^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) * linux/mm/memory.c
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5) * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9) * demand-loading started 01.12.91 - seems it is high on the list of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10) * things wanted, and it should be easy to implement. - Linus
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 12)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 13) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 14) * Ok, demand-loading was easy, shared pages a little bit tricker. Shared
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 15) * pages started 02.12.91, seems to work. - Linus.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 16) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 17) * Tested sharing by executing about 30 /bin/sh: under the old kernel it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 18) * would have taken more than the 6M I have free, but it worked well as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 19) * far as I could see.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 20) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 21) * Also corrected some "invalidate()"s - I wasn't doing enough of them.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 22) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 23)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 24) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 25) * Real VM (paging to/from disk) started 18.12.91. Much more work and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 26) * thought has to go into this. Oh, well..
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 27) * 19.12.91 - works, somewhat. Sometimes I get faults, don't know why.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 28) * Found it. Everything seems to work now.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 29) * 20.12.91 - Ok, making the swap-device changeable like the root.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 30) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 31)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 32) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 33) * 05.04.94 - Multi-page memory management added for v1.1.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 34) * Idea by Alex Bligh (alex@cconcepts.co.uk)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 35) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 36) * 16.07.99 - Support of BIGMEM added by Gerhard Wichert, Siemens AG
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 37) * (Gerhard.Wichert@pdb.siemens.de)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 38) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 39) * Aug/Sep 2004 Changed to four level page tables (Andi Kleen)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 40) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 41)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 42) #include <linux/kernel_stat.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 43) #include <linux/mm.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 44) #include <linux/sched/mm.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 45) #include <linux/sched/coredump.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 46) #include <linux/sched/numa_balancing.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 47) #include <linux/sched/task.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 48) #include <linux/hugetlb.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 49) #include <linux/mman.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 50) #include <linux/swap.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 51) #include <linux/highmem.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 52) #include <linux/pagemap.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 53) #include <linux/memremap.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 54) #include <linux/ksm.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 55) #include <linux/rmap.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 56) #include <linux/export.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 57) #include <linux/delayacct.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 58) #include <linux/init.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 59) #include <linux/pfn_t.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 60) #include <linux/writeback.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 61) #include <linux/memcontrol.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 62) #include <linux/mmu_notifier.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 63) #include <linux/swapops.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 64) #include <linux/elf.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 65) #include <linux/gfp.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 66) #include <linux/migrate.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 67) #include <linux/string.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 68) #include <linux/debugfs.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 69) #include <linux/userfaultfd_k.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 70) #include <linux/dax.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 71) #include <linux/oom.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 72) #include <linux/numa.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 73) #include <linux/perf_event.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 74) #include <linux/ptrace.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 75) #include <linux/vmalloc.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 76) #include <trace/hooks/mm.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 77)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 78) #include <trace/events/kmem.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 79)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 80) #include <asm/io.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 81) #include <asm/mmu_context.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 82) #include <asm/pgalloc.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 83) #include <linux/uaccess.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 84) #include <asm/tlb.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 85) #include <asm/tlbflush.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 86)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 87) #include "pgalloc-track.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 88) #include "internal.h"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 89) #include <trace/hooks/mm.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 90)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 91) #define CREATE_TRACE_POINTS
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 92) #include <trace/events/pagefault.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 93)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 94) #if defined(LAST_CPUPID_NOT_IN_PAGE_FLAGS) && !defined(CONFIG_COMPILE_TEST)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 95) #warning Unfortunate NUMA and NUMA Balancing config, growing page-frame for last_cpupid.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 96) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 97)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 98) #ifndef CONFIG_NEED_MULTIPLE_NODES
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 99) /* use the per-pgdat data instead for discontigmem - mbligh */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 100) unsigned long max_mapnr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 101) EXPORT_SYMBOL(max_mapnr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 102)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 103) struct page *mem_map;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 104) EXPORT_SYMBOL(mem_map);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 105) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 106)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 107) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 108) * A number of key systems in x86 including ioremap() rely on the assumption
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 109) * that high_memory defines the upper bound on direct map memory, then end
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 110) * of ZONE_NORMAL. Under CONFIG_DISCONTIG this means that max_low_pfn and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 111) * highstart_pfn must be the same; there must be no gap between ZONE_NORMAL
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 112) * and ZONE_HIGHMEM.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 113) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 114) void *high_memory;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 115) EXPORT_SYMBOL(high_memory);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 116)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 117) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 118) * Randomize the address space (stacks, mmaps, brk, etc.).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 119) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 120) * ( When CONFIG_COMPAT_BRK=y we exclude brk from randomization,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 121) * as ancient (libc5 based) binaries can segfault. )
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 122) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 123) int randomize_va_space __read_mostly =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 124) #ifdef CONFIG_COMPAT_BRK
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 125) 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 126) #else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 127) 2;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 128) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 129)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 130) #ifndef arch_faults_on_old_pte
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 131) static inline bool arch_faults_on_old_pte(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 132) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 133) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 134) * Those arches which don't have hw access flag feature need to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 135) * implement their own helper. By default, "true" means pagefault
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 136) * will be hit on old pte.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 137) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 138) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 139) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 140) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 141)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 142) #ifndef arch_wants_old_prefaulted_pte
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 143) static inline bool arch_wants_old_prefaulted_pte(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 144) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 145) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 146) * Transitioning a PTE from 'old' to 'young' can be expensive on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 147) * some architectures, even if it's performed in hardware. By
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 148) * default, "false" means prefaulted entries will be 'young'.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 149) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 150) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 151) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 152) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 153)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 154) static int __init disable_randmaps(char *s)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 155) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 156) randomize_va_space = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 157) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 158) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 159) __setup("norandmaps", disable_randmaps);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 160)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 161) unsigned long zero_pfn __read_mostly;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 162) EXPORT_SYMBOL(zero_pfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 163)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 164) unsigned long highest_memmap_pfn __read_mostly;
^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) * CONFIG_MMU architectures set up ZERO_PAGE in their paging_init()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 168) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 169) static int __init init_zero_pfn(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 170) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 171) zero_pfn = page_to_pfn(ZERO_PAGE(0));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 172) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 173) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 174) early_initcall(init_zero_pfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 175)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 176) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 177) * Only trace rss_stat when there is a 512kb cross over.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 178) * Smaller changes may be lost unless every small change is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 179) * crossing into or returning to a 512kb boundary.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 180) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 181) #define TRACE_MM_COUNTER_THRESHOLD 128
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 182)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 183) void mm_trace_rss_stat(struct mm_struct *mm, int member, long count,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 184) long value)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 185) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 186) long thresh_mask = ~(TRACE_MM_COUNTER_THRESHOLD - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 187)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 188) /* Threshold roll-over, trace it */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 189) if ((count & thresh_mask) != ((count - value) & thresh_mask))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 190) trace_rss_stat(mm, member, count);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 191) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 192) EXPORT_SYMBOL_GPL(mm_trace_rss_stat);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 193)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 194) #if defined(SPLIT_RSS_COUNTING)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 195)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 196) void sync_mm_rss(struct mm_struct *mm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 197) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 198) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 199)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 200) for (i = 0; i < NR_MM_COUNTERS; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 201) if (current->rss_stat.count[i]) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 202) add_mm_counter(mm, i, current->rss_stat.count[i]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 203) current->rss_stat.count[i] = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 204) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 205) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 206) current->rss_stat.events = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 207) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 208)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 209) static void add_mm_counter_fast(struct mm_struct *mm, int member, int val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 210) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 211) struct task_struct *task = current;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 212)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 213) if (likely(task->mm == mm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 214) task->rss_stat.count[member] += val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 215) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 216) add_mm_counter(mm, member, val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 217) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 218) #define inc_mm_counter_fast(mm, member) add_mm_counter_fast(mm, member, 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 219) #define dec_mm_counter_fast(mm, member) add_mm_counter_fast(mm, member, -1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 220)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 221) /* sync counter once per 64 page faults */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 222) #define TASK_RSS_EVENTS_THRESH (64)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 223) static void check_sync_rss_stat(struct task_struct *task)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 224) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 225) if (unlikely(task != current))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 226) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 227) if (unlikely(task->rss_stat.events++ > TASK_RSS_EVENTS_THRESH))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 228) sync_mm_rss(task->mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 229) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 230) #else /* SPLIT_RSS_COUNTING */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 231)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 232) #define inc_mm_counter_fast(mm, member) inc_mm_counter(mm, member)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 233) #define dec_mm_counter_fast(mm, member) dec_mm_counter(mm, member)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 234)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 235) static void check_sync_rss_stat(struct task_struct *task)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 236) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 237) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 238)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 239) #endif /* SPLIT_RSS_COUNTING */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 240)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 241) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 242) * Note: this doesn't free the actual pages themselves. That
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 243) * has been handled earlier when unmapping all the memory regions.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 244) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 245) static void free_pte_range(struct mmu_gather *tlb, pmd_t *pmd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 246) unsigned long addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 247) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 248) pgtable_t token = pmd_pgtable(*pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 249) pmd_clear(pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 250) pte_free_tlb(tlb, token, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 251) mm_dec_nr_ptes(tlb->mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 252) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 253)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 254) static inline void free_pmd_range(struct mmu_gather *tlb, pud_t *pud,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 255) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 256) unsigned long floor, unsigned long ceiling)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 257) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 258) pmd_t *pmd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 259) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 260) unsigned long start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 261)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 262) start = addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 263) pmd = pmd_offset(pud, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 264) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 265) next = pmd_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 266) if (pmd_none_or_clear_bad(pmd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 267) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 268) free_pte_range(tlb, pmd, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 269) } while (pmd++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 270)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 271) start &= PUD_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 272) if (start < floor)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 273) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 274) if (ceiling) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 275) ceiling &= PUD_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 276) if (!ceiling)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 277) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 278) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 279) if (end - 1 > ceiling - 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 280) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 281)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 282) pmd = pmd_offset(pud, start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 283) pud_clear(pud);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 284) pmd_free_tlb(tlb, pmd, start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 285) mm_dec_nr_pmds(tlb->mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 286) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 287)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 288) static inline void free_pud_range(struct mmu_gather *tlb, p4d_t *p4d,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 289) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 290) unsigned long floor, unsigned long ceiling)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 291) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 292) pud_t *pud;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 293) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 294) unsigned long start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 295)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 296) start = addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 297) pud = pud_offset(p4d, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 298) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 299) next = pud_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 300) if (pud_none_or_clear_bad(pud))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 301) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 302) free_pmd_range(tlb, pud, addr, next, floor, ceiling);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 303) } while (pud++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 304)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 305) start &= P4D_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 306) if (start < floor)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 307) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 308) if (ceiling) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 309) ceiling &= P4D_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 310) if (!ceiling)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 311) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 312) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 313) if (end - 1 > ceiling - 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 314) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 315)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 316) pud = pud_offset(p4d, start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 317) p4d_clear(p4d);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 318) pud_free_tlb(tlb, pud, start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 319) mm_dec_nr_puds(tlb->mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 320) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 321)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 322) static inline void free_p4d_range(struct mmu_gather *tlb, pgd_t *pgd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 323) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 324) unsigned long floor, unsigned long ceiling)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 325) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 326) p4d_t *p4d;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 327) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 328) unsigned long start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 329)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 330) start = addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 331) p4d = p4d_offset(pgd, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 332) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 333) next = p4d_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 334) if (p4d_none_or_clear_bad(p4d))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 335) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 336) free_pud_range(tlb, p4d, addr, next, floor, ceiling);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 337) } while (p4d++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 338)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 339) start &= PGDIR_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 340) if (start < floor)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 341) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 342) if (ceiling) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 343) ceiling &= PGDIR_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 344) if (!ceiling)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 345) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 346) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 347) if (end - 1 > ceiling - 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 348) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 349)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 350) p4d = p4d_offset(pgd, start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 351) pgd_clear(pgd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 352) p4d_free_tlb(tlb, p4d, start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 353) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 354)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 355) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 356) * This function frees user-level page tables of a process.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 357) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 358) void free_pgd_range(struct mmu_gather *tlb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 359) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 360) unsigned long floor, unsigned long ceiling)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 361) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 362) pgd_t *pgd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 363) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 364)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 365) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 366) * The next few lines have given us lots of grief...
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 367) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 368) * Why are we testing PMD* at this top level? Because often
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 369) * there will be no work to do at all, and we'd prefer not to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 370) * go all the way down to the bottom just to discover that.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 371) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 372) * Why all these "- 1"s? Because 0 represents both the bottom
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 373) * of the address space and the top of it (using -1 for the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 374) * top wouldn't help much: the masks would do the wrong thing).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 375) * The rule is that addr 0 and floor 0 refer to the bottom of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 376) * the address space, but end 0 and ceiling 0 refer to the top
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 377) * Comparisons need to use "end - 1" and "ceiling - 1" (though
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 378) * that end 0 case should be mythical).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 379) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 380) * Wherever addr is brought up or ceiling brought down, we must
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 381) * be careful to reject "the opposite 0" before it confuses the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 382) * subsequent tests. But what about where end is brought down
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 383) * by PMD_SIZE below? no, end can't go down to 0 there.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 384) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 385) * Whereas we round start (addr) and ceiling down, by different
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 386) * masks at different levels, in order to test whether a table
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 387) * now has no other vmas using it, so can be freed, we don't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 388) * bother to round floor or end up - the tests don't need that.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 389) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 390)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 391) addr &= PMD_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 392) if (addr < floor) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 393) addr += PMD_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 394) if (!addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 395) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 396) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 397) if (ceiling) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 398) ceiling &= PMD_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 399) if (!ceiling)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 400) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 401) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 402) if (end - 1 > ceiling - 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 403) end -= PMD_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 404) if (addr > end - 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 405) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 406) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 407) * We add page table cache pages with PAGE_SIZE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 408) * (see pte_free_tlb()), flush the tlb if we need
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 409) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 410) tlb_change_page_size(tlb, PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 411) pgd = pgd_offset(tlb->mm, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 412) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 413) next = pgd_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 414) if (pgd_none_or_clear_bad(pgd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 415) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 416) free_p4d_range(tlb, pgd, addr, next, floor, ceiling);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 417) } while (pgd++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 418) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 419)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 420) void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 421) unsigned long floor, unsigned long ceiling)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 422) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 423) while (vma) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 424) struct vm_area_struct *next = vma->vm_next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 425) unsigned long addr = vma->vm_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 426)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 427) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 428) * Hide vma from rmap and truncate_pagecache before freeing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 429) * pgtables
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 430) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 431) vm_write_begin(vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 432) unlink_anon_vmas(vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 433) vm_write_end(vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 434) unlink_file_vma(vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 435)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 436) if (is_vm_hugetlb_page(vma)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 437) hugetlb_free_pgd_range(tlb, addr, vma->vm_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 438) floor, next ? next->vm_start : ceiling);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 439) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 440) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 441) * Optimization: gather nearby vmas into one call down
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 442) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 443) while (next && next->vm_start <= vma->vm_end + PMD_SIZE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 444) && !is_vm_hugetlb_page(next)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 445) vma = next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 446) next = vma->vm_next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 447) vm_write_begin(vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 448) unlink_anon_vmas(vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 449) vm_write_end(vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 450) unlink_file_vma(vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 451) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 452) free_pgd_range(tlb, addr, vma->vm_end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 453) floor, next ? next->vm_start : ceiling);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 454) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 455) vma = next;
^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)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 459) int __pte_alloc(struct mm_struct *mm, pmd_t *pmd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 460) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 461) spinlock_t *ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 462) pgtable_t new = pte_alloc_one(mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 463) if (!new)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 464) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 465)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 466) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 467) * Ensure all pte setup (eg. pte page lock and page clearing) are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 468) * visible before the pte is made visible to other CPUs by being
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 469) * put into page tables.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 470) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 471) * The other side of the story is the pointer chasing in the page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 472) * table walking code (when walking the page table without locking;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 473) * ie. most of the time). Fortunately, these data accesses consist
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 474) * of a chain of data-dependent loads, meaning most CPUs (alpha
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 475) * being the notable exception) will already guarantee loads are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 476) * seen in-order. See the alpha page table accessors for the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 477) * smp_rmb() barriers in page table walking code.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 478) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 479) smp_wmb(); /* Could be smp_wmb__xxx(before|after)_spin_lock */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 480)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 481) ptl = pmd_lock(mm, pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 482) if (likely(pmd_none(*pmd))) { /* Has another populated it ? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 483) mm_inc_nr_ptes(mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 484) pmd_populate(mm, pmd, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 485) new = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 486) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 487) spin_unlock(ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 488) if (new)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 489) pte_free(mm, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 490) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 491) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 492)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 493) int __pte_alloc_kernel(pmd_t *pmd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 494) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 495) pte_t *new = pte_alloc_one_kernel(&init_mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 496) if (!new)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 497) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 498)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 499) smp_wmb(); /* See comment in __pte_alloc */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 500)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 501) spin_lock(&init_mm.page_table_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 502) if (likely(pmd_none(*pmd))) { /* Has another populated it ? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 503) pmd_populate_kernel(&init_mm, pmd, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 504) new = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 505) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 506) spin_unlock(&init_mm.page_table_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 507) if (new)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 508) pte_free_kernel(&init_mm, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 509) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 510) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 511)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 512) static inline void init_rss_vec(int *rss)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 513) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 514) memset(rss, 0, sizeof(int) * NR_MM_COUNTERS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 515) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 516)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 517) static inline void add_mm_rss_vec(struct mm_struct *mm, int *rss)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 518) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 519) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 520)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 521) if (current->mm == mm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 522) sync_mm_rss(mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 523) for (i = 0; i < NR_MM_COUNTERS; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 524) if (rss[i])
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 525) add_mm_counter(mm, i, rss[i]);
^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) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 529) * This function is called to print an error when a bad pte
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 530) * is found. For example, we might have a PFN-mapped pte in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 531) * a region that doesn't allow it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 532) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 533) * The calling function must still handle the error.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 534) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 535) static void print_bad_pte(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 536) pte_t pte, struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 537) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 538) pgd_t *pgd = pgd_offset(vma->vm_mm, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 539) p4d_t *p4d = p4d_offset(pgd, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 540) pud_t *pud = pud_offset(p4d, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 541) pmd_t *pmd = pmd_offset(pud, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 542) struct address_space *mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 543) pgoff_t index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 544) static unsigned long resume;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 545) static unsigned long nr_shown;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 546) static unsigned long nr_unshown;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 547)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 548) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 549) * Allow a burst of 60 reports, then keep quiet for that minute;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 550) * or allow a steady drip of one report per second.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 551) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 552) if (nr_shown == 60) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 553) if (time_before(jiffies, resume)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 554) nr_unshown++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 555) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 556) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 557) if (nr_unshown) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 558) pr_alert("BUG: Bad page map: %lu messages suppressed\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 559) nr_unshown);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 560) nr_unshown = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 561) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 562) nr_shown = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 563) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 564) if (nr_shown++ == 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 565) resume = jiffies + 60 * HZ;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 566)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 567) mapping = vma->vm_file ? vma->vm_file->f_mapping : NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 568) index = linear_page_index(vma, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 569)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 570) pr_alert("BUG: Bad page map in process %s pte:%08llx pmd:%08llx\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 571) current->comm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 572) (long long)pte_val(pte), (long long)pmd_val(*pmd));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 573) if (page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 574) dump_page(page, "bad pte");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 575) pr_alert("addr:%px vm_flags:%08lx anon_vma:%px mapping:%px index:%lx\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 576) (void *)addr, READ_ONCE(vma->vm_flags), vma->anon_vma, mapping, index);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 577) pr_alert("file:%pD fault:%ps mmap:%ps readpage:%ps\n",
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 578) vma->vm_file,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 579) vma->vm_ops ? vma->vm_ops->fault : NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 580) vma->vm_file ? vma->vm_file->f_op->mmap : NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 581) mapping ? mapping->a_ops->readpage : NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 582) dump_stack();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 583) add_taint(TAINT_BAD_PAGE, LOCKDEP_NOW_UNRELIABLE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 584) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 585)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 586) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 587) * __vm_normal_page -- This function gets the "struct page" associated with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 588) * a pte.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 589) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 590) * "Special" mappings do not wish to be associated with a "struct page" (either
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 591) * it doesn't exist, or it exists but they don't want to touch it). In this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 592) * case, NULL is returned here. "Normal" mappings do have a struct page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 593) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 594) * There are 2 broad cases. Firstly, an architecture may define a pte_special()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 595) * pte bit, in which case this function is trivial. Secondly, an architecture
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 596) * may not have a spare pte bit, which requires a more complicated scheme,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 597) * described below.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 598) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 599) * A raw VM_PFNMAP mapping (ie. one that is not COWed) is always considered a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 600) * special mapping (even if there are underlying and valid "struct pages").
