) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4809) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4810) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4811) if (copy_to_user(argp, &tac, sizeof(tac)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4812) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4813) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4814) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4815) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4816) case KVM_SET_VAPIC_ADDR: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4817) struct kvm_vapic_addr va;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4818) int idx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4819)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4820) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4821) if (!lapic_in_kernel(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4822) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4823) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4824) if (copy_from_user(&va, argp, sizeof(va)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4825) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4826) idx = srcu_read_lock(&vcpu->kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4827) r = kvm_lapic_set_vapic_addr(vcpu, va.vapic_addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4828) srcu_read_unlock(&vcpu->kvm->srcu, idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4829) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4830) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4831) case KVM_X86_SETUP_MCE: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4832) u64 mcg_cap;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4833)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4834) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4835) if (copy_from_user(&mcg_cap, argp, sizeof(mcg_cap)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4836) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4837) r = kvm_vcpu_ioctl_x86_setup_mce(vcpu, mcg_cap);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4838) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4839) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4840) case KVM_X86_SET_MCE: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4841) struct kvm_x86_mce mce;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4842)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4843) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4844) if (copy_from_user(&mce, argp, sizeof(mce)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4845) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4846) r = kvm_vcpu_ioctl_x86_set_mce(vcpu, &mce);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4847) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4848) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4849) case KVM_GET_VCPU_EVENTS: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4850) struct kvm_vcpu_events events;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4851)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4852) kvm_vcpu_ioctl_x86_get_vcpu_events(vcpu, &events);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4853)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4854) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4855) if (copy_to_user(argp, &events, sizeof(struct kvm_vcpu_events)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4856) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4857) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4858) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4859) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4860) case KVM_SET_VCPU_EVENTS: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4861) struct kvm_vcpu_events events;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4862)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4863) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4864) if (copy_from_user(&events, argp, sizeof(struct kvm_vcpu_events)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4865) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4866)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4867) r = kvm_vcpu_ioctl_x86_set_vcpu_events(vcpu, &events);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4868) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4869) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4870) case KVM_GET_DEBUGREGS: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4871) struct kvm_debugregs dbgregs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4872)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4873) kvm_vcpu_ioctl_x86_get_debugregs(vcpu, &dbgregs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4874)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4875) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4876) if (copy_to_user(argp, &dbgregs,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4877) sizeof(struct kvm_debugregs)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4878) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4879) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4880) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4881) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4882) case KVM_SET_DEBUGREGS: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4883) struct kvm_debugregs dbgregs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4884)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4885) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4886) if (copy_from_user(&dbgregs, argp,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4887) sizeof(struct kvm_debugregs)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4888) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4889)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4890) r = kvm_vcpu_ioctl_x86_set_debugregs(vcpu, &dbgregs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4891) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4892) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4893) case KVM_GET_XSAVE: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4894) u.xsave = kzalloc(sizeof(struct kvm_xsave), GFP_KERNEL_ACCOUNT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4895) r = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4896) if (!u.xsave)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4897) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4898)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4899) kvm_vcpu_ioctl_x86_get_xsave(vcpu, u.xsave);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4900)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4901) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4902) if (copy_to_user(argp, u.xsave, sizeof(struct kvm_xsave)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4903) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4904) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4905) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4906) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4907) case KVM_SET_XSAVE: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4908) u.xsave = memdup_user(argp, sizeof(*u.xsave));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4909) if (IS_ERR(u.xsave)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4910) r = PTR_ERR(u.xsave);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4911) goto out_nofree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4912) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4913)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4914) r = kvm_vcpu_ioctl_x86_set_xsave(vcpu, u.xsave);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4915) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4916) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4917) case KVM_GET_XCRS: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4918) u.xcrs = kzalloc(sizeof(struct kvm_xcrs), GFP_KERNEL_ACCOUNT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4919) r = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4920) if (!u.xcrs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4921) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4922)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4923) kvm_vcpu_ioctl_x86_get_xcrs(vcpu, u.xcrs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4924)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4925) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4926) if (copy_to_user(argp, u.xcrs,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4927) sizeof(struct kvm_xcrs)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4928) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4929) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4930) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4931) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4932) case KVM_SET_XCRS: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4933) u.xcrs = memdup_user(argp, sizeof(*u.xcrs));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4934) if (IS_ERR(u.xcrs)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4935) r = PTR_ERR(u.xcrs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4936) goto out_nofree;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4937) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4938)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4939) r = kvm_vcpu_ioctl_x86_set_xcrs(vcpu, u.xcrs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4940) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4941) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4942) case KVM_SET_TSC_KHZ: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4943) u32 user_tsc_khz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4944)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4945) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4946) user_tsc_khz = (u32)arg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4947)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4948) if (kvm_has_tsc_control &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4949) user_tsc_khz >= kvm_max_guest_tsc_khz)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4950) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4951)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4952) if (user_tsc_khz == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4953) user_tsc_khz = tsc_khz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4954)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4955) if (!kvm_set_tsc_khz(vcpu, user_tsc_khz))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4956) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4957)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4958) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4959) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4960) case KVM_GET_TSC_KHZ: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4961) r = vcpu->arch.virtual_tsc_khz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4962) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4963) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4964) case KVM_KVMCLOCK_CTRL: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4965) r = kvm_set_guest_paused(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4966) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4967) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4968) case KVM_ENABLE_CAP: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4969) struct kvm_enable_cap cap;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4970)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4971) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4972) if (copy_from_user(&cap, argp, sizeof(cap)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4973) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4974) r = kvm_vcpu_ioctl_enable_cap(vcpu, &cap);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4975) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4976) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4977) case KVM_GET_NESTED_STATE: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4978) struct kvm_nested_state __user *user_kvm_nested_state = argp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4979) u32 user_data_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4980)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4981) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4982) if (!kvm_x86_ops.nested_ops->get_state)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4983) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4984)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4985) BUILD_BUG_ON(sizeof(user_data_size) != sizeof(user_kvm_nested_state->size));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4986) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4987) if (get_user(user_data_size, &user_kvm_nested_state->size))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4988) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4989)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4990) r = kvm_x86_ops.nested_ops->get_state(vcpu, user_kvm_nested_state,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4991) user_data_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4992) if (r < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4993) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4994)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4995) if (r > user_data_size) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4996) if (put_user(r, &user_kvm_nested_state->size))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4997) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4998) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4999) r = -E2BIG;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5000) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5001) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5002)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5003) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5004) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5005) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5006) case KVM_SET_NESTED_STATE: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5007) struct kvm_nested_state __user *user_kvm_nested_state = argp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5008) struct kvm_nested_state kvm_state;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5009) int idx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5010)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5011) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5012) if (!kvm_x86_ops.nested_ops->set_state)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5013) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5014)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5015) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5016) if (copy_from_user(&kvm_state, user_kvm_nested_state, sizeof(kvm_state)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5017) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5018)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5019) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5020) if (kvm_state.size < sizeof(kvm_state))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5021) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5022)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5023) if (kvm_state.flags &
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5024) ~(KVM_STATE_NESTED_RUN_PENDING | KVM_STATE_NESTED_GUEST_MODE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5025) | KVM_STATE_NESTED_EVMCS | KVM_STATE_NESTED_MTF_PENDING
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5026) | KVM_STATE_NESTED_GIF_SET))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5027) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5028)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5029) /* nested_run_pending implies guest_mode. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5030) if ((kvm_state.flags & KVM_STATE_NESTED_RUN_PENDING)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5031) && !(kvm_state.flags & KVM_STATE_NESTED_GUEST_MODE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5032) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5033)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5034) idx = srcu_read_lock(&vcpu->kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5035) r = kvm_x86_ops.nested_ops->set_state(vcpu, user_kvm_nested_state, &kvm_state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5036) srcu_read_unlock(&vcpu->kvm->srcu, idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5037) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5038) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5039) case KVM_GET_SUPPORTED_HV_CPUID: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5040) struct kvm_cpuid2 __user *cpuid_arg = argp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5041) struct kvm_cpuid2 cpuid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5042)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5043) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5044) if (copy_from_user(&cpuid, cpuid_arg, sizeof(cpuid)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5045) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5046)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5047) r = kvm_vcpu_ioctl_get_hv_cpuid(vcpu, &cpuid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5048) cpuid_arg->entries);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5049) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5050) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5051)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5052) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5053) if (copy_to_user(cpuid_arg, &cpuid, sizeof(cpuid)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5054) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5055) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5056) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5057) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5058) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5059) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5060) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5061) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5062) kfree(u.buffer);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5063) out_nofree:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5064) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5065) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5066) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5067)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5068) vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5069) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5070) return VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5071) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5072)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5073) static int kvm_vm_ioctl_set_tss_addr(struct kvm *kvm, unsigned long addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5074) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5075) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5076)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5077) if (addr > (unsigned int)(-3 * PAGE_SIZE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5078) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5079) ret = kvm_x86_ops.set_tss_addr(kvm, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5080) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5081) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5082)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5083) static int kvm_vm_ioctl_set_identity_map_addr(struct kvm *kvm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5084) u64 ident_addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5085) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5086) return kvm_x86_ops.set_identity_map_addr(kvm, ident_addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5087) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5088)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5089) static int kvm_vm_ioctl_set_nr_mmu_pages(struct kvm *kvm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5090) unsigned long kvm_nr_mmu_pages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5091) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5092) if (kvm_nr_mmu_pages < KVM_MIN_ALLOC_MMU_PAGES)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5093) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5094)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5095) mutex_lock(&kvm->slots_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5096)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5097) kvm_mmu_change_mmu_pages(kvm, kvm_nr_mmu_pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5098) kvm->arch.n_requested_mmu_pages = kvm_nr_mmu_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5099)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5100) mutex_unlock(&kvm->slots_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5101) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5102) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5103)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5104) static unsigned long kvm_vm_ioctl_get_nr_mmu_pages(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5105) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5106) return kvm->arch.n_max_mmu_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5107) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5108)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5109) static int kvm_vm_ioctl_get_irqchip(struct kvm *kvm, struct kvm_irqchip *chip)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5110) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5111) struct kvm_pic *pic = kvm->arch.vpic;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5112) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5113)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5114) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5115) switch (chip->chip_id) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5116) case KVM_IRQCHIP_PIC_MASTER:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5117) memcpy(&chip->chip.pic, &pic->pics[0],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5118) sizeof(struct kvm_pic_state));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5119) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5120) case KVM_IRQCHIP_PIC_SLAVE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5121) memcpy(&chip->chip.pic, &pic->pics[1],
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5122) sizeof(struct kvm_pic_state));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5123) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5124) case KVM_IRQCHIP_IOAPIC:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5125) kvm_get_ioapic(kvm, &chip->chip.ioapic);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5126) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5127) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5128) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5129) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5130) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5131) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5132) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5133)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5134) static int kvm_vm_ioctl_set_irqchip(struct kvm *kvm, struct kvm_irqchip *chip)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5135) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5136) struct kvm_pic *pic = kvm->arch.vpic;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5137) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5138)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5139) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5140) switch (chip->chip_id) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5141) case KVM_IRQCHIP_PIC_MASTER:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5142) spin_lock(&pic->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5143) memcpy(&pic->pics[0], &chip->chip.pic,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5144) sizeof(struct kvm_pic_state));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5145) spin_unlock(&pic->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5146) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5147) case KVM_IRQCHIP_PIC_SLAVE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5148) spin_lock(&pic->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5149) memcpy(&pic->pics[1], &chip->chip.pic,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5150) sizeof(struct kvm_pic_state));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5151) spin_unlock(&pic->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5152) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5153) case KVM_IRQCHIP_IOAPIC:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5154) kvm_set_ioapic(kvm, &chip->chip.ioapic);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5155) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5156) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5157) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5158) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5159) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5160) kvm_pic_update_irq(pic);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5161) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5162) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5163)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5164) static int kvm_vm_ioctl_get_pit(struct kvm *kvm, struct kvm_pit_state *ps)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5165) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5166) struct kvm_kpit_state *kps = &kvm->arch.vpit->pit_state;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5167)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5168) BUILD_BUG_ON(sizeof(*ps) != sizeof(kps->channels));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5169)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5170) mutex_lock(&kps->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5171) memcpy(ps, &kps->channels, sizeof(*ps));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5172) mutex_unlock(&kps->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5173) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5174) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5175)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5176) static int kvm_vm_ioctl_set_pit(struct kvm *kvm, struct kvm_pit_state *ps)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5177) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5178) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5179) struct kvm_pit *pit = kvm->arch.vpit;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5180)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5181) mutex_lock(&pit->pit_state.lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5182) memcpy(&pit->pit_state.channels, ps, sizeof(*ps));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5183) for (i = 0; i < 3; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5184) kvm_pit_load_count(pit, i, ps->channels[i].count, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5185) mutex_unlock(&pit->pit_state.lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5186) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5187) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5188)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5189) static int kvm_vm_ioctl_get_pit2(struct kvm *kvm, struct kvm_pit_state2 *ps)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5190) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5191) mutex_lock(&kvm->arch.vpit->pit_state.lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5192) memcpy(ps->channels, &kvm->arch.vpit->pit_state.channels,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5193) sizeof(ps->channels));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5194) ps->flags = kvm->arch.vpit->pit_state.flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5195) mutex_unlock(&kvm->arch.vpit->pit_state.lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5196) memset(&ps->reserved, 0, sizeof(ps->reserved));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5197) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5198) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5199)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5200) static int kvm_vm_ioctl_set_pit2(struct kvm *kvm, struct kvm_pit_state2 *ps)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5201) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5202) int start = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5203) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5204) u32 prev_legacy, cur_legacy;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5205) struct kvm_pit *pit = kvm->arch.vpit;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5206)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5207) mutex_lock(&pit->pit_state.lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5208) prev_legacy = pit->pit_state.flags & KVM_PIT_FLAGS_HPET_LEGACY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5209) cur_legacy = ps->flags & KVM_PIT_FLAGS_HPET_LEGACY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5210) if (!prev_legacy && cur_legacy)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5211) start = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5212) memcpy(&pit->pit_state.channels, &ps->channels,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5213) sizeof(pit->pit_state.channels));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5214) pit->pit_state.flags = ps->flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5215) for (i = 0; i < 3; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5216) kvm_pit_load_count(pit, i, pit->pit_state.channels[i].count,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5217) start && i == 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5218) mutex_unlock(&pit->pit_state.lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5219) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5220) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5221)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5222) static int kvm_vm_ioctl_reinject(struct kvm *kvm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5223) struct kvm_reinject_control *control)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5224) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5225) struct kvm_pit *pit = kvm->arch.vpit;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5226)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5227) /* pit->pit_state.lock was overloaded to prevent userspace from getting
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5228) * an inconsistent state after running multiple KVM_REINJECT_CONTROL
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5229) * ioctls in parallel. Use a separate lock if that ioctl isn't rare.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5230) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5231) mutex_lock(&pit->pit_state.lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5232) kvm_pit_set_reinject(pit, control->pit_reinject);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5233) mutex_unlock(&pit->pit_state.lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5234)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5235) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5236) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5237)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5238) void kvm_arch_sync_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5239) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5240) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5241) * Flush potentially hardware-cached dirty pages to dirty_bitmap.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5242) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5243) if (kvm_x86_ops.flush_log_dirty)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5244) kvm_x86_ops.flush_log_dirty(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5245) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5246)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5247) int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_event,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5248) bool line_status)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5249) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5250) if (!irqchip_in_kernel(kvm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5251) return -ENXIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5252)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5253) irq_event->status = kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5254) irq_event->irq, irq_event->level,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5255) line_status);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5256) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5257) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5258)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5259) int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5260) struct kvm_enable_cap *cap)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5261) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5262) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5263)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5264) if (cap->flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5265) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5266)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5267) switch (cap->cap) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5268) case KVM_CAP_DISABLE_QUIRKS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5269) kvm->arch.disabled_quirks = cap->args[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5270) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5271) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5272) case KVM_CAP_SPLIT_IRQCHIP: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5273) mutex_lock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5274) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5275) if (cap->args[0] > MAX_NR_RESERVED_IOAPIC_PINS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5276) goto split_irqchip_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5277) r = -EEXIST;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5278) if (irqchip_in_kernel(kvm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5279) goto split_irqchip_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5280) if (kvm->created_vcpus)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5281) goto split_irqchip_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5282) r = kvm_setup_empty_irq_routing(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5283) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5284) goto split_irqchip_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5285) /* Pairs with irqchip_in_kernel. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5286) smp_wmb();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5287) kvm->arch.irqchip_mode = KVM_IRQCHIP_SPLIT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5288) kvm->arch.nr_reserved_ioapic_pins = cap->args[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5289) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5290) split_irqchip_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5291) mutex_unlock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5292) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5293) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5294) case KVM_CAP_X2APIC_API:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5295) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5296) if (cap->args[0] & ~KVM_X2APIC_API_VALID_FLAGS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5297) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5298)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5299) if (cap->args[0] & KVM_X2APIC_API_USE_32BIT_IDS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5300) kvm->arch.x2apic_format = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5301) if (cap->args[0] & KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5302) kvm->arch.x2apic_broadcast_quirk_disabled = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5303)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5304) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5305) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5306) case KVM_CAP_X86_DISABLE_EXITS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5307) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5308) if (cap->args[0] & ~KVM_X86_DISABLE_VALID_EXITS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5309) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5310)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5311) if ((cap->args[0] & KVM_X86_DISABLE_EXITS_MWAIT) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5312) kvm_can_mwait_in_guest())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5313) kvm->arch.mwait_in_guest = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5314) if (cap->args[0] & KVM_X86_DISABLE_EXITS_HLT)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5315) kvm->arch.hlt_in_guest = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5316) if (cap->args[0] & KVM_X86_DISABLE_EXITS_PAUSE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5317) kvm->arch.pause_in_guest = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5318) if (cap->args[0] & KVM_X86_DISABLE_EXITS_CSTATE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5319) kvm->arch.cstate_in_guest = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5320) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5321) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5322) case KVM_CAP_MSR_PLATFORM_INFO:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5323) kvm->arch.guest_can_read_msr_platform_info = cap->args[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5324) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5325) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5326) case KVM_CAP_EXCEPTION_PAYLOAD:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5327) kvm->arch.exception_payload_enabled = cap->args[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5328) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5329) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5330) case KVM_CAP_X86_USER_SPACE_MSR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5331) kvm->arch.user_space_msr_mask = cap->args[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5332) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5333) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5334) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5335) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5336) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5337) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5338) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5339) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5340)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5341) static struct kvm_x86_msr_filter *kvm_alloc_msr_filter(bool default_allow)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5342) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5343) struct kvm_x86_msr_filter *msr_filter;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5344)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5345) msr_filter = kzalloc(sizeof(*msr_filter), GFP_KERNEL_ACCOUNT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5346) if (!msr_filter)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5347) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5348)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5349) msr_filter->default_allow = default_allow;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5350) return msr_filter;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5351) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5352)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5353) static void kvm_free_msr_filter(struct kvm_x86_msr_filter *msr_filter)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5354) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5355) u32 i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5356)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5357) if (!msr_filter)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5358) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5359)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5360) for (i = 0; i < msr_filter->count; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5361) kfree(msr_filter->ranges[i].bitmap);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5362)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5363) kfree(msr_filter);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5364) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5365)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5366) static int kvm_add_msr_filter(struct kvm_x86_msr_filter *msr_filter,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5367) struct kvm_msr_filter_range *user_range)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5368) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5369) struct msr_bitmap_range range;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5370) unsigned long *bitmap = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5371) size_t bitmap_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5372) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5373)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5374) if (!user_range->nmsrs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5375) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5376)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5377) bitmap_size = BITS_TO_LONGS(user_range->nmsrs) * sizeof(long);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5378) if (!bitmap_size || bitmap_size > KVM_MSR_FILTER_MAX_BITMAP_SIZE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5379) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5380)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5381) bitmap = memdup_user((__user u8*)user_range->bitmap, bitmap_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5382) if (IS_ERR(bitmap))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5383) return PTR_ERR(bitmap);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5384)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5385) range = (struct msr_bitmap_range) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5386) .flags = user_range->flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5387) .base = user_range->base,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5388) .nmsrs = user_range->nmsrs,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5389) .bitmap = bitmap,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5390) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5391)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5392) if (range.flags & ~(KVM_MSR_FILTER_READ | KVM_MSR_FILTER_WRITE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5393) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5394) goto err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5395) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5396)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5397) if (!range.flags) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5398) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5399) goto err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5400) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5401)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5402) /* Everything ok, add this range identifier. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5403) msr_filter->ranges[msr_filter->count] = range;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5404) msr_filter->count++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5405)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5406) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5407) err:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5408) kfree(bitmap);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5409) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5410) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5411)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5412) static int kvm_vm_ioctl_set_msr_filter(struct kvm *kvm, void __user *argp)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5413) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5414) struct kvm_msr_filter __user *user_msr_filter = argp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5415) struct kvm_x86_msr_filter *new_filter, *old_filter;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5416) struct kvm_msr_filter filter;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5417) bool default_allow;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5418) bool empty = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5419) int r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5420) u32 i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5421)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5422) if (copy_from_user(&filter, user_msr_filter, sizeof(filter)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5423) return -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5424)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5425) for (i = 0; i < ARRAY_SIZE(filter.ranges); i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5426) empty &= !filter.ranges[i].nmsrs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5427)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5428) default_allow = !(filter.flags & KVM_MSR_FILTER_DEFAULT_DENY);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5429) if (empty && !default_allow)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5430) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5431)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5432) new_filter = kvm_alloc_msr_filter(default_allow);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5433) if (!new_filter)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5434) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5435)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5436) for (i = 0; i < ARRAY_SIZE(filter.ranges); i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5437) r = kvm_add_msr_filter(new_filter, &filter.ranges[i]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5438) if (r) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5439) kvm_free_msr_filter(new_filter);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5440) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5441) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5442) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5443)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5444) mutex_lock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5445)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5446) /* The per-VM filter is protected by kvm->lock... */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5447) old_filter = srcu_dereference_check(kvm->arch.msr_filter, &kvm->srcu, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5448)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5449) rcu_assign_pointer(kvm->arch.msr_filter, new_filter);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5450) synchronize_srcu(&kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5451)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5452) kvm_free_msr_filter(old_filter);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5453)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5454) kvm_make_all_cpus_request(kvm, KVM_REQ_MSR_FILTER_CHANGED);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5455) mutex_unlock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5456)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5457) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5458) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5459)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5460) long kvm_arch_vm_ioctl(struct file *filp,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5461) unsigned int ioctl, unsigned long arg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5462) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5463) struct kvm *kvm = filp->private_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5464) void __user *argp = (void __user *)arg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5465) int r = -ENOTTY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5466) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5467) * This union makes it completely explicit to gcc-3.x
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5468) * that these two variables' stack usage should be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5469) * combined, not added together.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5470) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5471) union {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5472) struct kvm_pit_state ps;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5473) struct kvm_pit_state2 ps2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5474) struct kvm_pit_config pit_config;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5475) } u;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5476)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5477) switch (ioctl) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5478) case KVM_SET_TSS_ADDR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5479) r = kvm_vm_ioctl_set_tss_addr(kvm, arg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5480) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5481) case KVM_SET_IDENTITY_MAP_ADDR: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5482) u64 ident_addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5483)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5484) mutex_lock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5485) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5486) if (kvm->created_vcpus)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5487) goto set_identity_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5488) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5489) if (copy_from_user(&ident_addr, argp, sizeof(ident_addr)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5490) goto set_identity_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5491) r = kvm_vm_ioctl_set_identity_map_addr(kvm, ident_addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5492) set_identity_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5493) mutex_unlock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5494) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5495) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5496) case KVM_SET_NR_MMU_PAGES:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5497) r = kvm_vm_ioctl_set_nr_mmu_pages(kvm, arg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5498) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5499) case KVM_GET_NR_MMU_PAGES:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5500) r = kvm_vm_ioctl_get_nr_mmu_pages(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5501) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5502) case KVM_CREATE_IRQCHIP: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5503) mutex_lock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5504)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5505) r = -EEXIST;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5506) if (irqchip_in_kernel(kvm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5507) goto create_irqchip_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5508)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5509) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5510) if (kvm->created_vcpus)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5511) goto create_irqchip_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5512)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5513) r = kvm_pic_init(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5514) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5515) goto create_irqchip_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5516)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5517) r = kvm_ioapic_init(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5518) if (r) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5519) kvm_pic_destroy(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5520) goto create_irqchip_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5521) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5522)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5523) r = kvm_setup_default_irq_routing(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5524) if (r) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5525) kvm_ioapic_destroy(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5526) kvm_pic_destroy(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5527) goto create_irqchip_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5528) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5529) /* Write kvm->irq_routing before enabling irqchip_in_kernel. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5530) smp_wmb();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5531) kvm->arch.irqchip_mode = KVM_IRQCHIP_KERNEL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5532) create_irqchip_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5533) mutex_unlock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5534) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5535) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5536) case KVM_CREATE_PIT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5537) u.pit_config.flags = KVM_PIT_SPEAKER_DUMMY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5538) goto create_pit;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5539) case KVM_CREATE_PIT2:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5540) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5541) if (copy_from_user(&u.pit_config, argp,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5542) sizeof(struct kvm_pit_config)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5543) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5544) create_pit:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5545) mutex_lock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5546) r = -EEXIST;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5547) if (kvm->arch.vpit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5548) goto create_pit_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5549) r = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5550) kvm->arch.vpit = kvm_create_pit(kvm, u.pit_config.flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5551) if (kvm->arch.vpit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5552) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5553) create_pit_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5554) mutex_unlock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5555) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5556) case KVM_GET_IRQCHIP: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5557) /* 0: PIC master, 1: PIC slave, 2: IOAPIC */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5558) struct kvm_irqchip *chip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5559)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5560) chip = memdup_user(argp, sizeof(*chip));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5561) if (IS_ERR(chip)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5562) r = PTR_ERR(chip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5563) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5564) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5565)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5566) r = -ENXIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5567) if (!irqchip_kernel(kvm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5568) goto get_irqchip_out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5569) r = kvm_vm_ioctl_get_irqchip(kvm, chip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5570) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5571) goto get_irqchip_out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5572) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5573) if (copy_to_user(argp, chip, sizeof(*chip)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5574) goto get_irqchip_out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5575) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5576) get_irqchip_out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5577) kfree(chip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5578) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5579) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5580) case KVM_SET_IRQCHIP: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5581) /* 0: PIC master, 1: PIC slave, 2: IOAPIC */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5582) struct kvm_irqchip *chip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5583)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5584) chip = memdup_user(argp, sizeof(*chip));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5585) if (IS_ERR(chip)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5586) r = PTR_ERR(chip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5587) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5588) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5589)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5590) r = -ENXIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5591) if (!irqchip_kernel(kvm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5592) goto set_irqchip_out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5593) r = kvm_vm_ioctl_set_irqchip(kvm, chip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5594) set_irqchip_out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5595) kfree(chip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5596) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5597) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5598) case KVM_GET_PIT: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5599) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5600) if (copy_from_user(&u.ps, argp, sizeof(struct kvm_pit_state)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5601) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5602) r = -ENXIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5603) if (!kvm->arch.vpit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5604) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5605) r = kvm_vm_ioctl_get_pit(kvm, &u.ps);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5606) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5607) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5608) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5609) if (copy_to_user(argp, &u.ps, sizeof(struct kvm_pit_state)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5610) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5611) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5612) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5613) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5614) case KVM_SET_PIT: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5615) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5616) if (copy_from_user(&u.ps, argp, sizeof(u.ps)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5617) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5618) mutex_lock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5619) r = -ENXIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5620) if (!kvm->arch.vpit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5621) goto set_pit_out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5622) r = kvm_vm_ioctl_set_pit(kvm, &u.ps);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5623) set_pit_out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5624) mutex_unlock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5625) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5626) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5627) case KVM_GET_PIT2: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5628) r = -ENXIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5629) if (!kvm->arch.vpit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5630) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5631) r = kvm_vm_ioctl_get_pit2(kvm, &u.ps2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5632) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5633) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5634) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5635) if (copy_to_user(argp, &u.ps2, sizeof(u.ps2)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5636) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5637) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5638) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5639) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5640) case KVM_SET_PIT2: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5641) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5642) if (copy_from_user(&u.ps2, argp, sizeof(u.ps2)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5643) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5644) mutex_lock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5645) r = -ENXIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5646) if (!kvm->arch.vpit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5647) goto set_pit2_out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5648) r = kvm_vm_ioctl_set_pit2(kvm, &u.ps2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5649) set_pit2_out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5650) mutex_unlock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5651) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5652) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5653) case KVM_REINJECT_CONTROL: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5654) struct kvm_reinject_control control;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5655) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5656) if (copy_from_user(&control, argp, sizeof(control)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5657) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5658) r = -ENXIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5659) if (!kvm->arch.vpit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5660) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5661) r = kvm_vm_ioctl_reinject(kvm, &control);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5662) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5663) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5664) case KVM_SET_BOOT_CPU_ID:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5665) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5666) mutex_lock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5667) if (kvm->created_vcpus)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5668) r = -EBUSY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5669) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5670) kvm->arch.bsp_vcpu_id = arg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5671) mutex_unlock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5672) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5673) case KVM_XEN_HVM_CONFIG: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5674) struct kvm_xen_hvm_config xhc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5675) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5676) if (copy_from_user(&xhc, argp, sizeof(xhc)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5677) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5678) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5679) if (xhc.flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5680) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5681) memcpy(&kvm->arch.xen_hvm_config, &xhc, sizeof(xhc));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5682) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5683) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5684) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5685) case KVM_SET_CLOCK: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5686) struct kvm_clock_data user_ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5687) u64 now_ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5688)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5689) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5690) if (copy_from_user(&user_ns, argp, sizeof(user_ns)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5691) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5692)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5693) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5694) if (user_ns.flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5695) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5696)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5697) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5698) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5699) * TODO: userspace has to take care of races with VCPU_RUN, so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5700) * kvm_gen_update_masterclock() can be cut down to locked
