blob: abb8abc1647b26c90618d3b2fd0fe1c51a5cb162 [file] [log] [blame]
// SPDX-License-Identifier: GPL-2.0
/*
* Performance events core code:
*
* Copyright (C) 2008 Thomas Gleixner <tglx@linutronix.de>
* Copyright (C) 2008-2011 Red Hat, Inc., Ingo Molnar
* Copyright (C) 2008-2011 Red Hat, Inc., Peter Zijlstra
* Copyright © 2009 Paul Mackerras, IBM Corp. <paulus@au1.ibm.com>
*/
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/cpu.h>
#include <linux/smp.h>
#include <linux/idr.h>
#include <linux/file.h>
#include <linux/poll.h>
#include <linux/slab.h>
#include <linux/hash.h>
#include <linux/tick.h>
#include <linux/sysfs.h>
#include <linux/dcache.h>
#include <linux/percpu.h>
#include <linux/ptrace.h>
#include <linux/reboot.h>
#include <linux/vmstat.h>
#include <linux/device.h>
#include <linux/export.h>
#include <linux/vmalloc.h>
#include <linux/hardirq.h>
#include <linux/rculist.h>
#include <linux/uaccess.h>
#include <linux/syscalls.h>
#include <linux/anon_inodes.h>
#include <linux/kernel_stat.h>
#include <linux/cgroup.h>
#include <linux/perf_event.h>
#include <linux/trace_events.h>
#include <linux/hw_breakpoint.h>
#include <linux/mm_types.h>
#include <linux/module.h>
#include <linux/mman.h>
#include <linux/compat.h>
#include <linux/bpf.h>
#include <linux/filter.h>
#include <linux/namei.h>
#include <linux/parser.h>
#include <linux/sched/clock.h>
#include <linux/sched/mm.h>
#include <linux/proc_ns.h>
#include <linux/mount.h>
#include "internal.h"
#include <asm/irq_regs.h>
typedef int (*remote_function_f)(void *);
struct remote_function_call {
struct task_struct *p;
remote_function_f func;
void *info;
int ret;
};
static void remote_function(void *data)
{
struct remote_function_call *tfc = data;
struct task_struct *p = tfc->p;
if (p) {
/* -EAGAIN */
if (task_cpu(p) != smp_processor_id())
return;
/*
* Now that we're on right CPU with IRQs disabled, we can test
* if we hit the right task without races.
*/
tfc->ret = -ESRCH; /* No such (running) process */
if (p != current)
return;
}
tfc->ret = tfc->func(tfc->info);
}
/**
* task_function_call - call a function on the cpu on which a task runs
* @p: the task to evaluate
* @func: the function to be called
* @info: the function call argument
*
* Calls the function @func when the task is currently running. This might
* be on the current CPU, which just calls the function directly. This will
* retry due to any failures in smp_call_function_single(), such as if the
* task_cpu() goes offline concurrently.
*
* returns @func return value or -ESRCH or -ENXIO when the process isn't running
*/
static int
task_function_call(struct task_struct *p, remote_function_f func, void *info)
{
struct remote_function_call data = {
.p = p,
.func = func,
.info = info,
.ret = -EAGAIN,
};
int ret;
for (;;) {
ret = smp_call_function_single(task_cpu(p), remote_function,
&data, 1);
if (!ret)
ret = data.ret;
if (ret != -EAGAIN)
break;
cond_resched();
}
return ret;
}
/**
* cpu_function_call - call a function on the cpu
* @func: the function to be called
* @info: the function call argument
*
* Calls the function @func on the remote cpu.
*
* returns: @func return value or -ENXIO when the cpu is offline
*/
static int cpu_function_call(int cpu, remote_function_f func, void *info)
{
struct remote_function_call data = {
.p = NULL,
.func = func,
.info = info,
.ret = -ENXIO, /* No such CPU */
};
smp_call_function_single(cpu, remote_function, &data, 1);
return data.ret;
}
static inline struct perf_cpu_context *
__get_cpu_context(struct perf_event_context *ctx)
{
return this_cpu_ptr(ctx->pmu->pmu_cpu_context);
}
static void perf_ctx_lock(struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx)
{
raw_spin_lock(&cpuctx->ctx.lock);
if (ctx)
raw_spin_lock(&ctx->lock);
}
static void perf_ctx_unlock(struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx)
{
if (ctx)
raw_spin_unlock(&ctx->lock);
raw_spin_unlock(&cpuctx->ctx.lock);
}
#define TASK_TOMBSTONE ((void *)-1L)
static bool is_kernel_event(struct perf_event *event)
{
return READ_ONCE(event->owner) == TASK_TOMBSTONE;
}
/*
* On task ctx scheduling...
*
* When !ctx->nr_events a task context will not be scheduled. This means
* we can disable the scheduler hooks (for performance) without leaving
* pending task ctx state.