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 601) * COWed pages of a VM_PFNMAP are always normal.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 602) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 603) * The way we recognize COWed pages within VM_PFNMAP mappings is through the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 604) * rules set up by "remap_pfn_range()": the vma will have the VM_PFNMAP bit
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 605) * set, and the vm_pgoff will point to the first PFN mapped: thus every special
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 606) * mapping will always honor the rule
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 607) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 608) * pfn_of_page == vma->vm_pgoff + ((addr - vma->vm_start) >> PAGE_SHIFT)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 609) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 610) * And for normal mappings this is false.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 611) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 612) * This restricts such mappings to be a linear translation from virtual address
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 613) * to pfn. To get around this restriction, we allow arbitrary mappings so long
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 614) * as the vma is not a COW mapping; in that case, we know that all ptes are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 615) * special (because none can have been COWed).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 616) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 617) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 618) * In order to support COW of arbitrary special mappings, we have VM_MIXEDMAP.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 619) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 620) * VM_MIXEDMAP mappings can likewise contain memory with or without "struct
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 621) * page" backing, however the difference is that _all_ pages with a struct
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 622) * page (that is, those where pfn_valid is true) are refcounted and considered
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 623) * normal pages by the VM. The disadvantage is that pages are refcounted
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 624) * (which can be slower and simply not an option for some PFNMAP users). The
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 625) * advantage is that we don't have to follow the strict linearity rule of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 626) * PFNMAP mappings in order to support COWable mappings.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 627) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 628) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 629) struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 630) pte_t pte, unsigned long vma_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 631) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 632) unsigned long pfn = pte_pfn(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 633)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 634) if (IS_ENABLED(CONFIG_ARCH_HAS_PTE_SPECIAL)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 635) if (likely(!pte_special(pte)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 636) goto check_pfn;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 637) if (vma->vm_ops && vma->vm_ops->find_special_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 638) return vma->vm_ops->find_special_page(vma, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 639) if (vma_flags & (VM_PFNMAP | VM_MIXEDMAP))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 640) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 641) if (is_zero_pfn(pfn))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 642) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 643) if (pte_devmap(pte))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 644) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 645)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 646) print_bad_pte(vma, addr, pte, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 647) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 648) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 649)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 650) /* !CONFIG_ARCH_HAS_PTE_SPECIAL case follows: */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 651) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 652) * This part should never get called when CONFIG_SPECULATIVE_PAGE_FAULT
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 653) * is set. This is mainly because we can't rely on vm_start.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 654) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 655)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 656) if (unlikely(vma_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 657) if (vma_flags & VM_MIXEDMAP) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 658) if (!pfn_valid(pfn))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 659) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 660) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 661) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 662) unsigned long off;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 663) off = (addr - vma->vm_start) >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 664) if (pfn == vma->vm_pgoff + off)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 665) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 666) if (!is_cow_mapping(vma_flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 667) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 668) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 669) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 670)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 671) if (is_zero_pfn(pfn))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 672) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 673)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 674) check_pfn:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 675) if (unlikely(pfn > highest_memmap_pfn)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 676) print_bad_pte(vma, addr, pte, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 677) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 678) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 679)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 680) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 681) * NOTE! We still have PageReserved() pages in the page tables.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 682) * eg. VDSO mappings can cause them to exist.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 683) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 684) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 685) return pfn_to_page(pfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 686) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 687)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 688) #ifdef CONFIG_TRANSPARENT_HUGEPAGE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 689) struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 690) pmd_t pmd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 691) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 692) unsigned long pfn = pmd_pfn(pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 693)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 694) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 695) * There is no pmd_special() but there may be special pmds, e.g.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 696) * in a direct-access (dax) mapping, so let's just replicate the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 697) * !CONFIG_ARCH_HAS_PTE_SPECIAL case from vm_normal_page() here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 698) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 699) if (unlikely(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 700) if (vma->vm_flags & VM_MIXEDMAP) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 701) if (!pfn_valid(pfn))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 702) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 703) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 704) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 705) unsigned long off;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 706) off = (addr - vma->vm_start) >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 707) if (pfn == vma->vm_pgoff + off)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 708) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 709) if (!is_cow_mapping(vma->vm_flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 710) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 711) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 712) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 713)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 714) if (pmd_devmap(pmd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 715) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 716) if (is_huge_zero_pmd(pmd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 717) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 718) if (unlikely(pfn > highest_memmap_pfn))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 719) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 720)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 721) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 722) * NOTE! We still have PageReserved() pages in the page tables.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 723) * eg. VDSO mappings can cause them to exist.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 724) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 725) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 726) return pfn_to_page(pfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 727) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 728) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 729)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 730) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 731) * copy one vm_area from one task to the other. Assumes the page tables
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 732) * already present in the new task to be cleared in the whole range
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 733) * covered by this vma.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 734) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 735)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 736) static unsigned long
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 737) copy_nonpresent_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 738) pte_t *dst_pte, pte_t *src_pte, struct vm_area_struct *dst_vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 739) struct vm_area_struct *src_vma, unsigned long addr, int *rss)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 740) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 741) unsigned long vm_flags = dst_vma->vm_flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 742) pte_t pte = *src_pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 743) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 744) swp_entry_t entry = pte_to_swp_entry(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 745)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 746) if (likely(!non_swap_entry(entry))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 747) if (swap_duplicate(entry) < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 748) return entry.val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 749)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 750) /* make sure dst_mm is on swapoff's mmlist. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 751) if (unlikely(list_empty(&dst_mm->mmlist))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 752) spin_lock(&mmlist_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 753) if (list_empty(&dst_mm->mmlist))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 754) list_add(&dst_mm->mmlist,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 755) &src_mm->mmlist);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 756) spin_unlock(&mmlist_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 757) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 758) rss[MM_SWAPENTS]++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 759) } else if (is_migration_entry(entry)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 760) page = migration_entry_to_page(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 761)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 762) rss[mm_counter(page)]++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 763)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 764) if (is_write_migration_entry(entry) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 765) is_cow_mapping(vm_flags)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 766) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 767) * COW mappings require pages in both
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 768) * parent and child to be set to read.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 769) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 770) make_migration_entry_read(&entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 771) pte = swp_entry_to_pte(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 772) if (pte_swp_soft_dirty(*src_pte))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 773) pte = pte_swp_mksoft_dirty(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 774) if (pte_swp_uffd_wp(*src_pte))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 775) pte = pte_swp_mkuffd_wp(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 776) set_pte_at(src_mm, addr, src_pte, pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 777) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 778) } else if (is_device_private_entry(entry)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 779) page = device_private_entry_to_page(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 780)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 781) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 782) * Update rss count even for unaddressable pages, as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 783) * they should treated just like normal pages in this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 784) * respect.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 785) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 786) * We will likely want to have some new rss counters
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 787) * for unaddressable pages, at some point. But for now
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 788) * keep things as they are.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 789) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 790) get_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 791) rss[mm_counter(page)]++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 792) page_dup_rmap(page, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 793)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 794) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 795) * We do not preserve soft-dirty information, because so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 796) * far, checkpoint/restore is the only feature that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 797) * requires that. And checkpoint/restore does not work
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 798) * when a device driver is involved (you cannot easily
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 799) * save and restore device driver state).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 800) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 801) if (is_write_device_private_entry(entry) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 802) is_cow_mapping(vm_flags)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 803) make_device_private_entry_read(&entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 804) pte = swp_entry_to_pte(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 805) if (pte_swp_uffd_wp(*src_pte))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 806) pte = pte_swp_mkuffd_wp(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 807) set_pte_at(src_mm, addr, src_pte, pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 808) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 809) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 810) if (!userfaultfd_wp(dst_vma))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 811) pte = pte_swp_clear_uffd_wp(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 812) set_pte_at(dst_mm, addr, dst_pte, pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 813) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 814) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 815)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 816) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 817) * Copy a present and normal page if necessary.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 818) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 819) * NOTE! The usual case is that this doesn't need to do
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 820) * anything, and can just return a positive value. That
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 821) * will let the caller know that it can just increase
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 822) * the page refcount and re-use the pte the traditional
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 823) * way.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 824) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 825) * But _if_ we need to copy it because it needs to be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 826) * pinned in the parent (and the child should get its own
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 827) * copy rather than just a reference to the same page),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 828) * we'll do that here and return zero to let the caller
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 829) * know we're done.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 830) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 831) * And if we need a pre-allocated page but don't yet have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 832) * one, return a negative error to let the preallocation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 833) * code know so that it can do so outside the page table
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 834) * lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 835) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 836) static inline int
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 837) copy_present_page(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 838) pte_t *dst_pte, pte_t *src_pte, unsigned long addr, int *rss,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 839) struct page **prealloc, pte_t pte, struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 840) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 841) struct mm_struct *src_mm = src_vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 842) struct page *new_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 843)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 844) if (!is_cow_mapping(src_vma->vm_flags))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 845) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 846)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 847) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 848) * What we want to do is to check whether this page may
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 849) * have been pinned by the parent process. If so,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 850) * instead of wrprotect the pte on both sides, we copy
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 851) * the page immediately so that we'll always guarantee
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 852) * the pinned page won't be randomly replaced in the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 853) * future.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 854) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 855) * The page pinning checks are just "has this mm ever
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 856) * seen pinning", along with the (inexact) check of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 857) * the page count. That might give false positives for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 858) * for pinning, but it will work correctly.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 859) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 860) if (likely(!atomic_read(&src_mm->has_pinned)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 861) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 862) if (likely(!page_maybe_dma_pinned(page)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 863) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 864)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 865) new_page = *prealloc;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 866) if (!new_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 867) return -EAGAIN;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 868)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 869) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 870) * We have a prealloc page, all good! Take it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 871) * over and copy the page & arm it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 872) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 873) *prealloc = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 874) copy_user_highpage(new_page, page, addr, src_vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 875) __SetPageUptodate(new_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 876) page_add_new_anon_rmap(new_page, dst_vma, addr, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 877) lru_cache_add_inactive_or_unevictable(new_page, dst_vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 878) rss[mm_counter(new_page)]++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 879)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 880) /* All done, just insert the new page copy in the child */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 881) pte = mk_pte(new_page, dst_vma->vm_page_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 882) pte = maybe_mkwrite(pte_mkdirty(pte), dst_vma->vm_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 883) if (userfaultfd_pte_wp(dst_vma, *src_pte))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 884) /* Uffd-wp needs to be delivered to dest pte as well */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 885) pte = pte_wrprotect(pte_mkuffd_wp(pte));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 886) set_pte_at(dst_vma->vm_mm, addr, dst_pte, pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 887) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 888) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 889)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 890) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 891) * Copy one pte. Returns 0 if succeeded, or -EAGAIN if one preallocated page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 892) * is required to copy this pte.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 893) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 894) static inline int
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 895) copy_present_pte(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 896) pte_t *dst_pte, pte_t *src_pte, unsigned long addr, int *rss,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 897) struct page **prealloc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 898) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 899) struct mm_struct *src_mm = src_vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 900) unsigned long vm_flags = src_vma->vm_flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 901) pte_t pte = *src_pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 902) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 903)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 904) page = vm_normal_page(src_vma, addr, pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 905) if (page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 906) int retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 907)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 908) retval = copy_present_page(dst_vma, src_vma, dst_pte, src_pte,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 909) addr, rss, prealloc, pte, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 910) if (retval <= 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 911) return retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 912)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 913) get_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 914) page_dup_rmap(page, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 915) rss[mm_counter(page)]++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 916) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 917)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 918) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 919) * If it's a COW mapping, write protect it both
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 920) * in the parent and the child
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 921) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 922) if (is_cow_mapping(vm_flags) && pte_write(pte)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 923) ptep_set_wrprotect(src_mm, addr, src_pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 924) pte = pte_wrprotect(pte);
^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) * If it's a shared mapping, mark it clean in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 929) * the child
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 930) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 931) if (vm_flags & VM_SHARED)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 932) pte = pte_mkclean(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 933) pte = pte_mkold(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 934)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 935) if (!userfaultfd_wp(dst_vma))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 936) pte = pte_clear_uffd_wp(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 937)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 938) set_pte_at(dst_vma->vm_mm, addr, dst_pte, pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 939) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 940) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 941)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 942) static inline struct page *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 943) page_copy_prealloc(struct mm_struct *src_mm, struct vm_area_struct *vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 944) unsigned long addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 945) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 946) struct page *new_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 947)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 948) new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 949) if (!new_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 950) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 951)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 952) if (mem_cgroup_charge(new_page, src_mm, GFP_KERNEL)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 953) put_page(new_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 954) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 955) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 956) cgroup_throttle_swaprate(new_page, GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 957)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 958) return new_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 959) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 960)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 961) static int
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 962) copy_pte_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 963) pmd_t *dst_pmd, pmd_t *src_pmd, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 964) unsigned long end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 965) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 966) struct mm_struct *dst_mm = dst_vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 967) struct mm_struct *src_mm = src_vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 968) pte_t *orig_src_pte, *orig_dst_pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 969) pte_t *src_pte, *dst_pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 970) spinlock_t *src_ptl, *dst_ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 971) int progress, ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 972) int rss[NR_MM_COUNTERS];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 973) swp_entry_t entry = (swp_entry_t){0};
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 974) struct page *prealloc = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 975)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 976) again:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 977) progress = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 978) init_rss_vec(rss);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 979)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 980) dst_pte = pte_alloc_map_lock(dst_mm, dst_pmd, addr, &dst_ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 981) if (!dst_pte) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 982) ret = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 983) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 984) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 985) src_pte = pte_offset_map(src_pmd, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 986) src_ptl = pte_lockptr(src_mm, src_pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 987) spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 988) orig_src_pte = src_pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 989) orig_dst_pte = dst_pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 990) arch_enter_lazy_mmu_mode();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 991)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 992) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 993) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 994) * We are holding two locks at this point - either of them
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 995) * could generate latencies in another task on another CPU.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 996) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 997) if (progress >= 32) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 998) progress = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 999) if (need_resched() ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1000) spin_needbreak(src_ptl) || spin_needbreak(dst_ptl))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1001) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1002) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1003) if (pte_none(*src_pte)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1004) progress++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1005) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1006) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1007) if (unlikely(!pte_present(*src_pte))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1008) entry.val = copy_nonpresent_pte(dst_mm, src_mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1009) dst_pte, src_pte,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1010) dst_vma, src_vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1011) addr, rss);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1012) if (entry.val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1013) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1014) progress += 8;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1015) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1016) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1017) /* copy_present_pte() will clear `*prealloc' if consumed */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1018) ret = copy_present_pte(dst_vma, src_vma, dst_pte, src_pte,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1019) addr, rss, &prealloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1020) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1021) * If we need a pre-allocated page for this pte, drop the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1022) * locks, allocate, and try again.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1023) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1024) if (unlikely(ret == -EAGAIN))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1025) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1026) if (unlikely(prealloc)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1027) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1028) * pre-alloc page cannot be reused by next time so as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1029) * to strictly follow mempolicy (e.g., alloc_page_vma()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1030) * will allocate page according to address). This
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1031) * could only happen if one pinned pte changed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1032) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1033) put_page(prealloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1034) prealloc = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1035) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1036) progress += 8;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1037) } while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1038)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1039) arch_leave_lazy_mmu_mode();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1040) spin_unlock(src_ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1041) pte_unmap(orig_src_pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1042) add_mm_rss_vec(dst_mm, rss);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1043) pte_unmap_unlock(orig_dst_pte, dst_ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1044) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1045)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1046) if (entry.val) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1047) if (add_swap_count_continuation(entry, GFP_KERNEL) < 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1048) ret = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1049) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1050) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1051) entry.val = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1052) } else if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1053) WARN_ON_ONCE(ret != -EAGAIN);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1054) prealloc = page_copy_prealloc(src_mm, src_vma, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1055) if (!prealloc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1056) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1057) /* We've captured and resolved the error. Reset, try again. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1058) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1059) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1060) if (addr != end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1061) goto again;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1062) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1063) if (unlikely(prealloc))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1064) put_page(prealloc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1065) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1066) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1067)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1068) static inline int
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1069) copy_pmd_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1070) pud_t *dst_pud, pud_t *src_pud, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1071) unsigned long end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1072) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1073) struct mm_struct *dst_mm = dst_vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1074) struct mm_struct *src_mm = src_vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1075) pmd_t *src_pmd, *dst_pmd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1076) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1077)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1078) dst_pmd = pmd_alloc(dst_mm, dst_pud, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1079) if (!dst_pmd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1080) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1081) src_pmd = pmd_offset(src_pud, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1082) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1083) next = pmd_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1084) if (is_swap_pmd(*src_pmd) || pmd_trans_huge(*src_pmd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1085) || pmd_devmap(*src_pmd)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1086) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1087) VM_BUG_ON_VMA(next-addr != HPAGE_PMD_SIZE, src_vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1088) err = copy_huge_pmd(dst_mm, src_mm, dst_pmd, src_pmd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1089) addr, dst_vma, src_vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1090) if (err == -ENOMEM)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1091) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1092) if (!err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1093) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1094) /* fall through */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1095) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1096) if (pmd_none_or_clear_bad(src_pmd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1097) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1098) if (copy_pte_range(dst_vma, src_vma, dst_pmd, src_pmd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1099) addr, next))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1100) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1101) } while (dst_pmd++, src_pmd++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1102) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1103) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1104)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1105) static inline int
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1106) copy_pud_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1107) p4d_t *dst_p4d, p4d_t *src_p4d, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1108) unsigned long end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1109) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1110) struct mm_struct *dst_mm = dst_vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1111) struct mm_struct *src_mm = src_vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1112) pud_t *src_pud, *dst_pud;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1113) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1114)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1115) dst_pud = pud_alloc(dst_mm, dst_p4d, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1116) if (!dst_pud)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1117) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1118) src_pud = pud_offset(src_p4d, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1119) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1120) next = pud_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1121) if (pud_trans_huge(*src_pud) || pud_devmap(*src_pud)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1122) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1123)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1124) VM_BUG_ON_VMA(next-addr != HPAGE_PUD_SIZE, src_vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1125) err = copy_huge_pud(dst_mm, src_mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1126) dst_pud, src_pud, addr, src_vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1127) if (err == -ENOMEM)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1128) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1129) if (!err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1130) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1131) /* fall through */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1132) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1133) if (pud_none_or_clear_bad(src_pud))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1134) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1135) if (copy_pmd_range(dst_vma, src_vma, dst_pud, src_pud,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1136) addr, next))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1137) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1138) } while (dst_pud++, src_pud++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1139) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1140) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1141)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1142) static inline int
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1143) copy_p4d_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1144) pgd_t *dst_pgd, pgd_t *src_pgd, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1145) unsigned long end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1146) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1147) struct mm_struct *dst_mm = dst_vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1148) p4d_t *src_p4d, *dst_p4d;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1149) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1150)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1151) dst_p4d = p4d_alloc(dst_mm, dst_pgd, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1152) if (!dst_p4d)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1153) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1154) src_p4d = p4d_offset(src_pgd, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1155) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1156) next = p4d_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1157) if (p4d_none_or_clear_bad(src_p4d))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1158) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1159) if (copy_pud_range(dst_vma, src_vma, dst_p4d, src_p4d,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1160) addr, next))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1161) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1162) } while (dst_p4d++, src_p4d++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1163) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1164) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1165)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1166) int
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1167) copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1168) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1169) pgd_t *src_pgd, *dst_pgd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1170) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1171) unsigned long addr = src_vma->vm_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1172) unsigned long end = src_vma->vm_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1173) struct mm_struct *dst_mm = dst_vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1174) struct mm_struct *src_mm = src_vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1175) struct mmu_notifier_range range;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1176) bool is_cow;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1177) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1178)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1179) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1180) * Don't copy ptes where a page fault will fill them correctly.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1181) * Fork becomes much lighter when there are big shared or private
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1182) * readonly mappings. The tradeoff is that copy_page_range is more
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1183) * efficient than faulting.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1184) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1185) if (!(src_vma->vm_flags & (VM_HUGETLB | VM_PFNMAP | VM_MIXEDMAP)) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1186) !src_vma->anon_vma)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1187) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1188)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1189) if (is_vm_hugetlb_page(src_vma))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1190) return copy_hugetlb_page_range(dst_mm, src_mm, src_vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1191)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1192) if (unlikely(src_vma->vm_flags & VM_PFNMAP)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1193) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1194) * We do not free on error cases below as remove_vma
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1195) * gets called on error from higher level routine
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1196) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1197) ret = track_pfn_copy(src_vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1198) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1199) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1200) }
^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) * We need to invalidate the secondary MMU mappings only when
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1204) * there could be a permission downgrade on the ptes of the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1205) * parent mm. And a permission downgrade will only happen if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1206) * is_cow_mapping() returns true.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1207) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1208) is_cow = is_cow_mapping(src_vma->vm_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1209)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1210) if (is_cow) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1211) mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_PAGE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1212) 0, src_vma, src_mm, addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1213) mmu_notifier_invalidate_range_start(&range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1214) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1215) * Disabling preemption is not needed for the write side, as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1216) * the read side doesn't spin, but goes to the mmap_lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1217) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1218) * Use the raw variant of the seqcount_t write API to avoid
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1219) * lockdep complaining about preemptibility.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1220) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1221) mmap_assert_write_locked(src_mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1222) raw_write_seqcount_begin(&src_mm->write_protect_seq);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1223) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1224)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1225) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1226) dst_pgd = pgd_offset(dst_mm, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1227) src_pgd = pgd_offset(src_mm, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1228) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1229) next = pgd_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1230) if (pgd_none_or_clear_bad(src_pgd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1231) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1232) if (unlikely(copy_p4d_range(dst_vma, src_vma, dst_pgd, src_pgd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1233) addr, next))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1234) ret = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1235) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1236) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1237) } while (dst_pgd++, src_pgd++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1238)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1239) if (is_cow) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1240) raw_write_seqcount_end(&src_mm->write_protect_seq);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1241) mmu_notifier_invalidate_range_end(&range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1242) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1243) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1244) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1245)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1246) static unsigned long zap_pte_range(struct mmu_gather *tlb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1247) struct vm_area_struct *vma, pmd_t *pmd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1248) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1249) struct zap_details *details)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1250) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1251) struct mm_struct *mm = tlb->mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1252) int force_flush = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1253) int rss[NR_MM_COUNTERS];
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1254) spinlock_t *ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1255) pte_t *start_pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1256) pte_t *pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1257) swp_entry_t entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1258)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1259) tlb_change_page_size(tlb, PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1260) again:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1261) init_rss_vec(rss);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1262) start_pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1263) pte = start_pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1264) flush_tlb_batched_pending(mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1265) arch_enter_lazy_mmu_mode();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1266) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1267) pte_t ptent = *pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1268) if (pte_none(ptent))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1269) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1270)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1271) if (need_resched())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1272) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1273)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1274) if (pte_present(ptent)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1275) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1276)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1277) page = vm_normal_page(vma, addr, ptent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1278) if (unlikely(details) && page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1279) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1280) * unmap_shared_mapping_pages() wants to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1281) * invalidate cache without truncating:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1282) * unmap shared but keep private pages.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1283) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1284) if (details->check_mapping &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1285) details->check_mapping != page_rmapping(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1286) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1287) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1288) ptent = ptep_get_and_clear_full(mm, addr, pte,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1289) tlb->fullmm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1290) tlb_remove_tlb_entry(tlb, pte, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1291) if (unlikely(!page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1292) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1293)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1294) if (!PageAnon(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1295) if (pte_dirty(ptent)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1296) force_flush = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1297) set_page_dirty(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1298) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1299) if (pte_young(ptent) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1300) likely(!(vma->vm_flags & VM_SEQ_READ)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1301) mark_page_accessed(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1302) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1303) rss[mm_counter(page)]--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1304) page_remove_rmap(page, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1305) if (unlikely(page_mapcount(page) < 0))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1306) print_bad_pte(vma, addr, ptent, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1307) if (unlikely(__tlb_remove_page(tlb, page)) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1308) lru_cache_disabled()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1309) force_flush = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1310) addr += PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1311) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1312) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1313) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1314) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1315)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1316) entry = pte_to_swp_entry(ptent);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1317) if (is_device_private_entry(entry)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1318) struct page *page = device_private_entry_to_page(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1319)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1320) if (unlikely(details && details->check_mapping)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1321) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1322) * unmap_shared_mapping_pages() wants to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1323) * invalidate cache without truncating:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1324) * unmap shared but keep private pages.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1325) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1326) if (details->check_mapping !=
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1327) page_rmapping(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1328) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1329) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1330)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1331) pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1332) rss[mm_counter(page)]--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1333) page_remove_rmap(page, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1334) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1335) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1336) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1337)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1338) /* If details->check_mapping, we leave swap entries. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1339) if (unlikely(details))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1340) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1341)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1342) if (!non_swap_entry(entry))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1343) rss[MM_SWAPENTS]--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1344) else if (is_migration_entry(entry)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1345) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1346)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1347) page = migration_entry_to_page(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1348) rss[mm_counter(page)]--;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1349) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1350) if (unlikely(!free_swap_and_cache(entry)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1351) print_bad_pte(vma, addr, ptent, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1352) pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1353) } while (pte++, addr += PAGE_SIZE, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1354)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1355) add_mm_rss_vec(mm, rss);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1356) arch_leave_lazy_mmu_mode();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1357)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1358) /* Do the actual TLB flush before dropping ptl */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1359) if (force_flush)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1360) tlb_flush_mmu_tlbonly(tlb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1361) pte_unmap_unlock(start_pte, ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1362)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1363) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1364) * If we forced a TLB flush (either due to running out of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1365) * batch buffers or because we needed to flush dirty TLB
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1366) * entries before releasing the ptl), free the batched
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1367) * memory too. Restart if we didn't do everything.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1368) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1369) if (force_flush) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1370) force_flush = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1371) tlb_flush_mmu(tlb);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1372) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1373)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1374) if (addr != end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1375) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1376) goto again;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1377) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1378)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1379) return addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1380) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1381)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1382) static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1383) struct vm_area_struct *vma, pud_t *pud,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1384) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1385) struct zap_details *details)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1386) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1387) pmd_t *pmd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1388) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1389)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1390) pmd = pmd_offset(pud, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1391) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1392) next = pmd_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1393) if (is_swap_pmd(*pmd) || pmd_trans_huge(*pmd) || pmd_devmap(*pmd)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1394) if (next - addr != HPAGE_PMD_SIZE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1395) __split_huge_pmd(vma, pmd, addr, false, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1396) else if (zap_huge_pmd(tlb, vma, pmd, addr))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1397) goto next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1398) /* fall through */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1399) } else if (details && details->single_page &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1400) PageTransCompound(details->single_page) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1401) next - addr == HPAGE_PMD_SIZE && pmd_none(*pmd)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1402) spinlock_t *ptl = pmd_lock(tlb->mm, pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1403) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1404) * Take and drop THP pmd lock so that we cannot return
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1405) * prematurely, while zap_huge_pmd() has cleared *pmd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1406) * but not yet decremented compound_mapcount().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1407) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1408) spin_unlock(ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1409) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1410)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1411) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1412) * Here there can be other concurrent MADV_DONTNEED or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1413) * trans huge page faults running, and if the pmd is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1414) * none or trans huge it can change under us. This is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1415) * because MADV_DONTNEED holds the mmap_lock in read
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1416) * mode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1417) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1418) if (pmd_none_or_trans_huge_or_clear_bad(pmd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1419) goto next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1420) next = zap_pte_range(tlb, vma, pmd, addr, next, details);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1421) next:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1422) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1423) } while (pmd++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1424)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1425) return addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1426) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1427)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1428) static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1429) struct vm_area_struct *vma, p4d_t *p4d,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1430) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1431) struct zap_details *details)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1432) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1433) pud_t *pud;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1434) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1435)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1436) pud = pud_offset(p4d, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1437) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1438) next = pud_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1439) if (pud_trans_huge(*pud) || pud_devmap(*pud)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1440) if (next - addr != HPAGE_PUD_SIZE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1441) mmap_assert_locked(tlb->mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1442) split_huge_pud(vma, pud, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1443) } else if (zap_huge_pud(tlb, vma, pud, addr))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1444) goto next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1445) /* fall through */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1446) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1447) if (pud_none_or_clear_bad(pud))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1448) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1449) next = zap_pmd_range(tlb, vma, pud, addr, next, details);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1450) next:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1451) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1452) } while (pud++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1453)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1454) return addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1455) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1456)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1457) static inline unsigned long zap_p4d_range(struct mmu_gather *tlb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1458) struct vm_area_struct *vma, pgd_t *pgd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1459) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1460) struct zap_details *details)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1461) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1462) p4d_t *p4d;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1463) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1464)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1465) p4d = p4d_offset(pgd, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1466) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1467) next = p4d_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1468) if (p4d_none_or_clear_bad(p4d))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1469) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1470) next = zap_pud_range(tlb, vma, p4d, addr, next, details);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1471) } while (p4d++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1472)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1473) return addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1474) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1475)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1476) void unmap_page_range(struct mmu_gather *tlb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1477) struct vm_area_struct *vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1478) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1479) struct zap_details *details)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1480) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1481) pgd_t *pgd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1482) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1483)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1484) BUG_ON(addr >= end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1485) vm_write_begin(vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1486) tlb_start_vma(tlb, vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1487) pgd = pgd_offset(vma->vm_mm, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1488) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1489) next = pgd_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1490) if (pgd_none_or_clear_bad(pgd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1491) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1492) next = zap_p4d_range(tlb, vma, pgd, addr, next, details);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1493) } while (pgd++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1494) tlb_end_vma(tlb, vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1495) vm_write_end(vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1496) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1497)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1498)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1499) static void unmap_single_vma(struct mmu_gather *tlb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1500) struct vm_area_struct *vma, unsigned long start_addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1501) unsigned long end_addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1502) struct zap_details *details)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1503) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1504) unsigned long start = max(vma->vm_start, start_addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1505) unsigned long end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1506)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1507) if (start >= vma->vm_end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1508) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1509) end = min(vma->vm_end, end_addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1510) if (end <= vma->vm_start)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1511) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1512)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1513) if (vma->vm_file)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1514) uprobe_munmap(vma, start, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1515)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1516) if (unlikely(vma->vm_flags & VM_PFNMAP))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1517) untrack_pfn(vma, 0, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1518)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1519) if (start != end) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1520) if (unlikely(is_vm_hugetlb_page(vma))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1521) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1522) * It is undesirable to test vma->vm_file as it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1523) * should be non-null for valid hugetlb area.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1524) * However, vm_file will be NULL in the error