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5701) * pvclock_update_vm_gtod_copy().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5702) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5703) kvm_gen_update_masterclock(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5704) now_ns = get_kvmclock_ns(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5705) kvm->arch.kvmclock_offset += user_ns.clock - now_ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5706) kvm_make_all_cpus_request(kvm, KVM_REQ_CLOCK_UPDATE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5707) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5708) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5709) case KVM_GET_CLOCK: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5710) struct kvm_clock_data user_ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5711) u64 now_ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5712)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5713) now_ns = get_kvmclock_ns(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5714) user_ns.clock = now_ns;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5715) user_ns.flags = kvm->arch.use_master_clock ? KVM_CLOCK_TSC_STABLE : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5716) memset(&user_ns.pad, 0, sizeof(user_ns.pad));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5717)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5718) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5719) if (copy_to_user(argp, &user_ns, sizeof(user_ns)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5720) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5721) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5722) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5723) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5724) case KVM_MEMORY_ENCRYPT_OP: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5725) r = -ENOTTY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5726) if (kvm_x86_ops.mem_enc_op)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5727) r = kvm_x86_ops.mem_enc_op(kvm, argp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5728) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5729) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5730) case KVM_MEMORY_ENCRYPT_REG_REGION: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5731) struct kvm_enc_region region;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5732)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5733) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5734) if (copy_from_user(®ion, argp, sizeof(region)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5735) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5736)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5737) r = -ENOTTY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5738) if (kvm_x86_ops.mem_enc_reg_region)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5739) r = kvm_x86_ops.mem_enc_reg_region(kvm, ®ion);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5740) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5741) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5742) case KVM_MEMORY_ENCRYPT_UNREG_REGION: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5743) struct kvm_enc_region region;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5744)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5745) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5746) if (copy_from_user(®ion, argp, sizeof(region)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5747) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5748)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5749) r = -ENOTTY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5750) if (kvm_x86_ops.mem_enc_unreg_region)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5751) r = kvm_x86_ops.mem_enc_unreg_region(kvm, ®ion);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5752) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5753) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5754) case KVM_HYPERV_EVENTFD: {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5755) struct kvm_hyperv_eventfd hvevfd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5756)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5757) r = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5758) if (copy_from_user(&hvevfd, argp, sizeof(hvevfd)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5759) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5760) r = kvm_vm_ioctl_hv_eventfd(kvm, &hvevfd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5761) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5762) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5763) case KVM_SET_PMU_EVENT_FILTER:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5764) r = kvm_vm_ioctl_set_pmu_event_filter(kvm, argp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5765) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5766) case KVM_X86_SET_MSR_FILTER:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5767) r = kvm_vm_ioctl_set_msr_filter(kvm, argp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5768) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5769) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5770) r = -ENOTTY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5771) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5772) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5773) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5774) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5775)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5776) static void kvm_init_msr_list(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5777) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5778) struct x86_pmu_capability x86_pmu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5779) u32 dummy[2];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5780) unsigned i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5781)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5782) BUILD_BUG_ON_MSG(INTEL_PMC_MAX_FIXED != 4,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5783) "Please update the fixed PMCs in msrs_to_saved_all[]");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5784)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5785) perf_get_x86_pmu_capability(&x86_pmu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5786)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5787) num_msrs_to_save = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5788) num_emulated_msrs = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5789) num_msr_based_features = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5790)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5791) for (i = 0; i < ARRAY_SIZE(msrs_to_save_all); i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5792) if (rdmsr_safe(msrs_to_save_all[i], &dummy[0], &dummy[1]) < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5793) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5794)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5795) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5796) * Even MSRs that are valid in the host may not be exposed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5797) * to the guests in some cases.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5798) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5799) switch (msrs_to_save_all[i]) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5800) case MSR_IA32_BNDCFGS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5801) if (!kvm_mpx_supported())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5802) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5803) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5804) case MSR_TSC_AUX:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5805) if (!kvm_cpu_cap_has(X86_FEATURE_RDTSCP))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5806) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5807) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5808) case MSR_IA32_UMWAIT_CONTROL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5809) if (!kvm_cpu_cap_has(X86_FEATURE_WAITPKG))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5810) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5811) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5812) case MSR_IA32_RTIT_CTL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5813) case MSR_IA32_RTIT_STATUS:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5814) if (!kvm_cpu_cap_has(X86_FEATURE_INTEL_PT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5815) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5816) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5817) case MSR_IA32_RTIT_CR3_MATCH:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5818) if (!kvm_cpu_cap_has(X86_FEATURE_INTEL_PT) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5819) !intel_pt_validate_hw_cap(PT_CAP_cr3_filtering))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5820) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5821) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5822) case MSR_IA32_RTIT_OUTPUT_BASE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5823) case MSR_IA32_RTIT_OUTPUT_MASK:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5824) if (!kvm_cpu_cap_has(X86_FEATURE_INTEL_PT) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5825) (!intel_pt_validate_hw_cap(PT_CAP_topa_output) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5826) !intel_pt_validate_hw_cap(PT_CAP_single_range_output)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5827) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5828) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5829) case MSR_IA32_RTIT_ADDR0_A ... MSR_IA32_RTIT_ADDR3_B:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5830) if (!kvm_cpu_cap_has(X86_FEATURE_INTEL_PT) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5831) msrs_to_save_all[i] - MSR_IA32_RTIT_ADDR0_A >=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5832) intel_pt_validate_hw_cap(PT_CAP_num_address_ranges) * 2)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5833) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5834) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5835) case MSR_ARCH_PERFMON_PERFCTR0 ... MSR_ARCH_PERFMON_PERFCTR0 + 17:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5836) if (msrs_to_save_all[i] - MSR_ARCH_PERFMON_PERFCTR0 >=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5837) min(INTEL_PMC_MAX_GENERIC, x86_pmu.num_counters_gp))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5838) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5839) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5840) case MSR_ARCH_PERFMON_EVENTSEL0 ... MSR_ARCH_PERFMON_EVENTSEL0 + 17:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5841) if (msrs_to_save_all[i] - MSR_ARCH_PERFMON_EVENTSEL0 >=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5842) min(INTEL_PMC_MAX_GENERIC, x86_pmu.num_counters_gp))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5843) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5844) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5845) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5846) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5847) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5848)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5849) msrs_to_save[num_msrs_to_save++] = msrs_to_save_all[i];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5850) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5851)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5852) for (i = 0; i < ARRAY_SIZE(emulated_msrs_all); i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5853) if (!kvm_x86_ops.has_emulated_msr(emulated_msrs_all[i]))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5854) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5855)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5856) emulated_msrs[num_emulated_msrs++] = emulated_msrs_all[i];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5857) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5858)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5859) for (i = 0; i < ARRAY_SIZE(msr_based_features_all); i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5860) struct kvm_msr_entry msr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5861)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5862) msr.index = msr_based_features_all[i];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5863) if (kvm_get_msr_feature(&msr))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5864) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5865)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5866) msr_based_features[num_msr_based_features++] = msr_based_features_all[i];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5867) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5868) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5869)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5870) static int vcpu_mmio_write(struct kvm_vcpu *vcpu, gpa_t addr, int len,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5871) const void *v)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5872) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5873) int handled = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5874) int n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5875)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5876) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5877) n = min(len, 8);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5878) if (!(lapic_in_kernel(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5879) !kvm_iodevice_write(vcpu, &vcpu->arch.apic->dev, addr, n, v))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5880) && kvm_io_bus_write(vcpu, KVM_MMIO_BUS, addr, n, v))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5881) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5882) handled += n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5883) addr += n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5884) len -= n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5885) v += n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5886) } while (len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5887)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5888) return handled;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5889) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5890)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5891) static int vcpu_mmio_read(struct kvm_vcpu *vcpu, gpa_t addr, int len, void *v)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5892) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5893) int handled = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5894) int n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5895)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5896) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5897) n = min(len, 8);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5898) if (!(lapic_in_kernel(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5899) !kvm_iodevice_read(vcpu, &vcpu->arch.apic->dev,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5900) addr, n, v))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5901) && kvm_io_bus_read(vcpu, KVM_MMIO_BUS, addr, n, v))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5902) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5903) trace_kvm_mmio(KVM_TRACE_MMIO_READ, n, addr, v);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5904) handled += n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5905) addr += n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5906) len -= n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5907) v += n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5908) } while (len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5909)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5910) return handled;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5911) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5912)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5913) static void kvm_set_segment(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5914) struct kvm_segment *var, int seg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5915) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5916) kvm_x86_ops.set_segment(vcpu, var, seg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5917) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5918)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5919) void kvm_get_segment(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5920) struct kvm_segment *var, int seg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5921) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5922) kvm_x86_ops.get_segment(vcpu, var, seg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5923) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5924)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5925) gpa_t translate_nested_gpa(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5926) struct x86_exception *exception)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5927) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5928) gpa_t t_gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5929)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5930) BUG_ON(!mmu_is_nested(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5931)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5932) /* NPT walks are always user-walks */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5933) access |= PFERR_USER_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5934) t_gpa = vcpu->arch.mmu->gva_to_gpa(vcpu, gpa, access, exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5935)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5936) return t_gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5937) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5938)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5939) gpa_t kvm_mmu_gva_to_gpa_read(struct kvm_vcpu *vcpu, gva_t gva,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5940) struct x86_exception *exception)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5941) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5942) u32 access = (kvm_x86_ops.get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5943) return vcpu->arch.walk_mmu->gva_to_gpa(vcpu, gva, access, exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5944) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5945)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5946) gpa_t kvm_mmu_gva_to_gpa_fetch(struct kvm_vcpu *vcpu, gva_t gva,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5947) struct x86_exception *exception)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5948) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5949) u32 access = (kvm_x86_ops.get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5950) access |= PFERR_FETCH_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5951) return vcpu->arch.walk_mmu->gva_to_gpa(vcpu, gva, access, exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5952) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5953)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5954) gpa_t kvm_mmu_gva_to_gpa_write(struct kvm_vcpu *vcpu, gva_t gva,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5955) struct x86_exception *exception)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5956) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5957) u32 access = (kvm_x86_ops.get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5958) access |= PFERR_WRITE_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5959) return vcpu->arch.walk_mmu->gva_to_gpa(vcpu, gva, access, exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5960) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5961)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5962) /* uses this to access any guest's mapped memory without checking CPL */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5963) gpa_t kvm_mmu_gva_to_gpa_system(struct kvm_vcpu *vcpu, gva_t gva,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5964) struct x86_exception *exception)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5965) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5966) return vcpu->arch.walk_mmu->gva_to_gpa(vcpu, gva, 0, exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5967) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5968)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5969) static int kvm_read_guest_virt_helper(gva_t addr, void *val, unsigned int bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5970) struct kvm_vcpu *vcpu, u32 access,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5971) struct x86_exception *exception)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5972) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5973) void *data = val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5974) int r = X86EMUL_CONTINUE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5975)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5976) while (bytes) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5977) gpa_t gpa = vcpu->arch.walk_mmu->gva_to_gpa(vcpu, addr, access,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5978) exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5979) unsigned offset = addr & (PAGE_SIZE-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5980) unsigned toread = min(bytes, (unsigned)PAGE_SIZE - offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5981) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5982)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5983) if (gpa == UNMAPPED_GVA)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5984) return X86EMUL_PROPAGATE_FAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5985) ret = kvm_vcpu_read_guest_page(vcpu, gpa >> PAGE_SHIFT, data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5986) offset, toread);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5987) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5988) r = X86EMUL_IO_NEEDED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5989) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5990) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5991)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5992) bytes -= toread;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5993) data += toread;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5994) addr += toread;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5995) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5996) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5997) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5998) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5999)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6000) /* used for instruction fetching */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6001) static int kvm_fetch_guest_virt(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6002) gva_t addr, void *val, unsigned int bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6003) struct x86_exception *exception)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6004) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6005) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6006) u32 access = (kvm_x86_ops.get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6007) unsigned offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6008) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6009)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6010) /* Inline kvm_read_guest_virt_helper for speed. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6011) gpa_t gpa = vcpu->arch.walk_mmu->gva_to_gpa(vcpu, addr, access|PFERR_FETCH_MASK,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6012) exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6013) if (unlikely(gpa == UNMAPPED_GVA))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6014) return X86EMUL_PROPAGATE_FAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6015)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6016) offset = addr & (PAGE_SIZE-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6017) if (WARN_ON(offset + bytes > PAGE_SIZE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6018) bytes = (unsigned)PAGE_SIZE - offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6019) ret = kvm_vcpu_read_guest_page(vcpu, gpa >> PAGE_SHIFT, val,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6020) offset, bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6021) if (unlikely(ret < 0))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6022) return X86EMUL_IO_NEEDED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6023)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6024) return X86EMUL_CONTINUE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6025) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6026)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6027) int kvm_read_guest_virt(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6028) gva_t addr, void *val, unsigned int bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6029) struct x86_exception *exception)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6030) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6031) u32 access = (kvm_x86_ops.get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6032)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6033) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6034) * FIXME: this should call handle_emulation_failure if X86EMUL_IO_NEEDED
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6035) * is returned, but our callers are not ready for that and they blindly
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6036) * call kvm_inject_page_fault. Ensure that they at least do not leak
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6037) * uninitialized kernel stack memory into cr2 and error code.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6038) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6039) memset(exception, 0, sizeof(*exception));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6040) return kvm_read_guest_virt_helper(addr, val, bytes, vcpu, access,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6041) exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6042) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6043) EXPORT_SYMBOL_GPL(kvm_read_guest_virt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6044)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6045) static int emulator_read_std(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6046) gva_t addr, void *val, unsigned int bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6047) struct x86_exception *exception, bool system)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6048) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6049) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6050) u32 access = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6051)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6052) if (!system && kvm_x86_ops.get_cpl(vcpu) == 3)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6053) access |= PFERR_USER_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6054)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6055) return kvm_read_guest_virt_helper(addr, val, bytes, vcpu, access, exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6056) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6057)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6058) static int kvm_read_guest_phys_system(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6059) unsigned long addr, void *val, unsigned int bytes)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6060) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6061) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6062) int r = kvm_vcpu_read_guest(vcpu, addr, val, bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6063)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6064) return r < 0 ? X86EMUL_IO_NEEDED : X86EMUL_CONTINUE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6065) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6066)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6067) static int kvm_write_guest_virt_helper(gva_t addr, void *val, unsigned int bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6068) struct kvm_vcpu *vcpu, u32 access,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6069) struct x86_exception *exception)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6070) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6071) void *data = val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6072) int r = X86EMUL_CONTINUE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6073)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6074) while (bytes) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6075) gpa_t gpa = vcpu->arch.walk_mmu->gva_to_gpa(vcpu, addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6076) access,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6077) exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6078) unsigned offset = addr & (PAGE_SIZE-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6079) unsigned towrite = min(bytes, (unsigned)PAGE_SIZE - offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6080) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6081)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6082) if (gpa == UNMAPPED_GVA)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6083) return X86EMUL_PROPAGATE_FAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6084) ret = kvm_vcpu_write_guest(vcpu, gpa, data, towrite);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6085) if (ret < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6086) r = X86EMUL_IO_NEEDED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6087) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6088) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6089)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6090) bytes -= towrite;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6091) data += towrite;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6092) addr += towrite;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6093) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6094) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6095) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6096) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6097)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6098) static int emulator_write_std(struct x86_emulate_ctxt *ctxt, gva_t addr, void *val,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6099) unsigned int bytes, struct x86_exception *exception,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6100) bool system)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6101) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6102) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6103) u32 access = PFERR_WRITE_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6104)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6105) if (!system && kvm_x86_ops.get_cpl(vcpu) == 3)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6106) access |= PFERR_USER_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6107)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6108) return kvm_write_guest_virt_helper(addr, val, bytes, vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6109) access, exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6110) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6111)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6112) int kvm_write_guest_virt_system(struct kvm_vcpu *vcpu, gva_t addr, void *val,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6113) unsigned int bytes, struct x86_exception *exception)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6114) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6115) /* kvm_write_guest_virt_system can pull in tons of pages. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6116) vcpu->arch.l1tf_flush_l1d = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6117)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6118) return kvm_write_guest_virt_helper(addr, val, bytes, vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6119) PFERR_WRITE_MASK, exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6120) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6121) EXPORT_SYMBOL_GPL(kvm_write_guest_virt_system);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6122)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6123) int handle_ud(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6124) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6125) static const char kvm_emulate_prefix[] = { __KVM_EMULATE_PREFIX };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6126) int emul_type = EMULTYPE_TRAP_UD;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6127) char sig[5]; /* ud2; .ascii "kvm" */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6128) struct x86_exception e;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6129)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6130) if (unlikely(!kvm_x86_ops.can_emulate_instruction(vcpu, NULL, 0)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6131) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6132)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6133) if (force_emulation_prefix &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6134) kvm_read_guest_virt(vcpu, kvm_get_linear_rip(vcpu),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6135) sig, sizeof(sig), &e) == 0 &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6136) memcmp(sig, kvm_emulate_prefix, sizeof(sig)) == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6137) kvm_rip_write(vcpu, kvm_rip_read(vcpu) + sizeof(sig));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6138) emul_type = EMULTYPE_TRAP_UD_FORCED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6139) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6140)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6141) return kvm_emulate_instruction(vcpu, emul_type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6142) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6143) EXPORT_SYMBOL_GPL(handle_ud);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6144)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6145) static int vcpu_is_mmio_gpa(struct kvm_vcpu *vcpu, unsigned long gva,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6146) gpa_t gpa, bool write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6147) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6148) /* For APIC access vmexit */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6149) if ((gpa & PAGE_MASK) == APIC_DEFAULT_PHYS_BASE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6150) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6151)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6152) if (vcpu_match_mmio_gpa(vcpu, gpa)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6153) trace_vcpu_match_mmio(gva, gpa, write, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6154) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6155) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6156)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6157) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6158) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6159)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6160) static int vcpu_mmio_gva_to_gpa(struct kvm_vcpu *vcpu, unsigned long gva,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6161) gpa_t *gpa, struct x86_exception *exception,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6162) bool write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6163) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6164) u32 access = ((kvm_x86_ops.get_cpl(vcpu) == 3) ? PFERR_USER_MASK : 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6165) | (write ? PFERR_WRITE_MASK : 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6166)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6167) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6168) * currently PKRU is only applied to ept enabled guest so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6169) * there is no pkey in EPT page table for L1 guest or EPT
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6170) * shadow page table for L2 guest.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6171) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6172) if (vcpu_match_mmio_gva(vcpu, gva)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6173) && !permission_fault(vcpu, vcpu->arch.walk_mmu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6174) vcpu->arch.mmio_access, 0, access)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6175) *gpa = vcpu->arch.mmio_gfn << PAGE_SHIFT |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6176) (gva & (PAGE_SIZE - 1));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6177) trace_vcpu_match_mmio(gva, *gpa, write, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6178) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6179) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6180)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6181) *gpa = vcpu->arch.walk_mmu->gva_to_gpa(vcpu, gva, access, exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6182)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6183) if (*gpa == UNMAPPED_GVA)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6184) return -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6185)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6186) return vcpu_is_mmio_gpa(vcpu, gva, *gpa, write);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6187) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6188)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6189) int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6190) const void *val, int bytes)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6191) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6192) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6193)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6194) ret = kvm_vcpu_write_guest(vcpu, gpa, val, bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6195) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6196) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6197) kvm_page_track_write(vcpu, gpa, val, bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6198) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6199) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6200)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6201) struct read_write_emulator_ops {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6202) int (*read_write_prepare)(struct kvm_vcpu *vcpu, void *val,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6203) int bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6204) int (*read_write_emulate)(struct kvm_vcpu *vcpu, gpa_t gpa,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6205) void *val, int bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6206) int (*read_write_mmio)(struct kvm_vcpu *vcpu, gpa_t gpa,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6207) int bytes, void *val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6208) int (*read_write_exit_mmio)(struct kvm_vcpu *vcpu, gpa_t gpa,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6209) void *val, int bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6210) bool write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6211) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6212)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6213) static int read_prepare(struct kvm_vcpu *vcpu, void *val, int bytes)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6214) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6215) if (vcpu->mmio_read_completed) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6216) trace_kvm_mmio(KVM_TRACE_MMIO_READ, bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6217) vcpu->mmio_fragments[0].gpa, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6218) vcpu->mmio_read_completed = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6219) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6220) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6221)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6222) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6223) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6224)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6225) static int read_emulate(struct kvm_vcpu *vcpu, gpa_t gpa,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6226) void *val, int bytes)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6227) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6228) return !kvm_vcpu_read_guest(vcpu, gpa, val, bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6229) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6230)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6231) static int write_emulate(struct kvm_vcpu *vcpu, gpa_t gpa,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6232) void *val, int bytes)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6233) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6234) return emulator_write_phys(vcpu, gpa, val, bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6235) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6236)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6237) static int write_mmio(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes, void *val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6238) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6239) trace_kvm_mmio(KVM_TRACE_MMIO_WRITE, bytes, gpa, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6240) return vcpu_mmio_write(vcpu, gpa, bytes, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6241) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6242)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6243) static int read_exit_mmio(struct kvm_vcpu *vcpu, gpa_t gpa,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6244) void *val, int bytes)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6245) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6246) trace_kvm_mmio(KVM_TRACE_MMIO_READ_UNSATISFIED, bytes, gpa, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6247) return X86EMUL_IO_NEEDED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6248) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6249)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6250) static int write_exit_mmio(struct kvm_vcpu *vcpu, gpa_t gpa,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6251) void *val, int bytes)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6252) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6253) struct kvm_mmio_fragment *frag = &vcpu->mmio_fragments[0];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6254)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6255) memcpy(vcpu->run->mmio.data, frag->data, min(8u, frag->len));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6256) return X86EMUL_CONTINUE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6257) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6258)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6259) static const struct read_write_emulator_ops read_emultor = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6260) .read_write_prepare = read_prepare,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6261) .read_write_emulate = read_emulate,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6262) .read_write_mmio = vcpu_mmio_read,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6263) .read_write_exit_mmio = read_exit_mmio,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6264) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6265)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6266) static const struct read_write_emulator_ops write_emultor = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6267) .read_write_emulate = write_emulate,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6268) .read_write_mmio = write_mmio,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6269) .read_write_exit_mmio = write_exit_mmio,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6270) .write = true,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6271) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6272)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6273) static int emulator_read_write_onepage(unsigned long addr, void *val,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6274) unsigned int bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6275) struct x86_exception *exception,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6276) struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6277) const struct read_write_emulator_ops *ops)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6278) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6279) gpa_t gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6280) int handled, ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6281) bool write = ops->write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6282) struct kvm_mmio_fragment *frag;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6283) struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6284)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6285) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6286) * If the exit was due to a NPF we may already have a GPA.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6287) * If the GPA is present, use it to avoid the GVA to GPA table walk.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6288) * Note, this cannot be used on string operations since string
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6289) * operation using rep will only have the initial GPA from the NPF
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6290) * occurred.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6291) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6292) if (ctxt->gpa_available && emulator_can_use_gpa(ctxt) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6293) (addr & ~PAGE_MASK) == (ctxt->gpa_val & ~PAGE_MASK)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6294) gpa = ctxt->gpa_val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6295) ret = vcpu_is_mmio_gpa(vcpu, addr, gpa, write);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6296) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6297) ret = vcpu_mmio_gva_to_gpa(vcpu, addr, &gpa, exception, write);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6298) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6299) return X86EMUL_PROPAGATE_FAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6300) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6301)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6302) if (!ret && ops->read_write_emulate(vcpu, gpa, val, bytes))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6303) return X86EMUL_CONTINUE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6304)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6305) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6306) * Is this MMIO handled locally?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6307) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6308) handled = ops->read_write_mmio(vcpu, gpa, bytes, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6309) if (handled == bytes)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6310) return X86EMUL_CONTINUE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6311)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6312) gpa += handled;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6313) bytes -= handled;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6314) val += handled;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6315)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6316) WARN_ON(vcpu->mmio_nr_fragments >= KVM_MAX_MMIO_FRAGMENTS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6317) frag = &vcpu->mmio_fragments[vcpu->mmio_nr_fragments++];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6318) frag->gpa = gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6319) frag->data = val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6320) frag->len = bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6321) return X86EMUL_CONTINUE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6322) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6323)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6324) static int emulator_read_write(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6325) unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6326) void *val, unsigned int bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6327) struct x86_exception *exception,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6328) const struct read_write_emulator_ops *ops)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6329) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6330) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6331) gpa_t gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6332) int rc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6333)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6334) if (ops->read_write_prepare &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6335) ops->read_write_prepare(vcpu, val, bytes))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6336) return X86EMUL_CONTINUE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6337)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6338) vcpu->mmio_nr_fragments = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6339)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6340) /* Crossing a page boundary? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6341) if (((addr + bytes - 1) ^ addr) & PAGE_MASK) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6342) int now;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6343)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6344) now = -addr & ~PAGE_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6345) rc = emulator_read_write_onepage(addr, val, now, exception,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6346) vcpu, ops);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6347)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6348) if (rc != X86EMUL_CONTINUE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6349) return rc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6350) addr += now;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6351) if (ctxt->mode != X86EMUL_MODE_PROT64)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6352) addr = (u32)addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6353) val += now;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6354) bytes -= now;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6355) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6356)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6357) rc = emulator_read_write_onepage(addr, val, bytes, exception,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6358) vcpu, ops);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6359) if (rc != X86EMUL_CONTINUE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6360) return rc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6361)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6362) if (!vcpu->mmio_nr_fragments)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6363) return rc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6364)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6365) gpa = vcpu->mmio_fragments[0].gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6366)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6367) vcpu->mmio_needed = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6368) vcpu->mmio_cur_fragment = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6369)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6370) vcpu->run->mmio.len = min(8u, vcpu->mmio_fragments[0].len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6371) vcpu->run->mmio.is_write = vcpu->mmio_is_write = ops->write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6372) vcpu->run->exit_reason = KVM_EXIT_MMIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6373) vcpu->run->mmio.phys_addr = gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6374)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6375) return ops->read_write_exit_mmio(vcpu, gpa, val, bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6376) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6377)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6378) static int emulator_read_emulated(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6379) unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6380) void *val,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6381) unsigned int bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6382) struct x86_exception *exception)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6383) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6384) return emulator_read_write(ctxt, addr, val, bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6385) exception, &read_emultor);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6386) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6387)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6388) static int emulator_write_emulated(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6389) unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6390) const void *val,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6391) unsigned int bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6392) struct x86_exception *exception)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6393) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6394) return emulator_read_write(ctxt, addr, (void *)val, bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6395) exception, &write_emultor);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6396) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6397)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6398) #define CMPXCHG_TYPE(t, ptr, old, new) \
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6399) (cmpxchg((t *)(ptr), *(t *)(old), *(t *)(new)) == *(t *)(old))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6400)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6401) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6402) # define CMPXCHG64(ptr, old, new) CMPXCHG_TYPE(u64, ptr, old, new)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6403) #else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6404) # define CMPXCHG64(ptr, old, new) \
