^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1) // SPDX-License-Identifier: GPL-2.0
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3) * tracing clocks
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5) * Copyright (C) 2009 Red Hat, Inc., Ingo Molnar <mingo@redhat.com>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7) * Implements 3 trace clock variants, with differing scalability/precision
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8) * tradeoffs:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10) * - local: CPU-local trace clock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11) * - medium: scalable global clock with some jitter
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 12) * - global: globally monotonic, serialized clock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 13) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 14) * Tracer plugins will chose a default from these clocks.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 15) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 16) #include <linux/spinlock.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 17) #include <linux/irqflags.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 18) #include <linux/hardirq.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 19) #include <linux/module.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 20) #include <linux/percpu.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 21) #include <linux/sched.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 22) #include <linux/sched/clock.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 23) #include <linux/ktime.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 24) #include <linux/trace_clock.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 25)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 26) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 27) * trace_clock_local(): the simplest and least coherent tracing clock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 28) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 29) * Useful for tracing that does not cross to other CPUs nor
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 30) * does it go through idle events.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 31) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 32) u64 notrace trace_clock_local(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 33) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 34) u64 clock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 35)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 36) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 37) * sched_clock() is an architecture implemented, fast, scalable,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 38) * lockless clock. It is not guaranteed to be coherent across
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 39) * CPUs, nor across CPU idle events.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 40) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 41) preempt_disable_notrace();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 42) clock = sched_clock();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 43) preempt_enable_notrace();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 44)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 45) return clock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 46) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 47) EXPORT_SYMBOL_GPL(trace_clock_local);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 48)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 49) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 50) * trace_clock(): 'between' trace clock. Not completely serialized,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 51) * but not completely incorrect when crossing CPUs either.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 52) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 53) * This is based on cpu_clock(), which will allow at most ~1 jiffy of
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 54) * jitter between CPUs. So it's a pretty scalable clock, but there
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 55) * can be offsets in the trace data.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 56) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 57) u64 notrace trace_clock(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 58) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 59) return local_clock();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 60) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 61) EXPORT_SYMBOL_GPL(trace_clock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 62)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 63) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 64) * trace_jiffy_clock(): Simply use jiffies as a clock counter.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 65) * Note that this use of jiffies_64 is not completely safe on
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 66) * 32-bit systems. But the window is tiny, and the effect if
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 67) * we are affected is that we will have an obviously bogus
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 68) * timestamp on a trace event - i.e. not life threatening.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 69) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 70) u64 notrace trace_clock_jiffies(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 71) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 72) return jiffies_64_to_clock_t(jiffies_64 - INITIAL_JIFFIES);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 73) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 74) EXPORT_SYMBOL_GPL(trace_clock_jiffies);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 75)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 76) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 77) * trace_clock_global(): special globally coherent trace clock
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 78) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 79) * It has higher overhead than the other trace clocks but is still
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 80) * an order of magnitude faster than GTOD derived hardware clocks.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 81) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 82) * Used by plugins that need globally coherent timestamps.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 83) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 84)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 85) /* keep prev_time and lock in the same cacheline. */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 86) static struct {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 87) u64 prev_time;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 88) arch_spinlock_t lock;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 89) } trace_clock_struct ____cacheline_aligned_in_smp =
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 90) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 91) .lock = (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED,
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 92) };
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 93)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 94) u64 notrace trace_clock_global(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 95) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 96) unsigned long flags;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 97) int this_cpu;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 98) u64 now, prev_time;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 99)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 100) raw_local_irq_save(flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 101)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 102) this_cpu = raw_smp_processor_id();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 103)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 104) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 105) * The global clock "guarantees" that the events are ordered
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 106) * between CPUs. But if two events on two different CPUS call
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 107) * trace_clock_global at roughly the same time, it really does
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 108) * not matter which one gets the earlier time. Just make sure
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 109) * that the same CPU will always show a monotonic clock.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 110) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 111) * Use a read memory barrier to get the latest written
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 112) * time that was recorded.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 113) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 114) smp_rmb();
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 115) prev_time = READ_ONCE(trace_clock_struct.prev_time);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 116) now = sched_clock_cpu(this_cpu);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 117)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 118) /* Make sure that now is always greater than or equal to prev_time */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 119) if ((s64)(now - prev_time) < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 120) now = prev_time;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 121)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 122) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 123) * If in an NMI context then dont risk lockups and simply return
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 124) * the current time.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 125) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 126) if (unlikely(in_nmi()))
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 127) goto out;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 128)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 129) /* Tracing can cause strange recursion, always use a try lock */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 130) if (arch_spin_trylock(&trace_clock_struct.lock)) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 131) /* Reread prev_time in case it was already updated */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 132) prev_time = READ_ONCE(trace_clock_struct.prev_time);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 133) if ((s64)(now - prev_time) < 0)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 134) now = prev_time;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 135)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 136) trace_clock_struct.prev_time = now;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 137)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 138) /* The unlock acts as the wmb for the above rmb */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 139) arch_spin_unlock(&trace_clock_struct.lock);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 140) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 141) out:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 142) raw_local_irq_restore(flags);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 143)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 144) return now;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 145) }
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 146) EXPORT_SYMBOL_GPL(trace_clock_global);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 147)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 148) static atomic64_t trace_counter;
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 149)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 150) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 151) * trace_clock_counter(): simply an atomic counter.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 152) * Use the trace_counter "counter" for cases where you do not care
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 153) * about timings, but are interested in strict ordering.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 154) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 155) u64 notrace trace_clock_counter(void)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 156) {
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 157) return atomic64_add_return(1, &trace_counter);
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 158) }