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1525) * cleanup path of mmap_region. When
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1526) * hugetlbfs ->mmap method fails,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1527) * mmap_region() nullifies vma->vm_file
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1528) * before calling this function to clean up.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1529) * Since no pte has actually been setup, it is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1530) * safe to do nothing in this case.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1531) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1532) if (vma->vm_file) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1533) i_mmap_lock_write(vma->vm_file->f_mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1534) __unmap_hugepage_range_final(tlb, vma, start, end, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1535) i_mmap_unlock_write(vma->vm_file->f_mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1536) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1537) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1538) unmap_page_range(tlb, vma, start, end, details);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1539) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1540) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1541)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1542) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1543) * unmap_vmas - unmap a range of memory covered by a list of vma's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1544) * @tlb: address of the caller's struct mmu_gather
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1545) * @vma: the starting vma
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1546) * @start_addr: virtual address at which to start unmapping
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1547) * @end_addr: virtual address at which to end unmapping
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1548) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1549) * Unmap all pages in the vma list.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1550) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1551) * Only addresses between `start' and `end' will be unmapped.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1552) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1553) * The VMA list must be sorted in ascending virtual address order.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1554) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1555) * unmap_vmas() assumes that the caller will flush the whole unmapped address
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1556) * range after unmap_vmas() returns. So the only responsibility here is to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1557) * ensure that any thus-far unmapped pages are flushed before unmap_vmas()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1558) * drops the lock and schedules.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1559) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1560) void unmap_vmas(struct mmu_gather *tlb,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1561) struct vm_area_struct *vma, unsigned long start_addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1562) unsigned long end_addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1563) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1564) struct mmu_notifier_range range;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1565)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1566) mmu_notifier_range_init(&range, MMU_NOTIFY_UNMAP, 0, vma, vma->vm_mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1567) start_addr, end_addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1568) mmu_notifier_invalidate_range_start(&range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1569) for ( ; vma && vma->vm_start < end_addr; vma = vma->vm_next)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1570) unmap_single_vma(tlb, vma, start_addr, end_addr, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1571) mmu_notifier_invalidate_range_end(&range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1572) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1573)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1574) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1575) * zap_page_range - remove user pages in a given range
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1576) * @vma: vm_area_struct holding the applicable pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1577) * @start: starting address of pages to zap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1578) * @size: number of bytes to zap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1579) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1580) * Caller must protect the VMA list
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1581) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1582) void zap_page_range(struct vm_area_struct *vma, unsigned long start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1583) unsigned long size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1584) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1585) struct mmu_notifier_range range;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1586) struct mmu_gather tlb;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1587)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1588) lru_add_drain();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1589) mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, vma->vm_mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1590) start, start + size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1591) tlb_gather_mmu(&tlb, vma->vm_mm, start, range.end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1592) update_hiwater_rss(vma->vm_mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1593) mmu_notifier_invalidate_range_start(&range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1594) for ( ; vma && vma->vm_start < range.end; vma = vma->vm_next)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1595) unmap_single_vma(&tlb, vma, start, range.end, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1596) mmu_notifier_invalidate_range_end(&range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1597) tlb_finish_mmu(&tlb, start, range.end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1598) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1599)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1600) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1601) * zap_page_range_single - remove user pages in a given range
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1602) * @vma: vm_area_struct holding the applicable pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1603) * @address: starting address of pages to zap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1604) * @size: number of bytes to zap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1605) * @details: details of shared cache invalidation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1606) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1607) * The range must fit into one VMA.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1608) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1609) static void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1610) unsigned long size, struct zap_details *details)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1611) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1612) struct mmu_notifier_range range;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1613) struct mmu_gather tlb;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1614)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1615) lru_add_drain();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1616) mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, vma->vm_mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1617) address, address + size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1618) tlb_gather_mmu(&tlb, vma->vm_mm, address, range.end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1619) update_hiwater_rss(vma->vm_mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1620) mmu_notifier_invalidate_range_start(&range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1621) unmap_single_vma(&tlb, vma, address, range.end, details);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1622) mmu_notifier_invalidate_range_end(&range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1623) tlb_finish_mmu(&tlb, address, range.end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1624) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1625)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1626) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1627) * zap_vma_ptes - remove ptes mapping the vma
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1628) * @vma: vm_area_struct holding ptes to be zapped
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1629) * @address: starting address of pages to zap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1630) * @size: number of bytes to zap
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1631) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1632) * This function only unmaps ptes assigned to VM_PFNMAP vmas.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1633) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1634) * The entire address range must be fully contained within the vma.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1635) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1636) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1637) void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1638) unsigned long size)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1639) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1640) if (address < vma->vm_start || address + size > vma->vm_end ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1641) !(vma->vm_flags & VM_PFNMAP))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1642) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1643)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1644) zap_page_range_single(vma, address, size, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1645) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1646) EXPORT_SYMBOL_GPL(zap_vma_ptes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1647)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1648) static pmd_t *walk_to_pmd(struct mm_struct *mm, unsigned long addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1649) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1650) pgd_t *pgd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1651) p4d_t *p4d;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1652) pud_t *pud;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1653) pmd_t *pmd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1654)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1655) pgd = pgd_offset(mm, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1656) p4d = p4d_alloc(mm, pgd, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1657) if (!p4d)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1658) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1659) pud = pud_alloc(mm, p4d, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1660) if (!pud)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1661) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1662) pmd = pmd_alloc(mm, pud, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1663) if (!pmd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1664) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1665)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1666) VM_BUG_ON(pmd_trans_huge(*pmd));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1667) return pmd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1668) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1669)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1670) pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1671) spinlock_t **ptl)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1672) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1673) pmd_t *pmd = walk_to_pmd(mm, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1674)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1675) if (!pmd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1676) return NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1677) return pte_alloc_map_lock(mm, pmd, addr, ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1678) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1679)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1680) static int validate_page_before_insert(struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1681) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1682) if (PageAnon(page) || PageSlab(page) || page_has_type(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1683) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1684) flush_dcache_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1685) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1686) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1687)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1688) static int insert_page_into_pte_locked(struct mm_struct *mm, pte_t *pte,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1689) unsigned long addr, struct page *page, pgprot_t prot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1690) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1691) if (!pte_none(*pte))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1692) return -EBUSY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1693) /* Ok, finally just insert the thing.. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1694) get_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1695) inc_mm_counter_fast(mm, mm_counter_file(page));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1696) page_add_file_rmap(page, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1697) set_pte_at(mm, addr, pte, mk_pte(page, prot));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1698) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1699) }
^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) * This is the old fallback for page remapping.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1703) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1704) * For historical reasons, it only allows reserved pages. Only
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1705) * old drivers should use this, and they needed to mark their
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1706) * pages reserved for the old functions anyway.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1707) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1708) static int insert_page(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1709) struct page *page, pgprot_t prot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1710) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1711) struct mm_struct *mm = vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1712) int retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1713) pte_t *pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1714) spinlock_t *ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1715)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1716) retval = validate_page_before_insert(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1717) if (retval)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1718) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1719) retval = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1720) pte = get_locked_pte(mm, addr, &ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1721) if (!pte)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1722) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1723) retval = insert_page_into_pte_locked(mm, pte, addr, page, prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1724) pte_unmap_unlock(pte, ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1725) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1726) return retval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1727) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1728)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1729) #ifdef pte_index
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1730) static int insert_page_in_batch_locked(struct mm_struct *mm, pte_t *pte,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1731) unsigned long addr, struct page *page, pgprot_t prot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1732) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1733) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1734)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1735) if (!page_count(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1736) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1737) err = validate_page_before_insert(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1738) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1739) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1740) return insert_page_into_pte_locked(mm, pte, addr, page, prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1741) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1742)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1743) /* insert_pages() amortizes the cost of spinlock operations
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1744) * when inserting pages in a loop. Arch *must* define pte_index.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1745) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1746) static int insert_pages(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1747) struct page **pages, unsigned long *num, pgprot_t prot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1748) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1749) pmd_t *pmd = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1750) pte_t *start_pte, *pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1751) spinlock_t *pte_lock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1752) struct mm_struct *const mm = vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1753) unsigned long curr_page_idx = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1754) unsigned long remaining_pages_total = *num;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1755) unsigned long pages_to_write_in_pmd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1756) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1757) more:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1758) ret = -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1759) pmd = walk_to_pmd(mm, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1760) if (!pmd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1761) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1762)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1763) pages_to_write_in_pmd = min_t(unsigned long,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1764) remaining_pages_total, PTRS_PER_PTE - pte_index(addr));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1765)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1766) /* Allocate the PTE if necessary; takes PMD lock once only. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1767) ret = -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1768) if (pte_alloc(mm, pmd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1769) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1770)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1771) while (pages_to_write_in_pmd) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1772) int pte_idx = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1773) const int batch_size = min_t(int, pages_to_write_in_pmd, 8);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1774)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1775) start_pte = pte_offset_map_lock(mm, pmd, addr, &pte_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1776) for (pte = start_pte; pte_idx < batch_size; ++pte, ++pte_idx) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1777) int err = insert_page_in_batch_locked(mm, pte,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1778) addr, pages[curr_page_idx], prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1779) if (unlikely(err)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1780) pte_unmap_unlock(start_pte, pte_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1781) ret = err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1782) remaining_pages_total -= pte_idx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1783) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1784) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1785) addr += PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1786) ++curr_page_idx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1787) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1788) pte_unmap_unlock(start_pte, pte_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1789) pages_to_write_in_pmd -= batch_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1790) remaining_pages_total -= batch_size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1791) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1792) if (remaining_pages_total)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1793) goto more;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1794) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1795) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1796) *num = remaining_pages_total;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1797) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1798) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1799) #endif /* ifdef pte_index */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1800)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1801) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1802) * vm_insert_pages - insert multiple pages into user vma, batching the pmd lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1803) * @vma: user vma to map to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1804) * @addr: target start user address of these pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1805) * @pages: source kernel pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1806) * @num: in: number of pages to map. out: number of pages that were *not*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1807) * mapped. (0 means all pages were successfully mapped).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1808) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1809) * Preferred over vm_insert_page() when inserting multiple pages.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1810) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1811) * In case of error, we may have mapped a subset of the provided
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1812) * pages. It is the caller's responsibility to account for this case.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1813) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1814) * The same restrictions apply as in vm_insert_page().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1815) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1816) int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1817) struct page **pages, unsigned long *num)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1818) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1819) #ifdef pte_index
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1820) const unsigned long end_addr = addr + (*num * PAGE_SIZE) - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1821)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1822) if (addr < vma->vm_start || end_addr >= vma->vm_end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1823) return -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1824) if (!(vma->vm_flags & VM_MIXEDMAP)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1825) BUG_ON(mmap_read_trylock(vma->vm_mm));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1826) BUG_ON(vma->vm_flags & VM_PFNMAP);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1827) vma->vm_flags |= VM_MIXEDMAP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1828) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1829) /* Defer page refcount checking till we're about to map that page. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1830) return insert_pages(vma, addr, pages, num, vma->vm_page_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1831) #else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1832) unsigned long idx = 0, pgcount = *num;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1833) int err = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1834)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1835) for (; idx < pgcount; ++idx) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1836) err = vm_insert_page(vma, addr + (PAGE_SIZE * idx), pages[idx]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1837) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1838) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1839) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1840) *num = pgcount - idx;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1841) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1842) #endif /* ifdef pte_index */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1843) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1844) EXPORT_SYMBOL(vm_insert_pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1845)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1846) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1847) * vm_insert_page - insert single page into user vma
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1848) * @vma: user vma to map to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1849) * @addr: target user address of this page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1850) * @page: source kernel page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1851) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1852) * This allows drivers to insert individual pages they've allocated
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1853) * into a user vma.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1854) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1855) * The page has to be a nice clean _individual_ kernel allocation.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1856) * If you allocate a compound page, you need to have marked it as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1857) * such (__GFP_COMP), or manually just split the page up yourself
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1858) * (see split_page()).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1859) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1860) * NOTE! Traditionally this was done with "remap_pfn_range()" which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1861) * took an arbitrary page protection parameter. This doesn't allow
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1862) * that. Your vma protection will have to be set up correctly, which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1863) * means that if you want a shared writable mapping, you'd better
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1864) * ask for a shared writable mapping!
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1865) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1866) * The page does not need to be reserved.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1867) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1868) * Usually this function is called from f_op->mmap() handler
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1869) * under mm->mmap_lock write-lock, so it can change vma->vm_flags.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1870) * Caller must set VM_MIXEDMAP on vma if it wants to call this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1871) * function from other places, for example from page-fault handler.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1872) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1873) * Return: %0 on success, negative error code otherwise.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1874) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1875) int vm_insert_page(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1876) struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1877) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1878) if (addr < vma->vm_start || addr >= vma->vm_end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1879) return -EFAULT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1880) if (!page_count(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1881) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1882) if (!(vma->vm_flags & VM_MIXEDMAP)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1883) BUG_ON(mmap_read_trylock(vma->vm_mm));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1884) BUG_ON(vma->vm_flags & VM_PFNMAP);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1885) vma->vm_flags |= VM_MIXEDMAP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1886) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1887) return insert_page(vma, addr, page, vma->vm_page_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1888) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1889) EXPORT_SYMBOL(vm_insert_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1890)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1891) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1892) * __vm_map_pages - maps range of kernel pages into user vma
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1893) * @vma: user vma to map to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1894) * @pages: pointer to array of source kernel pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1895) * @num: number of pages in page array
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1896) * @offset: user's requested vm_pgoff
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1897) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1898) * This allows drivers to map range of kernel pages into a user vma.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1899) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1900) * Return: 0 on success and error code otherwise.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1901) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1902) static int __vm_map_pages(struct vm_area_struct *vma, struct page **pages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1903) unsigned long num, unsigned long offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1904) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1905) unsigned long count = vma_pages(vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1906) unsigned long uaddr = vma->vm_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1907) int ret, i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1908)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1909) /* Fail if the user requested offset is beyond the end of the object */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1910) if (offset >= num)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1911) return -ENXIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1912)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1913) /* Fail if the user requested size exceeds available object size */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1914) if (count > num - offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1915) return -ENXIO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1916)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1917) for (i = 0; i < count; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1918) ret = vm_insert_page(vma, uaddr, pages[offset + i]);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1919) if (ret < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1920) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1921) uaddr += PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1922) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1923)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1924) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1925) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1926)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1927) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1928) * vm_map_pages - maps range of kernel pages starts with non zero offset
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1929) * @vma: user vma to map to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1930) * @pages: pointer to array of source kernel pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1931) * @num: number of pages in page array
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1932) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1933) * Maps an object consisting of @num pages, catering for the user's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1934) * requested vm_pgoff
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1935) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1936) * If we fail to insert any page into the vma, the function will return
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1937) * immediately leaving any previously inserted pages present. Callers
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1938) * from the mmap handler may immediately return the error as their caller
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1939) * will destroy the vma, removing any successfully inserted pages. Other
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1940) * callers should make their own arrangements for calling unmap_region().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1941) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1942) * Context: Process context. Called by mmap handlers.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1943) * Return: 0 on success and error code otherwise.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1944) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1945) int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1946) unsigned long num)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1947) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1948) return __vm_map_pages(vma, pages, num, vma->vm_pgoff);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1949) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1950) EXPORT_SYMBOL(vm_map_pages);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1951)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1952) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1953) * vm_map_pages_zero - map range of kernel pages starts with zero offset
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1954) * @vma: user vma to map to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1955) * @pages: pointer to array of source kernel pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1956) * @num: number of pages in page array
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1957) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1958) * Similar to vm_map_pages(), except that it explicitly sets the offset
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1959) * to 0. This function is intended for the drivers that did not consider
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1960) * vm_pgoff.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1961) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1962) * Context: Process context. Called by mmap handlers.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1963) * Return: 0 on success and error code otherwise.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1964) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1965) int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1966) unsigned long num)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1967) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1968) return __vm_map_pages(vma, pages, num, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1969) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1970) EXPORT_SYMBOL(vm_map_pages_zero);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1971)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1972) static vm_fault_t insert_pfn(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1973) pfn_t pfn, pgprot_t prot, bool mkwrite)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1974) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1975) struct mm_struct *mm = vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1976) pte_t *pte, entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1977) spinlock_t *ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1978)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1979) pte = get_locked_pte(mm, addr, &ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1980) if (!pte)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1981) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1982) if (!pte_none(*pte)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1983) if (mkwrite) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1984) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1985) * For read faults on private mappings the PFN passed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1986) * in may not match the PFN we have mapped if the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1987) * mapped PFN is a writeable COW page. In the mkwrite
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1988) * case we are creating a writable PTE for a shared
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1989) * mapping and we expect the PFNs to match. If they
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1990) * don't match, we are likely racing with block
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1991) * allocation and mapping invalidation so just skip the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1992) * update.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1993) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1994) if (pte_pfn(*pte) != pfn_t_to_pfn(pfn)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1995) WARN_ON_ONCE(!is_zero_pfn(pte_pfn(*pte)));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1996) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1997) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1998) entry = pte_mkyoung(*pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1999) entry = maybe_mkwrite(pte_mkdirty(entry),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2000) vma->vm_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2001) if (ptep_set_access_flags(vma, addr, pte, entry, 1))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2002) update_mmu_cache(vma, addr, pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2003) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2004) goto out_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2005) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2006)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2007) /* Ok, finally just insert the thing.. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2008) if (pfn_t_devmap(pfn))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2009) entry = pte_mkdevmap(pfn_t_pte(pfn, prot));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2010) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2011) entry = pte_mkspecial(pfn_t_pte(pfn, prot));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2012)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2013) if (mkwrite) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2014) entry = pte_mkyoung(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2015) entry = maybe_mkwrite(pte_mkdirty(entry), vma->vm_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2016) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2017)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2018) set_pte_at(mm, addr, pte, entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2019) update_mmu_cache(vma, addr, pte); /* XXX: why not for insert_page? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2020)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2021) out_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2022) pte_unmap_unlock(pte, ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2023) return VM_FAULT_NOPAGE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2024) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2025)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2026) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2027) * vmf_insert_pfn_prot - insert single pfn into user vma with specified pgprot
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2028) * @vma: user vma to map to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2029) * @addr: target user address of this page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2030) * @pfn: source kernel pfn
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2031) * @pgprot: pgprot flags for the inserted page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2032) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2033) * This is exactly like vmf_insert_pfn(), except that it allows drivers
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2034) * to override pgprot on a per-page basis.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2035) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2036) * This only makes sense for IO mappings, and it makes no sense for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2037) * COW mappings. In general, using multiple vmas is preferable;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2038) * vmf_insert_pfn_prot should only be used if using multiple VMAs is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2039) * impractical.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2040) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2041) * See vmf_insert_mixed_prot() for a discussion of the implication of using
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2042) * a value of @pgprot different from that of @vma->vm_page_prot.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2043) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2044) * Context: Process context. May allocate using %GFP_KERNEL.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2045) * Return: vm_fault_t value.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2046) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2047) vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2048) unsigned long pfn, pgprot_t pgprot)
^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) * Technically, architectures with pte_special can avoid all these
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2052) * restrictions (same for remap_pfn_range). However we would like
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2053) * consistency in testing and feature parity among all, so we should
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2054) * try to keep these invariants in place for everybody.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2055) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2056) BUG_ON(!(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP)));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2057) BUG_ON((vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP)) ==
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2058) (VM_PFNMAP|VM_MIXEDMAP));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2059) BUG_ON((vma->vm_flags & VM_PFNMAP) && is_cow_mapping(vma->vm_flags));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2060) BUG_ON((vma->vm_flags & VM_MIXEDMAP) && pfn_valid(pfn));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2061)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2062) if (addr < vma->vm_start || addr >= vma->vm_end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2063) return VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2064)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2065) if (!pfn_modify_allowed(pfn, pgprot))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2066) return VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2067)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2068) track_pfn_insert(vma, &pgprot, __pfn_to_pfn_t(pfn, PFN_DEV));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2069)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2070) return insert_pfn(vma, addr, __pfn_to_pfn_t(pfn, PFN_DEV), pgprot,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2071) false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2072) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2073) EXPORT_SYMBOL(vmf_insert_pfn_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2074)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2075) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2076) * vmf_insert_pfn - insert single pfn into user vma
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2077) * @vma: user vma to map to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2078) * @addr: target user address of this page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2079) * @pfn: source kernel pfn
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2080) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2081) * Similar to vm_insert_page, this allows drivers to insert individual pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2082) * they've allocated into a user vma. Same comments apply.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2083) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2084) * This function should only be called from a vm_ops->fault handler, and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2085) * in that case the handler should return the result of this function.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2086) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2087) * vma cannot be a COW mapping.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2088) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2089) * As this is called only for pages that do not currently exist, we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2090) * do not need to flush old virtual caches or the TLB.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2091) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2092) * Context: Process context. May allocate using %GFP_KERNEL.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2093) * Return: vm_fault_t value.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2094) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2095) vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2096) unsigned long pfn)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2097) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2098) return vmf_insert_pfn_prot(vma, addr, pfn, vma->vm_page_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2099) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2100) EXPORT_SYMBOL(vmf_insert_pfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2101)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2102) static bool vm_mixed_ok(struct vm_area_struct *vma, pfn_t pfn)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2103) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2104) /* these checks mirror the abort conditions in vm_normal_page */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2105) if (vma->vm_flags & VM_MIXEDMAP)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2106) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2107) if (pfn_t_devmap(pfn))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2108) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2109) if (pfn_t_special(pfn))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2110) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2111) if (is_zero_pfn(pfn_t_to_pfn(pfn)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2112) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2113) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2114) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2115)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2116) static vm_fault_t __vm_insert_mixed(struct vm_area_struct *vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2117) unsigned long addr, pfn_t pfn, pgprot_t pgprot,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2118) bool mkwrite)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2119) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2120) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2121)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2122) BUG_ON(!vm_mixed_ok(vma, pfn));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2123)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2124) if (addr < vma->vm_start || addr >= vma->vm_end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2125) return VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2126)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2127) track_pfn_insert(vma, &pgprot, pfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2128)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2129) if (!pfn_modify_allowed(pfn_t_to_pfn(pfn), pgprot))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2130) return VM_FAULT_SIGBUS;
^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) * If we don't have pte special, then we have to use the pfn_valid()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2134) * based VM_MIXEDMAP scheme (see vm_normal_page), and thus we *must*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2135) * refcount the page if pfn_valid is true (hence insert_page rather
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2136) * than insert_pfn). If a zero_pfn were inserted into a VM_MIXEDMAP
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2137) * without pte special, it would there be refcounted as a normal page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2138) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2139) if (!IS_ENABLED(CONFIG_ARCH_HAS_PTE_SPECIAL) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2140) !pfn_t_devmap(pfn) && pfn_t_valid(pfn)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2141) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2142)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2143) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2144) * At this point we are committed to insert_page()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2145) * regardless of whether the caller specified flags that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2146) * result in pfn_t_has_page() == false.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2147) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2148) page = pfn_to_page(pfn_t_to_pfn(pfn));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2149) err = insert_page(vma, addr, page, pgprot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2150) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2151) return insert_pfn(vma, addr, pfn, pgprot, mkwrite);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2152) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2153)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2154) if (err == -ENOMEM)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2155) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2156) if (err < 0 && err != -EBUSY)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2157) return VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2158)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2159) return VM_FAULT_NOPAGE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2160) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2161)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2162) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2163) * vmf_insert_mixed_prot - insert single pfn into user vma with specified pgprot
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2164) * @vma: user vma to map to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2165) * @addr: target user address of this page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2166) * @pfn: source kernel pfn
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2167) * @pgprot: pgprot flags for the inserted page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2168) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2169) * This is exactly like vmf_insert_mixed(), except that it allows drivers
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2170) * to override pgprot on a per-page basis.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2171) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2172) * Typically this function should be used by drivers to set caching- and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2173) * encryption bits different than those of @vma->vm_page_prot, because
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2174) * the caching- or encryption mode may not be known at mmap() time.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2175) * This is ok as long as @vma->vm_page_prot is not used by the core vm
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2176) * to set caching and encryption bits for those vmas (except for COW pages).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2177) * This is ensured by core vm only modifying these page table entries using
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2178) * functions that don't touch caching- or encryption bits, using pte_modify()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2179) * if needed. (See for example mprotect()).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2180) * Also when new page-table entries are created, this is only done using the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2181) * fault() callback, and never using the value of vma->vm_page_prot,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2182) * except for page-table entries that point to anonymous pages as the result
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2183) * of COW.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2184) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2185) * Context: Process context. May allocate using %GFP_KERNEL.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2186) * Return: vm_fault_t value.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2187) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2188) vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2189) pfn_t pfn, pgprot_t pgprot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2190) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2191) return __vm_insert_mixed(vma, addr, pfn, pgprot, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2192) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2193) EXPORT_SYMBOL(vmf_insert_mixed_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2194)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2195) vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2196) pfn_t pfn)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2197) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2198) return __vm_insert_mixed(vma, addr, pfn, vma->vm_page_prot, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2199) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2200) EXPORT_SYMBOL(vmf_insert_mixed);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2201)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2202) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2203) * If the insertion of PTE failed because someone else already added a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2204) * different entry in the mean time, we treat that as success as we assume
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2205) * the same entry was actually inserted.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2206) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2207) vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2208) unsigned long addr, pfn_t pfn)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2209) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2210) return __vm_insert_mixed(vma, addr, pfn, vma->vm_page_prot, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2211) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2212) EXPORT_SYMBOL(vmf_insert_mixed_mkwrite);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2213)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2214) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2215) * maps a range of physical memory into the requested pages. the old
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2216) * mappings are removed. any references to nonexistent pages results
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2217) * in null mappings (currently treated as "copy-on-access")
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2218) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2219) static int remap_pte_range(struct mm_struct *mm, pmd_t *pmd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2220) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2221) unsigned long pfn, pgprot_t prot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2222) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2223) pte_t *pte, *mapped_pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2224) spinlock_t *ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2225) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2226)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2227) mapped_pte = pte = pte_alloc_map_lock(mm, pmd, addr, &ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2228) if (!pte)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2229) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2230) arch_enter_lazy_mmu_mode();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2231) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2232) BUG_ON(!pte_none(*pte));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2233) if (!pfn_modify_allowed(pfn, prot)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2234) err = -EACCES;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2235) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2236) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2237) set_pte_at(mm, addr, pte, pte_mkspecial(pfn_pte(pfn, prot)));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2238) pfn++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2239) } while (pte++, addr += PAGE_SIZE, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2240) arch_leave_lazy_mmu_mode();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2241) pte_unmap_unlock(mapped_pte, ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2242) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2243) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2244)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2245) static inline int remap_pmd_range(struct mm_struct *mm, pud_t *pud,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2246) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2247) unsigned long pfn, pgprot_t prot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2248) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2249) pmd_t *pmd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2250) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2251) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2252)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2253) pfn -= addr >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2254) pmd = pmd_alloc(mm, pud, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2255) if (!pmd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2256) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2257) VM_BUG_ON(pmd_trans_huge(*pmd));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2258) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2259) next = pmd_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2260) err = remap_pte_range(mm, pmd, addr, next,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2261) pfn + (addr >> PAGE_SHIFT), prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2262) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2263) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2264) } while (pmd++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2265) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2266) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2267)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2268) static inline int remap_pud_range(struct mm_struct *mm, p4d_t *p4d,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2269) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2270) unsigned long pfn, pgprot_t prot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2271) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2272) pud_t *pud;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2273) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2274) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2275)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2276) pfn -= addr >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2277) pud = pud_alloc(mm, p4d, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2278) if (!pud)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2279) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2280) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2281) next = pud_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2282) err = remap_pmd_range(mm, pud, addr, next,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2283) pfn + (addr >> PAGE_SHIFT), prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2284) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2285) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2286) } while (pud++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2287) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2288) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2289)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2290) static inline int remap_p4d_range(struct mm_struct *mm, pgd_t *pgd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2291) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2292) unsigned long pfn, pgprot_t prot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2293) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2294) p4d_t *p4d;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2295) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2296) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2297)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2298) pfn -= addr >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2299) p4d = p4d_alloc(mm, pgd, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2300) if (!p4d)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2301) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2302) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2303) next = p4d_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2304) err = remap_pud_range(mm, p4d, addr, next,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2305) pfn + (addr >> PAGE_SHIFT), prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2306) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2307) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2308) } while (p4d++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2309) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2310) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2311)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2312) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2313) * remap_pfn_range - remap kernel memory to userspace
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2314) * @vma: user vma to map to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2315) * @addr: target page aligned user address to start at
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2316) * @pfn: page frame number of kernel physical memory address
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2317) * @size: size of mapping area
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2318) * @prot: page protection flags for this mapping
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2319) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2320) * Note: this is only safe if the mm semaphore is held when called.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2321) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2322) * Return: %0 on success, negative error code otherwise.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2323) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2324) int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2325) unsigned long pfn, unsigned long size, pgprot_t prot)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2326) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2327) pgd_t *pgd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2328) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2329) unsigned long end = addr + PAGE_ALIGN(size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2330) struct mm_struct *mm = vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2331) unsigned long remap_pfn = pfn;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2332) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2333)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2334) if (WARN_ON_ONCE(!PAGE_ALIGNED(addr)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2335) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2336)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2337) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2338) * Physically remapped pages are special. Tell the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2339) * rest of the world about it:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2340) * VM_IO tells people not to look at these pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2341) * (accesses can have side effects).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2342) * VM_PFNMAP tells the core MM that the base pages are just
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2343) * raw PFN mappings, and do not have a "struct page" associated
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2344) * with them.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2345) * VM_DONTEXPAND
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2346) * Disable vma merging and expanding with mremap().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2347) * VM_DONTDUMP
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2348) * Omit vma from core dump, even when VM_IO turned off.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2349) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2350) * There's a horrible special case to handle copy-on-write
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2351) * behaviour that some programs depend on. We mark the "original"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2352) * un-COW'ed pages by matching them up with "vma->vm_pgoff".