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6405) (cmpxchg64((u64 *)(ptr), *(u64 *)(old), *(u64 *)(new)) == *(u64 *)(old))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6406) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6407)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6408) static int emulator_cmpxchg_emulated(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6409) unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6410) const void *old,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6411) const void *new,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6412) unsigned int bytes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6413) struct x86_exception *exception)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6414) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6415) struct kvm_host_map map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6416) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6417) u64 page_line_mask;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6418) gpa_t gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6419) char *kaddr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6420) bool exchanged;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6421)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6422) /* guests cmpxchg8b have to be emulated atomically */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6423) if (bytes > 8 || (bytes & (bytes - 1)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6424) goto emul_write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6425)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6426) gpa = kvm_mmu_gva_to_gpa_write(vcpu, addr, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6427)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6428) if (gpa == UNMAPPED_GVA ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6429) (gpa & PAGE_MASK) == APIC_DEFAULT_PHYS_BASE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6430) goto emul_write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6431)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6432) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6433) * Emulate the atomic as a straight write to avoid #AC if SLD is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6434) * enabled in the host and the access splits a cache line.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6435) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6436) if (boot_cpu_has(X86_FEATURE_SPLIT_LOCK_DETECT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6437) page_line_mask = ~(cache_line_size() - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6438) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6439) page_line_mask = PAGE_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6440)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6441) if (((gpa + bytes - 1) & page_line_mask) != (gpa & page_line_mask))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6442) goto emul_write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6443)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6444) if (kvm_vcpu_map(vcpu, gpa_to_gfn(gpa), &map))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6445) goto emul_write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6446)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6447) kaddr = map.hva + offset_in_page(gpa);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6448)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6449) switch (bytes) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6450) case 1:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6451) exchanged = CMPXCHG_TYPE(u8, kaddr, old, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6452) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6453) case 2:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6454) exchanged = CMPXCHG_TYPE(u16, kaddr, old, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6455) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6456) case 4:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6457) exchanged = CMPXCHG_TYPE(u32, kaddr, old, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6458) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6459) case 8:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6460) exchanged = CMPXCHG64(kaddr, old, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6461) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6462) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6463) BUG();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6464) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6465)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6466) kvm_vcpu_unmap(vcpu, &map, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6467)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6468) if (!exchanged)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6469) return X86EMUL_CMPXCHG_FAILED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6470)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6471) kvm_page_track_write(vcpu, gpa, new, bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6472)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6473) return X86EMUL_CONTINUE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6474)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6475) emul_write:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6476) printk_once(KERN_WARNING "kvm: emulating exchange as write\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6477)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6478) return emulator_write_emulated(ctxt, addr, new, bytes, exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6479) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6480)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6481) static int kernel_pio(struct kvm_vcpu *vcpu, void *pd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6482) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6483) int r = 0, i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6484)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6485) for (i = 0; i < vcpu->arch.pio.count; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6486) if (vcpu->arch.pio.in)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6487) r = kvm_io_bus_read(vcpu, KVM_PIO_BUS, vcpu->arch.pio.port,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6488) vcpu->arch.pio.size, pd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6489) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6490) r = kvm_io_bus_write(vcpu, KVM_PIO_BUS,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6491) vcpu->arch.pio.port, vcpu->arch.pio.size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6492) pd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6493) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6494) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6495) pd += vcpu->arch.pio.size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6496) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6497) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6498) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6499)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6500) static int emulator_pio_in_out(struct kvm_vcpu *vcpu, int size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6501) unsigned short port, void *val,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6502) unsigned int count, bool in)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6503) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6504) vcpu->arch.pio.port = port;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6505) vcpu->arch.pio.in = in;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6506) vcpu->arch.pio.count = count;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6507) vcpu->arch.pio.size = size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6508)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6509) if (!kernel_pio(vcpu, vcpu->arch.pio_data)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6510) vcpu->arch.pio.count = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6511) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6512) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6513)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6514) vcpu->run->exit_reason = KVM_EXIT_IO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6515) vcpu->run->io.direction = in ? KVM_EXIT_IO_IN : KVM_EXIT_IO_OUT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6516) vcpu->run->io.size = size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6517) vcpu->run->io.data_offset = KVM_PIO_PAGE_OFFSET * PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6518) vcpu->run->io.count = count;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6519) vcpu->run->io.port = port;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6520)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6521) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6522) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6523)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6524) static int emulator_pio_in(struct kvm_vcpu *vcpu, int size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6525) unsigned short port, void *val, unsigned int count)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6526) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6527) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6528)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6529) if (vcpu->arch.pio.count)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6530) goto data_avail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6531)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6532) memset(vcpu->arch.pio_data, 0, size * count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6533)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6534) ret = emulator_pio_in_out(vcpu, size, port, val, count, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6535) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6536) data_avail:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6537) memcpy(val, vcpu->arch.pio_data, size * count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6538) trace_kvm_pio(KVM_PIO_IN, port, size, count, vcpu->arch.pio_data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6539) vcpu->arch.pio.count = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6540) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6541) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6542)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6543) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6544) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6545)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6546) static int emulator_pio_in_emulated(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6547) int size, unsigned short port, void *val,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6548) unsigned int count)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6549) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6550) return emulator_pio_in(emul_to_vcpu(ctxt), size, port, val, count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6551)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6552) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6553)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6554) static int emulator_pio_out(struct kvm_vcpu *vcpu, int size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6555) unsigned short port, const void *val,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6556) unsigned int count)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6557) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6558) memcpy(vcpu->arch.pio_data, val, size * count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6559) trace_kvm_pio(KVM_PIO_OUT, port, size, count, vcpu->arch.pio_data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6560) return emulator_pio_in_out(vcpu, size, port, (void *)val, count, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6561) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6562)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6563) static int emulator_pio_out_emulated(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6564) int size, unsigned short port,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6565) const void *val, unsigned int count)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6566) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6567) return emulator_pio_out(emul_to_vcpu(ctxt), size, port, val, count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6568) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6569)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6570) static unsigned long get_segment_base(struct kvm_vcpu *vcpu, int seg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6571) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6572) return kvm_x86_ops.get_segment_base(vcpu, seg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6573) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6574)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6575) static void emulator_invlpg(struct x86_emulate_ctxt *ctxt, ulong address)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6576) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6577) kvm_mmu_invlpg(emul_to_vcpu(ctxt), address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6578) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6579)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6580) static int kvm_emulate_wbinvd_noskip(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6581) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6582) if (!need_emulate_wbinvd(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6583) return X86EMUL_CONTINUE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6584)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6585) if (kvm_x86_ops.has_wbinvd_exit()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6586) int cpu = get_cpu();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6587)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6588) cpumask_set_cpu(cpu, vcpu->arch.wbinvd_dirty_mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6589) smp_call_function_many(vcpu->arch.wbinvd_dirty_mask,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6590) wbinvd_ipi, NULL, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6591) put_cpu();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6592) cpumask_clear(vcpu->arch.wbinvd_dirty_mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6593) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6594) wbinvd();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6595) return X86EMUL_CONTINUE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6596) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6597)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6598) int kvm_emulate_wbinvd(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6599) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6600) kvm_emulate_wbinvd_noskip(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6601) return kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6602) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6603) EXPORT_SYMBOL_GPL(kvm_emulate_wbinvd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6604)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6605)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6606)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6607) static void emulator_wbinvd(struct x86_emulate_ctxt *ctxt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6608) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6609) kvm_emulate_wbinvd_noskip(emul_to_vcpu(ctxt));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6610) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6611)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6612) static int emulator_get_dr(struct x86_emulate_ctxt *ctxt, int dr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6613) unsigned long *dest)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6614) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6615) return kvm_get_dr(emul_to_vcpu(ctxt), dr, dest);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6616) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6617)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6618) static int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6619) unsigned long value)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6620) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6621)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6622) return __kvm_set_dr(emul_to_vcpu(ctxt), dr, value);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6623) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6624)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6625) static u64 mk_cr_64(u64 curr_cr, u32 new_val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6626) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6627) return (curr_cr & ~((1ULL << 32) - 1)) | new_val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6628) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6629)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6630) static unsigned long emulator_get_cr(struct x86_emulate_ctxt *ctxt, int cr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6631) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6632) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6633) unsigned long value;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6634)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6635) switch (cr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6636) case 0:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6637) value = kvm_read_cr0(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6638) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6639) case 2:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6640) value = vcpu->arch.cr2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6641) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6642) case 3:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6643) value = kvm_read_cr3(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6644) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6645) case 4:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6646) value = kvm_read_cr4(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6647) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6648) case 8:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6649) value = kvm_get_cr8(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6650) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6651) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6652) kvm_err("%s: unexpected cr %u\n", __func__, cr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6653) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6654) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6655)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6656) return value;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6657) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6658)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6659) static int emulator_set_cr(struct x86_emulate_ctxt *ctxt, int cr, ulong val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6660) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6661) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6662) int res = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6663)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6664) switch (cr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6665) case 0:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6666) res = kvm_set_cr0(vcpu, mk_cr_64(kvm_read_cr0(vcpu), val));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6667) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6668) case 2:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6669) vcpu->arch.cr2 = val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6670) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6671) case 3:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6672) res = kvm_set_cr3(vcpu, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6673) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6674) case 4:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6675) res = kvm_set_cr4(vcpu, mk_cr_64(kvm_read_cr4(vcpu), val));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6676) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6677) case 8:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6678) res = kvm_set_cr8(vcpu, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6679) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6680) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6681) kvm_err("%s: unexpected cr %u\n", __func__, cr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6682) res = -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6683) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6684)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6685) return res;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6686) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6687)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6688) static int emulator_get_cpl(struct x86_emulate_ctxt *ctxt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6689) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6690) return kvm_x86_ops.get_cpl(emul_to_vcpu(ctxt));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6691) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6692)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6693) static void emulator_get_gdt(struct x86_emulate_ctxt *ctxt, struct desc_ptr *dt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6694) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6695) kvm_x86_ops.get_gdt(emul_to_vcpu(ctxt), dt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6696) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6697)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6698) static void emulator_get_idt(struct x86_emulate_ctxt *ctxt, struct desc_ptr *dt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6699) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6700) kvm_x86_ops.get_idt(emul_to_vcpu(ctxt), dt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6701) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6702)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6703) static void emulator_set_gdt(struct x86_emulate_ctxt *ctxt, struct desc_ptr *dt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6704) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6705) kvm_x86_ops.set_gdt(emul_to_vcpu(ctxt), dt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6706) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6707)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6708) static void emulator_set_idt(struct x86_emulate_ctxt *ctxt, struct desc_ptr *dt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6709) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6710) kvm_x86_ops.set_idt(emul_to_vcpu(ctxt), dt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6711) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6712)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6713) static unsigned long emulator_get_cached_segment_base(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6714) struct x86_emulate_ctxt *ctxt, int seg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6715) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6716) return get_segment_base(emul_to_vcpu(ctxt), seg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6717) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6718)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6719) static bool emulator_get_segment(struct x86_emulate_ctxt *ctxt, u16 *selector,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6720) struct desc_struct *desc, u32 *base3,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6721) int seg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6722) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6723) struct kvm_segment var;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6724)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6725) kvm_get_segment(emul_to_vcpu(ctxt), &var, seg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6726) *selector = var.selector;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6727)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6728) if (var.unusable) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6729) memset(desc, 0, sizeof(*desc));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6730) if (base3)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6731) *base3 = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6732) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6733) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6734)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6735) if (var.g)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6736) var.limit >>= 12;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6737) set_desc_limit(desc, var.limit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6738) set_desc_base(desc, (unsigned long)var.base);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6739) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6740) if (base3)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6741) *base3 = var.base >> 32;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6742) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6743) desc->type = var.type;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6744) desc->s = var.s;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6745) desc->dpl = var.dpl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6746) desc->p = var.present;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6747) desc->avl = var.avl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6748) desc->l = var.l;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6749) desc->d = var.db;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6750) desc->g = var.g;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6751)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6752) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6753) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6754)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6755) static void emulator_set_segment(struct x86_emulate_ctxt *ctxt, u16 selector,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6756) struct desc_struct *desc, u32 base3,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6757) int seg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6758) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6759) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6760) struct kvm_segment var;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6761)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6762) var.selector = selector;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6763) var.base = get_desc_base(desc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6764) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6765) var.base |= ((u64)base3) << 32;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6766) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6767) var.limit = get_desc_limit(desc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6768) if (desc->g)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6769) var.limit = (var.limit << 12) | 0xfff;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6770) var.type = desc->type;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6771) var.dpl = desc->dpl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6772) var.db = desc->d;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6773) var.s = desc->s;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6774) var.l = desc->l;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6775) var.g = desc->g;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6776) var.avl = desc->avl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6777) var.present = desc->p;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6778) var.unusable = !var.present;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6779) var.padding = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6780)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6781) kvm_set_segment(vcpu, &var, seg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6782) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6783) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6784)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6785) static int emulator_get_msr(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6786) u32 msr_index, u64 *pdata)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6787) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6788) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6789) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6790)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6791) r = kvm_get_msr(vcpu, msr_index, pdata);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6792)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6793) if (r && kvm_get_msr_user_space(vcpu, msr_index, r)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6794) /* Bounce to user space */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6795) return X86EMUL_IO_NEEDED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6796) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6797)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6798) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6799) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6800)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6801) static int emulator_set_msr(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6802) u32 msr_index, u64 data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6803) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6804) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6805) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6806)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6807) r = kvm_set_msr(vcpu, msr_index, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6808)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6809) if (r && kvm_set_msr_user_space(vcpu, msr_index, data, r)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6810) /* Bounce to user space */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6811) return X86EMUL_IO_NEEDED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6812) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6813)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6814) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6815) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6816)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6817) static u64 emulator_get_smbase(struct x86_emulate_ctxt *ctxt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6818) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6819) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6820)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6821) return vcpu->arch.smbase;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6822) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6823)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6824) static void emulator_set_smbase(struct x86_emulate_ctxt *ctxt, u64 smbase)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6825) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6826) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6827)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6828) vcpu->arch.smbase = smbase;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6829) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6830)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6831) static int emulator_check_pmc(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6832) u32 pmc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6833) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6834) return kvm_pmu_is_valid_rdpmc_ecx(emul_to_vcpu(ctxt), pmc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6835) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6836)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6837) static int emulator_read_pmc(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6838) u32 pmc, u64 *pdata)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6839) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6840) return kvm_pmu_rdpmc(emul_to_vcpu(ctxt), pmc, pdata);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6841) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6842)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6843) static void emulator_halt(struct x86_emulate_ctxt *ctxt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6844) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6845) emul_to_vcpu(ctxt)->arch.halt_request = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6846) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6847)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6848) static int emulator_intercept(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6849) struct x86_instruction_info *info,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6850) enum x86_intercept_stage stage)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6851) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6852) return kvm_x86_ops.check_intercept(emul_to_vcpu(ctxt), info, stage,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6853) &ctxt->exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6854) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6855)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6856) static bool emulator_get_cpuid(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6857) u32 *eax, u32 *ebx, u32 *ecx, u32 *edx,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6858) bool exact_only)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6859) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6860) return kvm_cpuid(emul_to_vcpu(ctxt), eax, ebx, ecx, edx, exact_only);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6861) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6862)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6863) static bool emulator_guest_has_long_mode(struct x86_emulate_ctxt *ctxt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6864) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6865) return guest_cpuid_has(emul_to_vcpu(ctxt), X86_FEATURE_LM);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6866) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6867)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6868) static bool emulator_guest_has_movbe(struct x86_emulate_ctxt *ctxt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6869) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6870) return guest_cpuid_has(emul_to_vcpu(ctxt), X86_FEATURE_MOVBE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6871) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6872)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6873) static bool emulator_guest_has_fxsr(struct x86_emulate_ctxt *ctxt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6874) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6875) return guest_cpuid_has(emul_to_vcpu(ctxt), X86_FEATURE_FXSR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6876) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6877)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6878) static ulong emulator_read_gpr(struct x86_emulate_ctxt *ctxt, unsigned reg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6879) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6880) return kvm_register_read(emul_to_vcpu(ctxt), reg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6881) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6882)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6883) static void emulator_write_gpr(struct x86_emulate_ctxt *ctxt, unsigned reg, ulong val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6884) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6885) kvm_register_write(emul_to_vcpu(ctxt), reg, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6886) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6887)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6888) static void emulator_set_nmi_mask(struct x86_emulate_ctxt *ctxt, bool masked)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6889) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6890) kvm_x86_ops.set_nmi_mask(emul_to_vcpu(ctxt), masked);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6891) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6892)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6893) static unsigned emulator_get_hflags(struct x86_emulate_ctxt *ctxt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6894) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6895) return emul_to_vcpu(ctxt)->arch.hflags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6896) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6897)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6898) static void emulator_set_hflags(struct x86_emulate_ctxt *ctxt, unsigned emul_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6899) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6900) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6901)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6902) vcpu->arch.hflags = emul_flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6903) kvm_mmu_reset_context(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6904) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6905)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6906) static int emulator_pre_leave_smm(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6907) const char *smstate)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6908) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6909) return kvm_x86_ops.pre_leave_smm(emul_to_vcpu(ctxt), smstate);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6910) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6911)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6912) static void emulator_post_leave_smm(struct x86_emulate_ctxt *ctxt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6913) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6914) kvm_smm_changed(emul_to_vcpu(ctxt));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6915) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6916)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6917) static int emulator_set_xcr(struct x86_emulate_ctxt *ctxt, u32 index, u64 xcr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6918) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6919) return __kvm_set_xcr(emul_to_vcpu(ctxt), index, xcr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6920) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6921)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6922) static const struct x86_emulate_ops emulate_ops = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6923) .read_gpr = emulator_read_gpr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6924) .write_gpr = emulator_write_gpr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6925) .read_std = emulator_read_std,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6926) .write_std = emulator_write_std,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6927) .read_phys = kvm_read_guest_phys_system,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6928) .fetch = kvm_fetch_guest_virt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6929) .read_emulated = emulator_read_emulated,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6930) .write_emulated = emulator_write_emulated,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6931) .cmpxchg_emulated = emulator_cmpxchg_emulated,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6932) .invlpg = emulator_invlpg,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6933) .pio_in_emulated = emulator_pio_in_emulated,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6934) .pio_out_emulated = emulator_pio_out_emulated,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6935) .get_segment = emulator_get_segment,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6936) .set_segment = emulator_set_segment,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6937) .get_cached_segment_base = emulator_get_cached_segment_base,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6938) .get_gdt = emulator_get_gdt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6939) .get_idt = emulator_get_idt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6940) .set_gdt = emulator_set_gdt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6941) .set_idt = emulator_set_idt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6942) .get_cr = emulator_get_cr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6943) .set_cr = emulator_set_cr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6944) .cpl = emulator_get_cpl,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6945) .get_dr = emulator_get_dr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6946) .set_dr = emulator_set_dr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6947) .get_smbase = emulator_get_smbase,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6948) .set_smbase = emulator_set_smbase,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6949) .set_msr = emulator_set_msr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6950) .get_msr = emulator_get_msr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6951) .check_pmc = emulator_check_pmc,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6952) .read_pmc = emulator_read_pmc,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6953) .halt = emulator_halt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6954) .wbinvd = emulator_wbinvd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6955) .fix_hypercall = emulator_fix_hypercall,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6956) .intercept = emulator_intercept,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6957) .get_cpuid = emulator_get_cpuid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6958) .guest_has_long_mode = emulator_guest_has_long_mode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6959) .guest_has_movbe = emulator_guest_has_movbe,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6960) .guest_has_fxsr = emulator_guest_has_fxsr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6961) .set_nmi_mask = emulator_set_nmi_mask,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6962) .get_hflags = emulator_get_hflags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6963) .set_hflags = emulator_set_hflags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6964) .pre_leave_smm = emulator_pre_leave_smm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6965) .post_leave_smm = emulator_post_leave_smm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6966) .set_xcr = emulator_set_xcr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6967) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6968)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6969) static void toggle_interruptibility(struct kvm_vcpu *vcpu, u32 mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6970) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6971) u32 int_shadow = kvm_x86_ops.get_interrupt_shadow(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6972) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6973) * an sti; sti; sequence only disable interrupts for the first
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6974) * instruction. So, if the last instruction, be it emulated or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6975) * not, left the system with the INT_STI flag enabled, it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6976) * means that the last instruction is an sti. We should not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6977) * leave the flag on in this case. The same goes for mov ss
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6978) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6979) if (int_shadow & mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6980) mask = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6981) if (unlikely(int_shadow || mask)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6982) kvm_x86_ops.set_interrupt_shadow(vcpu, mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6983) if (!mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6984) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6985) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6986) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6987)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6988) static bool inject_emulated_exception(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6989) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6990) struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6991) if (ctxt->exception.vector == PF_VECTOR)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6992) return kvm_inject_emulated_page_fault(vcpu, &ctxt->exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6993)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6994) if (ctxt->exception.error_code_valid)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6995) kvm_queue_exception_e(vcpu, ctxt->exception.vector,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6996) ctxt->exception.error_code);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6997) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6998) kvm_queue_exception(vcpu, ctxt->exception.vector);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6999) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7000) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7001)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7002) static struct x86_emulate_ctxt *alloc_emulate_ctxt(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7003) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7004) struct x86_emulate_ctxt *ctxt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7005)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7006) ctxt = kmem_cache_zalloc(x86_emulator_cache, GFP_KERNEL_ACCOUNT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7007) if (!ctxt) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7008) pr_err("kvm: failed to allocate vcpu's emulator\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7009) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7010) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7011)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7012) ctxt->vcpu = vcpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7013) ctxt->ops = &emulate_ops;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7014) vcpu->arch.emulate_ctxt = ctxt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7015)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7016) return ctxt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7017) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7018)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7019) static void init_emulate_ctxt(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7020) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7021) struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7022) int cs_db, cs_l;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7023)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7024) kvm_x86_ops.get_cs_db_l_bits(vcpu, &cs_db, &cs_l);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7025)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7026) ctxt->gpa_available = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7027) ctxt->eflags = kvm_get_rflags(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7028) ctxt->tf = (ctxt->eflags & X86_EFLAGS_TF) != 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7029)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7030) ctxt->eip = kvm_rip_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7031) ctxt->mode = (!is_protmode(vcpu)) ? X86EMUL_MODE_REAL :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7032) (ctxt->eflags & X86_EFLAGS_VM) ? X86EMUL_MODE_VM86 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7033) (cs_l && is_long_mode(vcpu)) ? X86EMUL_MODE_PROT64 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7034) cs_db ? X86EMUL_MODE_PROT32 :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7035) X86EMUL_MODE_PROT16;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7036) BUILD_BUG_ON(HF_GUEST_MASK != X86EMUL_GUEST_MASK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7037) BUILD_BUG_ON(HF_SMM_MASK != X86EMUL_SMM_MASK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7038) BUILD_BUG_ON(HF_SMM_INSIDE_NMI_MASK != X86EMUL_SMM_INSIDE_NMI_MASK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7039)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7040) ctxt->interruptibility = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7041) ctxt->have_exception = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7042) ctxt->exception.vector = -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7043) ctxt->perm_ok = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7044)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7045) init_decode_cache(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7046) vcpu->arch.emulate_regs_need_sync_from_vcpu = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7047) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7048)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7049) void kvm_inject_realmode_interrupt(struct kvm_vcpu *vcpu, int irq, int inc_eip)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7050) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7051) struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7052) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7053)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7054) init_emulate_ctxt(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7055)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7056) ctxt->op_bytes = 2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7057) ctxt->ad_bytes = 2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7058) ctxt->_eip = ctxt->eip + inc_eip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7059) ret = emulate_int_real(ctxt, irq);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7060)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7061) if (ret != X86EMUL_CONTINUE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7062) kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7063) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7064) ctxt->eip = ctxt->_eip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7065) kvm_rip_write(vcpu, ctxt->eip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7066) kvm_set_rflags(vcpu, ctxt->eflags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7067) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7068) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7069) EXPORT_SYMBOL_GPL(kvm_inject_realmode_interrupt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7070)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7071) static int handle_emulation_failure(struct kvm_vcpu *vcpu, int emulation_type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7072) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7073) ++vcpu->stat.insn_emulation_fail;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7074) trace_kvm_emulate_insn_failed(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7075)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7076) if (emulation_type & EMULTYPE_VMWARE_GP) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7077) kvm_queue_exception_e(vcpu, GP_VECTOR, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7078) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7079) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7080)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7081) if (emulation_type & EMULTYPE_SKIP) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7082) vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7083) vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_EMULATION;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7084) vcpu->run->internal.ndata = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7085) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7086) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7087)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7088) kvm_queue_exception(vcpu, UD_VECTOR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7089)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7090) if (!is_guest_mode(vcpu) && kvm_x86_ops.get_cpl(vcpu) == 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7091) vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7092) vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_EMULATION;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7093) vcpu->run->internal.ndata = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7094) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7095) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7096)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7097) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7098) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7099)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7100) static bool reexecute_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7101) bool write_fault_to_shadow_pgtable,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7102) int emulation_type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7103) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7104) gpa_t gpa = cr2_or_gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7105) kvm_pfn_t pfn;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7106)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7107) if (!(emulation_type & EMULTYPE_ALLOW_RETRY_PF))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7108) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7109)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7110) if (WARN_ON_ONCE(is_guest_mode(vcpu)) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7111) WARN_ON_ONCE(!(emulation_type & EMULTYPE_PF)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7112) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7113)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7114) if (!vcpu->arch.mmu->direct_map) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7115) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7116) * Write permission should be allowed since only
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7117) * write access need to be emulated.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7118) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7119) gpa = kvm_mmu_gva_to_gpa_write(vcpu, cr2_or_gpa, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7120)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7121) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7122) * If the mapping is invalid in guest, let cpu retry
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7123) * it to generate fault.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7124) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7125) if (gpa == UNMAPPED_GVA)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7126) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7127) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7128)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7129) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7130) * Do not retry the unhandleable instruction if it faults on the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7131) * readonly host memory, otherwise it will goto a infinite loop:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7132) * retry instruction -> write #PF -> emulation fail -> retry
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7133) * instruction -> ...
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7134) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7135) pfn = gfn_to_pfn(vcpu->kvm, gpa_to_gfn(gpa));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7136)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7137) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7138) * If the instruction failed on the error pfn, it can not be fixed,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7139) * report the error to userspace.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7140) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7141) if (is_error_noslot_pfn(pfn))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7142) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7143)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7144) kvm_release_pfn_clean(pfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7145)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7146) /* The instructions are well-emulated on direct mmu. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7147) if (vcpu->arch.mmu->direct_map) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7148) unsigned int indirect_shadow_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7149)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7150) spin_lock(&vcpu->kvm->mmu_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7151) indirect_shadow_pages = vcpu->kvm->arch.indirect_shadow_pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7152) spin_unlock(&vcpu->kvm->mmu_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7153)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7154) if (indirect_shadow_pages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7155) kvm_mmu_unprotect_page(vcpu->kvm, gpa_to_gfn(gpa));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7156)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7157) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7158) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7159)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7160) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7161) * if emulation was due to access to shadowed page table
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7162) * and it failed try to unshadow page and re-enter the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7163) * guest to let CPU execute the instruction.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7164) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7165) kvm_mmu_unprotect_page(vcpu->kvm, gpa_to_gfn(gpa));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7166)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7167) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7168) * If the access faults on its page table, it can not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7169) * be fixed by unprotecting shadow page and it should
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7170) * be reported to userspace.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7171) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7172) return !write_fault_to_shadow_pgtable;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7173) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7174)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7175) static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7176) gpa_t cr2_or_gpa, int emulation_type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7177) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7178) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7179) unsigned long last_retry_eip, last_retry_addr, gpa = cr2_or_gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7180)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7181) last_retry_eip = vcpu->arch.last_retry_eip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7182) last_retry_addr = vcpu->arch.last_retry_addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7183)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7184) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7185) * If the emulation is caused by #PF and it is non-page_table
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7186) * writing instruction, it means the VM-EXIT is caused by shadow
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7187) * page protected, we can zap the shadow page and retry this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7188) * instruction directly.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7189) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7190) * Note: if the guest uses a non-page-table modifying instruction
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7191) * on the PDE that points to the instruction, then we will unmap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7192) * the instruction and go to an infinite loop. So, we cache the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7193) * last retried eip and the last fault address, if we meet the eip
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7194) * and the address again, we can break out of the potential infinite
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7195) * loop.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7196) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7197) vcpu->arch.last_retry_eip = vcpu->arch.last_retry_addr = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7198)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7199) if (!(emulation_type & EMULTYPE_ALLOW_RETRY_PF))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7200) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7201)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7202) if (WARN_ON_ONCE(is_guest_mode(vcpu)) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7203) WARN_ON_ONCE(!(emulation_type & EMULTYPE_PF)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7204) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7205)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7206) if (x86_page_table_writing_insn(ctxt))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7207) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7208)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7209) if (ctxt->eip == last_retry_eip && last_retry_addr == cr2_or_gpa)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7210) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7211)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7212) vcpu->arch.last_retry_eip = ctxt->eip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7213) vcpu->arch.last_retry_addr = cr2_or_gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7214)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7215) if (!vcpu->arch.mmu->direct_map)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7216) gpa = kvm_mmu_gva_to_gpa_write(vcpu, cr2_or_gpa, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7217)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7218) kvm_mmu_unprotect_page(vcpu->kvm, gpa_to_gfn(gpa));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7219)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7220) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7221) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7222)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7223) static int complete_emulated_mmio(struct kvm_vcpu *vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7224) static int complete_emulated_pio(struct kvm_vcpu *vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7225)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7226) static void kvm_smm_changed(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7227) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7228) if (!(vcpu->arch.hflags & HF_SMM_MASK)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7229) /* This is a good place to trace that we are exiting SMM. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7230) trace_kvm_enter_smm(vcpu->vcpu_id, vcpu->arch.smbase, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7231)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7232) /* Process a latched INIT or SMI, if any. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7233) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7234) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7235)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7236) kvm_mmu_reset_context(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7237) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7238)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7239) static int kvm_vcpu_check_hw_bp(unsigned long addr, u32 type, u32 dr7,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7240) unsigned long *db)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7241) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7242) u32 dr6 = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7243) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7244) u32 enable, rwlen;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7245)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7246) enable = dr7;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7247) rwlen = dr7 >> 16;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7248) for (i = 0; i < 4; i++, enable >>= 2, rwlen >>= 4)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7249) if ((enable & 3) && (rwlen & 15) == type && db[i] == addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7250) dr6 |= (1 << i);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7251) return dr6;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7252) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7253)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7254) static int kvm_vcpu_do_singlestep(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7255) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7256) struct kvm_run *kvm_run = vcpu->run;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7257)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7258) if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7259) kvm_run->debug.arch.dr6 = DR6_BS | DR6_FIXED_1 | DR6_RTM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7260) kvm_run->debug.arch.pc = kvm_get_linear_rip(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7261) kvm_run->debug.arch.exception = DB_VECTOR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7262) kvm_run->exit_reason = KVM_EXIT_DEBUG;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7263) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7264) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7265) kvm_queue_exception_p(vcpu, DB_VECTOR, DR6_BS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7266) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7267) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7268)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7269) int kvm_skip_emulated_instruction(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7270) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7271) unsigned long rflags = kvm_x86_ops.get_rflags(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7272) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7273)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7274) r = kvm_x86_ops.skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7275) if (unlikely(!r))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7276) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7277)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7278) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7279) * rflags is the old, "raw" value of the flags. The new value has
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7280) * not been saved yet.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7281) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7282) * This is correct even for TF set by the guest, because "the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7283) * processor will not generate this exception after the instruction
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7284) * that sets the TF flag".
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7285) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7286) if (unlikely(rflags & X86_EFLAGS_TF))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7287) r = kvm_vcpu_do_singlestep(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7288) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7289) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7290) EXPORT_SYMBOL_GPL(kvm_skip_emulated_instruction);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7291)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7292) static bool kvm_vcpu_check_breakpoint(struct kvm_vcpu *vcpu, int *r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7293) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7294) if (unlikely(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7295) (vcpu->arch.guest_debug_dr7 & DR7_BP_EN_MASK)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7296) struct kvm_run *kvm_run = vcpu->run;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7297) unsigned long eip = kvm_get_linear_rip(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7298) u32 dr6 = kvm_vcpu_check_hw_bp(eip, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7299) vcpu->arch.guest_debug_dr7,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7300) vcpu->arch.eff_db);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7301)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7302) if (dr6 != 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7303) kvm_run->debug.arch.dr6 = dr6 | DR6_FIXED_1 | DR6_RTM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7304) kvm_run->debug.arch.pc = eip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7305) kvm_run->debug.arch.exception = DB_VECTOR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7306) kvm_run->exit_reason = KVM_EXIT_DEBUG;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7307) *r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7308) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7309) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7310) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7311)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7312) if (unlikely(vcpu->arch.dr7 & DR7_BP_EN_MASK) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7313) !(kvm_get_rflags(vcpu) & X86_EFLAGS_RF)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7314) unsigned long eip = kvm_get_linear_rip(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7315) u32 dr6 = kvm_vcpu_check_hw_bp(eip, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7316) vcpu->arch.dr7,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7317) vcpu->arch.db);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7318)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7319) if (dr6 != 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7320) kvm_queue_exception_p(vcpu, DB_VECTOR, dr6);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7321) *r = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7322) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7323) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7324) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7325)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7326) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7327) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7328)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7329) static bool is_vmware_backdoor_opcode(struct x86_emulate_ctxt *ctxt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7330) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7331) switch (ctxt->opcode_len) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7332) case 1:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7333) switch (ctxt->b) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7334) case 0xe4: /* IN */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7335) case 0xe5:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7336) case 0xec:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7337) case 0xed:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7338) case 0xe6: /* OUT */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7339) case 0xe7:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7340) case 0xee:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7341) case 0xef:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7342) case 0x6c: /* INS */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7343) case 0x6d:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7344) case 0x6e: /* OUTS */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7345) case 0x6f:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7346) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7347) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7348) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7349) case 2:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7350) switch (ctxt->b) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7351) case 0x33: /* RDPMC */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7352) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7353) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7354) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7355) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7356)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7357) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7358) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7359)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7360) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7361) * Decode to be emulated instruction. Return EMULATION_OK if success.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7362) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7363) int x86_decode_emulated_instruction(struct kvm_vcpu *vcpu, int emulation_type,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7364) void *insn, int insn_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7365) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7366) int r = EMULATION_OK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7367) struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7368)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7369) init_emulate_ctxt(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7370)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7371) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7372) * We will reenter on the same instruction since we do not set
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7373) * complete_userspace_io. This does not handle watchpoints yet,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7374) * those would be handled in the emulate_ops.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7375) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7376) if (!(emulation_type & EMULTYPE_SKIP) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7377) kvm_vcpu_check_breakpoint(vcpu, &r))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7378) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7379)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7380) ctxt->ud = emulation_type & EMULTYPE_TRAP_UD;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7381)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7382) r = x86_decode_insn(ctxt, insn, insn_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7383)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7384) trace_kvm_emulate_insn_start(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7385) ++vcpu->stat.insn_emulation;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7386)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7387) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7388) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7389) EXPORT_SYMBOL_GPL(x86_decode_emulated_instruction);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7390)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7391) int x86_emulate_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7392) int emulation_type, void *insn, int insn_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7393) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7394) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7395) struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7396) bool writeback = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7397) bool write_fault_to_spt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7398)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7399) if (unlikely(!kvm_x86_ops.can_emulate_instruction(vcpu, insn, insn_len)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7400) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7401)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7402) vcpu->arch.l1tf_flush_l1d = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7403)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7404) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7405) * Clear write_fault_to_shadow_pgtable here to ensure it is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7406) * never reused.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7407) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7408) write_fault_to_spt = vcpu->arch.write_fault_to_shadow_pgtable;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7409) vcpu->arch.write_fault_to_shadow_pgtable = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7410)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7411) if (!(emulation_type & EMULTYPE_NO_DECODE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7412) kvm_clear_exception_queue(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7413)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7414) r = x86_decode_emulated_instruction(vcpu, emulation_type,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7415) insn, insn_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7416) if (r != EMULATION_OK) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7417) if ((emulation_type & EMULTYPE_TRAP_UD) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7418) (emulation_type & EMULTYPE_TRAP_UD_FORCED)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7419) kvm_queue_exception(vcpu, UD_VECTOR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7420) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7421) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7422) if (reexecute_instruction(vcpu, cr2_or_gpa,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7423) write_fault_to_spt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7424) emulation_type))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7425) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7426) if (ctxt->have_exception) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7427) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7428) * #UD should result in just EMULATION_FAILED, and trap-like
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7429) * exception should not be encountered during decode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7430) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7431) WARN_ON_ONCE(ctxt->exception.vector == UD_VECTOR ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7432) exception_type(ctxt->exception.vector) == EXCPT_TRAP);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7433) inject_emulated_exception(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7434) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7435) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7436) return handle_emulation_failure(vcpu, emulation_type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7437) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7438) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7439)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7440) if ((emulation_type & EMULTYPE_VMWARE_GP) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7441) !is_vmware_backdoor_opcode(ctxt)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7442) kvm_queue_exception_e(vcpu, GP_VECTOR, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7443) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7444) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7445)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7446) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7447) * Note, EMULTYPE_SKIP is intended for use *only* by vendor callbacks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7448) * for kvm_skip_emulated_instruction(). The caller is responsible for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7449) * updating interruptibility state and injecting single-step #DBs.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7450) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7451) if (emulation_type & EMULTYPE_SKIP) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7452) kvm_rip_write(vcpu, ctxt->_eip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7453) if (ctxt->eflags & X86_EFLAGS_RF)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7454) kvm_set_rflags(vcpu, ctxt->eflags & ~X86_EFLAGS_RF);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7455) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7456) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7457)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7458) if (retry_instruction(ctxt, cr2_or_gpa, emulation_type))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7459) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7460)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7461) /* this is needed for vmware backdoor interface to work since it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7462) changes registers values during IO operation */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7463) if (vcpu->arch.emulate_regs_need_sync_from_vcpu) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7464) vcpu->arch.emulate_regs_need_sync_from_vcpu = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7465) emulator_invalidate_register_cache(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7466) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7467)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7468) restart:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7469) if (emulation_type & EMULTYPE_PF) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7470) /* Save the faulting GPA (cr2) in the address field */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7471) ctxt->exception.address = cr2_or_gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7472)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7473) /* With shadow page tables, cr2 contains a GVA or nGPA. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7474) if (vcpu->arch.mmu->direct_map) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7475) ctxt->gpa_available = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7476) ctxt->gpa_val = cr2_or_gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7477) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7478) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7479) /* Sanitize the address out of an abundance of paranoia. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7480) ctxt->exception.address = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7481) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7482)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7483) r = x86_emulate_insn(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7484)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7485) if (r == EMULATION_INTERCEPTED)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7486) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7487)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7488) if (r == EMULATION_FAILED) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7489) if (reexecute_instruction(vcpu, cr2_or_gpa, write_fault_to_spt,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7490) emulation_type))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7491) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7492)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7493) return handle_emulation_failure(vcpu, emulation_type);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7494) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7495)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7496) if (ctxt->have_exception) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7497) r = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7498) if (inject_emulated_exception(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7499) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7500) } else if (vcpu->arch.pio.count) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7501) if (!vcpu->arch.pio.in) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7502) /* FIXME: return into emulator if single-stepping. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7503) vcpu->arch.pio.count = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7504) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7505) writeback = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7506) vcpu->arch.complete_userspace_io = complete_emulated_pio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7507) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7508) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7509) } else if (vcpu->mmio_needed) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7510) ++vcpu->stat.mmio_exits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7511)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7512) if (!vcpu->mmio_is_write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7513) writeback = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7514) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7515) vcpu->arch.complete_userspace_io = complete_emulated_mmio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7516) } else if (r == EMULATION_RESTART)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7517) goto restart;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7518) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7519) r = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7520)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7521) if (writeback) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7522) unsigned long rflags = kvm_x86_ops.get_rflags(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7523) toggle_interruptibility(vcpu, ctxt->interruptibility);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7524) vcpu->arch.emulate_regs_need_sync_to_vcpu = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7525) if (!ctxt->have_exception ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7526) exception_type(ctxt->exception.vector) == EXCPT_TRAP) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7527) kvm_rip_write(vcpu, ctxt->eip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7528) if (r && (ctxt->tf || (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7529) r = kvm_vcpu_do_singlestep(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7530) if (kvm_x86_ops.update_emulated_instruction)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7531) kvm_x86_ops.update_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7532) __kvm_set_rflags(vcpu, ctxt->eflags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7533) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7534)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7535) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7536) * For STI, interrupts are shadowed; so KVM_REQ_EVENT will
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7537) * do nothing, and it will be requested again as soon as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7538) * the shadow expires. But we still need to check here,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7539) * because POPF has no interrupt shadow.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7540) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7541) if (unlikely((ctxt->eflags & ~rflags) & X86_EFLAGS_IF))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7542) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7543) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7544) vcpu->arch.emulate_regs_need_sync_to_vcpu = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7545)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7546) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7547) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7548)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7549) int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7550) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7551) return x86_emulate_instruction(vcpu, 0, emulation_type, NULL, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7552) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7553) EXPORT_SYMBOL_GPL(kvm_emulate_instruction);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7554)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7555) int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7556) void *insn, int insn_len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7557) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7558) return x86_emulate_instruction(vcpu, 0, 0, insn, insn_len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7559) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7560) EXPORT_SYMBOL_GPL(kvm_emulate_instruction_from_buffer);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7561)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7562) static int complete_fast_pio_out_port_0x7e(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7563) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7564) vcpu->arch.pio.count = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7565) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7566) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7567)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7568) static int complete_fast_pio_out(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7569) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7570) vcpu->arch.pio.count = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7571)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7572) if (unlikely(!kvm_is_linear_rip(vcpu, vcpu->arch.pio.linear_rip)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7573) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7574)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7575) return kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7576) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7577)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7578) static int kvm_fast_pio_out(struct kvm_vcpu *vcpu, int size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7579) unsigned short port)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7580) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7581) unsigned long val = kvm_rax_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7582) int ret = emulator_pio_out(vcpu, size, port, &val, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7583)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7584) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7585) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7586)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7587) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7588) * Workaround userspace that relies on old KVM behavior of %rip being
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7589) * incremented prior to exiting to userspace to handle "OUT 0x7e".
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7590) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7591) if (port == 0x7e &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7592) kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_OUT_7E_INC_RIP)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7593) vcpu->arch.complete_userspace_io =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7594) complete_fast_pio_out_port_0x7e;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7595) kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7596) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7597) vcpu->arch.pio.linear_rip = kvm_get_linear_rip(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7598) vcpu->arch.complete_userspace_io = complete_fast_pio_out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7599) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7600) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7601) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7602)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7603) static int complete_fast_pio_in(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7604) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7605) unsigned long val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7606)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7607) /* We should only ever be called with arch.pio.count equal to 1 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7608) BUG_ON(vcpu->arch.pio.count != 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7609)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7610) if (unlikely(!kvm_is_linear_rip(vcpu, vcpu->arch.pio.linear_rip))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7611) vcpu->arch.pio.count = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7612) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7613) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7614)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7615) /* For size less than 4 we merge, else we zero extend */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7616) val = (vcpu->arch.pio.size < 4) ? kvm_rax_read(vcpu) : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7617)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7618) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7619) * Since vcpu->arch.pio.count == 1 let emulator_pio_in perform
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7620) * the copy and tracing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7621) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7622) emulator_pio_in(vcpu, vcpu->arch.pio.size, vcpu->arch.pio.port, &val, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7623) kvm_rax_write(vcpu, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7624)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7625) return kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7626) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7627)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7628) static int kvm_fast_pio_in(struct kvm_vcpu *vcpu, int size,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7629) unsigned short port)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7630) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7631) unsigned long val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7632) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7633)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7634) /* For size less than 4 we merge, else we zero extend */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7635) val = (size < 4) ? kvm_rax_read(vcpu) : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7636)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7637) ret = emulator_pio_in(vcpu, size, port, &val, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7638) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7639) kvm_rax_write(vcpu, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7640) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7641) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7642)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7643) vcpu->arch.pio.linear_rip = kvm_get_linear_rip(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7644) vcpu->arch.complete_userspace_io = complete_fast_pio_in;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7645)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7646) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7647) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7648)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7649) int kvm_fast_pio(struct kvm_vcpu *vcpu, int size, unsigned short port, int in)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7650) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7651) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7652)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7653) if (in)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7654) ret = kvm_fast_pio_in(vcpu, size, port);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7655) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7656) ret = kvm_fast_pio_out(vcpu, size, port);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7657) return ret && kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7658) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7659) EXPORT_SYMBOL_GPL(kvm_fast_pio);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7660)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7661) static int kvmclock_cpu_down_prep(unsigned int cpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7662) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7663) __this_cpu_write(cpu_tsc_khz, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7664) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7665) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7666)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7667) static void tsc_khz_changed(void *data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7668) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7669) struct cpufreq_freqs *freq = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7670) unsigned long khz = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7671)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7672) if (data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7673) khz = freq->new;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7674) else if (!boot_cpu_has(X86_FEATURE_CONSTANT_TSC))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7675) khz = cpufreq_quick_get(raw_smp_processor_id());
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7676) if (!khz)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7677) khz = tsc_khz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7678) __this_cpu_write(cpu_tsc_khz, khz);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7679) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7680)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7681) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7682) static void kvm_hyperv_tsc_notifier(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7683) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7684) struct kvm *kvm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7685) struct kvm_vcpu *vcpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7686) int cpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7687)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7688) mutex_lock(&kvm_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7689) list_for_each_entry(kvm, &vm_list, vm_list)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7690) kvm_make_mclock_inprogress_request(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7691)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7692) hyperv_stop_tsc_emulation();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7693)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7694) /* TSC frequency always matches when on Hyper-V */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7695) for_each_present_cpu(cpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7696) per_cpu(cpu_tsc_khz, cpu) = tsc_khz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7697) kvm_max_guest_tsc_khz = tsc_khz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7698)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7699) list_for_each_entry(kvm, &vm_list, vm_list) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7700) struct kvm_arch *ka = &kvm->arch;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7701)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7702) spin_lock(&ka->pvclock_gtod_sync_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7703)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7704) pvclock_update_vm_gtod_copy(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7705)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7706) kvm_for_each_vcpu(cpu, vcpu, kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7707) kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7708)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7709) kvm_for_each_vcpu(cpu, vcpu, kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7710) kvm_clear_request(KVM_REQ_MCLOCK_INPROGRESS, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7711)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7712) spin_unlock(&ka->pvclock_gtod_sync_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7713) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7714) mutex_unlock(&kvm_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7715) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7716) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7717)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7718) static void __kvmclock_cpufreq_notifier(struct cpufreq_freqs *freq, int cpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7719) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7720) struct kvm *kvm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7721) struct kvm_vcpu *vcpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7722) int i, send_ipi = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7723)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7724) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7725) * We allow guests to temporarily run on slowing clocks,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7726) * provided we notify them after, or to run on accelerating
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7727) * clocks, provided we notify them before. Thus time never
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7728) * goes backwards.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7729) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7730) * However, we have a problem. We can't atomically update
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7731) * the frequency of a given CPU from this function; it is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7732) * merely a notifier, which can be called from any CPU.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7733) * Changing the TSC frequency at arbitrary points in time
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7734) * requires a recomputation of local variables related to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7735) * the TSC for each VCPU. We must flag these local variables
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7736) * to be updated and be sure the update takes place with the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7737) * new frequency before any guests proceed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7738) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7739) * Unfortunately, the combination of hotplug CPU and frequency
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7740) * change creates an intractable locking scenario; the order
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7741) * of when these callouts happen is undefined with respect to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7742) * CPU hotplug, and they can race with each other. As such,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7743) * merely setting per_cpu(cpu_tsc_khz) = X during a hotadd is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7744) * undefined; you can actually have a CPU frequency change take
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7745) * place in between the computation of X and the setting of the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7746) * variable. To protect against this problem, all updates of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7747) * the per_cpu tsc_khz variable are done in an interrupt
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7748) * protected IPI, and all callers wishing to update the value
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7749) * must wait for a synchronous IPI to complete (which is trivial
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7750) * if the caller is on the CPU already). This establishes the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7751) * necessary total order on variable updates.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7752) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7753) * Note that because a guest time update may take place
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7754) * anytime after the setting of the VCPU's request bit, the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7755) * correct TSC value must be set before the request. However,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7756) * to ensure the update actually makes it to any guest which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7757) * starts running in hardware virtualization between the set
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7758) * and the acquisition of the spinlock, we must also ping the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7759) * CPU after setting the request bit.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7760) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7761) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7762)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7763) smp_call_function_single(cpu, tsc_khz_changed, freq, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7764)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7765) mutex_lock(&kvm_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7766) list_for_each_entry(kvm, &vm_list, vm_list) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7767) kvm_for_each_vcpu(i, vcpu, kvm) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7768) if (vcpu->cpu != cpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7769) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7770) kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7771) if (vcpu->cpu != raw_smp_processor_id())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7772) send_ipi = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7773) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7774) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7775) mutex_unlock(&kvm_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7776)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7777) if (freq->old < freq->new && send_ipi) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7778) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7779) * We upscale the frequency. Must make the guest
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7780) * doesn't see old kvmclock values while running with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7781) * the new frequency, otherwise we risk the guest sees
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7782) * time go backwards.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7783) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7784) * In case we update the frequency for another cpu
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7785) * (which might be in guest context) send an interrupt
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7786) * to kick the cpu out of guest context. Next time
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7787) * guest context is entered kvmclock will be updated,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7788) * so the guest will not see stale values.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7789) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7790) smp_call_function_single(cpu, tsc_khz_changed, freq, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7791) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7792) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7793)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7794) static int kvmclock_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7795) void *data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7796) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7797) struct cpufreq_freqs *freq = data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7798) int cpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7799)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7800) if (val == CPUFREQ_PRECHANGE && freq->old > freq->new)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7801) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7802) if (val == CPUFREQ_POSTCHANGE && freq->old < freq->new)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7803) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7804)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7805) for_each_cpu(cpu, freq->policy->cpus)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7806) __kvmclock_cpufreq_notifier(freq, cpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7807)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7808) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7809) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7810)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7811) static struct notifier_block kvmclock_cpufreq_notifier_block = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7812) .notifier_call = kvmclock_cpufreq_notifier
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7813) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7814)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7815) static int kvmclock_cpu_online(unsigned int cpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7816) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7817) tsc_khz_changed(NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7818) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7819) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7820)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7821) static void kvm_timer_init(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7822) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7823) max_tsc_khz = tsc_khz;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7824)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7825) if (!boot_cpu_has(X86_FEATURE_CONSTANT_TSC)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7826) #ifdef CONFIG_CPU_FREQ
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7827) struct cpufreq_policy *policy;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7828) int cpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7829)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7830) cpu = get_cpu();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7831) policy = cpufreq_cpu_get(cpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7832) if (policy) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7833) if (policy->cpuinfo.max_freq)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7834) max_tsc_khz = policy->cpuinfo.max_freq;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7835) cpufreq_cpu_put(policy);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7836) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7837) put_cpu();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7838) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7839) cpufreq_register_notifier(&kvmclock_cpufreq_notifier_block,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7840) CPUFREQ_TRANSITION_NOTIFIER);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7841) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7842)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7843) cpuhp_setup_state(CPUHP_AP_X86_KVM_CLK_ONLINE, "x86/kvm/clk:online",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7844) kvmclock_cpu_online, kvmclock_cpu_down_prep);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7845) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7846)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7847) DEFINE_PER_CPU(struct kvm_vcpu *, current_vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7848) EXPORT_PER_CPU_SYMBOL_GPL(current_vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7849)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7850) int kvm_is_in_guest(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7851) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7852) return __this_cpu_read(current_vcpu) != NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7853) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7854)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7855) static int kvm_is_user_mode(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7856) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7857) int user_mode = 3;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7858)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7859) if (__this_cpu_read(current_vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7860) user_mode = kvm_x86_ops.get_cpl(__this_cpu_read(current_vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7861)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7862) return user_mode != 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7863) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7864)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7865) static unsigned long kvm_get_guest_ip(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7866) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7867) unsigned long ip = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7868)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7869) if (__this_cpu_read(current_vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7870) ip = kvm_rip_read(__this_cpu_read(current_vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7871)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7872) return ip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7873) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7874)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7875) static void kvm_handle_intel_pt_intr(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7876) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7877) struct kvm_vcpu *vcpu = __this_cpu_read(current_vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7878)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7879) kvm_make_request(KVM_REQ_PMI, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7880) __set_bit(MSR_CORE_PERF_GLOBAL_OVF_CTRL_TRACE_TOPA_PMI_BIT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7881) (unsigned long *)&vcpu->arch.pmu.global_status);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7882) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7883)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7884) static struct perf_guest_info_callbacks kvm_guest_cbs = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7885) .is_in_guest = kvm_is_in_guest,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7886) .is_user_mode = kvm_is_user_mode,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7887) .get_guest_ip = kvm_get_guest_ip,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7888) .handle_intel_pt_intr = NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7889) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7890)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7891) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7892) static void pvclock_gtod_update_fn(struct work_struct *work)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7893) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7894) struct kvm *kvm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7895)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7896) struct kvm_vcpu *vcpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7897) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7898)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7899) mutex_lock(&kvm_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7900) list_for_each_entry(kvm, &vm_list, vm_list)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7901) kvm_for_each_vcpu(i, vcpu, kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7902) kvm_make_request(KVM_REQ_MASTERCLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7903) atomic_set(&kvm_guest_has_master_clock, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7904) mutex_unlock(&kvm_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7905) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7906)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7907) static DECLARE_WORK(pvclock_gtod_work, pvclock_gtod_update_fn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7908)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7909) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7910) * Indirection to move queue_work() out of the tk_core.seq write held
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7911) * region to prevent possible deadlocks against time accessors which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7912) * are invoked with work related locks held.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7913) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7914) static void pvclock_irq_work_fn(struct irq_work *w)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7915) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7916) queue_work(system_long_wq, &pvclock_gtod_work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7917) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7918)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7919) static DEFINE_IRQ_WORK(pvclock_irq_work, pvclock_irq_work_fn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7920)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7921) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7922) * Notification about pvclock gtod data update.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7923) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7924) static int pvclock_gtod_notify(struct notifier_block *nb, unsigned long unused,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7925) void *priv)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7926) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7927) struct pvclock_gtod_data *gtod = &pvclock_gtod_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7928) struct timekeeper *tk = priv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7929)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7930) update_pvclock_gtod(tk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7931)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7932) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7933) * Disable master clock if host does not trust, or does not use,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7934) * TSC based clocksource. Delegate queue_work() to irq_work as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7935) * this is invoked with tk_core.seq write held.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7936) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7937) if (!gtod_is_based_on_tsc(gtod->clock.vclock_mode) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7938) atomic_read(&kvm_guest_has_master_clock) != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7939) irq_work_queue(&pvclock_irq_work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7940) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7941) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7942)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7943) static struct notifier_block pvclock_gtod_notifier = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7944) .notifier_call = pvclock_gtod_notify,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7945) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7946) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7947)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7948) int kvm_arch_init(void *opaque)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7949) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7950) struct kvm_x86_init_ops *ops = opaque;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7951) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7952)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7953) if (kvm_x86_ops.hardware_enable) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7954) printk(KERN_ERR "kvm: already loaded the other module\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7955) r = -EEXIST;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7956) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7957) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7958)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7959) if (!ops->cpu_has_kvm_support()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7960) pr_err_ratelimited("kvm: no hardware support\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7961) r = -EOPNOTSUPP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7962) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7963) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7964) if (ops->disabled_by_bios()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7965) pr_err_ratelimited("kvm: disabled by bios\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7966) r = -EOPNOTSUPP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7967) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7968) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7969)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7970) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7971) * KVM explicitly assumes that the guest has an FPU and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7972) * FXSAVE/FXRSTOR. For example, the KVM_GET_FPU explicitly casts the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7973) * vCPU's FPU state as a fxregs_state struct.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7974) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7975) if (!boot_cpu_has(X86_FEATURE_FPU) || !boot_cpu_has(X86_FEATURE_FXSR)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7976) printk(KERN_ERR "kvm: inadequate fpu\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7977) r = -EOPNOTSUPP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7978) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7979) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7980)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7981) r = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7982) x86_fpu_cache = kmem_cache_create("x86_fpu", sizeof(struct fpu),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7983) __alignof__(struct fpu), SLAB_ACCOUNT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7984) NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7985) if (!x86_fpu_cache) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7986) printk(KERN_ERR "kvm: failed to allocate cache for x86 fpu\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7987) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7988) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7989)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7990) x86_emulator_cache = kvm_alloc_emulator_cache();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7991) if (!x86_emulator_cache) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7992) pr_err("kvm: failed to allocate cache for x86 emulator\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7993) goto out_free_x86_fpu_cache;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7994) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7995)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7996) user_return_msrs = alloc_percpu(struct kvm_user_return_msrs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7997) if (!user_return_msrs) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7998) printk(KERN_ERR "kvm: failed to allocate percpu kvm_user_return_msrs\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7999) goto out_free_x86_emulator_cache;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8000) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8001)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8002) r = kvm_mmu_module_init();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8003) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8004) goto out_free_percpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8005)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8006) kvm_mmu_set_mask_ptes(PT_USER_MASK, PT_ACCESSED_MASK,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8007) PT_DIRTY_MASK, PT64_NX_MASK, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8008) PT_PRESENT_MASK, 0, sme_me_mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8009) kvm_timer_init();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8010)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8011) if (ops->intel_pt_intr_in_guest && ops->intel_pt_intr_in_guest())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8012) kvm_guest_cbs.handle_intel_pt_intr = kvm_handle_intel_pt_intr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8013) perf_register_guest_info_callbacks(&kvm_guest_cbs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8014)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8015) if (boot_cpu_has(X86_FEATURE_XSAVE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8016) host_xcr0 = xgetbv(XCR_XFEATURE_ENABLED_MASK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8017) supported_xcr0 = host_xcr0 & KVM_SUPPORTED_XCR0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8018) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8019)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8020) kvm_lapic_init();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8021) if (pi_inject_timer == -1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8022) pi_inject_timer = housekeeping_enabled(HK_FLAG_TIMER);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8023) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8024) pvclock_gtod_register_notifier(&pvclock_gtod_notifier);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8025)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8026) if (hypervisor_is_type(X86_HYPER_MS_HYPERV))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8027) set_hv_tscchange_cb(kvm_hyperv_tsc_notifier);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8028) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8029)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8030) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8031)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8032) out_free_percpu:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8033) free_percpu(user_return_msrs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8034) out_free_x86_emulator_cache:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8035) kmem_cache_destroy(x86_emulator_cache);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8036) out_free_x86_fpu_cache:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8037) kmem_cache_destroy(x86_fpu_cache);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8038) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8039) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8040) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8041)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8042) void kvm_arch_exit(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8043) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8044) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8045) if (hypervisor_is_type(X86_HYPER_MS_HYPERV))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8046) clear_hv_tscchange_cb();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8047) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8048) kvm_lapic_exit();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8049) perf_unregister_guest_info_callbacks(&kvm_guest_cbs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8050) kvm_guest_cbs.handle_intel_pt_intr = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8051)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8052) if (!boot_cpu_has(X86_FEATURE_CONSTANT_TSC))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8053) cpufreq_unregister_notifier(&kvmclock_cpufreq_notifier_block,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8054) CPUFREQ_TRANSITION_NOTIFIER);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8055) cpuhp_remove_state_nocalls(CPUHP_AP_X86_KVM_CLK_ONLINE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8056) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8057) pvclock_gtod_unregister_notifier(&pvclock_gtod_notifier);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8058) irq_work_sync(&pvclock_irq_work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8059) cancel_work_sync(&pvclock_gtod_work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8060) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8061) kvm_x86_ops.hardware_enable = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8062) kvm_mmu_module_exit();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8063) free_percpu(user_return_msrs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8064) kmem_cache_destroy(x86_emulator_cache);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8065) kmem_cache_destroy(x86_fpu_cache);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8066) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8067)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8068) int kvm_vcpu_halt(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8069) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8070) ++vcpu->stat.halt_exits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8071) if (lapic_in_kernel(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8072) vcpu->arch.mp_state = KVM_MP_STATE_HALTED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8073) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8074) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8075) vcpu->run->exit_reason = KVM_EXIT_HLT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8076) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8077) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8078) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8079) EXPORT_SYMBOL_GPL(kvm_vcpu_halt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8080)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8081) int kvm_emulate_halt(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8082) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8083) int ret = kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8084) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8085) * TODO: we might be squashing a GUESTDBG_SINGLESTEP-triggered