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2353) * See vm_normal_page() for details.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2354) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2355) if (is_cow_mapping(vma->vm_flags)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2356) if (addr != vma->vm_start || end != vma->vm_end)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2357) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2358) vma->vm_pgoff = pfn;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2359) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2360)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2361) err = track_pfn_remap(vma, &prot, remap_pfn, addr, PAGE_ALIGN(size));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2362) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2363) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2364)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2365) vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2366)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2367) BUG_ON(addr >= end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2368) pfn -= addr >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2369) pgd = pgd_offset(mm, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2370) flush_cache_range(vma, addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2371) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2372) next = pgd_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2373) err = remap_p4d_range(mm, pgd, addr, next,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2374) pfn + (addr >> PAGE_SHIFT), prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2375) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2376) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2377) } while (pgd++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2378)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2379) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2380) untrack_pfn(vma, remap_pfn, PAGE_ALIGN(size));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2381)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2382) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2383) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2384) EXPORT_SYMBOL(remap_pfn_range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2385)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2386) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2387) * vm_iomap_memory - remap memory to userspace
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2388) * @vma: user vma to map to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2389) * @start: start of the physical memory to be mapped
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2390) * @len: size of area
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2391) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2392) * This is a simplified io_remap_pfn_range() for common driver use. The
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2393) * driver just needs to give us the physical memory range to be mapped,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2394) * we'll figure out the rest from the vma information.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2395) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2396) * NOTE! Some drivers might want to tweak vma->vm_page_prot first to get
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2397) * whatever write-combining details or similar.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2398) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2399) * Return: %0 on success, negative error code otherwise.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2400) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2401) int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2402) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2403) unsigned long vm_len, pfn, pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2404)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2405) /* Check that the physical memory area passed in looks valid */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2406) if (start + len < start)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2407) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2408) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2409) * You *really* shouldn't map things that aren't page-aligned,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2410) * but we've historically allowed it because IO memory might
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2411) * just have smaller alignment.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2412) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2413) len += start & ~PAGE_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2414) pfn = start >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2415) pages = (len + ~PAGE_MASK) >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2416) if (pfn + pages < pfn)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2417) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2418)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2419) /* We start the mapping 'vm_pgoff' pages into the area */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2420) if (vma->vm_pgoff > pages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2421) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2422) pfn += vma->vm_pgoff;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2423) pages -= vma->vm_pgoff;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2424)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2425) /* Can we fit all of the mapping? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2426) vm_len = vma->vm_end - vma->vm_start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2427) if (vm_len >> PAGE_SHIFT > pages)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2428) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2429)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2430) /* Ok, let it rip */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2431) return io_remap_pfn_range(vma, vma->vm_start, pfn, vm_len, vma->vm_page_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2432) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2433) EXPORT_SYMBOL(vm_iomap_memory);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2434)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2435) static int apply_to_pte_range(struct mm_struct *mm, pmd_t *pmd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2436) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2437) pte_fn_t fn, void *data, bool create,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2438) pgtbl_mod_mask *mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2439) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2440) pte_t *pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2441) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2442) spinlock_t *ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2443)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2444) if (create) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2445) pte = (mm == &init_mm) ?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2446) pte_alloc_kernel_track(pmd, addr, mask) :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2447) pte_alloc_map_lock(mm, pmd, addr, &ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2448) if (!pte)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2449) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2450) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2451) pte = (mm == &init_mm) ?
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2452) pte_offset_kernel(pmd, addr) :
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2453) pte_offset_map_lock(mm, pmd, addr, &ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2454) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2455)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2456) BUG_ON(pmd_huge(*pmd));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2457)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2458) arch_enter_lazy_mmu_mode();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2459)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2460) if (fn) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2461) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2462) if (create || !pte_none(*pte)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2463) err = fn(pte++, addr, data);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2464) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2465) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2466) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2467) } while (addr += PAGE_SIZE, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2468) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2469) *mask |= PGTBL_PTE_MODIFIED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2470)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2471) arch_leave_lazy_mmu_mode();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2472)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2473) if (mm != &init_mm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2474) pte_unmap_unlock(pte-1, ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2475) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2476) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2477)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2478) static int apply_to_pmd_range(struct mm_struct *mm, pud_t *pud,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2479) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2480) pte_fn_t fn, void *data, bool create,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2481) pgtbl_mod_mask *mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2482) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2483) pmd_t *pmd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2484) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2485) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2486)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2487) BUG_ON(pud_huge(*pud));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2488)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2489) if (create) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2490) pmd = pmd_alloc_track(mm, pud, addr, mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2491) if (!pmd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2492) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2493) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2494) pmd = pmd_offset(pud, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2495) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2496) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2497) next = pmd_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2498) if (create || !pmd_none_or_clear_bad(pmd)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2499) err = apply_to_pte_range(mm, pmd, addr, next, fn, data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2500) create, mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2501) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2502) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2503) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2504) } while (pmd++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2505) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2506) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2507)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2508) static int apply_to_pud_range(struct mm_struct *mm, p4d_t *p4d,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2509) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2510) pte_fn_t fn, void *data, bool create,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2511) pgtbl_mod_mask *mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2512) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2513) pud_t *pud;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2514) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2515) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2516)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2517) if (create) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2518) pud = pud_alloc_track(mm, p4d, addr, mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2519) if (!pud)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2520) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2521) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2522) pud = pud_offset(p4d, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2523) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2524) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2525) next = pud_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2526) if (create || !pud_none_or_clear_bad(pud)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2527) err = apply_to_pmd_range(mm, pud, addr, next, fn, data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2528) create, mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2529) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2530) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2531) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2532) } while (pud++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2533) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2534) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2535)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2536) static int apply_to_p4d_range(struct mm_struct *mm, pgd_t *pgd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2537) unsigned long addr, unsigned long end,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2538) pte_fn_t fn, void *data, bool create,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2539) pgtbl_mod_mask *mask)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2540) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2541) p4d_t *p4d;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2542) unsigned long next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2543) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2544)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2545) if (create) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2546) p4d = p4d_alloc_track(mm, pgd, addr, mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2547) if (!p4d)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2548) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2549) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2550) p4d = p4d_offset(pgd, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2551) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2552) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2553) next = p4d_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2554) if (create || !p4d_none_or_clear_bad(p4d)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2555) err = apply_to_pud_range(mm, p4d, addr, next, fn, data,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2556) create, mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2557) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2558) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2559) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2560) } while (p4d++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2561) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2562) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2563)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2564) static int __apply_to_page_range(struct mm_struct *mm, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2565) unsigned long size, pte_fn_t fn,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2566) void *data, bool create)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2567) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2568) pgd_t *pgd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2569) unsigned long start = addr, next;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2570) unsigned long end = addr + size;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2571) pgtbl_mod_mask mask = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2572) int err = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2573)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2574) if (WARN_ON(addr >= end))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2575) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2576)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2577) pgd = pgd_offset(mm, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2578) do {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2579) next = pgd_addr_end(addr, end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2580) if (!create && pgd_none_or_clear_bad(pgd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2581) continue;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2582) err = apply_to_p4d_range(mm, pgd, addr, next, fn, data, create, &mask);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2583) if (err)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2584) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2585) } while (pgd++, addr = next, addr != end);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2586)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2587) if (mask & ARCH_PAGE_TABLE_SYNC_MASK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2588) arch_sync_kernel_mappings(start, start + size);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2589)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2590) return err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2591) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2592)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2593) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2594) * Scan a region of virtual memory, filling in page tables as necessary
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2595) * and calling a provided function on each leaf page table.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2596) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2597) int apply_to_page_range(struct mm_struct *mm, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2598) unsigned long size, pte_fn_t fn, void *data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2599) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2600) return __apply_to_page_range(mm, addr, size, fn, data, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2601) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2602) EXPORT_SYMBOL_GPL(apply_to_page_range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2603)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2604) #ifdef CONFIG_SPECULATIVE_PAGE_FAULT
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2605) static bool pte_spinlock(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2606) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2607) bool ret = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2608) #ifdef CONFIG_TRANSPARENT_HUGEPAGE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2609) pmd_t pmdval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2610) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2611)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2612) /* Check if vma is still valid */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2613) if (!(vmf->flags & FAULT_FLAG_SPECULATIVE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2614) vmf->ptl = pte_lockptr(vmf->vma->vm_mm, vmf->pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2615) spin_lock(vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2616) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2617) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2618)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2619) local_irq_disable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2620) if (vma_has_changed(vmf)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2621) trace_spf_vma_changed(_RET_IP_, vmf->vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2622) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2623) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2624)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2625) #ifdef CONFIG_TRANSPARENT_HUGEPAGE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2626) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2627) * We check if the pmd value is still the same to ensure that there
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2628) * is not a huge collapse operation in progress in our back.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2629) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2630) pmdval = READ_ONCE(*vmf->pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2631) if (!pmd_same(pmdval, vmf->orig_pmd)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2632) trace_spf_pmd_changed(_RET_IP_, vmf->vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2633) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2634) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2635) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2636)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2637) vmf->ptl = pte_lockptr(vmf->vma->vm_mm, vmf->pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2638) if (unlikely(!spin_trylock(vmf->ptl))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2639) trace_spf_pte_lock(_RET_IP_, vmf->vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2640) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2641) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2642)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2643) if (vma_has_changed(vmf)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2644) spin_unlock(vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2645) trace_spf_vma_changed(_RET_IP_, vmf->vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2646) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2647) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2648)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2649) ret = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2650) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2651) local_irq_enable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2652) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2653) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2654)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2655) static bool __pte_map_lock_speculative(struct vm_fault *vmf, unsigned long addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2656) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2657) bool ret = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2658) pte_t *pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2659) spinlock_t *ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2660) #ifdef CONFIG_TRANSPARENT_HUGEPAGE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2661) pmd_t pmdval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2662) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2663)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2664) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2665) * The first vma_has_changed() guarantees the page-tables are still
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2666) * valid, having IRQs disabled ensures they stay around, hence the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2667) * second vma_has_changed() to make sure they are still valid once
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2668) * we've got the lock. After that a concurrent zap_pte_range() will
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2669) * block on the PTL and thus we're safe.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2670) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2671) local_irq_disable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2672) if (vma_has_changed(vmf)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2673) trace_spf_vma_changed(_RET_IP_, vmf->vma, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2674) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2675) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2676)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2677) #ifdef CONFIG_TRANSPARENT_HUGEPAGE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2678) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2679) * We check if the pmd value is still the same to ensure that there
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2680) * is not a huge collapse operation in progress in our back.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2681) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2682) pmdval = READ_ONCE(*vmf->pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2683) if (!pmd_same(pmdval, vmf->orig_pmd)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2684) trace_spf_pmd_changed(_RET_IP_, vmf->vma, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2685) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2686) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2687) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2688)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2689) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2690) * Same as pte_offset_map_lock() except that we call
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2691) * spin_trylock() in place of spin_lock() to avoid race with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2692) * unmap path which may have the lock and wait for this CPU
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2693) * to invalidate TLB but this CPU has irq disabled.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2694) * Since we are in a speculative patch, accept it could fail
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2695) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2696) ptl = pte_lockptr(vmf->vma->vm_mm, vmf->pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2697) pte = pte_offset_map(vmf->pmd, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2698) if (unlikely(!spin_trylock(ptl))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2699) pte_unmap(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2700) trace_spf_pte_lock(_RET_IP_, vmf->vma, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2701) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2702) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2703)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2704) if (vma_has_changed(vmf)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2705) pte_unmap_unlock(pte, ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2706) trace_spf_vma_changed(_RET_IP_, vmf->vma, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2707) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2708) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2709)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2710) vmf->pte = pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2711) vmf->ptl = ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2712) ret = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2713) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2714) local_irq_enable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2715) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2716) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2717)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2718) static bool pte_map_lock(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2719) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2720) if (!(vmf->flags & FAULT_FLAG_SPECULATIVE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2721) vmf->pte = pte_offset_map_lock(vmf->vma->vm_mm, vmf->pmd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2722) vmf->address, &vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2723) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2724) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2725)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2726) return __pte_map_lock_speculative(vmf, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2727) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2728)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2729) bool pte_map_lock_addr(struct vm_fault *vmf, unsigned long addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2730) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2731) if (!(vmf->flags & FAULT_FLAG_SPECULATIVE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2732) vmf->pte = pte_offset_map_lock(vmf->vma->vm_mm, vmf->pmd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2733) addr, &vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2734) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2735) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2736)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2737) return __pte_map_lock_speculative(vmf, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2738) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2739)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2740) static bool __read_mostly allow_file_spec_access;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2741) static int __init allow_file_spec_access_setup(char *str)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2742) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2743) allow_file_spec_access = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2744) return 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2745) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2746) __setup("allow_file_spec_access", allow_file_spec_access_setup);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2747)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2748) static bool vmf_allows_speculation(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2749) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2750) if (vma_is_anonymous(vmf->vma)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2751) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2752) * __anon_vma_prepare() requires the mmap_sem to be held
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2753) * because vm_next and vm_prev must be safe. This can't be
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2754) * guaranteed in the speculative path.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2755) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2756) if (!vmf->vma->anon_vma) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2757) trace_spf_vma_notsup(_RET_IP_, vmf->vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2758) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2759) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2760) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2761) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2762)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2763) if (!allow_file_spec_access) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2764) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2765) * Can't call vm_ops service has we don't know what they would
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2766) * do with the VMA.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2767) * This include huge page from hugetlbfs.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2768) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2769) trace_spf_vma_notsup(_RET_IP_, vmf->vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2770) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2771) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2772)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2773) if (!(vmf->vma->vm_flags & VM_SHARED) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2774) (vmf->flags & FAULT_FLAG_WRITE) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2775) !vmf->vma->anon_vma) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2776) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2777) * non-anonymous private COW without anon_vma.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2778) * See above.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2779) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2780) trace_spf_vma_notsup(_RET_IP_, vmf->vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2781) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2782) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2783)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2784) if (vmf->vma->vm_ops->allow_speculation &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2785) vmf->vma->vm_ops->allow_speculation()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2786) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2787) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2788)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2789) trace_spf_vma_notsup(_RET_IP_, vmf->vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2790) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2791) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2792)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2793) #else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2794) static inline bool pte_spinlock(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2795) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2796) vmf->ptl = pte_lockptr(vmf->vma->vm_mm, vmf->pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2797) spin_lock(vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2798) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2799) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2800)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2801) static inline bool pte_map_lock(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2802) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2803) vmf->pte = pte_offset_map_lock(vmf->vma->vm_mm, vmf->pmd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2804) vmf->address, &vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2805) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2806) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2807)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2808) inline bool pte_map_lock_addr(struct vm_fault *vmf, unsigned long addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2809) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2810) vmf->pte = pte_offset_map_lock(vmf->vma->vm_mm, vmf->pmd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2811) addr, &vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2812) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2813) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2814)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2815) static inline bool vmf_allows_speculation(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2816) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2817) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2818) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2819) #endif /* CONFIG_SPECULATIVE_PAGE_FAULT */
^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) * Scan a region of virtual memory, calling a provided function on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2823) * each leaf page table where it exists.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2824) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2825) * Unlike apply_to_page_range, this does _not_ fill in page tables
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2826) * where they are absent.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2827) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2828) int apply_to_existing_page_range(struct mm_struct *mm, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2829) unsigned long size, pte_fn_t fn, void *data)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2830) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2831) return __apply_to_page_range(mm, addr, size, fn, data, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2832) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2833) EXPORT_SYMBOL_GPL(apply_to_existing_page_range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2834)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2835) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2836) * handle_pte_fault chooses page fault handler according to an entry which was
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2837) * read non-atomically. Before making any commitment, on those architectures
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2838) * or configurations (e.g. i386 with PAE) which might give a mix of unmatched
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2839) * parts, do_swap_page must check under lock before unmapping the pte and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2840) * proceeding (but do_wp_page is only called after already making such a check;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2841) * and do_anonymous_page can safely check later on).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2842) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2843) * pte_unmap_same() returns:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2844) * 0 if the PTE are the same
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2845) * VM_FAULT_PTNOTSAME if the PTE are different
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2846) * VM_FAULT_RETRY if the VMA has changed in our back during
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2847) * a speculative page fault handling.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2848) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2849) static inline int pte_unmap_same(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2850) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2851) int ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2852)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2853) #if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2854) if (sizeof(pte_t) > sizeof(unsigned long)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2855) if (pte_spinlock(vmf)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2856) if (!pte_same(*vmf->pte, vmf->orig_pte))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2857) ret = VM_FAULT_PTNOTSAME;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2858) spin_unlock(vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2859) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2860) ret = VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2861) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2862) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2863) pte_unmap(vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2864) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2865) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2866)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2867) static inline bool cow_user_page(struct page *dst, struct page *src,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2868) struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2869) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2870) bool ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2871) void *kaddr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2872) void __user *uaddr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2873) bool locked = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2874) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2875) struct mm_struct *mm = vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2876) unsigned long addr = vmf->address;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2877)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2878) if (likely(src)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2879) copy_user_highpage(dst, src, addr, vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2880) return true;
^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) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2884) * If the source page was a PFN mapping, we don't have
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2885) * a "struct page" for it. We do a best-effort copy by
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2886) * just copying from the original user address. If that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2887) * fails, we just zero-fill it. Live with it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2888) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2889) kaddr = kmap_atomic(dst);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2890) uaddr = (void __user *)(addr & PAGE_MASK);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2891)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2892) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2893) * On architectures with software "accessed" bits, we would
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2894) * take a double page fault, so mark it accessed here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2895) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2896) if (arch_faults_on_old_pte() && !pte_young(vmf->orig_pte)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2897) pte_t entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2898)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2899) vmf->pte = pte_offset_map_lock(mm, vmf->pmd, addr, &vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2900) locked = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2901) if (!likely(pte_same(*vmf->pte, vmf->orig_pte))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2902) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2903) * Other thread has already handled the fault
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2904) * and update local tlb only
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2905) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2906) update_mmu_tlb(vma, addr, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2907) ret = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2908) goto pte_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2909) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2910)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2911) entry = pte_mkyoung(vmf->orig_pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2912) if (ptep_set_access_flags(vma, addr, vmf->pte, entry, 0))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2913) update_mmu_cache(vma, addr, vmf->pte);
^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) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2917) * This really shouldn't fail, because the page is there
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2918) * in the page tables. But it might just be unreadable,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2919) * in which case we just give up and fill the result with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2920) * zeroes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2921) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2922) if (__copy_from_user_inatomic(kaddr, uaddr, PAGE_SIZE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2923) if (locked)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2924) goto warn;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2925)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2926) /* Re-validate under PTL if the page is still mapped */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2927) vmf->pte = pte_offset_map_lock(mm, vmf->pmd, addr, &vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2928) locked = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2929) if (!likely(pte_same(*vmf->pte, vmf->orig_pte))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2930) /* The PTE changed under us, update local tlb */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2931) update_mmu_tlb(vma, addr, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2932) ret = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2933) goto pte_unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2934) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2935)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2936) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2937) * The same page can be mapped back since last copy attempt.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2938) * Try to copy again under PTL.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2939) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2940) if (__copy_from_user_inatomic(kaddr, uaddr, PAGE_SIZE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2941) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2942) * Give a warn in case there can be some obscure
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2943) * use-case
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2944) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2945) warn:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2946) WARN_ON_ONCE(1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2947) clear_page(kaddr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2948) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2949) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2950)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2951) ret = true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2952)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2953) pte_unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2954) if (locked)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2955) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2956) kunmap_atomic(kaddr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2957) flush_dcache_page(dst);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2958)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2959) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2960) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2961)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2962) static gfp_t __get_fault_gfp_mask(struct vm_area_struct *vma)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2963) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2964) struct file *vm_file = vma->vm_file;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2965)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2966) if (vm_file)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2967) return mapping_gfp_mask(vm_file->f_mapping) | __GFP_FS | __GFP_IO;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2968)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2969) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2970) * Special mappings (e.g. VDSO) do not have any file so fake
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2971) * a default GFP_KERNEL for them.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2972) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2973) return GFP_KERNEL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2974) }
^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) * Notify the address space that the page is about to become writable so that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2978) * it can prohibit this or wait for the page to get into an appropriate state.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2979) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2980) * We do this without the lock held, so that it can sleep if it needs to.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2981) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2982) static vm_fault_t do_page_mkwrite(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2983) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2984) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2985) struct page *page = vmf->page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2986) unsigned int old_flags = vmf->flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2987)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2988) vmf->flags = FAULT_FLAG_WRITE|FAULT_FLAG_MKWRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2989)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2990) if (vmf->vma->vm_file &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2991) IS_SWAPFILE(vmf->vma->vm_file->f_mapping->host))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2992) return VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2993)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2994) ret = vmf->vma->vm_ops->page_mkwrite(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2995) /* Restore original flags so that caller is not surprised */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2996) vmf->flags = old_flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2997) if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2998) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2999) if (unlikely(!(ret & VM_FAULT_LOCKED))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3000) lock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3001) if (!page->mapping) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3002) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3003) return 0; /* retry */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3004) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3005) ret |= VM_FAULT_LOCKED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3006) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3007) VM_BUG_ON_PAGE(!PageLocked(page), page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3008) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3009) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3010)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3011) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3012) * Handle dirtying of a page in shared file mapping on a write fault.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3013) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3014) * The function expects the page to be locked and unlocks it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3015) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3016) static vm_fault_t fault_dirty_shared_page(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3017) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3018) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3019) struct address_space *mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3020) struct page *page = vmf->page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3021) bool dirtied;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3022) bool page_mkwrite = vma->vm_ops && vma->vm_ops->page_mkwrite;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3023)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3024) dirtied = set_page_dirty(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3025) VM_BUG_ON_PAGE(PageAnon(page), page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3026) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3027) * Take a local copy of the address_space - page.mapping may be zeroed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3028) * by truncate after unlock_page(). The address_space itself remains
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3029) * pinned by vma->vm_file's reference. We rely on unlock_page()'s
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3030) * release semantics to prevent the compiler from undoing this copying.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3031) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3032) mapping = page_rmapping(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3033) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3034)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3035) if (!page_mkwrite)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3036) file_update_time(vma->vm_file);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3037)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3038) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3039) * Throttle page dirtying rate down to writeback speed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3040) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3041) * mapping may be NULL here because some device drivers do not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3042) * set page.mapping but still dirty their pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3043) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3044) * Drop the mmap_lock before waiting on IO, if we can. The file
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3045) * is pinning the mapping, as per above.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3046) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3047) if ((dirtied || page_mkwrite) && mapping) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3048) struct file *fpin;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3049)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3050) fpin = maybe_unlock_mmap_for_io(vmf, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3051) balance_dirty_pages_ratelimited(mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3052) if (fpin) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3053) fput(fpin);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3054) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3055) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3056) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3057)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3058) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3059) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3060)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3061) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3062) * Handle write page faults for pages that can be reused in the current vma