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8086) * KVM_EXIT_DEBUG here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8087) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8088) return kvm_vcpu_halt(vcpu) && ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8089) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8090) EXPORT_SYMBOL_GPL(kvm_emulate_halt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8091)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8092) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8093) static int kvm_pv_clock_pairing(struct kvm_vcpu *vcpu, gpa_t paddr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8094) unsigned long clock_type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8095) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8096) struct kvm_clock_pairing clock_pairing;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8097) struct timespec64 ts;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8098) u64 cycle;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8099) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8100)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8101) if (clock_type != KVM_CLOCK_PAIRING_WALLCLOCK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8102) return -KVM_EOPNOTSUPP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8103)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8104) if (kvm_get_walltime_and_clockread(&ts, &cycle) == false)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8105) return -KVM_EOPNOTSUPP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8106)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8107) clock_pairing.sec = ts.tv_sec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8108) clock_pairing.nsec = ts.tv_nsec;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8109) clock_pairing.tsc = kvm_read_l1_tsc(vcpu, cycle);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8110) clock_pairing.flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8111) memset(&clock_pairing.pad, 0, sizeof(clock_pairing.pad));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8112)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8113) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8114) if (kvm_write_guest(vcpu->kvm, paddr, &clock_pairing,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8115) sizeof(struct kvm_clock_pairing)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8116) ret = -KVM_EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8117)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8118) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8119) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8120) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8121)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8122) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8123) * kvm_pv_kick_cpu_op: Kick a vcpu.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8124) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8125) * @apicid - apicid of vcpu to be kicked.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8126) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8127) static void kvm_pv_kick_cpu_op(struct kvm *kvm, unsigned long flags, int apicid)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8128) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8129) struct kvm_lapic_irq lapic_irq;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8130)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8131) lapic_irq.shorthand = APIC_DEST_NOSHORT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8132) lapic_irq.dest_mode = APIC_DEST_PHYSICAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8133) lapic_irq.level = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8134) lapic_irq.dest_id = apicid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8135) lapic_irq.msi_redir_hint = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8136)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8137) lapic_irq.delivery_mode = APIC_DM_REMRD;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8138) kvm_irq_delivery_to_apic(kvm, NULL, &lapic_irq, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8139) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8140)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8141) bool kvm_apicv_activated(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8142) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8143) return (READ_ONCE(kvm->arch.apicv_inhibit_reasons) == 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8144) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8145) EXPORT_SYMBOL_GPL(kvm_apicv_activated);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8146)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8147) void kvm_apicv_init(struct kvm *kvm, bool enable)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8148) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8149) if (enable)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8150) clear_bit(APICV_INHIBIT_REASON_DISABLE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8151) &kvm->arch.apicv_inhibit_reasons);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8152) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8153) set_bit(APICV_INHIBIT_REASON_DISABLE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8154) &kvm->arch.apicv_inhibit_reasons);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8155) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8156) EXPORT_SYMBOL_GPL(kvm_apicv_init);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8157)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8158) static void kvm_sched_yield(struct kvm *kvm, unsigned long dest_id)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8159) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8160) struct kvm_vcpu *target = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8161) struct kvm_apic_map *map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8162)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8163) rcu_read_lock();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8164) map = rcu_dereference(kvm->arch.apic_map);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8165)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8166) if (likely(map) && dest_id <= map->max_apic_id && map->phys_map[dest_id])
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8167) target = map->phys_map[dest_id]->vcpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8168)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8169) rcu_read_unlock();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8170)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8171) if (target && READ_ONCE(target->ready))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8172) kvm_vcpu_yield_to(target);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8173) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8174)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8175) int kvm_emulate_hypercall(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8176) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8177) unsigned long nr, a0, a1, a2, a3, ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8178) int op_64_bit;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8179)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8180) if (kvm_hv_hypercall_enabled(vcpu->kvm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8181) return kvm_hv_hypercall(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8182)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8183) nr = kvm_rax_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8184) a0 = kvm_rbx_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8185) a1 = kvm_rcx_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8186) a2 = kvm_rdx_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8187) a3 = kvm_rsi_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8188)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8189) trace_kvm_hypercall(nr, a0, a1, a2, a3);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8190)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8191) op_64_bit = is_64_bit_mode(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8192) if (!op_64_bit) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8193) nr &= 0xFFFFFFFF;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8194) a0 &= 0xFFFFFFFF;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8195) a1 &= 0xFFFFFFFF;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8196) a2 &= 0xFFFFFFFF;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8197) a3 &= 0xFFFFFFFF;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8198) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8199)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8200) if (kvm_x86_ops.get_cpl(vcpu) != 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8201) ret = -KVM_EPERM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8202) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8203) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8204)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8205) ret = -KVM_ENOSYS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8206)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8207) switch (nr) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8208) case KVM_HC_VAPIC_POLL_IRQ:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8209) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8210) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8211) case KVM_HC_KICK_CPU:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8212) if (!guest_pv_has(vcpu, KVM_FEATURE_PV_UNHALT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8213) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8214)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8215) kvm_pv_kick_cpu_op(vcpu->kvm, a0, a1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8216) kvm_sched_yield(vcpu->kvm, a1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8217) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8218) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8219) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8220) case KVM_HC_CLOCK_PAIRING:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8221) ret = kvm_pv_clock_pairing(vcpu, a0, a1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8222) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8223) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8224) case KVM_HC_SEND_IPI:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8225) if (!guest_pv_has(vcpu, KVM_FEATURE_PV_SEND_IPI))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8226) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8227)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8228) ret = kvm_pv_send_ipi(vcpu->kvm, a0, a1, a2, a3, op_64_bit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8229) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8230) case KVM_HC_SCHED_YIELD:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8231) if (!guest_pv_has(vcpu, KVM_FEATURE_PV_SCHED_YIELD))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8232) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8233)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8234) kvm_sched_yield(vcpu->kvm, a0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8235) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8236) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8237) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8238) ret = -KVM_ENOSYS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8239) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8240) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8241) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8242) if (!op_64_bit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8243) ret = (u32)ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8244) kvm_rax_write(vcpu, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8245)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8246) ++vcpu->stat.hypercalls;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8247) return kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8248) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8249) EXPORT_SYMBOL_GPL(kvm_emulate_hypercall);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8250)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8251) static int emulator_fix_hypercall(struct x86_emulate_ctxt *ctxt)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8252) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8253) struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8254) char instruction[3];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8255) unsigned long rip = kvm_rip_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8256)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8257) kvm_x86_ops.patch_hypercall(vcpu, instruction);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8258)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8259) return emulator_write_emulated(ctxt, rip, instruction, 3,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8260) &ctxt->exception);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8261) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8262)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8263) static int dm_request_for_irq_injection(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8264) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8265) return vcpu->run->request_interrupt_window &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8266) likely(!pic_in_kernel(vcpu->kvm));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8267) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8268)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8269) static void post_kvm_run_save(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8270) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8271) struct kvm_run *kvm_run = vcpu->run;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8272)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8273) kvm_run->if_flag = (kvm_get_rflags(vcpu) & X86_EFLAGS_IF) != 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8274) kvm_run->flags = is_smm(vcpu) ? KVM_RUN_X86_SMM : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8275) kvm_run->cr8 = kvm_get_cr8(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8276) kvm_run->apic_base = kvm_get_apic_base(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8277) kvm_run->ready_for_interrupt_injection =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8278) pic_in_kernel(vcpu->kvm) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8279) kvm_vcpu_ready_for_interrupt_injection(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8280) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8281)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8282) static void update_cr8_intercept(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8283) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8284) int max_irr, tpr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8285)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8286) if (!kvm_x86_ops.update_cr8_intercept)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8287) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8288)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8289) if (!lapic_in_kernel(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8290) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8291)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8292) if (vcpu->arch.apicv_active)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8293) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8294)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8295) if (!vcpu->arch.apic->vapic_addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8296) max_irr = kvm_lapic_find_highest_irr(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8297) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8298) max_irr = -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8299)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8300) if (max_irr != -1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8301) max_irr >>= 4;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8302)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8303) tpr = kvm_lapic_get_cr8(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8304)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8305) kvm_x86_ops.update_cr8_intercept(vcpu, tpr, max_irr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8306) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8307)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8308) static void kvm_inject_exception(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8309) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8310) if (vcpu->arch.exception.error_code && !is_protmode(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8311) vcpu->arch.exception.error_code = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8312) kvm_x86_ops.queue_exception(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8313) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8314)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8315) static void inject_pending_event(struct kvm_vcpu *vcpu, bool *req_immediate_exit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8316) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8317) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8318) bool can_inject = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8319)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8320) /* try to reinject previous events if any */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8321)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8322) if (vcpu->arch.exception.injected) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8323) kvm_inject_exception(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8324) can_inject = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8325) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8326) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8327) * Do not inject an NMI or interrupt if there is a pending
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8328) * exception. Exceptions and interrupts are recognized at
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8329) * instruction boundaries, i.e. the start of an instruction.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8330) * Trap-like exceptions, e.g. #DB, have higher priority than
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8331) * NMIs and interrupts, i.e. traps are recognized before an
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8332) * NMI/interrupt that's pending on the same instruction.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8333) * Fault-like exceptions, e.g. #GP and #PF, are the lowest
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8334) * priority, but are only generated (pended) during instruction
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8335) * execution, i.e. a pending fault-like exception means the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8336) * fault occurred on the *previous* instruction and must be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8337) * serviced prior to recognizing any new events in order to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8338) * fully complete the previous instruction.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8339) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8340) else if (!vcpu->arch.exception.pending) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8341) if (vcpu->arch.nmi_injected) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8342) kvm_x86_ops.set_nmi(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8343) can_inject = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8344) } else if (vcpu->arch.interrupt.injected) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8345) kvm_x86_ops.set_irq(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8346) can_inject = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8347) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8348) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8349)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8350) WARN_ON_ONCE(vcpu->arch.exception.injected &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8351) vcpu->arch.exception.pending);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8352)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8353) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8354) * Call check_nested_events() even if we reinjected a previous event
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8355) * in order for caller to determine if it should require immediate-exit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8356) * from L2 to L1 due to pending L1 events which require exit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8357) * from L2 to L1.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8358) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8359) if (is_guest_mode(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8360) r = kvm_x86_ops.nested_ops->check_events(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8361) if (r < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8362) goto busy;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8363) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8364)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8365) /* try to inject new event if pending */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8366) if (vcpu->arch.exception.pending) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8367) trace_kvm_inj_exception(vcpu->arch.exception.nr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8368) vcpu->arch.exception.has_error_code,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8369) vcpu->arch.exception.error_code);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8370)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8371) vcpu->arch.exception.pending = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8372) vcpu->arch.exception.injected = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8373)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8374) if (exception_type(vcpu->arch.exception.nr) == EXCPT_FAULT)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8375) __kvm_set_rflags(vcpu, kvm_get_rflags(vcpu) |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8376) X86_EFLAGS_RF);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8377)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8378) if (vcpu->arch.exception.nr == DB_VECTOR) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8379) kvm_deliver_exception_payload(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8380) if (vcpu->arch.dr7 & DR7_GD) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8381) vcpu->arch.dr7 &= ~DR7_GD;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8382) kvm_update_dr7(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8383) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8384) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8385)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8386) kvm_inject_exception(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8387) can_inject = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8388) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8389)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8390) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8391) * Finally, inject interrupt events. If an event cannot be injected
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8392) * due to architectural conditions (e.g. IF=0) a window-open exit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8393) * will re-request KVM_REQ_EVENT. Sometimes however an event is pending
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8394) * and can architecturally be injected, but we cannot do it right now:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8395) * an interrupt could have arrived just now and we have to inject it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8396) * as a vmexit, or there could already an event in the queue, which is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8397) * indicated by can_inject. In that case we request an immediate exit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8398) * in order to make progress and get back here for another iteration.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8399) * The kvm_x86_ops hooks communicate this by returning -EBUSY.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8400) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8401) if (vcpu->arch.smi_pending) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8402) r = can_inject ? kvm_x86_ops.smi_allowed(vcpu, true) : -EBUSY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8403) if (r < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8404) goto busy;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8405) if (r) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8406) vcpu->arch.smi_pending = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8407) ++vcpu->arch.smi_count;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8408) enter_smm(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8409) can_inject = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8410) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8411) kvm_x86_ops.enable_smi_window(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8412) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8413)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8414) if (vcpu->arch.nmi_pending) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8415) r = can_inject ? kvm_x86_ops.nmi_allowed(vcpu, true) : -EBUSY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8416) if (r < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8417) goto busy;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8418) if (r) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8419) --vcpu->arch.nmi_pending;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8420) vcpu->arch.nmi_injected = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8421) kvm_x86_ops.set_nmi(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8422) can_inject = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8423) WARN_ON(kvm_x86_ops.nmi_allowed(vcpu, true) < 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8424) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8425) if (vcpu->arch.nmi_pending)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8426) kvm_x86_ops.enable_nmi_window(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8427) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8428)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8429) if (kvm_cpu_has_injectable_intr(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8430) r = can_inject ? kvm_x86_ops.interrupt_allowed(vcpu, true) : -EBUSY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8431) if (r < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8432) goto busy;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8433) if (r) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8434) kvm_queue_interrupt(vcpu, kvm_cpu_get_interrupt(vcpu), false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8435) kvm_x86_ops.set_irq(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8436) WARN_ON(kvm_x86_ops.interrupt_allowed(vcpu, true) < 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8437) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8438) if (kvm_cpu_has_injectable_intr(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8439) kvm_x86_ops.enable_irq_window(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8440) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8441)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8442) if (is_guest_mode(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8443) kvm_x86_ops.nested_ops->hv_timer_pending &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8444) kvm_x86_ops.nested_ops->hv_timer_pending(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8445) *req_immediate_exit = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8446)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8447) WARN_ON(vcpu->arch.exception.pending);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8448) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8449)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8450) busy:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8451) *req_immediate_exit = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8452) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8453) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8454)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8455) static void process_nmi(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8456) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8457) unsigned limit = 2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8458)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8459) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8460) * x86 is limited to one NMI running, and one NMI pending after it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8461) * If an NMI is already in progress, limit further NMIs to just one.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8462) * Otherwise, allow two (and we'll inject the first one immediately).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8463) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8464) if (kvm_x86_ops.get_nmi_mask(vcpu) || vcpu->arch.nmi_injected)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8465) limit = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8466)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8467) vcpu->arch.nmi_pending += atomic_xchg(&vcpu->arch.nmi_queued, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8468) vcpu->arch.nmi_pending = min(vcpu->arch.nmi_pending, limit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8469) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8470) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8471)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8472) static u32 enter_smm_get_segment_flags(struct kvm_segment *seg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8473) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8474) u32 flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8475) flags |= seg->g << 23;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8476) flags |= seg->db << 22;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8477) flags |= seg->l << 21;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8478) flags |= seg->avl << 20;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8479) flags |= seg->present << 15;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8480) flags |= seg->dpl << 13;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8481) flags |= seg->s << 12;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8482) flags |= seg->type << 8;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8483) return flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8484) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8485)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8486) static void enter_smm_save_seg_32(struct kvm_vcpu *vcpu, char *buf, int n)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8487) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8488) struct kvm_segment seg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8489) int offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8490)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8491) kvm_get_segment(vcpu, &seg, n);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8492) put_smstate(u32, buf, 0x7fa8 + n * 4, seg.selector);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8493)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8494) if (n < 3)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8495) offset = 0x7f84 + n * 12;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8496) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8497) offset = 0x7f2c + (n - 3) * 12;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8498)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8499) put_smstate(u32, buf, offset + 8, seg.base);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8500) put_smstate(u32, buf, offset + 4, seg.limit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8501) put_smstate(u32, buf, offset, enter_smm_get_segment_flags(&seg));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8502) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8503)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8504) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8505) static void enter_smm_save_seg_64(struct kvm_vcpu *vcpu, char *buf, int n)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8506) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8507) struct kvm_segment seg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8508) int offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8509) u16 flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8510)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8511) kvm_get_segment(vcpu, &seg, n);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8512) offset = 0x7e00 + n * 16;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8513)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8514) flags = enter_smm_get_segment_flags(&seg) >> 8;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8515) put_smstate(u16, buf, offset, seg.selector);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8516) put_smstate(u16, buf, offset + 2, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8517) put_smstate(u32, buf, offset + 4, seg.limit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8518) put_smstate(u64, buf, offset + 8, seg.base);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8519) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8520) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8521)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8522) static void enter_smm_save_state_32(struct kvm_vcpu *vcpu, char *buf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8523) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8524) struct desc_ptr dt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8525) struct kvm_segment seg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8526) unsigned long val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8527) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8528)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8529) put_smstate(u32, buf, 0x7ffc, kvm_read_cr0(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8530) put_smstate(u32, buf, 0x7ff8, kvm_read_cr3(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8531) put_smstate(u32, buf, 0x7ff4, kvm_get_rflags(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8532) put_smstate(u32, buf, 0x7ff0, kvm_rip_read(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8533)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8534) for (i = 0; i < 8; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8535) put_smstate(u32, buf, 0x7fd0 + i * 4, kvm_register_read(vcpu, i));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8536)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8537) kvm_get_dr(vcpu, 6, &val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8538) put_smstate(u32, buf, 0x7fcc, (u32)val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8539) kvm_get_dr(vcpu, 7, &val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8540) put_smstate(u32, buf, 0x7fc8, (u32)val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8541)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8542) kvm_get_segment(vcpu, &seg, VCPU_SREG_TR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8543) put_smstate(u32, buf, 0x7fc4, seg.selector);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8544) put_smstate(u32, buf, 0x7f64, seg.base);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8545) put_smstate(u32, buf, 0x7f60, seg.limit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8546) put_smstate(u32, buf, 0x7f5c, enter_smm_get_segment_flags(&seg));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8547)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8548) kvm_get_segment(vcpu, &seg, VCPU_SREG_LDTR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8549) put_smstate(u32, buf, 0x7fc0, seg.selector);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8550) put_smstate(u32, buf, 0x7f80, seg.base);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8551) put_smstate(u32, buf, 0x7f7c, seg.limit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8552) put_smstate(u32, buf, 0x7f78, enter_smm_get_segment_flags(&seg));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8553)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8554) kvm_x86_ops.get_gdt(vcpu, &dt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8555) put_smstate(u32, buf, 0x7f74, dt.address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8556) put_smstate(u32, buf, 0x7f70, dt.size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8557)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8558) kvm_x86_ops.get_idt(vcpu, &dt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8559) put_smstate(u32, buf, 0x7f58, dt.address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8560) put_smstate(u32, buf, 0x7f54, dt.size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8561)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8562) for (i = 0; i < 6; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8563) enter_smm_save_seg_32(vcpu, buf, i);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8564)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8565) put_smstate(u32, buf, 0x7f14, kvm_read_cr4(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8566)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8567) /* revision id */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8568) put_smstate(u32, buf, 0x7efc, 0x00020000);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8569) put_smstate(u32, buf, 0x7ef8, vcpu->arch.smbase);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8570) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8571)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8572) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8573) static void enter_smm_save_state_64(struct kvm_vcpu *vcpu, char *buf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8574) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8575) struct desc_ptr dt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8576) struct kvm_segment seg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8577) unsigned long val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8578) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8579)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8580) for (i = 0; i < 16; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8581) put_smstate(u64, buf, 0x7ff8 - i * 8, kvm_register_read(vcpu, i));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8582)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8583) put_smstate(u64, buf, 0x7f78, kvm_rip_read(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8584) put_smstate(u32, buf, 0x7f70, kvm_get_rflags(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8585)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8586) kvm_get_dr(vcpu, 6, &val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8587) put_smstate(u64, buf, 0x7f68, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8588) kvm_get_dr(vcpu, 7, &val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8589) put_smstate(u64, buf, 0x7f60, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8590)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8591) put_smstate(u64, buf, 0x7f58, kvm_read_cr0(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8592) put_smstate(u64, buf, 0x7f50, kvm_read_cr3(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8593) put_smstate(u64, buf, 0x7f48, kvm_read_cr4(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8594)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8595) put_smstate(u32, buf, 0x7f00, vcpu->arch.smbase);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8596)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8597) /* revision id */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8598) put_smstate(u32, buf, 0x7efc, 0x00020064);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8599)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8600) put_smstate(u64, buf, 0x7ed0, vcpu->arch.efer);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8601)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8602) kvm_get_segment(vcpu, &seg, VCPU_SREG_TR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8603) put_smstate(u16, buf, 0x7e90, seg.selector);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8604) put_smstate(u16, buf, 0x7e92, enter_smm_get_segment_flags(&seg) >> 8);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8605) put_smstate(u32, buf, 0x7e94, seg.limit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8606) put_smstate(u64, buf, 0x7e98, seg.base);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8607)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8608) kvm_x86_ops.get_idt(vcpu, &dt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8609) put_smstate(u32, buf, 0x7e84, dt.size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8610) put_smstate(u64, buf, 0x7e88, dt.address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8611)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8612) kvm_get_segment(vcpu, &seg, VCPU_SREG_LDTR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8613) put_smstate(u16, buf, 0x7e70, seg.selector);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8614) put_smstate(u16, buf, 0x7e72, enter_smm_get_segment_flags(&seg) >> 8);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8615) put_smstate(u32, buf, 0x7e74, seg.limit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8616) put_smstate(u64, buf, 0x7e78, seg.base);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8617)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8618) kvm_x86_ops.get_gdt(vcpu, &dt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8619) put_smstate(u32, buf, 0x7e64, dt.size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8620) put_smstate(u64, buf, 0x7e68, dt.address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8621)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8622) for (i = 0; i < 6; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8623) enter_smm_save_seg_64(vcpu, buf, i);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8624) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8625) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8626)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8627) static void enter_smm(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8628) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8629) struct kvm_segment cs, ds;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8630) struct desc_ptr dt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8631) char buf[512];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8632) u32 cr0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8633)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8634) trace_kvm_enter_smm(vcpu->vcpu_id, vcpu->arch.smbase, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8635) memset(buf, 0, 512);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8636) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8637) if (guest_cpuid_has(vcpu, X86_FEATURE_LM))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8638) enter_smm_save_state_64(vcpu, buf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8639) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8640) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8641) enter_smm_save_state_32(vcpu, buf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8642)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8643) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8644) * Give pre_enter_smm() a chance to make ISA-specific changes to the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8645) * vCPU state (e.g. leave guest mode) after we've saved the state into
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8646) * the SMM state-save area.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8647) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8648) kvm_x86_ops.pre_enter_smm(vcpu, buf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8649)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8650) vcpu->arch.hflags |= HF_SMM_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8651) kvm_vcpu_write_guest(vcpu, vcpu->arch.smbase + 0xfe00, buf, sizeof(buf));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8652)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8653) if (kvm_x86_ops.get_nmi_mask(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8654) vcpu->arch.hflags |= HF_SMM_INSIDE_NMI_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8655) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8656) kvm_x86_ops.set_nmi_mask(vcpu, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8657)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8658) kvm_set_rflags(vcpu, X86_EFLAGS_FIXED);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8659) kvm_rip_write(vcpu, 0x8000);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8660)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8661) cr0 = vcpu->arch.cr0 & ~(X86_CR0_PE | X86_CR0_EM | X86_CR0_TS | X86_CR0_PG);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8662) kvm_x86_ops.set_cr0(vcpu, cr0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8663) vcpu->arch.cr0 = cr0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8664)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8665) kvm_x86_ops.set_cr4(vcpu, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8666)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8667) /* Undocumented: IDT limit is set to zero on entry to SMM. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8668) dt.address = dt.size = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8669) kvm_x86_ops.set_idt(vcpu, &dt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8670)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8671) __kvm_set_dr(vcpu, 7, DR7_FIXED_1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8672)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8673) cs.selector = (vcpu->arch.smbase >> 4) & 0xffff;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8674) cs.base = vcpu->arch.smbase;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8675)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8676) ds.selector = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8677) ds.base = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8678)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8679) cs.limit = ds.limit = 0xffffffff;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8680) cs.type = ds.type = 0x3;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8681) cs.dpl = ds.dpl = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8682) cs.db = ds.db = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8683) cs.s = ds.s = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8684) cs.l = ds.l = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8685) cs.g = ds.g = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8686) cs.avl = ds.avl = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8687) cs.present = ds.present = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8688) cs.unusable = ds.unusable = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8689) cs.padding = ds.padding = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8690)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8691) kvm_set_segment(vcpu, &cs, VCPU_SREG_CS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8692) kvm_set_segment(vcpu, &ds, VCPU_SREG_DS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8693) kvm_set_segment(vcpu, &ds, VCPU_SREG_ES);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8694) kvm_set_segment(vcpu, &ds, VCPU_SREG_FS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8695) kvm_set_segment(vcpu, &ds, VCPU_SREG_GS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8696) kvm_set_segment(vcpu, &ds, VCPU_SREG_SS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8697)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8698) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8699) if (guest_cpuid_has(vcpu, X86_FEATURE_LM))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8700) kvm_x86_ops.set_efer(vcpu, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8701) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8702)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8703) kvm_update_cpuid_runtime(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8704) kvm_mmu_reset_context(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8705) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8706)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8707) static void process_smi(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8708) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8709) vcpu->arch.smi_pending = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8710) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8711) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8712)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8713) void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8714) unsigned long *vcpu_bitmap)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8715) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8716) cpumask_var_t cpus;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8717)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8718) zalloc_cpumask_var(&cpus, GFP_ATOMIC);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8719)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8720) kvm_make_vcpus_request_mask(kvm, KVM_REQ_SCAN_IOAPIC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8721) NULL, vcpu_bitmap, cpus);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8722)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8723) free_cpumask_var(cpus);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8724) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8725)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8726) void kvm_make_scan_ioapic_request(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8727) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8728) kvm_make_all_cpus_request(kvm, KVM_REQ_SCAN_IOAPIC);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8729) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8730)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8731) void kvm_vcpu_update_apicv(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8732) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8733) if (!lapic_in_kernel(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8734) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8735)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8736) vcpu->arch.apicv_active = kvm_apicv_activated(vcpu->kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8737) kvm_apic_update_apicv(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8738) kvm_x86_ops.refresh_apicv_exec_ctrl(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8739) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8740) EXPORT_SYMBOL_GPL(kvm_vcpu_update_apicv);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8741)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8742) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8743) * NOTE: Do not hold any lock prior to calling this.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8744) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8745) * In particular, kvm_request_apicv_update() expects kvm->srcu not to be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8746) * locked, because it calls __x86_set_memory_region() which does
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8747) * synchronize_srcu(&kvm->srcu).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8748) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8749) void kvm_request_apicv_update(struct kvm *kvm, bool activate, ulong bit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8750) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8751) struct kvm_vcpu *except;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8752) unsigned long old, new, expected;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8753)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8754) if (!kvm_x86_ops.check_apicv_inhibit_reasons ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8755) !kvm_x86_ops.check_apicv_inhibit_reasons(bit))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8756) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8757)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8758) old = READ_ONCE(kvm->arch.apicv_inhibit_reasons);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8759) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8760) expected = new = old;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8761) if (activate)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8762) __clear_bit(bit, &new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8763) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8764) __set_bit(bit, &new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8765) if (new == old)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8766) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8767) old = cmpxchg(&kvm->arch.apicv_inhibit_reasons, expected, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8768) } while (old != expected);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8769)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8770) if (!!old == !!new)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8771) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8772)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8773) trace_kvm_apicv_update_request(activate, bit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8774) if (kvm_x86_ops.pre_update_apicv_exec_ctrl)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8775) kvm_x86_ops.pre_update_apicv_exec_ctrl(kvm, activate);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8776)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8777) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8778) * Sending request to update APICV for all other vcpus,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8779) * while update the calling vcpu immediately instead of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8780) * waiting for another #VMEXIT to handle the request.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8781) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8782) except = kvm_get_running_vcpu();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8783) kvm_make_all_cpus_request_except(kvm, KVM_REQ_APICV_UPDATE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8784) except);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8785) if (except)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8786) kvm_vcpu_update_apicv(except);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8787) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8788) EXPORT_SYMBOL_GPL(kvm_request_apicv_update);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8789)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8790) static void vcpu_scan_ioapic(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8791) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8792) if (!kvm_apic_present(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8793) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8794)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8795) bitmap_zero(vcpu->arch.ioapic_handled_vectors, 256);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8796)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8797) if (irqchip_split(vcpu->kvm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8798) kvm_scan_ioapic_routes(vcpu, vcpu->arch.ioapic_handled_vectors);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8799) else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8800) if (vcpu->arch.apicv_active)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8801) kvm_x86_ops.sync_pir_to_irr(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8802) if (ioapic_in_kernel(vcpu->kvm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8803) kvm_ioapic_scan_entry(vcpu, vcpu->arch.ioapic_handled_vectors);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8804) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8805)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8806) if (is_guest_mode(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8807) vcpu->arch.load_eoi_exitmap_pending = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8808) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8809) kvm_make_request(KVM_REQ_LOAD_EOI_EXITMAP, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8810) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8811)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8812) static void vcpu_load_eoi_exitmap(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8813) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8814) u64 eoi_exit_bitmap[4];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8815)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8816) if (!kvm_apic_hw_enabled(vcpu->arch.apic))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8817) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8818)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8819) bitmap_or((ulong *)eoi_exit_bitmap, vcpu->arch.ioapic_handled_vectors,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8820) vcpu_to_synic(vcpu)->vec_bitmap, 256);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8821) kvm_x86_ops.load_eoi_exitmap(vcpu, eoi_exit_bitmap);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8822) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8823)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8824) void kvm_arch_mmu_notifier_invalidate_range(struct kvm *kvm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8825) unsigned long start, unsigned long end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8826) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8827) unsigned long apic_address;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8828)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8829) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8830) * The physical address of apic access page is stored in the VMCS.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8831) * Update it when it becomes invalid.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8832) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8833) apic_address = gfn_to_hva(kvm, APIC_DEFAULT_PHYS_BASE >> PAGE_SHIFT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8834) if (start <= apic_address && apic_address < end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8835) kvm_make_all_cpus_request(kvm, KVM_REQ_APIC_PAGE_RELOAD);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8836) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8837)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8838) void kvm_vcpu_reload_apic_access_page(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8839) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8840) if (!lapic_in_kernel(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8841) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8842)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8843) if (!kvm_x86_ops.set_apic_access_page_addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8844) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8845)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8846) kvm_x86_ops.set_apic_access_page_addr(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8847) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8848)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8849) void __kvm_request_immediate_exit(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8850) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8851) smp_send_reschedule(vcpu->cpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8852) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8853) EXPORT_SYMBOL_GPL(__kvm_request_immediate_exit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8854)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8855) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8856) * Returns 1 to let vcpu_run() continue the guest execution loop without
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8857) * exiting to the userspace. Otherwise, the value will be returned to the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8858) * userspace.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8859) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8860) static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8861) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8862) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8863) bool req_int_win =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8864) dm_request_for_irq_injection(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8865) kvm_cpu_accept_dm_intr(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8866) fastpath_t exit_fastpath;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8867)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8868) bool req_immediate_exit = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8869)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8870) if (kvm_request_pending(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8871) if (kvm_check_request(KVM_REQ_GET_NESTED_STATE_PAGES, vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8872) if (unlikely(!kvm_x86_ops.nested_ops->get_nested_state_pages(vcpu))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8873) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8874) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8875) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8876) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8877) if (kvm_check_request(KVM_REQ_MMU_RELOAD, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8878) kvm_mmu_unload(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8879) if (kvm_check_request(KVM_REQ_MIGRATE_TIMER, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8880) __kvm_migrate_timers(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8881) if (kvm_check_request(KVM_REQ_MASTERCLOCK_UPDATE, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8882) kvm_gen_update_masterclock(vcpu->kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8883) if (kvm_check_request(KVM_REQ_GLOBAL_CLOCK_UPDATE, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8884) kvm_gen_kvmclock_update(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8885) if (kvm_check_request(KVM_REQ_CLOCK_UPDATE, vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8886) r = kvm_guest_time_update(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8887) if (unlikely(r))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8888) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8889) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8890) if (kvm_check_request(KVM_REQ_MMU_SYNC, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8891) kvm_mmu_sync_roots(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8892) if (kvm_check_request(KVM_REQ_LOAD_MMU_PGD, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8893) kvm_mmu_load_pgd(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8894) if (kvm_check_request(KVM_REQ_TLB_FLUSH, vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8895) kvm_vcpu_flush_tlb_all(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8896)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8897) /* Flushing all ASIDs flushes the current ASID... */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8898) kvm_clear_request(KVM_REQ_TLB_FLUSH_CURRENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8899) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8900) if (kvm_check_request(KVM_REQ_TLB_FLUSH_CURRENT, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8901) kvm_vcpu_flush_tlb_current(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8902) if (kvm_check_request(KVM_REQ_TLB_FLUSH_GUEST, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8903) kvm_vcpu_flush_tlb_guest(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8904)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8905) if (kvm_check_request(KVM_REQ_REPORT_TPR_ACCESS, vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8906) vcpu->run->exit_reason = KVM_EXIT_TPR_ACCESS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8907) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8908) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8909) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8910) if (kvm_check_request(KVM_REQ_TRIPLE_FAULT, vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8911) vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8912) vcpu->mmio_needed = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8913) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8914) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8915) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8916) if (kvm_check_request(KVM_REQ_APF_HALT, vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8917) /* Page is swapped out. Do synthetic halt */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8918) vcpu->arch.apf.halted = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8919) r = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8920) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8921) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8922) if (kvm_check_request(KVM_REQ_STEAL_UPDATE, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8923) record_steal_time(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8924) if (kvm_check_request(KVM_REQ_SMI, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8925) process_smi(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8926) if (kvm_check_request(KVM_REQ_NMI, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8927) process_nmi(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8928) if (kvm_check_request(KVM_REQ_PMU, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8929) kvm_pmu_handle_event(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8930) if (kvm_check_request(KVM_REQ_PMI, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8931) kvm_pmu_deliver_pmi(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8932) if (kvm_check_request(KVM_REQ_IOAPIC_EOI_EXIT, vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8933) BUG_ON(vcpu->arch.pending_ioapic_eoi > 255);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8934) if (test_bit(vcpu->arch.pending_ioapic_eoi,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8935) vcpu->arch.ioapic_handled_vectors)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8936) vcpu->run->exit_reason = KVM_EXIT_IOAPIC_EOI;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8937) vcpu->run->eoi.vector =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8938) vcpu->arch.pending_ioapic_eoi;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8939) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8940) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8941) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8942) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8943) if (kvm_check_request(KVM_REQ_SCAN_IOAPIC, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8944) vcpu_scan_ioapic(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8945) if (kvm_check_request(KVM_REQ_LOAD_EOI_EXITMAP, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8946) vcpu_load_eoi_exitmap(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8947) if (kvm_check_request(KVM_REQ_APIC_PAGE_RELOAD, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8948) kvm_vcpu_reload_apic_access_page(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8949) if (kvm_check_request(KVM_REQ_HV_CRASH, vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8950) vcpu->run->exit_reason = KVM_EXIT_SYSTEM_EVENT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8951) vcpu->run->system_event.type = KVM_SYSTEM_EVENT_CRASH;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8952) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8953) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8954) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8955) if (kvm_check_request(KVM_REQ_HV_RESET, vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8956) vcpu->run->exit_reason = KVM_EXIT_SYSTEM_EVENT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8957) vcpu->run->system_event.type = KVM_SYSTEM_EVENT_RESET;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8958) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8959) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8960) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8961) if (kvm_check_request(KVM_REQ_HV_EXIT, vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8962) vcpu->run->exit_reason = KVM_EXIT_HYPERV;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8963) vcpu->run->hyperv = vcpu->arch.hyperv.exit;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8964) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8965) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8966) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8967)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8968) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8969) * KVM_REQ_HV_STIMER has to be processed after