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3063) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3064) * This can happen either due to the mapping being with the VM_SHARED flag,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3065) * or due to us being the last reference standing to the page. In either
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3066) * case, all we need to do here is to mark the page as writable and update
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3067) * any related book-keeping.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3068) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3069) static inline void wp_page_reuse(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3070) __releases(vmf->ptl)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3071) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3072) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3073) struct page *page = vmf->page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3074) pte_t entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3075) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3076) * Clear the pages cpupid information as the existing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3077) * information potentially belongs to a now completely
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3078) * unrelated process.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3079) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3080) if (page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3081) page_cpupid_xchg_last(page, (1 << LAST_CPUPID_SHIFT) - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3082)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3083) flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3084) entry = pte_mkyoung(vmf->orig_pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3085) entry = maybe_mkwrite(pte_mkdirty(entry), vmf->vma_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3086) if (ptep_set_access_flags(vma, vmf->address, vmf->pte, entry, 1))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3087) update_mmu_cache(vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3088) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3089) count_vm_event(PGREUSE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3090) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3091)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3092) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3093) * Handle the case of a page which we actually need to copy to a new page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3094) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3095) * Called with mmap_lock locked and the old page referenced, but
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3096) * without the ptl held.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3097) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3098) * High level logic flow:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3099) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3100) * - Allocate a page, copy the content of the old page to the new one.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3101) * - Handle book keeping and accounting - cgroups, mmu-notifiers, etc.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3102) * - Take the PTL. If the pte changed, bail out and release the allocated page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3103) * - If the pte is still the way we remember it, update the page table and all
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3104) * relevant references. This includes dropping the reference the page-table
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3105) * held to the old page, as well as updating the rmap.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3106) * - In any case, unlock the PTL and drop the reference we took to the old page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3107) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3108) static vm_fault_t wp_page_copy(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3109) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3110) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3111) struct mm_struct *mm = vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3112) struct page *old_page = vmf->page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3113) struct page *new_page = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3114) pte_t entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3115) int page_copied = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3116) struct mmu_notifier_range range;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3117) vm_fault_t ret = VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3118)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3119) if (unlikely(anon_vma_prepare(vma)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3120) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3121)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3122) if (is_zero_pfn(pte_pfn(vmf->orig_pte))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3123) new_page = alloc_zeroed_user_highpage_movable(vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3124) vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3125) if (!new_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3126) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3127) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3128) new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3129) vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3130) if (!new_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3131) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3132)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3133) if (!cow_user_page(new_page, old_page, vmf)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3134) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3135) * COW failed, if the fault was solved by other,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3136) * it's fine. If not, userspace would re-fault on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3137) * the same address and we will handle the fault
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3138) * from the second attempt.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3139) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3140) put_page(new_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3141) if (old_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3142) put_page(old_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3143) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3144) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3145) trace_android_vh_cow_user_page(vmf, new_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3146) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3147)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3148) if (mem_cgroup_charge(new_page, mm, GFP_KERNEL))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3149) goto out_free_new;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3150) cgroup_throttle_swaprate(new_page, GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3151)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3152) __SetPageUptodate(new_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3153)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3154) mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3155) vmf->address & PAGE_MASK,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3156) (vmf->address & PAGE_MASK) + PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3157) mmu_notifier_invalidate_range_start(&range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3158)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3159) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3160) * Re-check the pte - we dropped the lock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3161) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3162) if (!pte_map_lock(vmf)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3163) ret = VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3164) goto out_invalidate_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3165) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3166) if (likely(pte_same(*vmf->pte, vmf->orig_pte))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3167) if (old_page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3168) if (!PageAnon(old_page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3169) dec_mm_counter_fast(mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3170) mm_counter_file(old_page));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3171) inc_mm_counter_fast(mm, MM_ANONPAGES);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3172) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3173) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3174) inc_mm_counter_fast(mm, MM_ANONPAGES);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3175) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3176) flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3177) entry = mk_pte(new_page, vmf->vma_page_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3178) entry = pte_sw_mkyoung(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3179) entry = maybe_mkwrite(pte_mkdirty(entry), vmf->vma_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3180) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3181) * Clear the pte entry and flush it first, before updating the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3182) * pte with the new entry. This will avoid a race condition
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3183) * seen in the presence of one thread doing SMC and another
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3184) * thread doing COW.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3185) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3186) ptep_clear_flush_notify(vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3187) __page_add_new_anon_rmap(new_page, vma, vmf->address, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3188) __lru_cache_add_inactive_or_unevictable(new_page, vmf->vma_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3189) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3190) * We call the notify macro here because, when using secondary
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3191) * mmu page tables (such as kvm shadow page tables), we want the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3192) * new page to be mapped directly into the secondary page table.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3193) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3194) set_pte_at_notify(mm, vmf->address, vmf->pte, entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3195) update_mmu_cache(vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3196) if (old_page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3197) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3198) * Only after switching the pte to the new page may
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3199) * we remove the mapcount here. Otherwise another
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3200) * process may come and find the rmap count decremented
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3201) * before the pte is switched to the new page, and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3202) * "reuse" the old page writing into it while our pte
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3203) * here still points into it and can be read by other
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3204) * threads.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3205) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3206) * The critical issue is to order this
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3207) * page_remove_rmap with the ptp_clear_flush above.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3208) * Those stores are ordered by (if nothing else,)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3209) * the barrier present in the atomic_add_negative
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3210) * in page_remove_rmap.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3211) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3212) * Then the TLB flush in ptep_clear_flush ensures that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3213) * no process can access the old page before the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3214) * decremented mapcount is visible. And the old page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3215) * cannot be reused until after the decremented
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3216) * mapcount is visible. So transitively, TLBs to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3217) * old page will be flushed before it can be reused.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3218) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3219) page_remove_rmap(old_page, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3220) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3221)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3222) /* Free the old page.. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3223) new_page = old_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3224) page_copied = 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3225) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3226) update_mmu_tlb(vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3227) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3228)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3229) if (new_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3230) put_page(new_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3231)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3232) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3233) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3234) * No need to double call mmu_notifier->invalidate_range() callback as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3235) * the above ptep_clear_flush_notify() did already call it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3236) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3237) mmu_notifier_invalidate_range_only_end(&range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3238) if (old_page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3239) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3240) * Don't let another task, with possibly unlocked vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3241) * keep the mlocked page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3242) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3243) if (page_copied && (vmf->vma_flags & VM_LOCKED)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3244) lock_page(old_page); /* LRU manipulation */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3245) if (PageMlocked(old_page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3246) munlock_vma_page(old_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3247) unlock_page(old_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3248) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3249) put_page(old_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3250) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3251) return page_copied ? VM_FAULT_WRITE : 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3252) out_invalidate_end:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3253) mmu_notifier_invalidate_range_only_end(&range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3254) out_free_new:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3255) put_page(new_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3256) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3257) if (old_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3258) put_page(old_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3259) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3260) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3261)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3262) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3263) * finish_mkwrite_fault - finish page fault for a shared mapping, making PTE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3264) * writeable once the page is prepared
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3265) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3266) * @vmf: structure describing the fault
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3267) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3268) * This function handles all that is needed to finish a write page fault in a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3269) * shared mapping due to PTE being read-only once the mapped page is prepared.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3270) * It handles locking of PTE and modifying it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3271) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3272) * The function expects the page to be locked or other protection against
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3273) * concurrent faults / writeback (such as DAX radix tree locks).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3274) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3275) * Return: %VM_FAULT_WRITE on success, %0 when PTE got changed before
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3276) * we acquired PTE lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3277) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3278) vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3279) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3280) WARN_ON_ONCE(!(vmf->vma_flags & VM_SHARED));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3281) if (!pte_map_lock(vmf))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3282) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3283) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3284) * We might have raced with another page fault while we released the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3285) * pte_offset_map_lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3286) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3287) if (!pte_same(*vmf->pte, vmf->orig_pte)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3288) update_mmu_tlb(vmf->vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3289) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3290) return VM_FAULT_NOPAGE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3291) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3292) wp_page_reuse(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3293) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3294) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3295)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3296) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3297) * Handle write page faults for VM_MIXEDMAP or VM_PFNMAP for a VM_SHARED
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3298) * mapping
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3299) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3300) static vm_fault_t wp_pfn_shared(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3301) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3302) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3303)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3304) if (vma->vm_ops && vma->vm_ops->pfn_mkwrite) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3305) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3306)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3307) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3308) vmf->flags |= FAULT_FLAG_MKWRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3309) ret = vma->vm_ops->pfn_mkwrite(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3310) if (ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3311) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3312) return finish_mkwrite_fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3313) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3314) wp_page_reuse(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3315) return VM_FAULT_WRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3316) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3317)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3318) static vm_fault_t wp_page_shared(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3319) __releases(vmf->ptl)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3320) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3321) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3322) vm_fault_t ret = VM_FAULT_WRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3323)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3324) get_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3325)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3326) if (vma->vm_ops && vma->vm_ops->page_mkwrite) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3327) vm_fault_t tmp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3328)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3329) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3330) tmp = do_page_mkwrite(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3331) if (unlikely(!tmp || (tmp &
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3332) (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3333) put_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3334) return tmp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3335) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3336) tmp = finish_mkwrite_fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3337) if (unlikely(tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3338) unlock_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3339) put_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3340) return tmp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3341) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3342) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3343) wp_page_reuse(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3344) lock_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3345) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3346) ret |= fault_dirty_shared_page(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3347) put_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3348)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3349) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3350) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3351)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3352) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3353) * This routine handles present pages, when users try to write
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3354) * to a shared page. It is done by copying the page to a new address
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3355) * and decrementing the shared-page counter for the old page.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3356) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3357) * Note that this routine assumes that the protection checks have been
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3358) * done by the caller (the low-level page fault routine in most cases).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3359) * Thus we can safely just mark it writable once we've done any necessary
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3360) * COW.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3361) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3362) * We also mark the page dirty at this point even though the page will
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3363) * change only once the write actually happens. This avoids a few races,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3364) * and potentially makes it more efficient.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3365) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3366) * We enter with non-exclusive mmap_lock (to exclude vma changes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3367) * but allow concurrent faults), with pte both mapped and locked.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3368) * We return with mmap_lock still held, but pte unmapped and unlocked.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3369) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3370) static vm_fault_t do_wp_page(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3371) __releases(vmf->ptl)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3372) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3373) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3374)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3375) if (userfaultfd_pte_wp(vma, *vmf->pte)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3376) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3377) return handle_userfault(vmf, VM_UFFD_WP);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3378) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3379)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3380) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3381) * Userfaultfd write-protect can defer flushes. Ensure the TLB
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3382) * is flushed in this case before copying.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3383) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3384) if (unlikely(userfaultfd_wp(vmf->vma) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3385) mm_tlb_flush_pending(vmf->vma->vm_mm)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3386) flush_tlb_page(vmf->vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3387)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3388) vmf->page = _vm_normal_page(vma, vmf->address, vmf->orig_pte,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3389) vmf->vma_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3390) if (!vmf->page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3391) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3392) * VM_MIXEDMAP !pfn_valid() case, or VM_SOFTDIRTY clear on a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3393) * VM_PFNMAP VMA.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3394) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3395) * We should not cow pages in a shared writeable mapping.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3396) * Just mark the pages writable and/or call ops->pfn_mkwrite.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3397) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3398) if ((vmf->vma_flags & (VM_WRITE|VM_SHARED)) ==
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3399) (VM_WRITE|VM_SHARED))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3400) return wp_pfn_shared(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3401)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3402) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3403) return wp_page_copy(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3404) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3405)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3406) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3407) * Take out anonymous pages first, anonymous shared vmas are
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3408) * not dirty accountable.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3409) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3410) if (PageAnon(vmf->page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3411) struct page *page = vmf->page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3412)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3413) /* PageKsm() doesn't necessarily raise the page refcount */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3414) if (PageKsm(page) || page_count(page) != 1)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3415) goto copy;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3416) if (!trylock_page(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3417) goto copy;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3418) if (PageKsm(page) || page_mapcount(page) != 1 || page_count(page) != 1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3419) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3420) goto copy;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3421) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3422) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3423) * Ok, we've got the only map reference, and the only
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3424) * page count reference, and the page is locked,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3425) * it's dark out, and we're wearing sunglasses. Hit it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3426) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3427) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3428) wp_page_reuse(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3429) return VM_FAULT_WRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3430) } else if (unlikely((vmf->vma_flags & (VM_WRITE|VM_SHARED)) ==
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3431) (VM_WRITE|VM_SHARED))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3432) return wp_page_shared(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3433) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3434) copy:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3435) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3436) * Ok, we need to copy. Oh, well..
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3437) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3438) get_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3439)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3440) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3441) return wp_page_copy(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3442) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3443)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3444) static void unmap_mapping_range_vma(struct vm_area_struct *vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3445) unsigned long start_addr, unsigned long end_addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3446) struct zap_details *details)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3447) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3448) zap_page_range_single(vma, start_addr, end_addr - start_addr, details);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3449) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3450)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3451) static inline void unmap_mapping_range_tree(struct rb_root_cached *root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3452) struct zap_details *details)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3453) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3454) struct vm_area_struct *vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3455) pgoff_t vba, vea, zba, zea;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3456)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3457) vma_interval_tree_foreach(vma, root,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3458) details->first_index, details->last_index) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3459)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3460) vba = vma->vm_pgoff;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3461) vea = vba + vma_pages(vma) - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3462) zba = details->first_index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3463) if (zba < vba)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3464) zba = vba;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3465) zea = details->last_index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3466) if (zea > vea)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3467) zea = vea;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3468)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3469) unmap_mapping_range_vma(vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3470) ((zba - vba) << PAGE_SHIFT) + vma->vm_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3471) ((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3472) details);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3473) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3474) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3475)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3476) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3477) * unmap_mapping_page() - Unmap single page from processes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3478) * @page: The locked page to be unmapped.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3479) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3480) * Unmap this page from any userspace process which still has it mmaped.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3481) * Typically, for efficiency, the range of nearby pages has already been
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3482) * unmapped by unmap_mapping_pages() or unmap_mapping_range(). But once
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3483) * truncation or invalidation holds the lock on a page, it may find that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3484) * the page has been remapped again: and then uses unmap_mapping_page()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3485) * to unmap it finally.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3486) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3487) void unmap_mapping_page(struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3488) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3489) struct address_space *mapping = page->mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3490) struct zap_details details = { };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3491)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3492) VM_BUG_ON(!PageLocked(page));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3493) VM_BUG_ON(PageTail(page));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3494)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3495) details.check_mapping = mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3496) details.first_index = page->index;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3497) details.last_index = page->index + thp_nr_pages(page) - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3498) details.single_page = page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3499)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3500) i_mmap_lock_write(mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3501) if (unlikely(!RB_EMPTY_ROOT(&mapping->i_mmap.rb_root)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3502) unmap_mapping_range_tree(&mapping->i_mmap, &details);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3503) i_mmap_unlock_write(mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3504) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3505)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3506) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3507) * unmap_mapping_pages() - Unmap pages from processes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3508) * @mapping: The address space containing pages to be unmapped.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3509) * @start: Index of first page to be unmapped.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3510) * @nr: Number of pages to be unmapped. 0 to unmap to end of file.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3511) * @even_cows: Whether to unmap even private COWed pages.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3512) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3513) * Unmap the pages in this address space from any userspace process which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3514) * has them mmaped. Generally, you want to remove COWed pages as well when
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3515) * a file is being truncated, but not when invalidating pages from the page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3516) * cache.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3517) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3518) void unmap_mapping_pages(struct address_space *mapping, pgoff_t start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3519) pgoff_t nr, bool even_cows)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3520) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3521) struct zap_details details = { };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3522)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3523) details.check_mapping = even_cows ? NULL : mapping;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3524) details.first_index = start;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3525) details.last_index = start + nr - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3526) if (details.last_index < details.first_index)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3527) details.last_index = ULONG_MAX;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3528)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3529) i_mmap_lock_write(mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3530) if (unlikely(!RB_EMPTY_ROOT(&mapping->i_mmap.rb_root)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3531) unmap_mapping_range_tree(&mapping->i_mmap, &details);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3532) i_mmap_unlock_write(mapping);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3533) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3534)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3535) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3536) * unmap_mapping_range - unmap the portion of all mmaps in the specified
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3537) * address_space corresponding to the specified byte range in the underlying
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3538) * file.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3539) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3540) * @mapping: the address space containing mmaps to be unmapped.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3541) * @holebegin: byte in first page to unmap, relative to the start of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3542) * the underlying file. This will be rounded down to a PAGE_SIZE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3543) * boundary. Note that this is different from truncate_pagecache(), which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3544) * must keep the partial page. In contrast, we must get rid of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3545) * partial pages.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3546) * @holelen: size of prospective hole in bytes. This will be rounded
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3547) * up to a PAGE_SIZE boundary. A holelen of zero truncates to the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3548) * end of the file.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3549) * @even_cows: 1 when truncating a file, unmap even private COWed pages;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3550) * but 0 when invalidating pagecache, don't throw away private data.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3551) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3552) void unmap_mapping_range(struct address_space *mapping,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3553) loff_t const holebegin, loff_t const holelen, int even_cows)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3554) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3555) pgoff_t hba = holebegin >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3556) pgoff_t hlen = (holelen + PAGE_SIZE - 1) >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3557)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3558) /* Check for overflow. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3559) if (sizeof(holelen) > sizeof(hlen)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3560) long long holeend =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3561) (holebegin + holelen + PAGE_SIZE - 1) >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3562) if (holeend & ~(long long)ULONG_MAX)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3563) hlen = ULONG_MAX - hba + 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3564) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3565)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3566) unmap_mapping_pages(mapping, hba, hlen, even_cows);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3567) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3568) EXPORT_SYMBOL(unmap_mapping_range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3569)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3570) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3571) * We enter with non-exclusive mmap_lock (to exclude vma changes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3572) * but allow concurrent faults), and pte mapped but not yet locked.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3573) * We return with pte unmapped and unlocked.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3574) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3575) * We return with the mmap_lock locked or unlocked in the same cases
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3576) * as does filemap_fault().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3577) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3578) vm_fault_t do_swap_page(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3579) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3580) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3581) struct page *page = NULL, *swapcache;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3582) swp_entry_t entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3583) pte_t pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3584) int locked;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3585) int exclusive = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3586) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3587) void *shadow = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3588)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3589) ret = pte_unmap_same(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3590) if (ret) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3591) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3592) * If pte != orig_pte, this means another thread did the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3593) * swap operation in our back.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3594) * So nothing else to do.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3595) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3596) if (ret == VM_FAULT_PTNOTSAME)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3597) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3598) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3599) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3600)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3601) entry = pte_to_swp_entry(vmf->orig_pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3602) if (unlikely(non_swap_entry(entry))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3603) if (is_migration_entry(entry)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3604) migration_entry_wait(vma->vm_mm, vmf->pmd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3605) vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3606) } else if (is_device_private_entry(entry)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3607) vmf->page = device_private_entry_to_page(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3608) ret = vmf->page->pgmap->ops->migrate_to_ram(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3609) } else if (is_hwpoison_entry(entry)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3610) ret = VM_FAULT_HWPOISON;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3611) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3612) print_bad_pte(vma, vmf->address, vmf->orig_pte, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3613) ret = VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3614) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3615) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3616) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3617)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3618)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3619) delayacct_set_flag(DELAYACCT_PF_SWAPIN);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3620) page = lookup_swap_cache(entry, vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3621) swapcache = page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3622)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3623) if (!page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3624) struct swap_info_struct *si = swp_swap_info(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3625)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3626) if (data_race(si->flags & SWP_SYNCHRONOUS_IO) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3627) __swap_count(entry) == 1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3628) /* skip swapcache */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3629) gfp_t flags = GFP_HIGHUSER_MOVABLE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3630)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3631) trace_android_rvh_set_skip_swapcache_flags(&flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3632) page = alloc_page_vma(flags, vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3633) if (page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3634) int err;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3635)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3636) __SetPageLocked(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3637) __SetPageSwapBacked(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3638) set_page_private(page, entry.val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3639)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3640) /* Tell memcg to use swap ownership records */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3641) SetPageSwapCache(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3642) err = mem_cgroup_charge(page, vma->vm_mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3643) GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3644) ClearPageSwapCache(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3645) if (err) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3646) ret = VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3647) goto out_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3648) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3649)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3650) shadow = get_shadow_from_swap_cache(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3651) if (shadow)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3652) workingset_refault(page, shadow);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3653)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3654) lru_cache_add(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3655) swap_readpage(page, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3656) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3657) } else if (vmf->flags & FAULT_FLAG_SPECULATIVE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3658) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3659) * Don't try readahead during a speculative page fault
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3660) * as the VMA's boundaries may change in our back.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3661) * If the page is not in the swap cache and synchronous
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3662) * read is disabled, fall back to the regular page fault
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3663) * mechanism.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3664) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3665) delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3666) ret = VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3667) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3668) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3669) page = swapin_readahead(entry, GFP_HIGHUSER_MOVABLE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3670) vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3671) swapcache = page;
^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) if (!page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3675) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3676) * Back out if the VMA has changed in our back during
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3677) * a speculative page fault or if somebody else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3678) * faulted in this pte while we released the pte lock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3679) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3680) if (!pte_map_lock(vmf)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3681) delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3682) ret = VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3683) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3684) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3685)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3686) if (likely(pte_same(*vmf->pte, vmf->orig_pte)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3687) ret = VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3688) delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3689) goto unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3690) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3691)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3692) /* Had to read the page from swap area: Major fault */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3693) ret = VM_FAULT_MAJOR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3694) count_vm_event(PGMAJFAULT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3695) count_memcg_event_mm(vma->vm_mm, PGMAJFAULT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3696) } else if (PageHWPoison(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3697) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3698) * hwpoisoned dirty swapcache pages are kept for killing
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3699) * owner processes (which may be unknown at hwpoison time)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3700) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3701) ret = VM_FAULT_HWPOISON;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3702) delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3703) goto out_release;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3704) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3705)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3706) locked = lock_page_or_retry(page, vma->vm_mm, vmf->flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3707)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3708) delayacct_clear_flag(DELAYACCT_PF_SWAPIN);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3709) if (!locked) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3710) ret |= VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3711) goto out_release;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3712) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3713)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3714) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3715) * Make sure try_to_free_swap or reuse_swap_page or swapoff did not
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3716) * release the swapcache from under us. The page pin, and pte_same
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3717) * test below, are not enough to exclude that. Even if it is still