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8970) * KVM_REQ_CLOCK_UPDATE, because Hyper-V SynIC timers
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8971) * depend on the guest clock being up-to-date
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8972) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8973) if (kvm_check_request(KVM_REQ_HV_STIMER, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8974) kvm_hv_process_stimers(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8975) if (kvm_check_request(KVM_REQ_APICV_UPDATE, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8976) kvm_vcpu_update_apicv(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8977) if (kvm_check_request(KVM_REQ_APF_READY, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8978) kvm_check_async_pf_completion(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8979) if (kvm_check_request(KVM_REQ_MSR_FILTER_CHANGED, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8980) kvm_x86_ops.msr_filter_changed(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8981) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8982)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8983) if (kvm_check_request(KVM_REQ_EVENT, vcpu) || req_int_win) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8984) ++vcpu->stat.req_event;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8985) kvm_apic_accept_events(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8986) if (vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8987) r = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8988) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8989) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8990)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8991) inject_pending_event(vcpu, &req_immediate_exit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8992) if (req_int_win)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8993) kvm_x86_ops.enable_irq_window(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8994)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8995) if (kvm_lapic_enabled(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8996) update_cr8_intercept(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8997) kvm_lapic_sync_to_vapic(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8998) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8999) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9000)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9001) r = kvm_mmu_reload(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9002) if (unlikely(r)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9003) goto cancel_injection;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9004) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9005)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9006) preempt_disable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9007)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9008) kvm_x86_ops.prepare_guest_switch(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9009)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9010) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9011) * Disable IRQs before setting IN_GUEST_MODE. Posted interrupt
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9012) * IPI are then delayed after guest entry, which ensures that they
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9013) * result in virtual interrupt delivery.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9014) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9015) local_irq_disable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9016) vcpu->mode = IN_GUEST_MODE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9017)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9018) srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9019)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9020) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9021) * 1) We should set ->mode before checking ->requests. Please see
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9022) * the comment in kvm_vcpu_exiting_guest_mode().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9023) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9024) * 2) For APICv, we should set ->mode before checking PID.ON. This
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9025) * pairs with the memory barrier implicit in pi_test_and_set_on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9026) * (see vmx_deliver_posted_interrupt).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9027) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9028) * 3) This also orders the write to mode from any reads to the page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9029) * tables done while the VCPU is running. Please see the comment
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9030) * in kvm_flush_remote_tlbs.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9031) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9032) smp_mb__after_srcu_read_unlock();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9033)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9034) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9035) * This handles the case where a posted interrupt was
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9036) * notified with kvm_vcpu_kick.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9037) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9038) if (kvm_lapic_enabled(vcpu) && vcpu->arch.apicv_active)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9039) kvm_x86_ops.sync_pir_to_irr(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9040)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9041) if (kvm_vcpu_exit_request(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9042) vcpu->mode = OUTSIDE_GUEST_MODE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9043) smp_wmb();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9044) local_irq_enable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9045) preempt_enable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9046) vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9047) r = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9048) goto cancel_injection;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9049) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9050)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9051) if (req_immediate_exit) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9052) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9053) kvm_x86_ops.request_immediate_exit(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9054) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9055)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9056) trace_kvm_entry(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9057)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9058) fpregs_assert_state_consistent();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9059) if (test_thread_flag(TIF_NEED_FPU_LOAD))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9060) switch_fpu_return();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9061)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9062) if (unlikely(vcpu->arch.switch_db_regs)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9063) set_debugreg(0, 7);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9064) set_debugreg(vcpu->arch.eff_db[0], 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9065) set_debugreg(vcpu->arch.eff_db[1], 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9066) set_debugreg(vcpu->arch.eff_db[2], 2);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9067) set_debugreg(vcpu->arch.eff_db[3], 3);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9068) set_debugreg(vcpu->arch.dr6, 6);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9069) vcpu->arch.switch_db_regs &= ~KVM_DEBUGREG_RELOAD;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9070) } else if (unlikely(hw_breakpoint_active())) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9071) set_debugreg(0, 7);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9072) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9073)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9074) exit_fastpath = kvm_x86_ops.run(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9075)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9076) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9077) * Do this here before restoring debug registers on the host. And
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9078) * since we do this before handling the vmexit, a DR access vmexit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9079) * can (a) read the correct value of the debug registers, (b) set
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9080) * KVM_DEBUGREG_WONT_EXIT again.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9081) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9082) if (unlikely(vcpu->arch.switch_db_regs & KVM_DEBUGREG_WONT_EXIT)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9083) WARN_ON(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9084) kvm_x86_ops.sync_dirty_debug_regs(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9085) kvm_update_dr0123(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9086) kvm_update_dr7(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9087) vcpu->arch.switch_db_regs &= ~KVM_DEBUGREG_RELOAD;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9088) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9089)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9090) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9091) * If the guest has used debug registers, at least dr7
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9092) * will be disabled while returning to the host.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9093) * If we don't have active breakpoints in the host, we don't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9094) * care about the messed up debug address registers. But if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9095) * we have some of them active, restore the old state.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9096) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9097) if (hw_breakpoint_active())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9098) hw_breakpoint_restore();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9099)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9100) vcpu->arch.last_vmentry_cpu = vcpu->cpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9101) vcpu->arch.last_guest_tsc = kvm_read_l1_tsc(vcpu, rdtsc());
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9102)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9103) vcpu->mode = OUTSIDE_GUEST_MODE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9104) smp_wmb();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9105)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9106) kvm_x86_ops.handle_exit_irqoff(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9107)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9108) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9109) * Consume any pending interrupts, including the possible source of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9110) * VM-Exit on SVM and any ticks that occur between VM-Exit and now.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9111) * An instruction is required after local_irq_enable() to fully unblock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9112) * interrupts on processors that implement an interrupt shadow, the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9113) * stat.exits increment will do nicely.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9114) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9115) kvm_before_interrupt(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9116) local_irq_enable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9117) ++vcpu->stat.exits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9118) local_irq_disable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9119) kvm_after_interrupt(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9120)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9121) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9122) * Wait until after servicing IRQs to account guest time so that any
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9123) * ticks that occurred while running the guest are properly accounted
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9124) * to the guest. Waiting until IRQs are enabled degrades the accuracy
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9125) * of accounting via context tracking, but the loss of accuracy is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9126) * acceptable for all known use cases.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9127) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9128) vtime_account_guest_exit();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9129)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9130) if (lapic_in_kernel(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9131) s64 delta = vcpu->arch.apic->lapic_timer.advance_expire_delta;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9132) if (delta != S64_MIN) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9133) trace_kvm_wait_lapic_expire(vcpu->vcpu_id, delta);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9134) vcpu->arch.apic->lapic_timer.advance_expire_delta = S64_MIN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9135) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9136) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9137)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9138) local_irq_enable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9139) preempt_enable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9140)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9141) vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9142)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9143) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9144) * Profile KVM exit RIPs:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9145) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9146) if (unlikely(prof_on == KVM_PROFILING)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9147) unsigned long rip = kvm_rip_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9148) profile_hit(KVM_PROFILING, (void *)rip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9149) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9150)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9151) if (unlikely(vcpu->arch.tsc_always_catchup))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9152) kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9153)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9154) if (vcpu->arch.apic_attention)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9155) kvm_lapic_sync_from_vapic(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9156)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9157) r = kvm_x86_ops.handle_exit(vcpu, exit_fastpath);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9158) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9159)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9160) cancel_injection:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9161) if (req_immediate_exit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9162) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9163) kvm_x86_ops.cancel_injection(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9164) if (unlikely(vcpu->arch.apic_attention))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9165) kvm_lapic_sync_from_vapic(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9166) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9167) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9168) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9169)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9170) static inline int vcpu_block(struct kvm *kvm, struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9171) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9172) if (!kvm_arch_vcpu_runnable(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9173) (!kvm_x86_ops.pre_block || kvm_x86_ops.pre_block(vcpu) == 0)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9174) srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9175) kvm_vcpu_block(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9176) vcpu->srcu_idx = srcu_read_lock(&kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9177)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9178) if (kvm_x86_ops.post_block)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9179) kvm_x86_ops.post_block(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9180)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9181) if (!kvm_check_request(KVM_REQ_UNHALT, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9182) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9183) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9184)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9185) kvm_apic_accept_events(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9186) switch(vcpu->arch.mp_state) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9187) case KVM_MP_STATE_HALTED:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9188) vcpu->arch.pv.pv_unhalted = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9189) vcpu->arch.mp_state =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9190) KVM_MP_STATE_RUNNABLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9191) fallthrough;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9192) case KVM_MP_STATE_RUNNABLE:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9193) vcpu->arch.apf.halted = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9194) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9195) case KVM_MP_STATE_INIT_RECEIVED:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9196) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9197) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9198) return -EINTR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9199) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9200) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9201) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9202)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9203) static inline bool kvm_vcpu_running(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9204) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9205) if (is_guest_mode(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9206) kvm_x86_ops.nested_ops->check_events(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9207)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9208) return (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9209) !vcpu->arch.apf.halted);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9210) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9211)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9212) static int vcpu_run(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9213) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9214) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9215) struct kvm *kvm = vcpu->kvm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9216)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9217) vcpu->srcu_idx = srcu_read_lock(&kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9218) vcpu->arch.l1tf_flush_l1d = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9219)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9220) for (;;) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9221) if (kvm_vcpu_running(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9222) r = vcpu_enter_guest(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9223) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9224) r = vcpu_block(kvm, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9225) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9226)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9227) if (r <= 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9228) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9229)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9230) kvm_clear_request(KVM_REQ_PENDING_TIMER, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9231) if (kvm_cpu_has_pending_timer(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9232) kvm_inject_pending_timer_irqs(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9233)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9234) if (dm_request_for_irq_injection(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9235) kvm_vcpu_ready_for_interrupt_injection(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9236) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9237) vcpu->run->exit_reason = KVM_EXIT_IRQ_WINDOW_OPEN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9238) ++vcpu->stat.request_irq_exits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9239) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9240) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9241)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9242) if (__xfer_to_guest_mode_work_pending()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9243) srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9244) r = xfer_to_guest_mode_handle_work(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9245) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9246) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9247) vcpu->srcu_idx = srcu_read_lock(&kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9248) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9249) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9250)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9251) srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9252)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9253) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9254) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9255)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9256) static inline int complete_emulated_io(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9257) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9258) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9259)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9260) vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9261) r = kvm_emulate_instruction(vcpu, EMULTYPE_NO_DECODE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9262) srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9263) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9264) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9265)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9266) static int complete_emulated_pio(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9267) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9268) BUG_ON(!vcpu->arch.pio.count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9269)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9270) return complete_emulated_io(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9271) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9272)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9273) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9274) * Implements the following, as a state machine:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9275) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9276) * read:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9277) * for each fragment
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9278) * for each mmio piece in the fragment
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9279) * write gpa, len
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9280) * exit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9281) * copy data
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9282) * execute insn
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9283) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9284) * write:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9285) * for each fragment
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9286) * for each mmio piece in the fragment
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9287) * write gpa, len
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9288) * copy data
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9289) * exit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9290) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9291) static int complete_emulated_mmio(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9292) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9293) struct kvm_run *run = vcpu->run;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9294) struct kvm_mmio_fragment *frag;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9295) unsigned len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9296)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9297) BUG_ON(!vcpu->mmio_needed);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9298)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9299) /* Complete previous fragment */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9300) frag = &vcpu->mmio_fragments[vcpu->mmio_cur_fragment];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9301) len = min(8u, frag->len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9302) if (!vcpu->mmio_is_write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9303) memcpy(frag->data, run->mmio.data, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9304)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9305) if (frag->len <= 8) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9306) /* Switch to the next fragment. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9307) frag++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9308) vcpu->mmio_cur_fragment++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9309) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9310) /* Go forward to the next mmio piece. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9311) frag->data += len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9312) frag->gpa += len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9313) frag->len -= len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9314) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9315)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9316) if (vcpu->mmio_cur_fragment >= vcpu->mmio_nr_fragments) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9317) vcpu->mmio_needed = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9318)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9319) /* FIXME: return into emulator if single-stepping. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9320) if (vcpu->mmio_is_write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9321) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9322) vcpu->mmio_read_completed = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9323) return complete_emulated_io(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9324) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9325)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9326) run->exit_reason = KVM_EXIT_MMIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9327) run->mmio.phys_addr = frag->gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9328) if (vcpu->mmio_is_write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9329) memcpy(run->mmio.data, frag->data, min(8u, frag->len));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9330) run->mmio.len = min(8u, frag->len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9331) run->mmio.is_write = vcpu->mmio_is_write;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9332) vcpu->arch.complete_userspace_io = complete_emulated_mmio;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9333) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9334) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9335)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9336) static void kvm_save_current_fpu(struct fpu *fpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9337) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9338) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9339) * If the target FPU state is not resident in the CPU registers, just
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9340) * memcpy() from current, else save CPU state directly to the target.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9341) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9342) if (test_thread_flag(TIF_NEED_FPU_LOAD))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9343) memcpy(&fpu->state, ¤t->thread.fpu.state,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9344) fpu_kernel_xstate_size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9345) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9346) copy_fpregs_to_fpstate(fpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9347) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9348)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9349) /* Swap (qemu) user FPU context for the guest FPU context. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9350) static void kvm_load_guest_fpu(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9351) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9352) fpregs_lock();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9353)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9354) kvm_save_current_fpu(vcpu->arch.user_fpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9355)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9356) /* PKRU is separately restored in kvm_x86_ops.run. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9357) __copy_kernel_to_fpregs(&vcpu->arch.guest_fpu->state,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9358) ~XFEATURE_MASK_PKRU);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9359)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9360) fpregs_mark_activate();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9361) fpregs_unlock();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9362)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9363) trace_kvm_fpu(1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9364) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9365)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9366) /* When vcpu_run ends, restore user space FPU context. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9367) static void kvm_put_guest_fpu(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9368) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9369) fpregs_lock();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9370)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9371) kvm_save_current_fpu(vcpu->arch.guest_fpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9372)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9373) copy_kernel_to_fpregs(&vcpu->arch.user_fpu->state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9374)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9375) fpregs_mark_activate();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9376) fpregs_unlock();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9377)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9378) ++vcpu->stat.fpu_reload;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9379) trace_kvm_fpu(0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9380) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9381)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9382) int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9383) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9384) struct kvm_run *kvm_run = vcpu->run;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9385) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9386)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9387) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9388) kvm_sigset_activate(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9389) kvm_load_guest_fpu(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9390)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9391) if (unlikely(vcpu->arch.mp_state == KVM_MP_STATE_UNINITIALIZED)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9392) if (kvm_run->immediate_exit) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9393) r = -EINTR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9394) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9395) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9396) kvm_vcpu_block(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9397) kvm_apic_accept_events(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9398) kvm_clear_request(KVM_REQ_UNHALT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9399) r = -EAGAIN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9400) if (signal_pending(current)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9401) r = -EINTR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9402) kvm_run->exit_reason = KVM_EXIT_INTR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9403) ++vcpu->stat.signal_exits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9404) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9405) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9406) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9407)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9408) if (kvm_run->kvm_valid_regs & ~KVM_SYNC_X86_VALID_FIELDS) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9409) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9410) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9411) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9412)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9413) if (kvm_run->kvm_dirty_regs) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9414) r = sync_regs(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9415) if (r != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9416) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9417) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9418)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9419) /* re-sync apic's tpr */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9420) if (!lapic_in_kernel(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9421) if (kvm_set_cr8(vcpu, kvm_run->cr8) != 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9422) r = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9423) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9424) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9425) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9426)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9427) if (unlikely(vcpu->arch.complete_userspace_io)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9428) int (*cui)(struct kvm_vcpu *) = vcpu->arch.complete_userspace_io;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9429) vcpu->arch.complete_userspace_io = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9430) r = cui(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9431) if (r <= 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9432) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9433) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9434) WARN_ON(vcpu->arch.pio.count || vcpu->mmio_needed);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9435)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9436) if (kvm_run->immediate_exit)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9437) r = -EINTR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9438) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9439) r = vcpu_run(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9440)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9441) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9442) kvm_put_guest_fpu(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9443) if (kvm_run->kvm_valid_regs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9444) store_regs(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9445) post_kvm_run_save(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9446) kvm_sigset_deactivate(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9447)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9448) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9449) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9450) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9451)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9452) static void __get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9453) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9454) if (vcpu->arch.emulate_regs_need_sync_to_vcpu) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9455) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9456) * We are here if userspace calls get_regs() in the middle of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9457) * instruction emulation. Registers state needs to be copied
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9458) * back from emulation context to vcpu. Userspace shouldn't do
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9459) * that usually, but some bad designed PV devices (vmware
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9460) * backdoor interface) need this to work
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9461) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9462) emulator_writeback_register_cache(vcpu->arch.emulate_ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9463) vcpu->arch.emulate_regs_need_sync_to_vcpu = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9464) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9465) regs->rax = kvm_rax_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9466) regs->rbx = kvm_rbx_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9467) regs->rcx = kvm_rcx_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9468) regs->rdx = kvm_rdx_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9469) regs->rsi = kvm_rsi_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9470) regs->rdi = kvm_rdi_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9471) regs->rsp = kvm_rsp_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9472) regs->rbp = kvm_rbp_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9473) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9474) regs->r8 = kvm_r8_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9475) regs->r9 = kvm_r9_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9476) regs->r10 = kvm_r10_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9477) regs->r11 = kvm_r11_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9478) regs->r12 = kvm_r12_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9479) regs->r13 = kvm_r13_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9480) regs->r14 = kvm_r14_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9481) regs->r15 = kvm_r15_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9482) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9483)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9484) regs->rip = kvm_rip_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9485) regs->rflags = kvm_get_rflags(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9486) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9487)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9488) int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9489) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9490) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9491) __get_regs(vcpu, regs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9492) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9493) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9494) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9495)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9496) static void __set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9497) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9498) vcpu->arch.emulate_regs_need_sync_from_vcpu = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9499) vcpu->arch.emulate_regs_need_sync_to_vcpu = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9500)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9501) kvm_rax_write(vcpu, regs->rax);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9502) kvm_rbx_write(vcpu, regs->rbx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9503) kvm_rcx_write(vcpu, regs->rcx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9504) kvm_rdx_write(vcpu, regs->rdx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9505) kvm_rsi_write(vcpu, regs->rsi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9506) kvm_rdi_write(vcpu, regs->rdi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9507) kvm_rsp_write(vcpu, regs->rsp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9508) kvm_rbp_write(vcpu, regs->rbp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9509) #ifdef CONFIG_X86_64
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9510) kvm_r8_write(vcpu, regs->r8);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9511) kvm_r9_write(vcpu, regs->r9);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9512) kvm_r10_write(vcpu, regs->r10);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9513) kvm_r11_write(vcpu, regs->r11);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9514) kvm_r12_write(vcpu, regs->r12);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9515) kvm_r13_write(vcpu, regs->r13);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9516) kvm_r14_write(vcpu, regs->r14);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9517) kvm_r15_write(vcpu, regs->r15);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9518) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9519)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9520) kvm_rip_write(vcpu, regs->rip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9521) kvm_set_rflags(vcpu, regs->rflags | X86_EFLAGS_FIXED);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9522)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9523) vcpu->arch.exception.pending = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9524)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9525) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9526) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9527)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9528) int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9529) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9530) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9531) __set_regs(vcpu, regs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9532) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9533) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9534) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9535)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9536) void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9537) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9538) struct kvm_segment cs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9539)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9540) kvm_get_segment(vcpu, &cs, VCPU_SREG_CS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9541) *db = cs.db;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9542) *l = cs.l;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9543) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9544) EXPORT_SYMBOL_GPL(kvm_get_cs_db_l_bits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9545)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9546) static void __get_sregs(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9547) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9548) struct desc_ptr dt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9549)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9550) kvm_get_segment(vcpu, &sregs->cs, VCPU_SREG_CS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9551) kvm_get_segment(vcpu, &sregs->ds, VCPU_SREG_DS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9552) kvm_get_segment(vcpu, &sregs->es, VCPU_SREG_ES);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9553) kvm_get_segment(vcpu, &sregs->fs, VCPU_SREG_FS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9554) kvm_get_segment(vcpu, &sregs->gs, VCPU_SREG_GS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9555) kvm_get_segment(vcpu, &sregs->ss, VCPU_SREG_SS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9556)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9557) kvm_get_segment(vcpu, &sregs->tr, VCPU_SREG_TR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9558) kvm_get_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9559)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9560) kvm_x86_ops.get_idt(vcpu, &dt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9561) sregs->idt.limit = dt.size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9562) sregs->idt.base = dt.address;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9563) kvm_x86_ops.get_gdt(vcpu, &dt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9564) sregs->gdt.limit = dt.size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9565) sregs->gdt.base = dt.address;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9566)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9567) sregs->cr0 = kvm_read_cr0(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9568) sregs->cr2 = vcpu->arch.cr2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9569) sregs->cr3 = kvm_read_cr3(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9570) sregs->cr4 = kvm_read_cr4(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9571) sregs->cr8 = kvm_get_cr8(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9572) sregs->efer = vcpu->arch.efer;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9573) sregs->apic_base = kvm_get_apic_base(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9574)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9575) memset(sregs->interrupt_bitmap, 0, sizeof(sregs->interrupt_bitmap));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9576)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9577) if (vcpu->arch.interrupt.injected && !vcpu->arch.interrupt.soft)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9578) set_bit(vcpu->arch.interrupt.nr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9579) (unsigned long *)sregs->interrupt_bitmap);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9580) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9581)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9582) int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9583) struct kvm_sregs *sregs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9584) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9585) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9586) __get_sregs(vcpu, sregs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9587) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9588) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9589) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9590)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9591) int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9592) struct kvm_mp_state *mp_state)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9593) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9594) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9595) if (kvm_mpx_supported())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9596) kvm_load_guest_fpu(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9597)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9598) kvm_apic_accept_events(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9599) if (vcpu->arch.mp_state == KVM_MP_STATE_HALTED &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9600) vcpu->arch.pv.pv_unhalted)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9601) mp_state->mp_state = KVM_MP_STATE_RUNNABLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9602) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9603) mp_state->mp_state = vcpu->arch.mp_state;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9604)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9605) if (kvm_mpx_supported())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9606) kvm_put_guest_fpu(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9607) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9608) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9609) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9610)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9611) int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9612) struct kvm_mp_state *mp_state)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9613) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9614) int ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9615)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9616) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9617)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9618) if (!lapic_in_kernel(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9619) mp_state->mp_state != KVM_MP_STATE_RUNNABLE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9620) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9621)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9622) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9623) * KVM_MP_STATE_INIT_RECEIVED means the processor is in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9624) * INIT state; latched init should be reported using
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9625) * KVM_SET_VCPU_EVENTS, so reject it here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9626) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9627) if ((kvm_vcpu_latch_init(vcpu) || vcpu->arch.smi_pending) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9628) (mp_state->mp_state == KVM_MP_STATE_SIPI_RECEIVED ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9629) mp_state->mp_state == KVM_MP_STATE_INIT_RECEIVED))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9630) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9631)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9632) if (mp_state->mp_state == KVM_MP_STATE_SIPI_RECEIVED) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9633) vcpu->arch.mp_state = KVM_MP_STATE_INIT_RECEIVED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9634) set_bit(KVM_APIC_SIPI, &vcpu->arch.apic->pending_events);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9635) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9636) vcpu->arch.mp_state = mp_state->mp_state;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9637) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9638)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9639) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9640) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9641) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9642) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9643) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9644)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9645) int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int idt_index,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9646) int reason, bool has_error_code, u32 error_code)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9647) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9648) struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9649) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9650)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9651) init_emulate_ctxt(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9652)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9653) ret = emulator_task_switch(ctxt, tss_selector, idt_index, reason,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9654) has_error_code, error_code);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9655) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9656) vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9657) vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_EMULATION;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9658) vcpu->run->internal.ndata = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9659) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9660) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9661)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9662) kvm_rip_write(vcpu, ctxt->eip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9663) kvm_set_rflags(vcpu, ctxt->eflags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9664) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9665) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9666) EXPORT_SYMBOL_GPL(kvm_task_switch);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9667)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9668) static int kvm_valid_sregs(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9669) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9670) if ((sregs->efer & EFER_LME) && (sregs->cr0 & X86_CR0_PG)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9671) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9672) * When EFER.LME and CR0.PG are set, the processor is in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9673) * 64-bit mode (though maybe in a 32-bit code segment).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9674) * CR4.PAE and EFER.LMA must be set.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9675) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9676) if (!(sregs->cr4 & X86_CR4_PAE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9677) || !(sregs->efer & EFER_LMA))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9678) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9679) if (sregs->cr3 & vcpu->arch.cr3_lm_rsvd_bits)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9680) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9681) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9682) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9683) * Not in 64-bit mode: EFER.LMA is clear and the code
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9684) * segment cannot be 64-bit.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9685) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9686) if (sregs->efer & EFER_LMA || sregs->cs.l)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9687) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9688) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9689)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9690) return kvm_valid_cr4(vcpu, sregs->cr4);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9691) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9692)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9693) static int __set_sregs(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9694) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9695) struct msr_data apic_base_msr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9696) int mmu_reset_needed = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9697) int cpuid_update_needed = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9698) int pending_vec, max_bits, idx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9699) struct desc_ptr dt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9700) int ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9701)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9702) if (kvm_valid_sregs(vcpu, sregs))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9703) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9704)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9705) apic_base_msr.data = sregs->apic_base;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9706) apic_base_msr.host_initiated = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9707) if (kvm_set_apic_base(vcpu, &apic_base_msr))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9708) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9709)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9710) dt.size = sregs->idt.limit;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9711) dt.address = sregs->idt.base;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9712) kvm_x86_ops.set_idt(vcpu, &dt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9713) dt.size = sregs->gdt.limit;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9714) dt.address = sregs->gdt.base;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9715) kvm_x86_ops.set_gdt(vcpu, &dt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9716)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9717) vcpu->arch.cr2 = sregs->cr2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9718) mmu_reset_needed |= kvm_read_cr3(vcpu) != sregs->cr3;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9719) vcpu->arch.cr3 = sregs->cr3;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9720) kvm_register_mark_available(vcpu, VCPU_EXREG_CR3);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9721)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9722) kvm_set_cr8(vcpu, sregs->cr8);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9723)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9724) mmu_reset_needed |= vcpu->arch.efer != sregs->efer;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9725) kvm_x86_ops.set_efer(vcpu, sregs->efer);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9726)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9727) mmu_reset_needed |= kvm_read_cr0(vcpu) != sregs->cr0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9728) kvm_x86_ops.set_cr0(vcpu, sregs->cr0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9729) vcpu->arch.cr0 = sregs->cr0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9730)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9731) mmu_reset_needed |= kvm_read_cr4(vcpu) != sregs->cr4;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9732) cpuid_update_needed |= ((kvm_read_cr4(vcpu) ^ sregs->cr4) &
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9733) (X86_CR4_OSXSAVE | X86_CR4_PKE));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9734) kvm_x86_ops.set_cr4(vcpu, sregs->cr4);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9735) if (cpuid_update_needed)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9736) kvm_update_cpuid_runtime(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9737)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9738) idx = srcu_read_lock(&vcpu->kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9739) if (is_pae_paging(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9740) load_pdptrs(vcpu, vcpu->arch.walk_mmu, kvm_read_cr3(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9741) mmu_reset_needed = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9742) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9743) srcu_read_unlock(&vcpu->kvm->srcu, idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9744)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9745) if (mmu_reset_needed)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9746) kvm_mmu_reset_context(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9747)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9748) max_bits = KVM_NR_INTERRUPTS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9749) pending_vec = find_first_bit(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9750) (const unsigned long *)sregs->interrupt_bitmap, max_bits);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9751) if (pending_vec < max_bits) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9752) kvm_queue_interrupt(vcpu, pending_vec, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9753) pr_debug("Set back pending irq %d\n", pending_vec);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9754) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9755)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9756) kvm_set_segment(vcpu, &sregs->cs, VCPU_SREG_CS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9757) kvm_set_segment(vcpu, &sregs->ds, VCPU_SREG_DS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9758) kvm_set_segment(vcpu, &sregs->es, VCPU_SREG_ES);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9759) kvm_set_segment(vcpu, &sregs->fs, VCPU_SREG_FS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9760) kvm_set_segment(vcpu, &sregs->gs, VCPU_SREG_GS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9761) kvm_set_segment(vcpu, &sregs->ss, VCPU_SREG_SS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9762)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9763) kvm_set_segment(vcpu, &sregs->tr, VCPU_SREG_TR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9764) kvm_set_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9765)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9766) update_cr8_intercept(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9767)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9768) /* Older userspace won't unhalt the vcpu on reset. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9769) if (kvm_vcpu_is_bsp(vcpu) && kvm_rip_read(vcpu) == 0xfff0 &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9770) sregs->cs.selector == 0xf000 && sregs->cs.base == 0xffff0000 &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9771) !is_protmode(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9772) vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9773)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9774) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9775)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9776) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9777) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9778) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9779) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9780)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9781) int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9782) struct kvm_sregs *sregs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9783) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9784) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9785)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9786) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9787) ret = __set_sregs(vcpu, sregs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9788) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9789) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9790) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9791)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9792) int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9793) struct kvm_guest_debug *dbg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9794) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9795) unsigned long rflags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9796) int i, r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9797)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9798) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9799)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9800) if (dbg->control & (KVM_GUESTDBG_INJECT_DB | KVM_GUESTDBG_INJECT_BP)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9801) r = -EBUSY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9802) if (vcpu->arch.exception.pending)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9803) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9804) if (dbg->control & KVM_GUESTDBG_INJECT_DB)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9805) kvm_queue_exception(vcpu, DB_VECTOR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9806) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9807) kvm_queue_exception(vcpu, BP_VECTOR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9808) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9809)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9810) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9811) * Read rflags as long as potentially injected trace flags are still
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9812) * filtered out.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9813) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9814) rflags = kvm_get_rflags(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9815)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9816) vcpu->guest_debug = dbg->control;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9817) if (!(vcpu->guest_debug & KVM_GUESTDBG_ENABLE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9818) vcpu->guest_debug = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9819)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9820) if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9821) for (i = 0; i < KVM_NR_DB_REGS; ++i)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9822) vcpu->arch.eff_db[i] = dbg->arch.debugreg[i];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9823) vcpu->arch.guest_debug_dr7 = dbg->arch.debugreg[7];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9824) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9825) for (i = 0; i < KVM_NR_DB_REGS; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9826) vcpu->arch.eff_db[i] = vcpu->arch.db[i];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9827) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9828) kvm_update_dr7(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9829)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9830) if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9831) vcpu->arch.singlestep_rip = kvm_rip_read(vcpu) +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9832) get_segment_base(vcpu, VCPU_SREG_CS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9833)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9834) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9835) * Trigger an rflags update that will inject or remove the trace
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9836) * flags.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9837) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9838) kvm_set_rflags(vcpu, rflags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9839)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9840) kvm_x86_ops.update_exception_bitmap(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9841)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9842) r = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9843)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9844) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9845) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9846) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9847) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9848)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9849) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9850) * Translate a guest virtual address to a guest physical address.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9851) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9852) int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9853) struct kvm_translation *tr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9854) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9855) unsigned long vaddr = tr->linear_address;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9856) gpa_t gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9857) int idx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9858)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9859) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9860)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9861) idx = srcu_read_lock(&vcpu->kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9862) gpa = kvm_mmu_gva_to_gpa_system(vcpu, vaddr, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9863) srcu_read_unlock(&vcpu->kvm->srcu, idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9864) tr->physical_address = gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9865) tr->valid = gpa != UNMAPPED_GVA;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9866) tr->writeable = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9867) tr->usermode = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9868)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9869) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9870) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9871) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9872)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9873) int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9874) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9875) struct fxregs_state *fxsave;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9876)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9877) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9878)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9879) fxsave = &vcpu->arch.guest_fpu->state.fxsave;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9880) memcpy(fpu->fpr, fxsave->st_space, 128);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9881) fpu->fcw = fxsave->cwd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9882) fpu->fsw = fxsave->swd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9883) fpu->ftwx = fxsave->twd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9884) fpu->last_opcode = fxsave->fop;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9885) fpu->last_ip = fxsave->rip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9886) fpu->last_dp = fxsave->rdp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9887) memcpy(fpu->xmm, fxsave->xmm_space, sizeof(fxsave->xmm_space));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9888)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9889) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9890) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9891) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9892)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9893) int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9894) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9895) struct fxregs_state *fxsave;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9896)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9897) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9898)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9899) fxsave = &vcpu->arch.guest_fpu->state.fxsave;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9900)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9901) memcpy(fxsave->st_space, fpu->fpr, 128);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9902) fxsave->cwd = fpu->fcw;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9903) fxsave->swd = fpu->fsw;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9904) fxsave->twd = fpu->ftwx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9905) fxsave->fop = fpu->last_opcode;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9906) fxsave->rip = fpu->last_ip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9907) fxsave->rdp = fpu->last_dp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9908) memcpy(fxsave->xmm_space, fpu->xmm, sizeof(fxsave->xmm_space));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9909)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9910) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9911) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9912) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9913)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9914) static void store_regs(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9915) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9916) BUILD_BUG_ON(sizeof(struct kvm_sync_regs) > SYNC_REGS_SIZE_BYTES);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9917)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9918) if (vcpu->run->kvm_valid_regs & KVM_SYNC_X86_REGS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9919) __get_regs(vcpu, &vcpu->run->s.regs.regs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9920)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9921) if (vcpu->run->kvm_valid_regs & KVM_SYNC_X86_SREGS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9922) __get_sregs(vcpu, &vcpu->run->s.regs.sregs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9923)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9924) if (vcpu->run->kvm_valid_regs & KVM_SYNC_X86_EVENTS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9925) kvm_vcpu_ioctl_x86_get_vcpu_events(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9926) vcpu, &vcpu->run->s.regs.events);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9927) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9928)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9929) static int sync_regs(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9930) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9931) if (vcpu->run->kvm_dirty_regs & ~KVM_SYNC_X86_VALID_FIELDS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9932) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9933)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9934) if (vcpu->run->kvm_dirty_regs & KVM_SYNC_X86_REGS) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9935) __set_regs(vcpu, &vcpu->run->s.regs.regs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9936) vcpu->run->kvm_dirty_regs &= ~KVM_SYNC_X86_REGS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9937) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9938) if (vcpu->run->kvm_dirty_regs & KVM_SYNC_X86_SREGS) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9939) if (__set_sregs(vcpu, &vcpu->run->s.regs.sregs))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9940) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9941) vcpu->run->kvm_dirty_regs &= ~KVM_SYNC_X86_SREGS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9942) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9943) if (vcpu->run->kvm_dirty_regs & KVM_SYNC_X86_EVENTS) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9944) if (kvm_vcpu_ioctl_x86_set_vcpu_events(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9945) vcpu, &vcpu->run->s.regs.events))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9946) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9947) vcpu->run->kvm_dirty_regs &= ~KVM_SYNC_X86_EVENTS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9948) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9949)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9950) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9951) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9952)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9953) static void fx_init(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9954) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9955) fpstate_init(&vcpu->arch.guest_fpu->state);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9956) if (boot_cpu_has(X86_FEATURE_XSAVES))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9957) vcpu->arch.guest_fpu->state.xsave.header.xcomp_bv =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9958) host_xcr0 | XSTATE_COMPACTION_ENABLED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9959)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9960) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9961) * Ensure guest xcr0 is valid for loading
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9962) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9963) vcpu->arch.xcr0 = XFEATURE_MASK_FP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9964)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9965) vcpu->arch.cr0 |= X86_CR0_ET;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9966) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9967)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9968) int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9969) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9970) if (kvm_check_tsc_unstable() && atomic_read(&kvm->online_vcpus) != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9971) pr_warn_once("kvm: SMP vm created on host with unstable TSC; "
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9972) "guest TSC will not be reliable\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9973)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9974) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9975) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9976)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9977) int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9978) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9979) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9980) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9981)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9982) if (!irqchip_in_kernel(vcpu->kvm) || kvm_vcpu_is_reset_bsp(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9983) vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9984) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9985) vcpu->arch.mp_state = KVM_MP_STATE_UNINITIALIZED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9986)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9987) kvm_set_tsc_khz(vcpu, max_tsc_khz);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9988)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9989) r = kvm_mmu_create(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9990) if (r < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9991) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9992)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9993) if (irqchip_in_kernel(vcpu->kvm)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9994) r = kvm_create_lapic(vcpu, lapic_timer_advance_ns);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9995) if (r < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9996) goto fail_mmu_destroy;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9997) if (kvm_apicv_activated(vcpu->kvm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9998) vcpu->arch.apicv_active = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9999) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10000) static_key_slow_inc(&kvm_no_apic_vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10001)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10002) r = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10003)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10004) page = alloc_page(GFP_KERNEL | __GFP_ZERO);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10005) if (!page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10006) goto fail_free_lapic;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10007) vcpu->arch.pio_data = page_address(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10008)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10009) vcpu->arch.mce_banks = kzalloc(KVM_MAX_MCE_BANKS * sizeof(u64) * 4,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10010) GFP_KERNEL_ACCOUNT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10011) if (!vcpu->arch.mce_banks)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10012) goto fail_free_pio_data;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10013) vcpu->arch.mcg_cap = KVM_MAX_MCE_BANKS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10014)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10015) if (!zalloc_cpumask_var(&vcpu->arch.wbinvd_dirty_mask,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10016) GFP_KERNEL_ACCOUNT))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10017) goto fail_free_mce_banks;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10018)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10019) if (!alloc_emulate_ctxt(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10020) goto free_wbinvd_dirty_mask;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10021)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10022) vcpu->arch.user_fpu = kmem_cache_zalloc(x86_fpu_cache,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10023) GFP_KERNEL_ACCOUNT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10024) if (!vcpu->arch.user_fpu) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10025) pr_err("kvm: failed to allocate userspace's fpu\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10026) goto free_emulate_ctxt;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10027) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10028)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10029) vcpu->arch.guest_fpu = kmem_cache_zalloc(x86_fpu_cache,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10030) GFP_KERNEL_ACCOUNT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10031) if (!vcpu->arch.guest_fpu) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10032) pr_err("kvm: failed to allocate vcpu's fpu\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10033) goto free_user_fpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10034) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10035) fx_init(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10036)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10037) vcpu->arch.maxphyaddr = cpuid_query_maxphyaddr(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10038) vcpu->arch.cr3_lm_rsvd_bits = rsvd_bits(cpuid_maxphyaddr(vcpu), 63);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10039)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10040) vcpu->arch.pat = MSR_IA32_CR_PAT_DEFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10041)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10042) kvm_async_pf_hash_reset(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10043) kvm_pmu_init(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10044)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10045) vcpu->arch.pending_external_vector = -1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10046) vcpu->arch.preempted_in_kernel = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10047)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10048) kvm_hv_vcpu_init(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10049)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10050) r = kvm_x86_ops.vcpu_create(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10051) if (r)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10052) goto free_guest_fpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10053)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10054) vcpu->arch.arch_capabilities = kvm_get_arch_capabilities();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10055) vcpu->arch.msr_platform_info = MSR_PLATFORM_INFO_CPUID_FAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10056) kvm_vcpu_mtrr_init(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10057) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10058) kvm_vcpu_reset(vcpu, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10059) kvm_init_mmu(vcpu, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10060) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10061) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10062)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10063) free_guest_fpu:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10064) kmem_cache_free(x86_fpu_cache, vcpu->arch.guest_fpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10065) free_user_fpu:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10066) kmem_cache_free(x86_fpu_cache, vcpu->arch.user_fpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10067) free_emulate_ctxt:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10068) kmem_cache_free(x86_emulator_cache, vcpu->arch.emulate_ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10069) free_wbinvd_dirty_mask:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10070) free_cpumask_var(vcpu->arch.wbinvd_dirty_mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10071) fail_free_mce_banks:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10072) kfree(vcpu->arch.mce_banks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10073) fail_free_pio_data:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10074) free_page((unsigned long)vcpu->arch.pio_data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10075) fail_free_lapic:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10076) kvm_free_lapic(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10077) fail_mmu_destroy:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10078) kvm_mmu_destroy(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10079) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10080) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10081)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10082) void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10083) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10084) struct kvm *kvm = vcpu->kvm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10085)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10086) kvm_hv_vcpu_postcreate(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10087)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10088) if (mutex_lock_killable(&vcpu->mutex))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10089) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10090) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10091) kvm_synchronize_tsc(vcpu, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10092) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10093)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10094) /* poll control enabled by default */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10095) vcpu->arch.msr_kvm_poll_control = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10096)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10097) mutex_unlock(&vcpu->mutex);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10098)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10099) if (kvmclock_periodic_sync && vcpu->vcpu_idx == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10100) schedule_delayed_work(&kvm->arch.kvmclock_sync_work,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10101) KVMCLOCK_SYNC_PERIOD);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10102) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10103)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10104) void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10105) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10106) struct gfn_to_pfn_cache *cache = &vcpu->arch.st.cache;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10107) int idx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10108)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10109) kvm_release_pfn(cache->pfn, cache->dirty, cache);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10110)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10111) kvmclock_reset(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10112)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10113) kvm_x86_ops.vcpu_free(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10114)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10115) kmem_cache_free(x86_emulator_cache, vcpu->arch.emulate_ctxt);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10116) free_cpumask_var(vcpu->arch.wbinvd_dirty_mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10117) kmem_cache_free(x86_fpu_cache, vcpu->arch.user_fpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10118) kmem_cache_free(x86_fpu_cache, vcpu->arch.guest_fpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10119)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10120) kvm_hv_vcpu_uninit(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10121) kvm_pmu_destroy(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10122) kfree(vcpu->arch.mce_banks);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10123) kvm_free_lapic(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10124) idx = srcu_read_lock(&vcpu->kvm->srcu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10125) kvm_mmu_destroy(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10126) srcu_read_unlock(&vcpu->kvm->srcu, idx);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10127) free_page((unsigned long)vcpu->arch.pio_data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10128) kvfree(vcpu->arch.cpuid_entries);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10129) if (!lapic_in_kernel(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10130) static_key_slow_dec(&kvm_no_apic_vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10131) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10132)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10133) void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10134) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10135) kvm_lapic_reset(vcpu, init_event);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10136)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10137) vcpu->arch.hflags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10138)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10139) vcpu->arch.smi_pending = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10140) vcpu->arch.smi_count = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10141) atomic_set(&vcpu->arch.nmi_queued, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10142) vcpu->arch.nmi_pending = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10143) vcpu->arch.nmi_injected = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10144) kvm_clear_interrupt_queue(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10145) kvm_clear_exception_queue(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10146)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10147) memset(vcpu->arch.db, 0, sizeof(vcpu->arch.db));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10148) kvm_update_dr0123(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10149) vcpu->arch.dr6 = DR6_INIT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10150) vcpu->arch.dr7 = DR7_FIXED_1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10151) kvm_update_dr7(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10152)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10153) vcpu->arch.cr2 = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10154)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10155) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10156) vcpu->arch.apf.msr_en_val = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10157) vcpu->arch.apf.msr_int_val = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10158) vcpu->arch.st.msr_val = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10159)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10160) kvmclock_reset(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10161)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10162) kvm_clear_async_pf_completion_queue(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10163) kvm_async_pf_hash_reset(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10164) vcpu->arch.apf.halted = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10165)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10166) if (kvm_mpx_supported()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10167) void *mpx_state_buffer;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10168)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10169) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10170) * To avoid have the INIT path from kvm_apic_has_events() that be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10171) * called with loaded FPU and does not let userspace fix the state.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10172) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10173) if (init_event)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10174) kvm_put_guest_fpu(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10175) mpx_state_buffer = get_xsave_addr(&vcpu->arch.guest_fpu->state.xsave,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10176) XFEATURE_BNDREGS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10177) if (mpx_state_buffer)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10178) memset(mpx_state_buffer, 0, sizeof(struct mpx_bndreg_state));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10179) mpx_state_buffer = get_xsave_addr(&vcpu->arch.guest_fpu->state.xsave,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10180) XFEATURE_BNDCSR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10181) if (mpx_state_buffer)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10182) memset(mpx_state_buffer, 0, sizeof(struct mpx_bndcsr));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10183) if (init_event)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10184) kvm_load_guest_fpu(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10185) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10186)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10187) if (!init_event) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10188) kvm_pmu_reset(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10189) vcpu->arch.smbase = 0x30000;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10190)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10191) vcpu->arch.msr_misc_features_enables = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10192)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10193) vcpu->arch.xcr0 = XFEATURE_MASK_FP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10194) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10195)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10196) memset(vcpu->arch.regs, 0, sizeof(vcpu->arch.regs));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10197) vcpu->arch.regs_avail = ~0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10198) vcpu->arch.regs_dirty = ~0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10199)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10200) vcpu->arch.ia32_xss = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10201)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10202) kvm_x86_ops.vcpu_reset(vcpu, init_event);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10203) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10204)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10205) void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10206) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10207) struct kvm_segment cs;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10208)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10209) kvm_get_segment(vcpu, &cs, VCPU_SREG_CS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10210) cs.selector = vector << 8;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10211) cs.base = vector << 12;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10212) kvm_set_segment(vcpu, &cs, VCPU_SREG_CS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10213) kvm_rip_write(vcpu, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10214) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10215)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10216) int kvm_arch_hardware_enable(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10217) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10218) struct kvm *kvm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10219) struct kvm_vcpu *vcpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10220) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10221) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10222) u64 local_tsc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10223) u64 max_tsc = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10224) bool stable, backwards_tsc = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10225)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10226) kvm_user_return_msr_cpu_online();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10227) ret = kvm_x86_ops.hardware_enable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10228) if (ret != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10229) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10230)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10231) local_tsc = rdtsc();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10232) stable = !kvm_check_tsc_unstable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10233) list_for_each_entry(kvm, &vm_list, vm_list) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10234) kvm_for_each_vcpu(i, vcpu, kvm) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10235) if (!stable && vcpu->cpu == smp_processor_id())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10236) kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10237) if (stable && vcpu->arch.last_host_tsc > local_tsc) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10238) backwards_tsc = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10239) if (vcpu->arch.last_host_tsc > max_tsc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10240) max_tsc = vcpu->arch.last_host_tsc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10241) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10242) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10243) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10244)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10245) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10246) * Sometimes, even reliable TSCs go backwards. This happens on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10247) * platforms that reset TSC during suspend or hibernate actions, but
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10248) * maintain synchronization. We must compensate. Fortunately, we can
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10249) * detect that condition here, which happens early in CPU bringup,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10250) * before any KVM threads can be running. Unfortunately, we can't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10251) * bring the TSCs fully up to date with real time, as we aren't yet far
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10252) * enough into CPU bringup that we know how much real time has actually
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10253) * elapsed; our helper function, ktime_get_boottime_ns() will be using boot
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10254) * variables that haven't been updated yet.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10255) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10256) * So we simply find the maximum observed TSC above, then record the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10257) * adjustment to TSC in each VCPU. When the VCPU later gets loaded,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10258) * the adjustment will be applied. Note that we accumulate
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10259) * adjustments, in case multiple suspend cycles happen before some VCPU
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10260) * gets a chance to run again. In the event that no KVM threads get a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10261) * chance to run, we will miss the entire elapsed period, as we'll have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10262) * reset last_host_tsc, so VCPUs will not have the TSC adjusted and may
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10263) * loose cycle time. This isn't too big a deal, since the loss will be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10264) * uniform across all VCPUs (not to mention the scenario is extremely
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10265) * unlikely). It is possible that a second hibernate recovery happens
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10266) * much faster than a first, causing the observed TSC here to be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10267) * smaller; this would require additional padding adjustment, which is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10268) * why we set last_host_tsc to the local tsc observed here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10269) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10270) * N.B. - this code below runs only on platforms with reliable TSC,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10271) * as that is the only way backwards_tsc is set above. Also note
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10272) * that this runs for ALL vcpus, which is not a bug; all VCPUs should
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10273) * have the same delta_cyc adjustment applied if backwards_tsc
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10274) * is detected. Note further, this adjustment is only done once,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10275) * as we reset last_host_tsc on all VCPUs to stop this from being
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10276) * called multiple times (one for each physical CPU bringup).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10277) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10278) * Platforms with unreliable TSCs don't have to deal with this, they
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10279) * will be compensated by the logic in vcpu_load, which sets the TSC to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10280) * catchup mode. This will catchup all VCPUs to real time, but cannot
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10281) * guarantee that they stay in perfect synchronization.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10282) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10283) if (backwards_tsc) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10284) u64 delta_cyc = max_tsc - local_tsc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10285) list_for_each_entry(kvm, &vm_list, vm_list) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10286) kvm->arch.backwards_tsc_observed = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10287) kvm_for_each_vcpu(i, vcpu, kvm) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10288) vcpu->arch.tsc_offset_adjustment += delta_cyc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10289) vcpu->arch.last_host_tsc = local_tsc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10290) kvm_make_request(KVM_REQ_MASTERCLOCK_UPDATE, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10291) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10292)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10293) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10294) * We have to disable TSC offset matching.. if you were
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10295) * booting a VM while issuing an S4 host suspend....
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10296) * you may have some problem. Solving this issue is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10297) * left as an exercise to the reader.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10298) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10299) kvm->arch.last_tsc_nsec = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10300) kvm->arch.last_tsc_write = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10301) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10302)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10303) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10304) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10305) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10306)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10307) void kvm_arch_hardware_disable(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10308) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10309) kvm_x86_ops.hardware_disable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10310) drop_user_return_notifiers();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10311) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10312)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10313) int kvm_arch_hardware_setup(void *opaque)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10314) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10315) struct kvm_x86_init_ops *ops = opaque;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10316) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10317)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10318) rdmsrl_safe(MSR_EFER, &host_efer);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10319)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10320) if (boot_cpu_has(X86_FEATURE_XSAVES))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10321) rdmsrl(MSR_IA32_XSS, host_xss);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10322)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10323) r = ops->hardware_setup();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10324) if (r != 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10325) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10326)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10327) memcpy(&kvm_x86_ops, ops->runtime_ops, sizeof(kvm_x86_ops));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10328)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10329) if (!kvm_cpu_cap_has(X86_FEATURE_XSAVES))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10330) supported_xss = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10331)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10332) #define __kvm_cpu_cap_has(UNUSED_, f) kvm_cpu_cap_has(f)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10333) cr4_reserved_bits = __cr4_reserved_bits(__kvm_cpu_cap_has, UNUSED_);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10334) #undef __kvm_cpu_cap_has
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10335)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10336) if (kvm_has_tsc_control) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10337) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10338) * Make sure the user can only configure tsc_khz values that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10339) * fit into a signed integer.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10340) * A min value is not calculated because it will always
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10341) * be 1 on all machines.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10342) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10343) u64 max = min(0x7fffffffULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10344) __scale_tsc(kvm_max_tsc_scaling_ratio, tsc_khz));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10345) kvm_max_guest_tsc_khz = max;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10346)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10347) kvm_default_tsc_scaling_ratio = 1ULL << kvm_tsc_scaling_ratio_frac_bits;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10348) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10349)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10350) kvm_init_msr_list();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10351) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10352) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10353)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10354) void kvm_arch_hardware_unsetup(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10355) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10356) kvm_x86_ops.hardware_unsetup();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10357) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10358)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10359) int kvm_arch_check_processor_compat(void *opaque)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10360) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10361) struct cpuinfo_x86 *c = &cpu_data(smp_processor_id());
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10362) struct kvm_x86_init_ops *ops = opaque;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10363)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10364) WARN_ON(!irqs_disabled());
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10365)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10366) if (__cr4_reserved_bits(cpu_has, c) !=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10367) __cr4_reserved_bits(cpu_has, &boot_cpu_data))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10368) return -EIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10369)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10370) return ops->check_processor_compatibility();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10371) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10372)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10373) bool kvm_vcpu_is_reset_bsp(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10374) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10375) return vcpu->kvm->arch.bsp_vcpu_id == vcpu->vcpu_id;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10376) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10377) EXPORT_SYMBOL_GPL(kvm_vcpu_is_reset_bsp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10378)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10379) bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10380) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10381) return (vcpu->arch.apic_base & MSR_IA32_APICBASE_BSP) != 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10382) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10383)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10384) struct static_key kvm_no_apic_vcpu __read_mostly;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10385) EXPORT_SYMBOL_GPL(kvm_no_apic_vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10386)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10387) void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10388) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10389) struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10390)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10391) vcpu->arch.l1tf_flush_l1d = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10392) if (pmu->version && unlikely(pmu->event_count)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10393) pmu->need_cleanup = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10394) kvm_make_request(KVM_REQ_PMU, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10395) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10396) kvm_x86_ops.sched_in(vcpu, cpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10397) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10398)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10399) void kvm_arch_free_vm(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10400) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10401) kfree(kvm->arch.hyperv.hv_pa_pg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10402) vfree(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10403) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10404)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10405)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10406) int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10407) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10408) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10409)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10410) if (type)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10411) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10412)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10413) ret = kvm_page_track_init(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10414) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10415) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10416)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10417) INIT_HLIST_HEAD(&kvm->arch.mask_notifier_list);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10418) INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10419) INIT_LIST_HEAD(&kvm->arch.zapped_obsolete_pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10420) INIT_LIST_HEAD(&kvm->arch.lpage_disallowed_mmu_pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10421) INIT_LIST_HEAD(&kvm->arch.assigned_dev_head);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10422) atomic_set(&kvm->arch.noncoherent_dma_count, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10423)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10424) /* Reserve bit 0 of irq_sources_bitmap for userspace irq source */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10425) set_bit(KVM_USERSPACE_IRQ_SOURCE_ID, &kvm->arch.irq_sources_bitmap);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10426) /* Reserve bit 1 of irq_sources_bitmap for irqfd-resampler */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10427) set_bit(KVM_IRQFD_RESAMPLE_IRQ_SOURCE_ID,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10428) &kvm->arch.irq_sources_bitmap);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10429)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10430) raw_spin_lock_init(&kvm->arch.tsc_write_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10431) mutex_init(&kvm->arch.apic_map_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10432) spin_lock_init(&kvm->arch.pvclock_gtod_sync_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10433)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10434) kvm->arch.kvmclock_offset = -get_kvmclock_base_ns();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10435) pvclock_update_vm_gtod_copy(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10436)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10437) kvm->arch.guest_can_read_msr_platform_info = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10438)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10439) INIT_DELAYED_WORK(&kvm->arch.kvmclock_update_work, kvmclock_update_fn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10440) INIT_DELAYED_WORK(&kvm->arch.kvmclock_sync_work, kvmclock_sync_fn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10441)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10442) kvm_hv_init_vm(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10443) kvm_mmu_init_vm(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10444)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10445) return kvm_x86_ops.vm_init(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10446) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10447)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10448) int kvm_arch_post_init_vm(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10449) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10450) return kvm_mmu_post_init_vm(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10451) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10452)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10453) static void kvm_unload_vcpu_mmu(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10454) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10455) vcpu_load(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10456) kvm_mmu_unload(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10457) vcpu_put(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10458) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10459)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10460) static void kvm_free_vcpus(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10461) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10462) unsigned int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10463) struct kvm_vcpu *vcpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10464)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10465) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10466) * Unpin any mmu pages first.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10467) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10468) kvm_for_each_vcpu(i, vcpu, kvm) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10469) kvm_clear_async_pf_completion_queue(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10470) kvm_unload_vcpu_mmu(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10471) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10472) kvm_for_each_vcpu(i, vcpu, kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10473) kvm_vcpu_destroy(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10474)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10475) mutex_lock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10476) for (i = 0; i < atomic_read(&kvm->online_vcpus); i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10477) kvm->vcpus[i] = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10478)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10479) atomic_set(&kvm->online_vcpus, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10480) mutex_unlock(&kvm->lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10481) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10482)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10483) void kvm_arch_sync_events(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10484) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10485) cancel_delayed_work_sync(&kvm->arch.kvmclock_sync_work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10486) cancel_delayed_work_sync(&kvm->arch.kvmclock_update_work);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10487) kvm_free_pit(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10488) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10489)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10490) int __x86_set_memory_region(struct kvm *kvm, int id, gpa_t gpa, u32 size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10491) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10492) int i, r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10493) unsigned long hva, old_npages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10494) struct kvm_memslots *slots = kvm_memslots(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10495) struct kvm_memory_slot *slot;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10496)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10497) /* Called with kvm->slots_lock held. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10498) if (WARN_ON(id >= KVM_MEM_SLOTS_NUM))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10499) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10500)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10501) slot = id_to_memslot(slots, id);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10502) if (size) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10503) if (slot && slot->npages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10504) return -EEXIST;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10505)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10506) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10507) * MAP_SHARED to prevent internal slot pages from being moved
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10508) * by fork()/COW.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10509) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10510) hva = vm_mmap(NULL, 0, size, PROT_READ | PROT_WRITE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10511) MAP_SHARED | MAP_ANONYMOUS, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10512) if (IS_ERR((void *)hva))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10513) return PTR_ERR((void *)hva);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10514) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10515) if (!slot || !slot->npages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10516) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10517)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10518) old_npages = slot->npages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10519) hva = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10520) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10521)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10522) for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10523) struct kvm_userspace_memory_region m;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10524)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10525) m.slot = id | (i << 16);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10526) m.flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10527) m.guest_phys_addr = gpa;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10528) m.userspace_addr = hva;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10529) m.memory_size = size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10530) r = __kvm_set_memory_region(kvm, &m);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10531) if (r < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10532) return r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10533) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10534)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10535) if (!size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10536) vm_munmap(hva, old_npages * PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10537)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10538) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10539) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10540) EXPORT_SYMBOL_GPL(__x86_set_memory_region);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10541)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10542) void kvm_arch_pre_destroy_vm(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10543) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10544) kvm_mmu_pre_destroy_vm(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10545) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10546)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10547) void kvm_arch_destroy_vm(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10548) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10549) if (current->mm == kvm->mm) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10550) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10551) * Free memory regions allocated on behalf of userspace,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10552) * unless the the memory map has changed due to process exit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10553) * or fd copying.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10554) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10555) mutex_lock(&kvm->slots_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10556) __x86_set_memory_region(kvm, APIC_ACCESS_PAGE_PRIVATE_MEMSLOT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10557) 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10558) __x86_set_memory_region(kvm, IDENTITY_PAGETABLE_PRIVATE_MEMSLOT,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10559) 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10560) __x86_set_memory_region(kvm, TSS_PRIVATE_MEMSLOT, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10561) mutex_unlock(&kvm->slots_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10562) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10563) if (kvm_x86_ops.vm_destroy)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10564) kvm_x86_ops.vm_destroy(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10565) kvm_free_msr_filter(srcu_dereference_check(kvm->arch.msr_filter, &kvm->srcu, 1));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10566) kvm_pic_destroy(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10567) kvm_ioapic_destroy(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10568) kvm_free_vcpus(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10569) kvfree(rcu_dereference_check(kvm->arch.apic_map, 1));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10570) kfree(srcu_dereference_check(kvm->arch.pmu_event_filter, &kvm->srcu, 1));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10571) kvm_mmu_uninit_vm(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10572) kvm_page_track_cleanup(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10573) kvm_hv_destroy_vm(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10574) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10575)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10576) void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *slot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10577) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10578) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10579)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10580) for (i = 0; i < KVM_NR_PAGE_SIZES; ++i) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10581) kvfree(slot->arch.rmap[i]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10582) slot->arch.rmap[i] = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10583)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10584) if (i == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10585) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10586)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10587) kvfree(slot->arch.lpage_info[i - 1]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10588) slot->arch.lpage_info[i - 1] = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10589) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10590)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10591) kvm_page_track_free_memslot(slot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10592) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10593)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10594) static int kvm_alloc_memslot_metadata(struct kvm_memory_slot *slot,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10595) unsigned long npages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10596) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10597) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10598)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10599) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10600) * Clear out the previous array pointers for the KVM_MR_MOVE case. The
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10601) * old arrays will be freed by __kvm_set_memory_region() if installing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10602) * the new memslot is successful.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10603) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10604) memset(&slot->arch, 0, sizeof(slot->arch));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10605)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10606) for (i = 0; i < KVM_NR_PAGE_SIZES; ++i) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10607) struct kvm_lpage_info *linfo;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10608) unsigned long ugfn;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10609) int lpages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10610) int level = i + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10611)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10612) lpages = gfn_to_index(slot->base_gfn + npages - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10613) slot->base_gfn, level) + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10614)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10615) slot->arch.rmap[i] =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10616) kvcalloc(lpages, sizeof(*slot->arch.rmap[i]),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10617) GFP_KERNEL_ACCOUNT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10618) if (!slot->arch.rmap[i])