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3718) * swapcache, we need to check that the page's swap has not changed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3719) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3720) if (unlikely((!PageSwapCache(page) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3721) page_private(page) != entry.val)) && swapcache)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3722) goto out_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3723)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3724) page = ksm_might_need_to_copy(page, vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3725) if (unlikely(!page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3726) ret = VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3727) page = swapcache;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3728) goto out_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3729) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3730)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3731) cgroup_throttle_swaprate(page, GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3732)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3733) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3734) * Back out if the VMA has changed in our back during a speculative
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3735) * page fault or if somebody else already faulted in this pte.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3736) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3737) if (!pte_map_lock(vmf)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3738) ret = VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3739) goto out_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3740) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3741) if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3742) goto out_nomap;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3743)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3744) if (unlikely(!PageUptodate(page))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3745) ret = VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3746) goto out_nomap;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3747) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3748)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3749) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3750) * The page isn't present yet, go ahead with the fault.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3751) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3752) * Be careful about the sequence of operations here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3753) * To get its accounting right, reuse_swap_page() must be called
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3754) * while the page is counted on swap but not yet in mapcount i.e.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3755) * before page_add_anon_rmap() and swap_free(); try_to_free_swap()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3756) * must be called after the swap_free(), or it will never succeed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3757) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3758)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3759) inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3760) dec_mm_counter_fast(vma->vm_mm, MM_SWAPENTS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3761) pte = mk_pte(page, vmf->vma_page_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3762) if ((vmf->flags & FAULT_FLAG_WRITE) && reuse_swap_page(page, NULL)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3763) pte = maybe_mkwrite(pte_mkdirty(pte), vmf->vma_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3764) vmf->flags &= ~FAULT_FLAG_WRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3765) ret |= VM_FAULT_WRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3766) exclusive = RMAP_EXCLUSIVE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3767) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3768) flush_icache_page(vma, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3769) if (pte_swp_soft_dirty(vmf->orig_pte))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3770) pte = pte_mksoft_dirty(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3771) if (pte_swp_uffd_wp(vmf->orig_pte)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3772) pte = pte_mkuffd_wp(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3773) pte = pte_wrprotect(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3774) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3775) set_pte_at(vma->vm_mm, vmf->address, vmf->pte, pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3776) arch_do_swap_page(vma->vm_mm, vma, vmf->address, pte, vmf->orig_pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3777) vmf->orig_pte = pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3778)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3779) /* ksm created a completely new copy */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3780) if (unlikely(page != swapcache && swapcache)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3781) __page_add_new_anon_rmap(page, vma, vmf->address, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3782) __lru_cache_add_inactive_or_unevictable(page, vmf->vma_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3783) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3784) do_page_add_anon_rmap(page, vma, vmf->address, exclusive);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3785) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3786)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3787) trace_android_vh_swapin_add_anon_rmap(vmf, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3788) swap_free(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3789) if (mem_cgroup_swap_full(page) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3790) (vmf->vma_flags & VM_LOCKED) || PageMlocked(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3791) try_to_free_swap(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3792) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3793) if (page != swapcache && swapcache) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3794) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3795) * Hold the lock to avoid the swap entry to be reused
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3796) * until we take the PT lock for the pte_same() check
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3797) * (to avoid false positives from pte_same). For
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3798) * further safety release the lock after the swap_free
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3799) * so that the swap count won't change under a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3800) * parallel locked swapcache.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3801) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3802) unlock_page(swapcache);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3803) put_page(swapcache);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3804) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3805)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3806) if (vmf->flags & FAULT_FLAG_WRITE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3807) ret |= do_wp_page(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3808) if (ret & VM_FAULT_ERROR)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3809) ret &= VM_FAULT_ERROR;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3810) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3811) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3812)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3813) /* No need to invalidate - it was non-present before */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3814) update_mmu_cache(vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3815) unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3816) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3817) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3818) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3819) out_nomap:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3820) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3821) out_page:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3822) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3823) out_release:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3824) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3825) if (page != swapcache && swapcache) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3826) unlock_page(swapcache);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3827) put_page(swapcache);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3828) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3829) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3830) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3831)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3832) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3833) * We enter with non-exclusive mmap_lock (to exclude vma changes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3834) * but allow concurrent faults), and pte mapped but not yet locked.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3835) * We return with mmap_lock still held, but pte unmapped and unlocked.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3836) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3837) static vm_fault_t do_anonymous_page(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3838) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3839) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3840) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3841) vm_fault_t ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3842) pte_t entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3843)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3844) /* File mapping without ->vm_ops ? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3845) if (vmf->vma_flags & VM_SHARED)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3846) return VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3847)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3848) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3849) * Use pte_alloc() instead of pte_alloc_map(). We can't run
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3850) * pte_offset_map() on pmds where a huge pmd might be created
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3851) * from a different thread.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3852) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3853) * pte_alloc_map() is safe to use under mmap_write_lock(mm) or when
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3854) * parallel threads are excluded by other means.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3855) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3856) * Here we only have mmap_read_lock(mm).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3857) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3858) if (pte_alloc(vma->vm_mm, vmf->pmd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3859) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3860)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3861) /* See comment in handle_pte_fault() */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3862) if (unlikely(pmd_trans_unstable(vmf->pmd)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3863) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3864)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3865) /* Use the zero-page for reads */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3866) if (!(vmf->flags & FAULT_FLAG_WRITE) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3867) !mm_forbids_zeropage(vma->vm_mm)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3868) entry = pte_mkspecial(pfn_pte(my_zero_pfn(vmf->address),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3869) vmf->vma_page_prot));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3870) if (!pte_map_lock(vmf))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3871) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3872) if (!pte_none(*vmf->pte)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3873) update_mmu_tlb(vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3874) goto unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3875) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3876) ret = check_stable_address_space(vma->vm_mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3877) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3878) goto unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3879) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3880) * Don't call the userfaultfd during the speculative path.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3881) * We already checked for the VMA to not be managed through
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3882) * userfaultfd, but it may be set in our back once we have lock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3883) * the pte. In such a case we can ignore it this time.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3884) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3885) if (vmf->flags & FAULT_FLAG_SPECULATIVE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3886) goto setpte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3887) /* Deliver the page fault to userland, check inside PT lock */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3888) if (userfaultfd_missing(vma)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3889) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3890) return handle_userfault(vmf, VM_UFFD_MISSING);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3891) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3892) goto setpte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3893) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3894)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3895) /* Allocate our own private page. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3896) if (unlikely(anon_vma_prepare(vma)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3897) goto oom;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3898) page = alloc_zeroed_user_highpage_movable(vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3899) if (!page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3900) goto oom;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3901)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3902) if (mem_cgroup_charge(page, vma->vm_mm, GFP_KERNEL))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3903) goto oom_free_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3904) cgroup_throttle_swaprate(page, GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3905)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3906) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3907) * The memory barrier inside __SetPageUptodate makes sure that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3908) * preceding stores to the page contents become visible before
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3909) * the set_pte_at() write.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3910) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3911) __SetPageUptodate(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3912)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3913) entry = mk_pte(page, vmf->vma_page_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3914) entry = pte_sw_mkyoung(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3915) if (vmf->vma_flags & VM_WRITE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3916) entry = pte_mkwrite(pte_mkdirty(entry));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3917)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3918) if (!pte_map_lock(vmf)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3919) ret = VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3920) goto release;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3921) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3922)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3923) if (!pte_none(*vmf->pte)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3924) update_mmu_cache(vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3925) goto unlock_and_release;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3926) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3927)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3928) ret = check_stable_address_space(vma->vm_mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3929) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3930) goto unlock_and_release;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3931)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3932) /* Deliver the page fault to userland, check inside PT lock */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3933) if (!(vmf->flags & FAULT_FLAG_SPECULATIVE) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3934) userfaultfd_missing(vma)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3935) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3936) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3937) return handle_userfault(vmf, VM_UFFD_MISSING);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3938) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3939)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3940) inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3941) __page_add_new_anon_rmap(page, vma, vmf->address, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3942) __lru_cache_add_inactive_or_unevictable(page, vmf->vma_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3943) setpte:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3944) set_pte_at(vma->vm_mm, vmf->address, vmf->pte, entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3945)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3946) /* No need to invalidate - it was non-present before */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3947) update_mmu_cache(vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3948) unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3949) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3950) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3951) unlock_and_release:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3952) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3953) release:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3954) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3955) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3956) oom_free_page:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3957) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3958) oom:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3959) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3960) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3961)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3962) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3963) * The mmap_lock must have been held on entry, and may have been
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3964) * released depending on flags and vma->vm_ops->fault() return value.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3965) * See filemap_fault() and __lock_page_retry().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3966) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3967) static vm_fault_t __do_fault(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3968) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3969) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3970) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3971)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3972) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3973) * Preallocate pte before we take page_lock because this might lead to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3974) * deadlocks for memcg reclaim which waits for pages under writeback:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3975) * lock_page(A)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3976) * SetPageWriteback(A)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3977) * unlock_page(A)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3978) * lock_page(B)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3979) * lock_page(B)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3980) * pte_alloc_one
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3981) * shrink_page_list
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3982) * wait_on_page_writeback(A)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3983) * SetPageWriteback(B)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3984) * unlock_page(B)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3985) * # flush A, B to clear the writeback
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3986) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3987) if (pmd_none(*vmf->pmd) && !vmf->prealloc_pte) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3988) vmf->prealloc_pte = pte_alloc_one(vma->vm_mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3989) if (!vmf->prealloc_pte)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3990) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3991) smp_wmb(); /* See comment in __pte_alloc() */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3992) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3993)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3994) ret = vma->vm_ops->fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3995) if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3996) VM_FAULT_DONE_COW)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3997) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3998)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3999) if (unlikely(PageHWPoison(vmf->page))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4000) struct page *page = vmf->page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4001) vm_fault_t poisonret = VM_FAULT_HWPOISON;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4002) if (ret & VM_FAULT_LOCKED) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4003) if (page_mapped(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4004) unmap_mapping_pages(page_mapping(page),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4005) page->index, 1, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4006) /* Retry if a clean page was removed from the cache. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4007) if (invalidate_inode_page(page))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4008) poisonret = VM_FAULT_NOPAGE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4009) unlock_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4010) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4011) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4012) vmf->page = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4013) return poisonret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4014) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4015)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4016) if (unlikely(!(ret & VM_FAULT_LOCKED)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4017) lock_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4018) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4019) VM_BUG_ON_PAGE(!PageLocked(vmf->page), vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4020)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4021) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4022) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4023)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4024) #ifdef CONFIG_TRANSPARENT_HUGEPAGE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4025) static void deposit_prealloc_pte(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4026) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4027) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4028)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4029) pgtable_trans_huge_deposit(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4030) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4031) * We are going to consume the prealloc table,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4032) * count that as nr_ptes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4033) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4034) mm_inc_nr_ptes(vma->vm_mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4035) vmf->prealloc_pte = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4036) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4037)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4038) vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4039) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4040) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4041) bool write = vmf->flags & FAULT_FLAG_WRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4042) unsigned long haddr = vmf->address & HPAGE_PMD_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4043) pmd_t entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4044) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4045) vm_fault_t ret = VM_FAULT_FALLBACK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4046)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4047) if (!transhuge_vma_suitable(vma, haddr))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4048) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4049)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4050) page = compound_head(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4051) if (compound_order(page) != HPAGE_PMD_ORDER)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4052) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4053)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4054) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4055) * Archs like ppc64 need additonal space to store information
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4056) * related to pte entry. Use the preallocated table for that.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4057) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4058) if (arch_needs_pgtable_deposit() && !vmf->prealloc_pte) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4059) vmf->prealloc_pte = pte_alloc_one(vma->vm_mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4060) if (!vmf->prealloc_pte)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4061) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4062) smp_wmb(); /* See comment in __pte_alloc() */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4063) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4064)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4065) vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4066) if (unlikely(!pmd_none(*vmf->pmd)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4067) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4068)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4069) for (i = 0; i < HPAGE_PMD_NR; i++)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4070) flush_icache_page(vma, page + i);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4071)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4072) entry = mk_huge_pmd(page, vmf->vma_page_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4073) if (write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4074) entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4075)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4076) add_mm_counter(vma->vm_mm, mm_counter_file(page), HPAGE_PMD_NR);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4077) page_add_file_rmap(page, true);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4078) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4079) * deposit and withdraw with pmd lock held
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4080) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4081) if (arch_needs_pgtable_deposit())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4082) deposit_prealloc_pte(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4083)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4084) set_pmd_at(vma->vm_mm, haddr, vmf->pmd, entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4085)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4086) update_mmu_cache_pmd(vma, haddr, vmf->pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4087)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4088) /* fault is handled */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4089) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4090) count_vm_event(THP_FILE_MAPPED);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4091) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4092) spin_unlock(vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4093) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4094) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4095) #else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4096) vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4097) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4098) return VM_FAULT_FALLBACK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4099) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4100) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4101)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4102) void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4103) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4104) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4105) bool write = vmf->flags & FAULT_FLAG_WRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4106) bool prefault = vmf->address != addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4107) pte_t entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4108)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4109) flush_icache_page(vma, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4110) entry = mk_pte(page, vmf->vma_page_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4111)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4112) if (prefault && arch_wants_old_prefaulted_pte())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4113) entry = pte_mkold(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4114) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4115) entry = pte_sw_mkyoung(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4116)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4117) if (write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4118) entry = maybe_mkwrite(pte_mkdirty(entry), vmf->vma_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4119) /* copy-on-write page */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4120) if (write && !(vmf->vma_flags & VM_SHARED)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4121) inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4122) __page_add_new_anon_rmap(page, vma, addr, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4123) __lru_cache_add_inactive_or_unevictable(page, vmf->vma_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4124) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4125) inc_mm_counter_fast(vma->vm_mm, mm_counter_file(page));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4126) page_add_file_rmap(page, false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4127) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4128) set_pte_at(vma->vm_mm, addr, vmf->pte, entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4129) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4130)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4131) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4132) * finish_fault - finish page fault once we have prepared the page to fault
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4133) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4134) * @vmf: structure describing the fault
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4135) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4136) * This function handles all that is needed to finish a page fault once the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4137) * page to fault in is prepared. It handles locking of PTEs, inserts PTE for
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4138) * given page, adds reverse page mapping, handles memcg charges and LRU
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4139) * addition.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4140) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4141) * The function expects the page to be locked and on success it consumes a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4142) * reference of a page being mapped (for the PTE which maps it).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4143) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4144) * Return: %0 on success, %VM_FAULT_ code in case of error.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4145) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4146) vm_fault_t finish_fault(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4147) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4148) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4149) struct page *page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4150) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4151)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4152) /* Did we COW the page? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4153) if ((vmf->flags & FAULT_FLAG_WRITE) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4154) !(vmf->vma_flags & VM_SHARED))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4155) page = vmf->cow_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4156) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4157) page = vmf->page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4158)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4159) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4160) * check even for read faults because we might have lost our CoWed
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4161) * page
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4162) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4163) if (!(vma->vm_flags & VM_SHARED)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4164) ret = check_stable_address_space(vma->vm_mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4165) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4166) return ret;
^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) if (pmd_none(*vmf->pmd) && !(vmf->flags & FAULT_FLAG_SPECULATIVE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4170) if (PageTransCompound(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4171) ret = do_set_pmd(vmf, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4172) if (ret != VM_FAULT_FALLBACK)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4173) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4174) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4175)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4176) if (vmf->prealloc_pte) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4177) vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4178) if (likely(pmd_none(*vmf->pmd))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4179) mm_inc_nr_ptes(vma->vm_mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4180) pmd_populate(vma->vm_mm, vmf->pmd, vmf->prealloc_pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4181) vmf->prealloc_pte = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4182) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4183) spin_unlock(vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4184) } else if (unlikely(pte_alloc(vma->vm_mm, vmf->pmd))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4185) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4186) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4187) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4188)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4189) /* See comment in handle_pte_fault() */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4190) if (pmd_devmap_trans_unstable(vmf->pmd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4191) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4192)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4193) if (!pte_map_lock(vmf))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4194) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4195)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4196) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4197) /* Re-check under ptl */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4198) if (likely(pte_none(*vmf->pte)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4199) do_set_pte(vmf, page, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4200) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4201) ret = VM_FAULT_NOPAGE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4202)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4203) update_mmu_tlb(vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4204) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4205) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4206) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4207)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4208) static unsigned long fault_around_bytes __read_mostly =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4209) rounddown_pow_of_two(65536);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4210)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4211) #ifdef CONFIG_DEBUG_FS
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4212) static int fault_around_bytes_get(void *data, u64 *val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4213) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4214) *val = fault_around_bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4215) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4216) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4217)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4218) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4219) * fault_around_bytes must be rounded down to the nearest page order as it's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4220) * what do_fault_around() expects to see.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4221) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4222) static int fault_around_bytes_set(void *data, u64 val)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4223) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4224) if (val / PAGE_SIZE > PTRS_PER_PTE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4225) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4226) if (val > PAGE_SIZE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4227) fault_around_bytes = rounddown_pow_of_two(val);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4228) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4229) fault_around_bytes = PAGE_SIZE; /* rounddown_pow_of_two(0) is undefined */
^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) DEFINE_DEBUGFS_ATTRIBUTE(fault_around_bytes_fops,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4233) fault_around_bytes_get, fault_around_bytes_set, "%llu\n");
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4234)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4235) static int __init fault_around_debugfs(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4236) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4237) debugfs_create_file_unsafe("fault_around_bytes", 0644, NULL, NULL,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4238) &fault_around_bytes_fops);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4239) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4240) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4241) late_initcall(fault_around_debugfs);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4242) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4243)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4244) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4245) * do_fault_around() tries to map few pages around the fault address. The hope
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4246) * is that the pages will be needed soon and this will lower the number of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4247) * faults to handle.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4248) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4249) * It uses vm_ops->map_pages() to map the pages, which skips the page if it's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4250) * not ready to be mapped: not up-to-date, locked, etc.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4251) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4252) * This function is called with the page table lock taken. In the split ptlock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4253) * case the page table lock only protects only those entries which belong to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4254) * the page table corresponding to the fault address.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4255) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4256) * This function doesn't cross the VMA boundaries, in order to call map_pages()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4257) * only once.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4258) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4259) * fault_around_bytes defines how many bytes we'll try to map.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4260) * do_fault_around() expects it to be set to a power of two less than or equal
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4261) * to PTRS_PER_PTE.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4262) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4263) * The virtual address of the area that we map is naturally aligned to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4264) * fault_around_bytes rounded down to the machine page size
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4265) * (and therefore to page order). This way it's easier to guarantee
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4266) * that we don't cross page table boundaries.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4267) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4268) static vm_fault_t do_fault_around(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4269) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4270) unsigned long address = vmf->address, nr_pages, mask;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4271) pgoff_t start_pgoff = vmf->pgoff;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4272) pgoff_t end_pgoff;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4273) int off;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4274)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4275) nr_pages = READ_ONCE(fault_around_bytes) >> PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4276) mask = ~(nr_pages * PAGE_SIZE - 1) & PAGE_MASK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4277)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4278) address = max(address & mask, vmf->vma->vm_start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4279) off = ((vmf->address - address) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4280) start_pgoff -= off;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4281)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4282) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4283) * end_pgoff is either the end of the page table, the end of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4284) * the vma or nr_pages from start_pgoff, depending what is nearest.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4285) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4286) end_pgoff = start_pgoff -
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4287) ((address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) +
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4288) PTRS_PER_PTE - 1;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4289) end_pgoff = min3(end_pgoff, vma_pages(vmf->vma) + vmf->vma->vm_pgoff - 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4290) start_pgoff + nr_pages - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4291)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4292) if (pmd_none(*vmf->pmd)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4293) vmf->prealloc_pte = pte_alloc_one(vmf->vma->vm_mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4294) if (!vmf->prealloc_pte)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4295) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4296) smp_wmb(); /* See comment in __pte_alloc() */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4297) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4298)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4299) return vmf->vma->vm_ops->map_pages(vmf, start_pgoff, end_pgoff);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4300) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4301)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4302) static vm_fault_t do_read_fault(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4303) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4304) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4305) vm_fault_t ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4306)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4307) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4308) * Let's call ->map_pages() first and use ->fault() as fallback
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4309) * if page by the offset is not ready to be mapped (cold cache or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4310) * something).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4311) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4312) if (vma->vm_ops->map_pages && fault_around_bytes >> PAGE_SHIFT > 1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4313) if (likely(!userfaultfd_minor(vmf->vma))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4314) ret = do_fault_around(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4315) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4316) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4317) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4318) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4319)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4320) ret = __do_fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4321) if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4322) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4323)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4324) ret |= finish_fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4325) unlock_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4326) if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4327) put_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4328) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4329) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4330)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4331) static vm_fault_t do_cow_fault(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4332) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4333) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4334) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4335)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4336) if (unlikely(anon_vma_prepare(vma)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4337) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4338)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4339) vmf->cow_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4340) if (!vmf->cow_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4341) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4342)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4343) if (mem_cgroup_charge(vmf->cow_page, vma->vm_mm, GFP_KERNEL)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4344) put_page(vmf->cow_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4345) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4346) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4347) cgroup_throttle_swaprate(vmf->cow_page, GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4348)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4349) ret = __do_fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4350) if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4351) goto uncharge_out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4352) if (ret & VM_FAULT_DONE_COW)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4353) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4354)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4355) copy_user_highpage(vmf->cow_page, vmf->page, vmf->address, vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4356) __SetPageUptodate(vmf->cow_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4357)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4358) ret |= finish_fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4359) unlock_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4360) put_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4361) if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4362) goto uncharge_out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4363) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4364) uncharge_out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4365) put_page(vmf->cow_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4366) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4367) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4368)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4369) static vm_fault_t do_shared_fault(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4370) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4371) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4372) vm_fault_t ret, tmp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4373)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4374) ret = __do_fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4375) if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE | VM_FAULT_RETRY)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4376) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4377)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4378) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4379) * Check if the backing address space wants to know that the page is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4380) * about to become writable
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4381) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4382) if (vma->vm_ops->page_mkwrite) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4383) unlock_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4384) tmp = do_page_mkwrite(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4385) if (unlikely(!tmp ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4386) (tmp & (VM_FAULT_ERROR | VM_FAULT_NOPAGE)))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4387) put_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4388) return tmp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4389) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4390) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4391)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4392) ret |= finish_fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4393) if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE |
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4394) VM_FAULT_RETRY))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4395) unlock_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4396) put_page(vmf->page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4397) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4398) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4399)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4400) ret |= fault_dirty_shared_page(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4401) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4402) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4403)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4404) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4405) * We enter with non-exclusive mmap_lock (to exclude vma changes,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4406) * but allow concurrent faults).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4407) * The mmap_lock may have been released depending on flags and our
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4408) * return value. See filemap_fault() and __lock_page_or_retry().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4409) * If mmap_lock is released, vma may become invalid (for example
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4410) * by other thread calling munmap()).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4411) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4412) static vm_fault_t do_fault(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4413) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4414) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4415) struct mm_struct *vm_mm = vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4416) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4417)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4418) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4419) * The VMA was not fully populated on mmap() or missing VM_DONTEXPAND