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10619) goto out_free;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10620) if (i == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10621) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10622)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10623) linfo = kvcalloc(lpages, sizeof(*linfo), GFP_KERNEL_ACCOUNT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10624) if (!linfo)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10625) goto out_free;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10626)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10627) slot->arch.lpage_info[i - 1] = linfo;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10628)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10629) if (slot->base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10630) linfo[0].disallow_lpage = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10631) if ((slot->base_gfn + npages) & (KVM_PAGES_PER_HPAGE(level) - 1))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10632) linfo[lpages - 1].disallow_lpage = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10633) ugfn = slot->userspace_addr >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10634) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10635) * If the gfn and userspace address are not aligned wrt each
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10636) * other, disable large page support for this slot.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10637) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10638) if ((slot->base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10639) unsigned long j;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10640)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10641) for (j = 0; j < lpages; ++j)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10642) linfo[j].disallow_lpage = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10643) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10644) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10645)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10646) if (kvm_page_track_create_memslot(slot, npages))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10647) goto out_free;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10648)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10649) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10650)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10651) out_free:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10652) for (i = 0; i < KVM_NR_PAGE_SIZES; ++i) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10653) kvfree(slot->arch.rmap[i]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10654) slot->arch.rmap[i] = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10655) if (i == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10656) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10657)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10658) kvfree(slot->arch.lpage_info[i - 1]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10659) slot->arch.lpage_info[i - 1] = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10660) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10661) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10662) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10663)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10664) void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10665) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10666) struct kvm_vcpu *vcpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10667) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10668)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10669) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10670) * memslots->generation has been incremented.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10671) * mmio generation may have reached its maximum value.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10672) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10673) kvm_mmu_invalidate_mmio_sptes(kvm, gen);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10674)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10675) /* Force re-initialization of steal_time cache */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10676) kvm_for_each_vcpu(i, vcpu, kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10677) kvm_vcpu_kick(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10678) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10679)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10680) int kvm_arch_prepare_memory_region(struct kvm *kvm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10681) struct kvm_memory_slot *memslot,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10682) const struct kvm_userspace_memory_region *mem,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10683) enum kvm_mr_change change)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10684) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10685) if (change == KVM_MR_CREATE || change == KVM_MR_MOVE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10686) return kvm_alloc_memslot_metadata(memslot,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10687) mem->memory_size >> PAGE_SHIFT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10688) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10689) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10690)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10691) static void kvm_mmu_slot_apply_flags(struct kvm *kvm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10692) struct kvm_memory_slot *old,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10693) struct kvm_memory_slot *new,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10694) enum kvm_mr_change change)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10695) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10696) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10697) * Nothing to do for RO slots or CREATE/MOVE/DELETE of a slot.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10698) * See comments below.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10699) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10700) if ((change != KVM_MR_FLAGS_ONLY) || (new->flags & KVM_MEM_READONLY))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10701) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10702)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10703) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10704) * Dirty logging tracks sptes in 4k granularity, meaning that large
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10705) * sptes have to be split. If live migration is successful, the guest
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10706) * in the source machine will be destroyed and large sptes will be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10707) * created in the destination. However, if the guest continues to run
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10708) * in the source machine (for example if live migration fails), small
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10709) * sptes will remain around and cause bad performance.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10710) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10711) * Scan sptes if dirty logging has been stopped, dropping those
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10712) * which can be collapsed into a single large-page spte. Later
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10713) * page faults will create the large-page sptes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10714) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10715) * There is no need to do this in any of the following cases:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10716) * CREATE: No dirty mappings will already exist.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10717) * MOVE/DELETE: The old mappings will already have been cleaned up by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10718) * kvm_arch_flush_shadow_memslot()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10719) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10720) if ((old->flags & KVM_MEM_LOG_DIRTY_PAGES) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10721) !(new->flags & KVM_MEM_LOG_DIRTY_PAGES))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10722) kvm_mmu_zap_collapsible_sptes(kvm, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10723)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10724) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10725) * Enable or disable dirty logging for the slot.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10726) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10727) * For KVM_MR_DELETE and KVM_MR_MOVE, the shadow pages of the old
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10728) * slot have been zapped so no dirty logging updates are needed for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10729) * the old slot.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10730) * For KVM_MR_CREATE and KVM_MR_MOVE, once the new slot is visible
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10731) * any mappings that might be created in it will consume the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10732) * properties of the new slot and do not need to be updated here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10733) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10734) * When PML is enabled, the kvm_x86_ops dirty logging hooks are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10735) * called to enable/disable dirty logging.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10736) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10737) * When disabling dirty logging with PML enabled, the D-bit is set
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10738) * for sptes in the slot in order to prevent unnecessary GPA
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10739) * logging in the PML buffer (and potential PML buffer full VMEXIT).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10740) * This guarantees leaving PML enabled for the guest's lifetime
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10741) * won't have any additional overhead from PML when the guest is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10742) * running with dirty logging disabled.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10743) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10744) * When enabling dirty logging, large sptes are write-protected
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10745) * so they can be split on first write. New large sptes cannot
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10746) * be created for this slot until the end of the logging.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10747) * See the comments in fast_page_fault().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10748) * For small sptes, nothing is done if the dirty log is in the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10749) * initial-all-set state. Otherwise, depending on whether pml
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10750) * is enabled the D-bit or the W-bit will be cleared.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10751) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10752) if (new->flags & KVM_MEM_LOG_DIRTY_PAGES) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10753) if (kvm_x86_ops.slot_enable_log_dirty) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10754) kvm_x86_ops.slot_enable_log_dirty(kvm, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10755) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10756) int level =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10757) kvm_dirty_log_manual_protect_and_init_set(kvm) ?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10758) PG_LEVEL_2M : PG_LEVEL_4K;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10759)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10760) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10761) * If we're with initial-all-set, we don't need
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10762) * to write protect any small page because
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10763) * they're reported as dirty already. However
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10764) * we still need to write-protect huge pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10765) * so that the page split can happen lazily on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10766) * the first write to the huge page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10767) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10768) kvm_mmu_slot_remove_write_access(kvm, new, level);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10769) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10770) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10771) if (kvm_x86_ops.slot_disable_log_dirty)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10772) kvm_x86_ops.slot_disable_log_dirty(kvm, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10773) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10774) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10775)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10776) void kvm_arch_commit_memory_region(struct kvm *kvm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10777) const struct kvm_userspace_memory_region *mem,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10778) struct kvm_memory_slot *old,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10779) const struct kvm_memory_slot *new,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10780) enum kvm_mr_change change)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10781) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10782) if (!kvm->arch.n_requested_mmu_pages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10783) kvm_mmu_change_mmu_pages(kvm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10784) kvm_mmu_calculate_default_mmu_pages(kvm));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10785)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10786) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10787) * FIXME: const-ify all uses of struct kvm_memory_slot.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10788) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10789) kvm_mmu_slot_apply_flags(kvm, old, (struct kvm_memory_slot *) new, change);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10790)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10791) /* Free the arrays associated with the old memslot. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10792) if (change == KVM_MR_MOVE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10793) kvm_arch_free_memslot(kvm, old);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10794) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10795)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10796) void kvm_arch_flush_shadow_all(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10797) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10798) kvm_mmu_zap_all(kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10799) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10800)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10801) void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10802) struct kvm_memory_slot *slot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10803) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10804) kvm_page_track_flush_slot(kvm, slot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10805) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10806)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10807) static inline bool kvm_guest_apic_has_interrupt(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10808) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10809) return (is_guest_mode(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10810) kvm_x86_ops.guest_apic_has_interrupt &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10811) kvm_x86_ops.guest_apic_has_interrupt(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10812) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10813)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10814) static inline bool kvm_vcpu_has_events(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10815) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10816) if (!list_empty_careful(&vcpu->async_pf.done))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10817) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10818)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10819) if (kvm_apic_has_events(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10820) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10821)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10822) if (vcpu->arch.pv.pv_unhalted)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10823) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10824)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10825) if (vcpu->arch.exception.pending)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10826) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10827)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10828) if (kvm_test_request(KVM_REQ_NMI, vcpu) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10829) (vcpu->arch.nmi_pending &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10830) kvm_x86_ops.nmi_allowed(vcpu, false)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10831) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10832)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10833) if (kvm_test_request(KVM_REQ_SMI, vcpu) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10834) (vcpu->arch.smi_pending &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10835) kvm_x86_ops.smi_allowed(vcpu, false)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10836) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10837)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10838) if (kvm_arch_interrupt_allowed(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10839) (kvm_cpu_has_interrupt(vcpu) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10840) kvm_guest_apic_has_interrupt(vcpu)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10841) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10842)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10843) if (kvm_hv_has_stimer_pending(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10844) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10845)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10846) if (is_guest_mode(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10847) kvm_x86_ops.nested_ops->hv_timer_pending &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10848) kvm_x86_ops.nested_ops->hv_timer_pending(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10849) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10850)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10851) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10852) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10853)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10854) int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10855) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10856) return kvm_vcpu_running(vcpu) || kvm_vcpu_has_events(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10857) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10858)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10859) bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10860) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10861) if (READ_ONCE(vcpu->arch.pv.pv_unhalted))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10862) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10863)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10864) if (kvm_test_request(KVM_REQ_NMI, vcpu) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10865) kvm_test_request(KVM_REQ_SMI, vcpu) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10866) kvm_test_request(KVM_REQ_EVENT, vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10867) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10868)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10869) if (vcpu->arch.apicv_active && kvm_x86_ops.dy_apicv_has_pending_interrupt(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10870) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10871)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10872) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10873) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10874)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10875) bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10876) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10877) return vcpu->arch.preempted_in_kernel;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10878) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10879)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10880) int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10881) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10882) return kvm_vcpu_exiting_guest_mode(vcpu) == IN_GUEST_MODE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10883) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10884)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10885) int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10886) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10887) return kvm_x86_ops.interrupt_allowed(vcpu, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10888) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10889)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10890) unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10891) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10892) if (is_64_bit_mode(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10893) return kvm_rip_read(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10894) return (u32)(get_segment_base(vcpu, VCPU_SREG_CS) +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10895) kvm_rip_read(vcpu));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10896) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10897) EXPORT_SYMBOL_GPL(kvm_get_linear_rip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10898)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10899) bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10900) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10901) return kvm_get_linear_rip(vcpu) == linear_rip;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10902) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10903) EXPORT_SYMBOL_GPL(kvm_is_linear_rip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10904)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10905) unsigned long kvm_get_rflags(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10906) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10907) unsigned long rflags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10908)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10909) rflags = kvm_x86_ops.get_rflags(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10910) if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10911) rflags &= ~X86_EFLAGS_TF;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10912) return rflags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10913) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10914) EXPORT_SYMBOL_GPL(kvm_get_rflags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10915)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10916) static void __kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10917) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10918) if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10919) kvm_is_linear_rip(vcpu, vcpu->arch.singlestep_rip))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10920) rflags |= X86_EFLAGS_TF;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10921) kvm_x86_ops.set_rflags(vcpu, rflags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10922) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10923)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10924) void kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10925) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10926) __kvm_set_rflags(vcpu, rflags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10927) kvm_make_request(KVM_REQ_EVENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10928) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10929) EXPORT_SYMBOL_GPL(kvm_set_rflags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10930)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10931) void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu, struct kvm_async_pf *work)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10932) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10933) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10934)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10935) if ((vcpu->arch.mmu->direct_map != work->arch.direct_map) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10936) work->wakeup_all)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10937) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10938)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10939) r = kvm_mmu_reload(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10940) if (unlikely(r))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10941) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10942)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10943) if (!vcpu->arch.mmu->direct_map &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10944) work->arch.cr3 != vcpu->arch.mmu->get_guest_pgd(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10945) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10946)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10947) kvm_mmu_do_page_fault(vcpu, work->cr2_or_gpa, 0, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10948) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10949)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10950) static inline u32 kvm_async_pf_hash_fn(gfn_t gfn)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10951) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10952) BUILD_BUG_ON(!is_power_of_2(ASYNC_PF_PER_VCPU));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10953)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10954) return hash_32(gfn & 0xffffffff, order_base_2(ASYNC_PF_PER_VCPU));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10955) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10956)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10957) static inline u32 kvm_async_pf_next_probe(u32 key)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10958) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10959) return (key + 1) & (ASYNC_PF_PER_VCPU - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10960) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10961)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10962) static void kvm_add_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10963) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10964) u32 key = kvm_async_pf_hash_fn(gfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10965)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10966) while (vcpu->arch.apf.gfns[key] != ~0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10967) key = kvm_async_pf_next_probe(key);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10968)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10969) vcpu->arch.apf.gfns[key] = gfn;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10970) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10971)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10972) static u32 kvm_async_pf_gfn_slot(struct kvm_vcpu *vcpu, gfn_t gfn)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10973) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10974) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10975) u32 key = kvm_async_pf_hash_fn(gfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10976)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10977) for (i = 0; i < ASYNC_PF_PER_VCPU &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10978) (vcpu->arch.apf.gfns[key] != gfn &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10979) vcpu->arch.apf.gfns[key] != ~0); i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10980) key = kvm_async_pf_next_probe(key);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10981)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10982) return key;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10983) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10984)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10985) bool kvm_find_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10986) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10987) return vcpu->arch.apf.gfns[kvm_async_pf_gfn_slot(vcpu, gfn)] == gfn;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10988) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10989)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10990) static void kvm_del_async_pf_gfn(struct kvm_vcpu *vcpu, gfn_t gfn)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10991) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10992) u32 i, j, k;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10993)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10994) i = j = kvm_async_pf_gfn_slot(vcpu, gfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10995)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10996) if (WARN_ON_ONCE(vcpu->arch.apf.gfns[i] != gfn))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10997) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10998)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10999) while (true) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11000) vcpu->arch.apf.gfns[i] = ~0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11001) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11002) j = kvm_async_pf_next_probe(j);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11003) if (vcpu->arch.apf.gfns[j] == ~0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11004) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11005) k = kvm_async_pf_hash_fn(vcpu->arch.apf.gfns[j]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11006) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11007) * k lies cyclically in ]i,j]
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11008) * | i.k.j |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11009) * |....j i.k.| or |.k..j i...|
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11010) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11011) } while ((i <= j) ? (i < k && k <= j) : (i < k || k <= j));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11012) vcpu->arch.apf.gfns[i] = vcpu->arch.apf.gfns[j];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11013) i = j;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11014) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11015) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11016)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11017) static inline int apf_put_user_notpresent(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11018) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11019) u32 reason = KVM_PV_REASON_PAGE_NOT_PRESENT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11020)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11021) return kvm_write_guest_cached(vcpu->kvm, &vcpu->arch.apf.data, &reason,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11022) sizeof(reason));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11023) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11024)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11025) static inline int apf_put_user_ready(struct kvm_vcpu *vcpu, u32 token)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11026) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11027) unsigned int offset = offsetof(struct kvm_vcpu_pv_apf_data, token);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11028)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11029) return kvm_write_guest_offset_cached(vcpu->kvm, &vcpu->arch.apf.data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11030) &token, offset, sizeof(token));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11031) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11032)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11033) static inline bool apf_pageready_slot_free(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11034) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11035) unsigned int offset = offsetof(struct kvm_vcpu_pv_apf_data, token);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11036) u32 val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11037)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11038) if (kvm_read_guest_offset_cached(vcpu->kvm, &vcpu->arch.apf.data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11039) &val, offset, sizeof(val)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11040) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11041)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11042) return !val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11043) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11044)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11045) static bool kvm_can_deliver_async_pf(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11046) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11047) if (!vcpu->arch.apf.delivery_as_pf_vmexit && is_guest_mode(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11048) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11049)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11050) if (!kvm_pv_async_pf_enabled(vcpu) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11051) (vcpu->arch.apf.send_user_only && kvm_x86_ops.get_cpl(vcpu) == 0))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11052) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11053)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11054) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11055) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11056)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11057) bool kvm_can_do_async_pf(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11058) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11059) if (unlikely(!lapic_in_kernel(vcpu) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11060) kvm_event_needs_reinjection(vcpu) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11061) vcpu->arch.exception.pending))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11062) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11063)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11064) if (kvm_hlt_in_guest(vcpu->kvm) && !kvm_can_deliver_async_pf(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11065) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11066)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11067) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11068) * If interrupts are off we cannot even use an artificial
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11069) * halt state.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11070) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11071) return kvm_arch_interrupt_allowed(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11072) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11073)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11074) bool kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11075) struct kvm_async_pf *work)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11076) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11077) struct x86_exception fault;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11078)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11079) trace_kvm_async_pf_not_present(work->arch.token, work->cr2_or_gpa);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11080) kvm_add_async_pf_gfn(vcpu, work->arch.gfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11081)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11082) if (kvm_can_deliver_async_pf(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11083) !apf_put_user_notpresent(vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11084) fault.vector = PF_VECTOR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11085) fault.error_code_valid = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11086) fault.error_code = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11087) fault.nested_page_fault = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11088) fault.address = work->arch.token;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11089) fault.async_page_fault = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11090) kvm_inject_page_fault(vcpu, &fault);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11091) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11092) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11093) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11094) * It is not possible to deliver a paravirtualized asynchronous
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11095) * page fault, but putting the guest in an artificial halt state
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11096) * can be beneficial nevertheless: if an interrupt arrives, we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11097) * can deliver it timely and perhaps the guest will schedule
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11098) * another process. When the instruction that triggered a page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11099) * fault is retried, hopefully the page will be ready in the host.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11100) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11101) kvm_make_request(KVM_REQ_APF_HALT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11102) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11103) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11104) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11105)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11106) void kvm_arch_async_page_present(struct kvm_vcpu *vcpu,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11107) struct kvm_async_pf *work)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11108) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11109) struct kvm_lapic_irq irq = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11110) .delivery_mode = APIC_DM_FIXED,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11111) .vector = vcpu->arch.apf.vec
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11112) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11113)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11114) if (work->wakeup_all)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11115) work->arch.token = ~0; /* broadcast wakeup */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11116) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11117) kvm_del_async_pf_gfn(vcpu, work->arch.gfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11118) trace_kvm_async_pf_ready(work->arch.token, work->cr2_or_gpa);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11119)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11120) if ((work->wakeup_all || work->notpresent_injected) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11121) kvm_pv_async_pf_enabled(vcpu) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11122) !apf_put_user_ready(vcpu, work->arch.token)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11123) vcpu->arch.apf.pageready_pending = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11124) kvm_apic_set_irq(vcpu, &irq, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11125) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11126)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11127) vcpu->arch.apf.halted = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11128) vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11129) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11130)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11131) void kvm_arch_async_page_present_queued(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11132) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11133) kvm_make_request(KVM_REQ_APF_READY, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11134) if (!vcpu->arch.apf.pageready_pending)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11135) kvm_vcpu_kick(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11136) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11137)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11138) bool kvm_arch_can_dequeue_async_page_present(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11139) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11140) if (!kvm_pv_async_pf_enabled(vcpu))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11141) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11142) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11143) return apf_pageready_slot_free(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11144) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11145)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11146) void kvm_arch_start_assignment(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11147) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11148) atomic_inc(&kvm->arch.assigned_device_count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11149) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11150) EXPORT_SYMBOL_GPL(kvm_arch_start_assignment);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11151)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11152) void kvm_arch_end_assignment(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11153) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11154) atomic_dec(&kvm->arch.assigned_device_count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11155) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11156) EXPORT_SYMBOL_GPL(kvm_arch_end_assignment);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11157)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11158) bool kvm_arch_has_assigned_device(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11159) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11160) return atomic_read(&kvm->arch.assigned_device_count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11161) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11162) EXPORT_SYMBOL_GPL(kvm_arch_has_assigned_device);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11163)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11164) void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11165) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11166) atomic_inc(&kvm->arch.noncoherent_dma_count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11167) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11168) EXPORT_SYMBOL_GPL(kvm_arch_register_noncoherent_dma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11169)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11170) void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11171) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11172) atomic_dec(&kvm->arch.noncoherent_dma_count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11173) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11174) EXPORT_SYMBOL_GPL(kvm_arch_unregister_noncoherent_dma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11175)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11176) bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11177) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11178) return atomic_read(&kvm->arch.noncoherent_dma_count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11179) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11180) EXPORT_SYMBOL_GPL(kvm_arch_has_noncoherent_dma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11181)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11182) bool kvm_arch_has_irq_bypass(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11183) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11184) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11185) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11186)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11187) int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *cons,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11188) struct irq_bypass_producer *prod)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11189) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11190) struct kvm_kernel_irqfd *irqfd =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11191) container_of(cons, struct kvm_kernel_irqfd, consumer);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11192) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11193)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11194) irqfd->producer = prod;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11195) kvm_arch_start_assignment(irqfd->kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11196) ret = kvm_x86_ops.update_pi_irte(irqfd->kvm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11197) prod->irq, irqfd->gsi, 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11198)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11199) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11200) kvm_arch_end_assignment(irqfd->kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11201)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11202) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11203) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11204)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11205) void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *cons,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11206) struct irq_bypass_producer *prod)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11207) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11208) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11209) struct kvm_kernel_irqfd *irqfd =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11210) container_of(cons, struct kvm_kernel_irqfd, consumer);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11211)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11212) WARN_ON(irqfd->producer != prod);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11213) irqfd->producer = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11214)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11215) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11216) * When producer of consumer is unregistered, we change back to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11217) * remapped mode, so we can re-use the current implementation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11218) * when the irq is masked/disabled or the consumer side (KVM
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11219) * int this case doesn't want to receive the interrupts.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11220) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11221) ret = kvm_x86_ops.update_pi_irte(irqfd->kvm, prod->irq, irqfd->gsi, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11222) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11223) printk(KERN_INFO "irq bypass consumer (token %p) unregistration"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11224) " fails: %d\n", irqfd->consumer.token, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11225)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11226) kvm_arch_end_assignment(irqfd->kvm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11227) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11228)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11229) int kvm_arch_update_irqfd_routing(struct kvm *kvm, unsigned int host_irq,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11230) uint32_t guest_irq, bool set)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11231) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11232) return kvm_x86_ops.update_pi_irte(kvm, host_irq, guest_irq, set);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11233) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11234)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11235) bool kvm_vector_hashing_enabled(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11236) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11237) return vector_hashing;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11238) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11239)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11240) bool kvm_arch_no_poll(struct kvm_vcpu *vcpu)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11241) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11242) return (vcpu->arch.msr_kvm_poll_control & 1) == 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11243) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11244) EXPORT_SYMBOL_GPL(kvm_arch_no_poll);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11245)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11246)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11247) int kvm_spec_ctrl_test_value(u64 value)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11248) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11249) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11250) * test that setting IA32_SPEC_CTRL to given value
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11251) * is allowed by the host processor
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11252) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11253)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11254) u64 saved_value;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11255) unsigned long flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11256) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11257)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11258) local_irq_save(flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11259)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11260) if (rdmsrl_safe(MSR_IA32_SPEC_CTRL, &saved_value))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11261) ret = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11262) else if (wrmsrl_safe(MSR_IA32_SPEC_CTRL, value))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11263) ret = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11264) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11265) wrmsrl(MSR_IA32_SPEC_CTRL, saved_value);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11266)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11267) local_irq_restore(flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11268)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11269) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11270) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11271) EXPORT_SYMBOL_GPL(kvm_spec_ctrl_test_value);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11272)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11273) void kvm_fixup_and_inject_pf_error(struct kvm_vcpu *vcpu, gva_t gva, u16 error_code)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11274) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11275) struct x86_exception fault;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11276) u32 access = error_code &
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11277) (PFERR_WRITE_MASK | PFERR_FETCH_MASK | PFERR_USER_MASK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11278)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11279) if (!(error_code & PFERR_PRESENT_MASK) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11280) vcpu->arch.walk_mmu->gva_to_gpa(vcpu, gva, access, &fault) != UNMAPPED_GVA) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11281) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11282) * If vcpu->arch.walk_mmu->gva_to_gpa succeeded, the page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11283) * tables probably do not match the TLB. Just proceed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11284) * with the error code that the processor gave.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11285) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11286) fault.vector = PF_VECTOR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11287) fault.error_code_valid = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11288) fault.error_code = error_code;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11289) fault.nested_page_fault = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11290) fault.address = gva;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11291) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11292) vcpu->arch.walk_mmu->inject_page_fault(vcpu, &fault);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11293) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11294) EXPORT_SYMBOL_GPL(kvm_fixup_and_inject_pf_error);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11295)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11296) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11297) * Handles kvm_read/write_guest_virt*() result and either injects #PF or returns
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11298) * KVM_EXIT_INTERNAL_ERROR for cases not currently handled by KVM. Return value
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11299) * indicates whether exit to userspace is needed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11300) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11301) int kvm_handle_memory_failure(struct kvm_vcpu *vcpu, int r,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11302) struct x86_exception *e)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11303) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11304) if (r == X86EMUL_PROPAGATE_FAULT) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11305) kvm_inject_emulated_page_fault(vcpu, e);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11306) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11307) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11308)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11309) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11310) * In case kvm_read/write_guest_virt*() failed with X86EMUL_IO_NEEDED
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11311) * while handling a VMX instruction KVM could've handled the request
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11312) * correctly by exiting to userspace and performing I/O but there
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11313) * doesn't seem to be a real use-case behind such requests, just return
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11314) * KVM_EXIT_INTERNAL_ERROR for now.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11315) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11316) vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11317) vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_EMULATION;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11318) vcpu->run->internal.ndata = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11319)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11320) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11321) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11322) EXPORT_SYMBOL_GPL(kvm_handle_memory_failure);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11323)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11324) int kvm_handle_invpcid(struct kvm_vcpu *vcpu, unsigned long type, gva_t gva)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11325) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11326) bool pcid_enabled;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11327) struct x86_exception e;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11328) unsigned i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11329) unsigned long roots_to_free = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11330) struct {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11331) u64 pcid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11332) u64 gla;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11333) } operand;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11334) int r;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11335)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11336) r = kvm_read_guest_virt(vcpu, gva, &operand, sizeof(operand), &e);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11337) if (r != X86EMUL_CONTINUE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11338) return kvm_handle_memory_failure(vcpu, r, &e);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11339)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11340) if (operand.pcid >> 12 != 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11341) kvm_inject_gp(vcpu, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11342) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11343) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11344)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11345) pcid_enabled = kvm_read_cr4_bits(vcpu, X86_CR4_PCIDE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11346)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11347) switch (type) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11348) case INVPCID_TYPE_INDIV_ADDR:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11349) if ((!pcid_enabled && (operand.pcid != 0)) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11350) is_noncanonical_address(operand.gla, vcpu)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11351) kvm_inject_gp(vcpu, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11352) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11353) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11354) kvm_mmu_invpcid_gva(vcpu, operand.gla, operand.pcid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11355) return kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11356)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11357) case INVPCID_TYPE_SINGLE_CTXT:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11358) if (!pcid_enabled && (operand.pcid != 0)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11359) kvm_inject_gp(vcpu, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11360) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11361) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11362)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11363) if (kvm_get_active_pcid(vcpu) == operand.pcid) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11364) kvm_mmu_sync_roots(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11365) kvm_make_request(KVM_REQ_TLB_FLUSH_CURRENT, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11366) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11367)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11368) for (i = 0; i < KVM_MMU_NUM_PREV_ROOTS; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11369) if (kvm_get_pcid(vcpu, vcpu->arch.mmu->prev_roots[i].pgd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11370) == operand.pcid)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11371) roots_to_free |= KVM_MMU_ROOT_PREVIOUS(i);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11372)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11373) kvm_mmu_free_roots(vcpu, vcpu->arch.mmu, roots_to_free);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11374) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11375) * If neither the current cr3 nor any of the prev_roots use the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11376) * given PCID, then nothing needs to be done here because a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11377) * resync will happen anyway before switching to any other CR3.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11378) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11379)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11380) return kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11381)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11382) case INVPCID_TYPE_ALL_NON_GLOBAL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11383) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11384) * Currently, KVM doesn't mark global entries in the shadow
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11385) * page tables, so a non-global flush just degenerates to a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11386) * global flush. If needed, we could optimize this later by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11387) * keeping track of global entries in shadow page tables.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11388) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11389)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11390) fallthrough;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11391) case INVPCID_TYPE_ALL_INCL_GLOBAL:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11392) kvm_make_request(KVM_REQ_MMU_RELOAD, vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11393) return kvm_skip_emulated_instruction(vcpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11394)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11395) default:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11396) BUG(); /* We have already checked above that type <= 3 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11397) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11398) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11399) EXPORT_SYMBOL_GPL(kvm_handle_invpcid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11400)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11401) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_exit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11402) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_fast_mmio);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11403) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_inj_virq);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11404) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_page_fault);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11405) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_msr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11406) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_cr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11407) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_nested_vmrun);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11408) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_nested_vmexit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11409) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_nested_vmexit_inject);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11410) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_nested_intr_vmexit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11411) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_nested_vmenter_failed);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11412) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_invlpga);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11413) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_skinit);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11414) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_nested_intercepts);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11415) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_write_tsc_offset);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11416) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_ple_window_update);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11417) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_pml_full);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11418) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_pi_irte_update);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11419) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_avic_unaccelerated_access);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11420) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_avic_incomplete_ipi);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11421) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_avic_ga_log);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11422) EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_apicv_update_request);
Orange Pi5 kernel
Deprecated Linux kernel 5.10.110 for OrangePi 5/5B/5+ boards
3 Commits
0 Branches
0 Tags