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4420) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4421) if (!vma->vm_ops->fault) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4422) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4423) * If we find a migration pmd entry or a none pmd entry, which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4424) * should never happen, return SIGBUS
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4425) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4426) if (unlikely(!pmd_present(*vmf->pmd)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4427) ret = VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4428) else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4429) vmf->pte = pte_offset_map_lock(vmf->vma->vm_mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4430) vmf->pmd,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4431) vmf->address,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4432) &vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4433) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4434) * Make sure this is not a temporary clearing of pte
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4435) * by holding ptl and checking again. A R/M/W update
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4436) * of pte involves: take ptl, clearing the pte so that
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4437) * we don't have concurrent modification by hardware
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4438) * followed by an update.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4439) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4440) if (unlikely(pte_none(*vmf->pte)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4441) ret = VM_FAULT_SIGBUS;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4442) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4443) ret = VM_FAULT_NOPAGE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4444)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4445) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4446) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4447) } else if (!(vmf->flags & FAULT_FLAG_WRITE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4448) ret = do_read_fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4449) else if (!(vmf->vma_flags & VM_SHARED))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4450) ret = do_cow_fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4451) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4452) ret = do_shared_fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4453)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4454) /* preallocated pagetable is unused: free it */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4455) if (vmf->prealloc_pte) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4456) pte_free(vm_mm, vmf->prealloc_pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4457) vmf->prealloc_pte = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4458) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4459) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4460) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4461)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4462) static int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4463) unsigned long addr, int page_nid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4464) int *flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4465) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4466) get_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4467)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4468) count_vm_numa_event(NUMA_HINT_FAULTS);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4469) if (page_nid == numa_node_id()) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4470) count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4471) *flags |= TNF_FAULT_LOCAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4472) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4473)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4474) return mpol_misplaced(page, vma, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4475) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4476)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4477) static vm_fault_t do_numa_page(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4478) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4479) struct vm_area_struct *vma = vmf->vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4480) struct page *page = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4481) int page_nid = NUMA_NO_NODE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4482) int last_cpupid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4483) int target_nid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4484) bool migrated = false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4485) pte_t pte, old_pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4486) bool was_writable = pte_savedwrite(vmf->orig_pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4487) int flags = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4488)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4489) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4490) * The "pte" at this point cannot be used safely without
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4491) * validation through pte_unmap_same(). It's of NUMA type but
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4492) * the pfn may be screwed if the read is non atomic.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4493) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4494) if (!pte_spinlock(vmf))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4495) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4496) if (unlikely(!pte_same(*vmf->pte, vmf->orig_pte))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4497) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4498) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4499) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4500)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4501) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4502) * Make it present again, Depending on how arch implementes non
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4503) * accessible ptes, some can allow access by kernel mode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4504) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4505) old_pte = ptep_modify_prot_start(vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4506) pte = pte_modify(old_pte, vmf->vma_page_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4507) pte = pte_mkyoung(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4508) if (was_writable)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4509) pte = pte_mkwrite(pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4510) ptep_modify_prot_commit(vma, vmf->address, vmf->pte, old_pte, pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4511) update_mmu_cache(vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4512)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4513) page = _vm_normal_page(vma, vmf->address, pte, vmf->vma_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4514) if (!page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4515) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4516) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4517) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4518)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4519) /* TODO: handle PTE-mapped THP */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4520) if (PageCompound(page)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4521) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4522) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4523) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4524)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4525) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4526) * Avoid grouping on RO pages in general. RO pages shouldn't hurt as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4527) * much anyway since they can be in shared cache state. This misses
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4528) * the case where a mapping is writable but the process never writes
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4529) * to it but pte_write gets cleared during protection updates and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4530) * pte_dirty has unpredictable behaviour between PTE scan updates,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4531) * background writeback, dirty balancing and application behaviour.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4532) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4533) if (!pte_write(pte))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4534) flags |= TNF_NO_GROUP;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4535)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4536) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4537) * Flag if the page is shared between multiple address spaces. This
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4538) * is later used when determining whether to group tasks together
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4539) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4540) if (page_mapcount(page) > 1 && (vmf->vma_flags & VM_SHARED))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4541) flags |= TNF_SHARED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4542)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4543) last_cpupid = page_cpupid_last(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4544) page_nid = page_to_nid(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4545) target_nid = numa_migrate_prep(page, vma, vmf->address, page_nid,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4546) &flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4547) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4548) if (target_nid == NUMA_NO_NODE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4549) put_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4550) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4551) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4552)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4553) /* Migrate to the requested node */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4554) migrated = migrate_misplaced_page(page, vmf, target_nid);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4555) if (migrated) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4556) page_nid = target_nid;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4557) flags |= TNF_MIGRATED;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4558) } else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4559) flags |= TNF_MIGRATE_FAIL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4560)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4561) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4562) if (page_nid != NUMA_NO_NODE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4563) task_numa_fault(last_cpupid, page_nid, 1, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4564) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4565) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4566)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4567) static inline vm_fault_t create_huge_pmd(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4568) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4569) if (vma_is_anonymous(vmf->vma))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4570) return do_huge_pmd_anonymous_page(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4571) if (vmf->vma->vm_ops->huge_fault)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4572) return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4573) return VM_FAULT_FALLBACK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4574) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4575)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4576) /* `inline' is required to avoid gcc 4.1.2 build error */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4577) static inline vm_fault_t wp_huge_pmd(struct vm_fault *vmf, pmd_t orig_pmd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4578) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4579) if (vma_is_anonymous(vmf->vma)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4580) if (userfaultfd_huge_pmd_wp(vmf->vma, orig_pmd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4581) return handle_userfault(vmf, VM_UFFD_WP);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4582) return do_huge_pmd_wp_page(vmf, orig_pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4583) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4584) if (vmf->vma->vm_ops->huge_fault) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4585) vm_fault_t ret = vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4586)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4587) if (!(ret & VM_FAULT_FALLBACK))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4588) return ret;
^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) /* COW or write-notify handled on pte level: split pmd. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4592) __split_huge_pmd(vmf->vma, vmf->pmd, vmf->address, false, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4593)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4594) return VM_FAULT_FALLBACK;
^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 vm_fault_t create_huge_pud(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4598) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4599) #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && \
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4600) defined(CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4601) /* No support for anonymous transparent PUD pages yet */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4602) if (vma_is_anonymous(vmf->vma))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4603) goto split;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4604) if (vmf->vma->vm_ops->huge_fault) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4605) vm_fault_t ret = vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4606)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4607) if (!(ret & VM_FAULT_FALLBACK))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4608) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4609) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4610) split:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4611) /* COW or write-notify not handled on PUD level: split pud.*/
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4612) __split_huge_pud(vmf->vma, vmf->pud, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4613) #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4614) return VM_FAULT_FALLBACK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4615) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4616)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4617) static vm_fault_t wp_huge_pud(struct vm_fault *vmf, pud_t orig_pud)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4618) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4619) #ifdef CONFIG_TRANSPARENT_HUGEPAGE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4620) /* No support for anonymous transparent PUD pages yet */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4621) if (vma_is_anonymous(vmf->vma))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4622) return VM_FAULT_FALLBACK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4623) if (vmf->vma->vm_ops->huge_fault)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4624) return vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PUD);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4625) #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4626) return VM_FAULT_FALLBACK;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4627) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4628)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4629) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4630) * These routines also need to handle stuff like marking pages dirty
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4631) * and/or accessed for architectures that don't do it in hardware (most
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4632) * RISC architectures). The early dirtying is also good on the i386.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4633) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4634) * There is also a hook called "update_mmu_cache()" that architectures
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4635) * with external mmu caches can use to update those (ie the Sparc or
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4636) * PowerPC hashed page tables that act as extended TLBs).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4637) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4638) * We enter with non-exclusive mmap_lock (to exclude vma changes, but allow
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4639) * concurrent faults).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4640) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4641) * The mmap_lock may have been released depending on flags and our return value.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4642) * See filemap_fault() and __lock_page_or_retry().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4643) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4644) static vm_fault_t handle_pte_fault(struct vm_fault *vmf)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4645) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4646) pte_t entry;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4647) vm_fault_t ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4648)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4649) if (unlikely(pmd_none(*vmf->pmd))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4650) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4651) * In the case of the speculative page fault handler we abort
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4652) * the speculative path immediately as the pmd is probably
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4653) * in the way to be converted in a huge one. We will try
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4654) * again holding the mmap_sem (which implies that the collapse
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4655) * operation is done).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4656) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4657) if (vmf->flags & FAULT_FLAG_SPECULATIVE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4658) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4659) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4660) * Leave __pte_alloc() until later: because vm_ops->fault may
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4661) * want to allocate huge page, and if we expose page table
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4662) * for an instant, it will be difficult to retract from
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4663) * concurrent faults and from rmap lookups.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4664) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4665) vmf->pte = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4666) } else if (!(vmf->flags & FAULT_FLAG_SPECULATIVE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4667) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4668) * If a huge pmd materialized under us just retry later. Use
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4669) * pmd_trans_unstable() via pmd_devmap_trans_unstable() instead
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4670) * of pmd_trans_huge() to ensure the pmd didn't become
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4671) * pmd_trans_huge under us and then back to pmd_none, as a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4672) * result of MADV_DONTNEED running immediately after a huge pmd
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4673) * fault in a different thread of this mm, in turn leading to a
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4674) * misleading pmd_trans_huge() retval. All we have to ensure is
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4675) * that it is a regular pmd that we can walk with
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4676) * pte_offset_map() and we can do that through an atomic read
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4677) * in C, which is what pmd_trans_unstable() provides.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4678) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4679) if (pmd_devmap_trans_unstable(vmf->pmd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4680) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4681) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4682) * A regular pmd is established and it can't morph into a huge
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4683) * pmd from under us anymore at this point because we hold the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4684) * mmap_lock read mode and khugepaged takes it in write mode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4685) * So now it's safe to run pte_offset_map().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4686) * This is not applicable to the speculative page fault handler
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4687) * but in that case, the pte is fetched earlier in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4688) * handle_speculative_fault().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4689) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4690) vmf->pte = pte_offset_map(vmf->pmd, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4691) vmf->orig_pte = *vmf->pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4692)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4693) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4694) * some architectures can have larger ptes than wordsize,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4695) * e.g.ppc44x-defconfig has CONFIG_PTE_64BIT=y and
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4696) * CONFIG_32BIT=y, so READ_ONCE cannot guarantee atomic
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4697) * accesses. The code below just needs a consistent view
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4698) * for the ifs and we later double check anyway with the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4699) * ptl lock held. So here a barrier will do.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4700) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4701) barrier();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4702) if (pte_none(vmf->orig_pte)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4703) pte_unmap(vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4704) vmf->pte = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4705) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4706) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4707)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4708) if (!vmf->pte) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4709) if (vma_is_anonymous(vmf->vma))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4710) return do_anonymous_page(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4711) else if ((vmf->flags & FAULT_FLAG_SPECULATIVE) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4712) !vmf_allows_speculation(vmf))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4713) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4714) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4715) return do_fault(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4716) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4717)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4718) if (!pte_present(vmf->orig_pte))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4719) return do_swap_page(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4720)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4721) if (pte_protnone(vmf->orig_pte) && vma_is_accessible(vmf->vma))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4722) return do_numa_page(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4723)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4724) if (!pte_spinlock(vmf))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4725) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4726) entry = vmf->orig_pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4727) if (unlikely(!pte_same(*vmf->pte, entry))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4728) update_mmu_tlb(vmf->vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4729) goto unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4730) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4731) if (vmf->flags & FAULT_FLAG_WRITE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4732) if (!pte_write(entry)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4733) if (!(vmf->flags & FAULT_FLAG_SPECULATIVE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4734) return do_wp_page(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4735)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4736) if (!mmu_notifier_trylock(vmf->vma->vm_mm)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4737) ret = VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4738) goto unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4739) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4740)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4741) ret = do_wp_page(vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4742) mmu_notifier_unlock(vmf->vma->vm_mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4743) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4744) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4745) entry = pte_mkdirty(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4746) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4747) entry = pte_mkyoung(entry);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4748) if (ptep_set_access_flags(vmf->vma, vmf->address, vmf->pte, entry,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4749) vmf->flags & FAULT_FLAG_WRITE)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4750) update_mmu_cache(vmf->vma, vmf->address, vmf->pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4751) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4752) /* Skip spurious TLB flush for retried page fault */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4753) if (vmf->flags & FAULT_FLAG_TRIED)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4754) goto unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4755) if (vmf->flags & FAULT_FLAG_SPECULATIVE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4756) ret = VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4757) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4758) * This is needed only for protection faults but the arch code
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4759) * is not yet telling us if this is a protection fault or not.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4760) * This still avoids useless tlb flushes for .text page faults
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4761) * with threads.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4762) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4763) if (vmf->flags & FAULT_FLAG_WRITE)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4764) flush_tlb_fix_spurious_fault(vmf->vma, vmf->address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4765) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4766) trace_android_vh_handle_pte_fault_end(vmf, highest_memmap_pfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4767) unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4768) pte_unmap_unlock(vmf->pte, vmf->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4769) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4770) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4771)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4772) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4773) * By the time we get here, we already hold the mm semaphore
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4774) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4775) * The mmap_lock may have been released depending on flags and our
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4776) * return value. See filemap_fault() and __lock_page_or_retry().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4777) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4778) static vm_fault_t __handle_mm_fault(struct vm_area_struct *vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4779) unsigned long address, unsigned int flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4780) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4781) struct vm_fault vmf = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4782) .vma = vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4783) .address = address & PAGE_MASK,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4784) .flags = flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4785) .pgoff = linear_page_index(vma, address),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4786) .gfp_mask = __get_fault_gfp_mask(vma),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4787) .vma_flags = vma->vm_flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4788) .vma_page_prot = vma->vm_page_prot,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4789) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4790) unsigned int dirty = flags & FAULT_FLAG_WRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4791) struct mm_struct *mm = vma->vm_mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4792) pgd_t *pgd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4793) p4d_t *p4d;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4794) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4795)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4796) pgd = pgd_offset(mm, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4797) p4d = p4d_alloc(mm, pgd, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4798) if (!p4d)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4799) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4800)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4801) vmf.pud = pud_alloc(mm, p4d, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4802) if (!vmf.pud)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4803) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4804) retry_pud:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4805) if (pud_none(*vmf.pud) && __transparent_hugepage_enabled(vma)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4806) ret = create_huge_pud(&vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4807) if (!(ret & VM_FAULT_FALLBACK))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4808) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4809) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4810) pud_t orig_pud = *vmf.pud;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4811)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4812) barrier();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4813) if (pud_trans_huge(orig_pud) || pud_devmap(orig_pud)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4814)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4815) /* NUMA case for anonymous PUDs would go here */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4816)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4817) if (dirty && !pud_write(orig_pud)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4818) ret = wp_huge_pud(&vmf, orig_pud);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4819) if (!(ret & VM_FAULT_FALLBACK))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4820) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4821) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4822) huge_pud_set_accessed(&vmf, orig_pud);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4823) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4824) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4825) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4826) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4827)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4828) vmf.pmd = pmd_alloc(mm, vmf.pud, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4829) if (!vmf.pmd)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4830) return VM_FAULT_OOM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4831)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4832) /* Huge pud page fault raced with pmd_alloc? */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4833) if (pud_trans_unstable(vmf.pud))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4834) goto retry_pud;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4835)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4836) #ifdef CONFIG_SPECULATIVE_PAGE_FAULT
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4837) vmf.sequence = raw_read_seqcount(&vma->vm_sequence);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4838) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4839) if (pmd_none(*vmf.pmd) && __transparent_hugepage_enabled(vma)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4840) ret = create_huge_pmd(&vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4841) if (!(ret & VM_FAULT_FALLBACK))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4842) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4843) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4844) pmd_t orig_pmd = *vmf.pmd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4845)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4846) barrier();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4847) if (unlikely(is_swap_pmd(orig_pmd))) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4848) VM_BUG_ON(thp_migration_supported() &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4849) !is_pmd_migration_entry(orig_pmd));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4850) if (is_pmd_migration_entry(orig_pmd))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4851) pmd_migration_entry_wait(mm, vmf.pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4852) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4853) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4854) if (pmd_trans_huge(orig_pmd) || pmd_devmap(orig_pmd)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4855) if (pmd_protnone(orig_pmd) && vma_is_accessible(vma))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4856) return do_huge_pmd_numa_page(&vmf, orig_pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4857)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4858) if (dirty && !pmd_write(orig_pmd)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4859) ret = wp_huge_pmd(&vmf, orig_pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4860) if (!(ret & VM_FAULT_FALLBACK))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4861) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4862) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4863) huge_pmd_set_accessed(&vmf, orig_pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4864) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4865) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4866) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4867) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4868)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4869) return handle_pte_fault(&vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4870) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4871)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4872) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4873) * mm_account_fault - Do page fault accountings
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4874) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4875) * @regs: the pt_regs struct pointer. When set to NULL, will skip accounting
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4876) * of perf event counters, but we'll still do the per-task accounting to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4877) * the task who triggered this page fault.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4878) * @address: the faulted address.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4879) * @flags: the fault flags.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4880) * @ret: the fault retcode.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4881) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4882) * This will take care of most of the page fault accountings. Meanwhile, it
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4883) * will also include the PERF_COUNT_SW_PAGE_FAULTS_[MAJ|MIN] perf counter
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4884) * updates. However note that the handling of PERF_COUNT_SW_PAGE_FAULTS should
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4885) * still be in per-arch page fault handlers at the entry of page fault.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4886) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4887) static inline void mm_account_fault(struct pt_regs *regs,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4888) unsigned long address, unsigned int flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4889) vm_fault_t ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4890) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4891) bool major;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4892)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4893) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4894) * We don't do accounting for some specific faults:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4895) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4896) * - Unsuccessful faults (e.g. when the address wasn't valid). That
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4897) * includes arch_vma_access_permitted() failing before reaching here.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4898) * So this is not a "this many hardware page faults" counter. We
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4899) * should use the hw profiling for that.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4900) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4901) * - Incomplete faults (VM_FAULT_RETRY). They will only be counted
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4902) * once they're completed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4903) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4904) if (ret & (VM_FAULT_ERROR | VM_FAULT_RETRY))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4905) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4906)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4907) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4908) * We define the fault as a major fault when the final successful fault
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4909) * is VM_FAULT_MAJOR, or if it retried (which implies that we couldn't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4910) * handle it immediately previously).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4911) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4912) major = (ret & VM_FAULT_MAJOR) || (flags & FAULT_FLAG_TRIED);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4913)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4914) if (major)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4915) current->maj_flt++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4916) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4917) current->min_flt++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4918)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4919) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4920) * If the fault is done for GUP, regs will be NULL. We only do the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4921) * accounting for the per thread fault counters who triggered the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4922) * fault, and we skip the perf event updates.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4923) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4924) if (!regs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4925) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4926)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4927) if (major)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4928) perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4929) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4930) perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4931) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4932) #ifdef CONFIG_SPECULATIVE_PAGE_FAULT
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4933)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4934) #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4935) /* This is required by vm_normal_page() */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4936) #error "Speculative page fault handler requires CONFIG_ARCH_HAS_PTE_SPECIAL"
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4937) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4938) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4939) * vm_normal_page() adds some processing which should be done while
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4940) * hodling the mmap_sem.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4941) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4942)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4943) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4944) * Tries to handle the page fault in a speculative way, without grabbing the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4945) * mmap_sem.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4946) * When VM_FAULT_RETRY is returned, the vma pointer is valid and this vma must
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4947) * be checked later when the mmap_sem has been grabbed by calling
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4948) * can_reuse_spf_vma().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4949) * This is needed as the returned vma is kept in memory until the call to
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4950) * can_reuse_spf_vma() is made.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4951) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4952) static vm_fault_t ___handle_speculative_fault(struct mm_struct *mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4953) unsigned long address, unsigned int flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4954) struct vm_area_struct *vma)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4955) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4956) struct vm_fault vmf = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4957) .address = address,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4958) .pgoff = linear_page_index(vma, address),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4959) .vma = vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4960) .gfp_mask = __get_fault_gfp_mask(vma),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4961) .flags = flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4962) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4963) #ifdef CONFIG_NUMA
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4964) struct mempolicy *pol;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4965) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4966) pgd_t *pgd, pgdval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4967) p4d_t *p4d, p4dval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4968) pud_t pudval;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4969) int seq;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4970) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4971)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4972) /* Clear flags that may lead to release the mmap_sem to retry */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4973) flags &= ~(FAULT_FLAG_ALLOW_RETRY|FAULT_FLAG_KILLABLE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4974) flags |= FAULT_FLAG_SPECULATIVE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4975)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4976) /* rmb <-> seqlock,vma_rb_erase() */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4977) seq = raw_read_seqcount(&vmf.vma->vm_sequence);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4978) if (seq & 1) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4979) trace_spf_vma_changed(_RET_IP_, vmf.vma, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4980) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4981) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4982)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4983) if (!vmf_allows_speculation(&vmf))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4984) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4985)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4986) vmf.vma_flags = READ_ONCE(vmf.vma->vm_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4987) vmf.vma_page_prot = READ_ONCE(vmf.vma->vm_page_prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4988)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4989) /* Can't call userland page fault handler in the speculative path */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4990) if (unlikely(vmf.vma_flags & VM_UFFD_MISSING)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4991) trace_spf_vma_notsup(_RET_IP_, vmf.vma, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4992) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4993) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4994)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4995) if (vmf.vma_flags & VM_GROWSDOWN || vmf.vma_flags & VM_GROWSUP) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4996) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4997) * This could be detected by the check address against VMA's
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4998) * boundaries but we want to trace it as not supported instead
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4999) * of changed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5000) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5001) trace_spf_vma_notsup(_RET_IP_, vmf.vma, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5002) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5003) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5004)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5005) if (address < READ_ONCE(vmf.vma->vm_start)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5006) || READ_ONCE(vmf.vma->vm_end) <= address) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5007) trace_spf_vma_changed(_RET_IP_, vmf.vma, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5008) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5009) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5010)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5011) if (!arch_vma_access_permitted(vmf.vma, flags & FAULT_FLAG_WRITE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5012) flags & FAULT_FLAG_INSTRUCTION,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5013) flags & FAULT_FLAG_REMOTE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5014) goto out_segv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5015)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5016) /* This is one is required to check that the VMA has write access set */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5017) if (flags & FAULT_FLAG_WRITE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5018) if (unlikely(!(vmf.vma_flags & VM_WRITE)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5019) goto out_segv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5020) } else if (unlikely(!(vmf.vma_flags & (VM_READ|VM_EXEC|VM_WRITE))))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5021) goto out_segv;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5022)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5023) #ifdef CONFIG_NUMA
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5024) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5025) * MPOL_INTERLEAVE implies additional checks in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5026) * mpol_misplaced() which are not compatible with the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5027) *speculative page fault processing.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5028) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5029) pol = __get_vma_policy(vmf.vma, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5030) if (!pol)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5031) pol = get_task_policy(current);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5032) if (!pol)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5033) if (pol && pol->mode == MPOL_INTERLEAVE) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5034) trace_spf_vma_notsup(_RET_IP_, vmf.vma, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5035) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5036) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5037) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5038)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5039) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5040) * Do a speculative lookup of the PTE entry.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5041) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5042) local_irq_disable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5043) pgd = pgd_offset(mm, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5044) pgdval = READ_ONCE(*pgd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5045) if (pgd_none(pgdval) || unlikely(pgd_bad(pgdval)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5046) goto out_walk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5047)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5048) p4d = p4d_offset(pgd, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5049) if (pgd_val(READ_ONCE(*pgd)) != pgd_val(pgdval))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5050) goto out_walk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5051) p4dval = READ_ONCE(*p4d);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5052) if (p4d_none(p4dval) || unlikely(p4d_bad(p4dval)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5053) goto out_walk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5054)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5055) vmf.pud = pud_offset(p4d, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5056) if (p4d_val(READ_ONCE(*p4d)) != p4d_val(p4dval))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5057) goto out_walk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5058) pudval = READ_ONCE(*vmf.pud);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5059) if (pud_none(pudval) || unlikely(pud_bad(pudval)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5060) goto out_walk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5061)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5062) /* Huge pages at PUD level are not supported. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5063) if (unlikely(pud_trans_huge(pudval)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5064) goto out_walk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5065)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5066) vmf.pmd = pmd_offset(vmf.pud, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5067) if (pud_val(READ_ONCE(*vmf.pud)) != pud_val(pudval))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5068) goto out_walk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5069) vmf.orig_pmd = READ_ONCE(*vmf.pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5070) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5071) * pmd_none could mean that a hugepage collapse is in progress
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5072) * in our back as collapse_huge_page() mark it before
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5073) * invalidating the pte (which is done once the IPI is catched
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5074) * by all CPU and we have interrupt disabled).
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5075) * For this reason we cannot handle THP in a speculative way since we
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5076) * can't safely indentify an in progress collapse operation done in our
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5077) * back on that PMD.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5078) * Regarding the order of the following checks, see comment in
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5079) * pmd_devmap_trans_unstable()
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5080) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5081) if (unlikely(pmd_devmap(vmf.orig_pmd) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5082) pmd_none(vmf.orig_pmd) || pmd_trans_huge(vmf.orig_pmd) ||
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5083) is_swap_pmd(vmf.orig_pmd)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5084) goto out_walk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5085)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5086) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5087) * The above does not allocate/instantiate page-tables because doing so
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5088) * would lead to the possibility of instantiating page-tables after
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5089) * free_pgtables() -- and consequently leaking them.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5090) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5091) * The result is that we take at least one !speculative fault per PMD
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5092) * in order to instantiate it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5093) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5094)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5095) vmf.pte = pte_offset_map(vmf.pmd, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5096) if (pmd_val(READ_ONCE(*vmf.pmd)) != pmd_val(vmf.orig_pmd)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5097) pte_unmap(vmf.pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5098) vmf.pte = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5099) goto out_walk;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5100) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5101) vmf.orig_pte = READ_ONCE(*vmf.pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5102) barrier(); /* See comment in handle_pte_fault() */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5103) if (pte_none(vmf.orig_pte)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5104) pte_unmap(vmf.pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5105) vmf.pte = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5106) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5107)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5108) vmf.sequence = seq;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5109) vmf.flags = flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5110)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5111) local_irq_enable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5112)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5113) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5114) * We need to re-validate the VMA after checking the bounds, otherwise
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5115) * we might have a false positive on the bounds.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5116) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5117) if (read_seqcount_retry(&vmf.vma->vm_sequence, seq)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5118) trace_spf_vma_changed(_RET_IP_, vmf.vma, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5119) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5120) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5121)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5122) mem_cgroup_enter_user_fault();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5123) ret = handle_pte_fault(&vmf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5124) mem_cgroup_exit_user_fault();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5125)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5126) if (ret != VM_FAULT_RETRY) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5127) if (vma_is_anonymous(vmf.vma))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5128) count_vm_event(SPECULATIVE_PGFAULT_ANON);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5129) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5130) count_vm_event(SPECULATIVE_PGFAULT_FILE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5131) }
^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) * The task may have entered a memcg OOM situation but
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5135) * if the allocation error was handled gracefully (no
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5136) * VM_FAULT_OOM), there is no need to kill anything.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5137) * Just clean up the OOM state peacefully.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5138) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5139) if (task_in_memcg_oom(current) && !(ret & VM_FAULT_OOM))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5140) mem_cgroup_oom_synchronize(false);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5141) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5142)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5143) out_walk:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5144) trace_spf_vma_notsup(_RET_IP_, vmf.vma, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5145) local_irq_enable();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5146) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5147)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5148) out_segv:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5149) trace_spf_vma_access(_RET_IP_, vmf.vma, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5150) return VM_FAULT_SIGSEGV;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5151) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5152)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5153) vm_fault_t __handle_speculative_fault(struct mm_struct *mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5154) unsigned long address, unsigned int flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5155) struct vm_area_struct **vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5156) struct pt_regs *regs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5157) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5158) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5159)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5160) check_sync_rss_stat(current);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5161)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5162) *vma = get_vma(mm, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5163) if (!*vma)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5164) return VM_FAULT_RETRY;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5165)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5166) ret = ___handle_speculative_fault(mm, address, flags, *vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5167)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5168) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5169) * If there is no need to retry, don't return the vma to the caller.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5170) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5171) if (ret != VM_FAULT_RETRY) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5172) put_vma(*vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5173) *vma = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5174) mm_account_fault(regs, address, flags, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5175) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5176)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5177) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5178) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5179)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5180) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5181) * This is used to know if the vma fetch in the speculative page fault handler
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5182) * is still valid when trying the regular fault path while holding the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5183) * mmap_sem.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5184) * The call to put_vma(vma) must be made after checking the vma's fields, as
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5185) * the vma may be freed by put_vma(). In such a case it is expected that false
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5186) * is returned.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5187) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5188) bool can_reuse_spf_vma(struct vm_area_struct *vma, unsigned long address)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5189) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5190) bool ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5191)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5192) ret = !RB_EMPTY_NODE(&vma->vm_rb) &&
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5193) vma->vm_start <= address && address < vma->vm_end;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5194) put_vma(vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5195) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5196) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5197) #endif /* CONFIG_SPECULATIVE_PAGE_FAULT */
^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) * By the time we get here, we already hold the mm semaphore
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5201) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5202) * The mmap_lock may have been released depending on flags and our
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5203) * return value. See filemap_fault() and __lock_page_or_retry().
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5204) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5205) vm_fault_t handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5206) unsigned int flags, struct pt_regs *regs)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5207) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5208) vm_fault_t ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5209)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5210) __set_current_state(TASK_RUNNING);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5211)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5212) count_vm_event(PGFAULT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5213) count_memcg_event_mm(vma->vm_mm, PGFAULT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5214)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5215) /* do counter updates before entering really critical section. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5216) check_sync_rss_stat(current);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5217)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5218) if (!arch_vma_access_permitted(vma, flags & FAULT_FLAG_WRITE,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5219) flags & FAULT_FLAG_INSTRUCTION,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5220) flags & FAULT_FLAG_REMOTE))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5221) return VM_FAULT_SIGSEGV;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5222)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5223) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5224) * Enable the memcg OOM handling for faults triggered in user
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5225) * space. Kernel faults are handled more gracefully.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5226) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5227) if (flags & FAULT_FLAG_USER)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5228) mem_cgroup_enter_user_fault();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5229)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5230) if (unlikely(is_vm_hugetlb_page(vma)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5231) ret = hugetlb_fault(vma->vm_mm, vma, address, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5232) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5233) ret = __handle_mm_fault(vma, address, flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5234)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5235) if (flags & FAULT_FLAG_USER) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5236) mem_cgroup_exit_user_fault();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5237) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5238) * The task may have entered a memcg OOM situation but
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5239) * if the allocation error was handled gracefully (no
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5240) * VM_FAULT_OOM), there is no need to kill anything.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5241) * Just clean up the OOM state peacefully.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5242) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5243) if (task_in_memcg_oom(current) && !(ret & VM_FAULT_OOM))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5244) mem_cgroup_oom_synchronize(false);
^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) mm_account_fault(regs, address, flags, ret);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5248)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5249) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5250) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5251) EXPORT_SYMBOL_GPL(handle_mm_fault);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5252)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5253) #ifndef __PAGETABLE_P4D_FOLDED
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5254) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5255) * Allocate p4d page table.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5256) * We've already handled the fast-path in-line.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5257) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5258) int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5259) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5260) p4d_t *new = p4d_alloc_one(mm, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5261) if (!new)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5262) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5263)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5264) smp_wmb(); /* See comment in __pte_alloc */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5265)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5266) spin_lock(&mm->page_table_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5267) if (pgd_present(*pgd)) /* Another has populated it */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5268) p4d_free(mm, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5269) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5270) pgd_populate(mm, pgd, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5271) spin_unlock(&mm->page_table_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5272) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5273) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5274) #endif /* __PAGETABLE_P4D_FOLDED */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5275)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5276) #ifndef __PAGETABLE_PUD_FOLDED
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5277) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5278) * Allocate page upper directory.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5279) * We've already handled the fast-path in-line.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5280) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5281) int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5282) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5283) pud_t *new = pud_alloc_one(mm, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5284) if (!new)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5285) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5286)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5287) smp_wmb(); /* See comment in __pte_alloc */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5288)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5289) spin_lock(&mm->page_table_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5290) if (!p4d_present(*p4d)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5291) mm_inc_nr_puds(mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5292) p4d_populate(mm, p4d, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5293) } else /* Another has populated it */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5294) pud_free(mm, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5295) spin_unlock(&mm->page_table_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5296) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5297) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5298) #endif /* __PAGETABLE_PUD_FOLDED */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5299)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5300) #ifndef __PAGETABLE_PMD_FOLDED
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5301) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5302) * Allocate page middle directory.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5303) * We've already handled the fast-path in-line.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5304) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5305) int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5306) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5307) spinlock_t *ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5308) pmd_t *new = pmd_alloc_one(mm, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5309) if (!new)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5310) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5311)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5312) smp_wmb(); /* See comment in __pte_alloc */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5313)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5314) ptl = pud_lock(mm, pud);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5315) if (!pud_present(*pud)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5316) mm_inc_nr_pmds(mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5317) pud_populate(mm, pud, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5318) } else /* Another has populated it */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5319) pmd_free(mm, new);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5320) spin_unlock(ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5321) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5322) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5323) #endif /* __PAGETABLE_PMD_FOLDED */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5324)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5325) int follow_invalidate_pte(struct mm_struct *mm, unsigned long address,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5326) struct mmu_notifier_range *range, pte_t **ptepp,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5327) pmd_t **pmdpp, spinlock_t **ptlp)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5328) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5329) pgd_t *pgd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5330) p4d_t *p4d;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5331) pud_t *pud;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5332) pmd_t *pmd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5333) pte_t *ptep;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5334)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5335) pgd = pgd_offset(mm, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5336) if (pgd_none(*pgd) || unlikely(pgd_bad(*pgd)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5337) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5338)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5339) p4d = p4d_offset(pgd, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5340) if (p4d_none(*p4d) || unlikely(p4d_bad(*p4d)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5341) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5342)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5343) pud = pud_offset(p4d, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5344) if (pud_none(*pud) || unlikely(pud_bad(*pud)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5345) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5346)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5347) pmd = pmd_offset(pud, address);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5348) VM_BUG_ON(pmd_trans_huge(*pmd));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5349)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5350) if (pmd_huge(*pmd)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5351) if (!pmdpp)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5352) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5353)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5354) if (range) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5355) mmu_notifier_range_init(range, MMU_NOTIFY_CLEAR, 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5356) NULL, mm, address & PMD_MASK,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5357) (address & PMD_MASK) + PMD_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5358) mmu_notifier_invalidate_range_start(range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5359) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5360) *ptlp = pmd_lock(mm, pmd);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5361) if (pmd_huge(*pmd)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5362) *pmdpp = pmd;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5363) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5364) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5365) spin_unlock(*ptlp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5366) if (range)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5367) mmu_notifier_invalidate_range_end(range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5368) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5369)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5370) if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5371) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5372)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5373) if (range) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5374) mmu_notifier_range_init(range, MMU_NOTIFY_CLEAR, 0, NULL, mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5375) address & PAGE_MASK,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5376) (address & PAGE_MASK) + PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5377) mmu_notifier_invalidate_range_start(range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5378) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5379) ptep = pte_offset_map_lock(mm, pmd, address, ptlp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5380) if (!pte_present(*ptep))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5381) goto unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5382) *ptepp = ptep;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5383) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5384) unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5385) pte_unmap_unlock(ptep, *ptlp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5386) if (range)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5387) mmu_notifier_invalidate_range_end(range);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5388) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5389) return -EINVAL;
^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) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5393) * follow_pte - look up PTE at a user virtual address
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5394) * @mm: the mm_struct of the target address space
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5395) * @address: user virtual address
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5396) * @ptepp: location to store found PTE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5397) * @ptlp: location to store the lock for the PTE
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5398) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5399) * On a successful return, the pointer to the PTE is stored in @ptepp;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5400) * the corresponding lock is taken and its location is stored in @ptlp.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5401) * The contents of the PTE are only stable until @ptlp is released;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5402) * any further use, if any, must be protected against invalidation
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5403) * with MMU notifiers.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5404) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5405) * Only IO mappings and raw PFN mappings are allowed. The mmap semaphore
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5406) * should be taken for read.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5407) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5408) * KVM uses this function. While it is arguably less bad than ``follow_pfn``,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5409) * it is not a good general-purpose API.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5410) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5411) * Return: zero on success, -ve otherwise.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5412) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5413) int follow_pte(struct mm_struct *mm, unsigned long address,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5414) pte_t **ptepp, spinlock_t **ptlp)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5415) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5416) return follow_invalidate_pte(mm, address, NULL, ptepp, NULL, ptlp);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5417) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5418) EXPORT_SYMBOL_GPL(follow_pte);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5419)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5420) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5421) * follow_pfn - look up PFN at a user virtual address
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5422) * @vma: memory mapping
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5423) * @address: user virtual address
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5424) * @pfn: location to store found PFN
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5425) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5426) * Only IO mappings and raw PFN mappings are allowed.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5427) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5428) * This function does not allow the caller to read the permissions
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5429) * of the PTE. Do not use it.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5430) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5431) * Return: zero and the pfn at @pfn on success, -ve otherwise.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5432) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5433) int follow_pfn(struct vm_area_struct *vma, unsigned long address,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5434) unsigned long *pfn)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5435) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5436) int ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5437) spinlock_t *ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5438) pte_t *ptep;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5439)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5440) if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5441) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5442)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5443) ret = follow_pte(vma->vm_mm, address, &ptep, &ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5444) if (ret)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5445) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5446) *pfn = pte_pfn(*ptep);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5447) pte_unmap_unlock(ptep, ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5448) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5449) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5450) EXPORT_SYMBOL(follow_pfn);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5451)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5452) #ifdef CONFIG_HAVE_IOREMAP_PROT
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5453) int follow_phys(struct vm_area_struct *vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5454) unsigned long address, unsigned int flags,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5455) unsigned long *prot, resource_size_t *phys)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5456) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5457) int ret = -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5458) pte_t *ptep, pte;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5459) spinlock_t *ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5460)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5461) if (!(vma->vm_flags & (VM_IO | VM_PFNMAP)))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5462) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5463)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5464) if (follow_pte(vma->vm_mm, address, &ptep, &ptl))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5465) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5466) pte = *ptep;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5467)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5468) if ((flags & FOLL_WRITE) && !pte_write(pte))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5469) goto unlock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5470)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5471) *prot = pgprot_val(pte_pgprot(pte));
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5472) *phys = (resource_size_t)pte_pfn(pte) << PAGE_SHIFT;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5473)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5474) ret = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5475) unlock:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5476) pte_unmap_unlock(ptep, ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5477) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5478) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5479) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5480)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5481) int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5482) void *buf, int len, int write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5483) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5484) resource_size_t phys_addr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5485) unsigned long prot = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5486) void __iomem *maddr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5487) int offset = addr & (PAGE_SIZE-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5488)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5489) if (follow_phys(vma, addr, write, &prot, &phys_addr))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5490) return -EINVAL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5491)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5492) maddr = ioremap_prot(phys_addr, PAGE_ALIGN(len + offset), prot);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5493) if (!maddr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5494) return -ENOMEM;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5495)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5496) if (write)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5497) memcpy_toio(maddr + offset, buf, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5498) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5499) memcpy_fromio(buf, maddr + offset, len);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5500) iounmap(maddr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5501)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5502) return len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5503) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5504) EXPORT_SYMBOL_GPL(generic_access_phys);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5505) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5506)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5507) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5508) * Access another process' address space as given in mm. If non-NULL, use the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5509) * given task for page fault accounting.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5510) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5511) int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5512) unsigned long addr, void *buf, int len, unsigned int gup_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5513) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5514) struct vm_area_struct *vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5515) void *old_buf = buf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5516) int write = gup_flags & FOLL_WRITE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5517)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5518) if (mmap_read_lock_killable(mm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5519) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5520)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5521) /* ignore errors, just check how much was successfully transferred */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5522) while (len) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5523) int bytes, ret, offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5524) void *maddr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5525) struct page *page = NULL;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5526)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5527) ret = get_user_pages_remote(mm, addr, 1,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5528) gup_flags, &page, &vma, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5529) if (ret <= 0) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5530) #ifndef CONFIG_HAVE_IOREMAP_PROT
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5531) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5532) #else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5533) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5534) * Check if this is a VM_IO | VM_PFNMAP VMA, which
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5535) * we can access using slightly different code.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5536) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5537) vma = find_vma(mm, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5538) if (!vma || vma->vm_start > addr)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5539) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5540) if (vma->vm_ops && vma->vm_ops->access)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5541) ret = vma->vm_ops->access(vma, addr, buf,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5542) len, write);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5543) if (ret <= 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5544) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5545) bytes = ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5546) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5547) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5548) bytes = len;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5549) offset = addr & (PAGE_SIZE-1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5550) if (bytes > PAGE_SIZE-offset)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5551) bytes = PAGE_SIZE-offset;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5552)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5553) maddr = kmap(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5554) if (write) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5555) copy_to_user_page(vma, page, addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5556) maddr + offset, buf, bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5557) set_page_dirty_lock(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5558) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5559) copy_from_user_page(vma, page, addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5560) buf, maddr + offset, bytes);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5561) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5562) kunmap(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5563) put_user_page(page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5564) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5565) len -= bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5566) buf += bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5567) addr += bytes;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5568) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5569) mmap_read_unlock(mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5570)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5571) return buf - old_buf;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5572) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5573)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5574) /**
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5575) * access_remote_vm - access another process' address space
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5576) * @mm: the mm_struct of the target address space
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5577) * @addr: start address to access
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5578) * @buf: source or destination buffer
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5579) * @len: number of bytes to transfer
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5580) * @gup_flags: flags modifying lookup behaviour
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5581) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5582) * The caller must hold a reference on @mm.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5583) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5584) * Return: number of bytes copied from source to destination.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5585) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5586) int access_remote_vm(struct mm_struct *mm, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5587) void *buf, int len, unsigned int gup_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5588) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5589) return __access_remote_vm(NULL, mm, addr, buf, len, gup_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5590) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5591)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5592) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5593) * Access another process' address space.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5594) * Source/target buffer must be kernel space,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5595) * Do not walk the page table directly, use get_user_pages
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5596) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5597) int access_process_vm(struct task_struct *tsk, unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5598) void *buf, int len, unsigned int gup_flags)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5599) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5600) struct mm_struct *mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5601) int ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5602)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5603) mm = get_task_mm(tsk);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5604) if (!mm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5605) return 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5606)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5607) ret = __access_remote_vm(tsk, mm, addr, buf, len, gup_flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5608)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5609) mmput(mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5610)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5611) return ret;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5612) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5613) EXPORT_SYMBOL_GPL(access_process_vm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5614)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5615) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5616) * Print the name of a VMA.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5617) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5618) void print_vma_addr(char *prefix, unsigned long ip)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5619) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5620) struct mm_struct *mm = current->mm;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5621) struct vm_area_struct *vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5622)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5623) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5624) * we might be running from an atomic context so we cannot sleep
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5625) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5626) if (!mmap_read_trylock(mm))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5627) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5628)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5629) vma = find_vma(mm, ip);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5630) if (vma && vma->vm_file) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5631) struct file *f = vma->vm_file;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5632) char *buf = (char *)__get_free_page(GFP_NOWAIT);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5633) if (buf) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5634) char *p;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5635)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5636) p = file_path(f, buf, PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5637) if (IS_ERR(p))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5638) p = "?";
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5639) printk("%s%s[%lx+%lx]", prefix, kbasename(p),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5640) vma->vm_start,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5641) vma->vm_end - vma->vm_start);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5642) free_page((unsigned long)buf);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5643) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5644) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5645) mmap_read_unlock(mm);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5646) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5647)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5648) #if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_DEBUG_ATOMIC_SLEEP)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5649) void __might_fault(const char *file, int line)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5650) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5651) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5652) * Some code (nfs/sunrpc) uses socket ops on kernel memory while
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5653) * holding the mmap_lock, this is safe because kernel memory doesn't
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5654) * get paged out, therefore we'll never actually fault, and the
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5655) * below annotations will generate false positives.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5656) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5657) if (uaccess_kernel())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5658) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5659) if (pagefault_disabled())
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5660) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5661) __might_sleep(file, line, 0);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5662) #if defined(CONFIG_DEBUG_ATOMIC_SLEEP)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5663) if (current->mm)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5664) might_lock_read(¤t->mm->mmap_lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5665) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5666) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5667) EXPORT_SYMBOL(__might_fault);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5668) #endif
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5669)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5670) #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5671) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5672) * Process all subpages of the specified huge page with the specified
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5673) * operation. The target subpage will be processed last to keep its
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5674) * cache lines hot.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5675) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5676) static inline void process_huge_page(
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5677) unsigned long addr_hint, unsigned int pages_per_huge_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5678) void (*process_subpage)(unsigned long addr, int idx, void *arg),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5679) void *arg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5680) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5681) int i, n, base, l;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5682) unsigned long addr = addr_hint &
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5683) ~(((unsigned long)pages_per_huge_page << PAGE_SHIFT) - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5684)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5685) /* Process target subpage last to keep its cache lines hot */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5686) might_sleep();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5687) n = (addr_hint - addr) / PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5688) if (2 * n <= pages_per_huge_page) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5689) /* If target subpage in first half of huge page */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5690) base = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5691) l = n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5692) /* Process subpages at the end of huge page */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5693) for (i = pages_per_huge_page - 1; i >= 2 * n; i--) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5694) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5695) process_subpage(addr + i * PAGE_SIZE, i, arg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5696) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5697) } else {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5698) /* If target subpage in second half of huge page */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5699) base = pages_per_huge_page - 2 * (pages_per_huge_page - n);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5700) l = pages_per_huge_page - n;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5701) /* Process subpages at the begin of huge page */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5702) for (i = 0; i < base; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5703) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5704) process_subpage(addr + i * PAGE_SIZE, i, arg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5705) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5706) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5707) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5708) * Process remaining subpages in left-right-left-right pattern
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5709) * towards the target subpage
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5710) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5711) for (i = 0; i < l; i++) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5712) int left_idx = base + i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5713) int right_idx = base + 2 * l - 1 - i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5714)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5715) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5716) process_subpage(addr + left_idx * PAGE_SIZE, left_idx, arg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5717) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5718) process_subpage(addr + right_idx * PAGE_SIZE, right_idx, arg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5719) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5720) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5721)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5722) static void clear_gigantic_page(struct page *page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5723) unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5724) unsigned int pages_per_huge_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5725) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5726) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5727) struct page *p = page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5728)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5729) might_sleep();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5730) for (i = 0; i < pages_per_huge_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5731) i++, p = mem_map_next(p, page, i)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5732) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5733) clear_user_highpage(p, addr + i * PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5734) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5735) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5736)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5737) static void clear_subpage(unsigned long addr, int idx, void *arg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5738) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5739) struct page *page = arg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5740)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5741) clear_user_highpage(page + idx, addr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5742) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5743)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5744) void clear_huge_page(struct page *page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5745) unsigned long addr_hint, unsigned int pages_per_huge_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5746) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5747) unsigned long addr = addr_hint &
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5748) ~(((unsigned long)pages_per_huge_page << PAGE_SHIFT) - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5749)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5750) if (unlikely(pages_per_huge_page > MAX_ORDER_NR_PAGES)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5751) clear_gigantic_page(page, addr, pages_per_huge_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5752) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5753) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5754)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5755) process_huge_page(addr_hint, pages_per_huge_page, clear_subpage, page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5756) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5757)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5758) static void copy_user_gigantic_page(struct page *dst, struct page *src,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5759) unsigned long addr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5760) struct vm_area_struct *vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5761) unsigned int pages_per_huge_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5762) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5763) int i;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5764) struct page *dst_base = dst;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5765) struct page *src_base = src;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5766)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5767) for (i = 0; i < pages_per_huge_page; ) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5768) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5769) copy_user_highpage(dst, src, addr + i*PAGE_SIZE, vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5770)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5771) i++;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5772) dst = mem_map_next(dst, dst_base, i);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5773) src = mem_map_next(src, src_base, i);
^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)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5777) struct copy_subpage_arg {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5778) struct page *dst;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5779) struct page *src;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5780) struct vm_area_struct *vma;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5781) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5782)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5783) static void copy_subpage(unsigned long addr, int idx, void *arg)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5784) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5785) struct copy_subpage_arg *copy_arg = arg;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5786)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5787) copy_user_highpage(copy_arg->dst + idx, copy_arg->src + idx,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5788) addr, copy_arg->vma);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5789) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5790)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5791) void copy_user_huge_page(struct page *dst, struct page *src,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5792) unsigned long addr_hint, struct vm_area_struct *vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5793) unsigned int pages_per_huge_page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5794) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5795) unsigned long addr = addr_hint &
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5796) ~(((unsigned long)pages_per_huge_page << PAGE_SHIFT) - 1);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5797) struct copy_subpage_arg arg = {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5798) .dst = dst,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5799) .src = src,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5800) .vma = vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5801) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5802)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5803) if (unlikely(pages_per_huge_page > MAX_ORDER_NR_PAGES)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5804) copy_user_gigantic_page(dst, src, addr, vma,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5805) pages_per_huge_page);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5806) return;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5807) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5808)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5809) process_huge_page(addr_hint, pages_per_huge_page, copy_subpage, &arg);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5810) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5811)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5812) long copy_huge_page_from_user(struct page *dst_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5813) const void __user *usr_src,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5814) unsigned int pages_per_huge_page,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5815) bool allow_pagefault)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5816) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5817) void *src = (void *)usr_src;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5818) void *page_kaddr;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5819) unsigned long i, rc = 0;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5820) unsigned long ret_val = pages_per_huge_page * PAGE_SIZE;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5821) struct page *subpage = dst_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5822)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5823) for (i = 0; i < pages_per_huge_page;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5824) i++, subpage = mem_map_next(subpage, dst_page, i)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5825) if (allow_pagefault)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5826) page_kaddr = kmap(subpage);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5827) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5828) page_kaddr = kmap_atomic(subpage);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5829) rc = copy_from_user(page_kaddr,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5830) (const void __user *)(src + i * PAGE_SIZE),
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5831) PAGE_SIZE);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5832) if (allow_pagefault)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5833) kunmap(subpage);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5834) else
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5835) kunmap_atomic(page_kaddr);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5836)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5837) ret_val -= (PAGE_SIZE - rc);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5838) if (rc)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5839) break;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5840)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5841) cond_resched();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5842) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5843) return ret_val;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5844) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5845) #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5846)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5847) #if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5848)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5849) static struct kmem_cache *page_ptl_cachep;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5850)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5851) void __init ptlock_cache_init(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5852) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5853) page_ptl_cachep = kmem_cache_create("page->ptl", sizeof(spinlock_t), 0,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5854) SLAB_PANIC, NULL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5855) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5856)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5857) bool ptlock_alloc(struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5858) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5859) spinlock_t *ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5860)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5861) ptl = kmem_cache_alloc(page_ptl_cachep, GFP_KERNEL);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5862) if (!ptl)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5863) return false;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5864) page->ptl = ptl;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5865) return true;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5866) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5867)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5868) void ptlock_free(struct page *page)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5869) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5870) kmem_cache_free(page_ptl_cachep, page->ptl);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5871) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5872) #endif
Orange Pi5 kernel
Deprecated Linux kernel 5.10.110 for OrangePi 5/5B/5+ boards
3 Commits
0 Branches
0 Tags