*
* This however results in two special cases:
*
* - removing the last event from a task ctx; this is relatively straight
* forward and is done in __perf_remove_from_context.
*
* - adding the first event to a task ctx; this is tricky because we cannot
* rely on ctx->is_active and therefore cannot use event_function_call().
* See perf_install_in_context().
*
* If ctx->nr_events, then ctx->is_active and cpuctx->task_ctx are set.
*/
typedef void (*event_f)(struct perf_event *, struct perf_cpu_context *,
struct perf_event_context *, void *);
struct event_function_struct {
struct perf_event *event;
event_f func;
void *data;
};
static int event_function(void *info)
{
struct event_function_struct *efs = info;
struct perf_event *event = efs->event;
struct perf_event_context *ctx = event->ctx;
struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
struct perf_event_context *task_ctx = cpuctx->task_ctx;
int ret = 0;
lockdep_assert_irqs_disabled();
perf_ctx_lock(cpuctx, task_ctx);
/*
* Since we do the IPI call without holding ctx->lock things can have
* changed, double check we hit the task we set out to hit.
*/
if (ctx->task) {
if (ctx->task != current) {
ret = -ESRCH;
goto unlock;
}
/*
* We only use event_function_call() on established contexts,
* and event_function() is only ever called when active (or
* rather, we'll have bailed in task_function_call() or the
* above ctx->task != current test), therefore we must have
* ctx->is_active here.
*/
WARN_ON_ONCE(!ctx->is_active);
/*
* And since we have ctx->is_active, cpuctx->task_ctx must
* match.
*/
WARN_ON_ONCE(task_ctx != ctx);
} else {
WARN_ON_ONCE(&cpuctx->ctx != ctx);
}
efs->func(event, cpuctx, ctx, efs->data);
unlock:
perf_ctx_unlock(cpuctx, task_ctx);
return ret;
}
static void event_function_call(struct perf_event *event, event_f func, void *data)
{
struct perf_event_context *ctx = event->ctx;
struct task_struct *task = READ_ONCE(ctx->task); /* verified in event_function */
struct event_function_struct efs = {
.event = event,
.func = func,
.data = data,
};
if (!event->parent) {
/*
* If this is a !child event, we must hold ctx::mutex to
* stabilize the the event->ctx relation. See
* perf_event_ctx_lock().
*/
lockdep_assert_held(&ctx->mutex);
}
if (!task) {
cpu_function_call(event->cpu, event_function, &efs);
return;
}
if (task == TASK_TOMBSTONE)
return;
again:
if (!task_function_call(task, event_function, &efs))
return;
raw_spin_lock_irq(&ctx->lock);
/*
* Reload the task pointer, it might have been changed by
* a concurrent perf_event_context_sched_out().
*/
task = ctx->task;
if (task == TASK_TOMBSTONE) {
raw_spin_unlock_irq(&ctx->lock);
return;
}
if (ctx->is_active) {
raw_spin_unlock_irq(&ctx->lock);
goto again;
}
func(event, NULL, ctx, data);
raw_spin_unlock_irq(&ctx->lock);
}
/*
* Similar to event_function_call() + event_function(), but hard assumes IRQs
* are already disabled and we're on the right CPU.
*/
static void event_function_local(struct perf_event *event, event_f func, void *data)
{
struct perf_event_context *ctx = event->ctx;
struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
struct task_struct *task = READ_ONCE(ctx->task);
struct perf_event_context *task_ctx = NULL;
lockdep_assert_irqs_disabled();
if (task) {
if (task == TASK_TOMBSTONE)
return;
task_ctx = ctx;
}
perf_ctx_lock(cpuctx, task_ctx);
task = ctx->task;
if (task == TASK_TOMBSTONE)
goto unlock;
if (task) {
/*
* We must be either inactive or active and the right task,
* otherwise we're screwed, since we cannot IPI to somewhere
* else.
*/
if (ctx->is_active) {
if (WARN_ON_ONCE(task != current))
goto unlock;
if (WARN_ON_ONCE(cpuctx->task_ctx != ctx))
goto unlock;
}
} else {
WARN_ON_ONCE(&cpuctx->ctx != ctx);
}
func(event, cpuctx, ctx, data);
unlock:
perf_ctx_unlock(cpuctx, task_ctx);
}
#define PERF_FLAG_ALL (PERF_FLAG_FD_NO_GROUP |\
PERF_FLAG_FD_OUTPUT |\
PERF_FLAG_PID_CGROUP |\
PERF_FLAG_FD_CLOEXEC)
/*
* branch priv levels that need permission checks
*/
#define PERF_SAMPLE_BRANCH_PERM_PLM \
(PERF_SAMPLE_BRANCH_KERNEL |\
PERF_SAMPLE_BRANCH_HV)
enum event_type_t {
EVENT_FLEXIBLE = 0x1,
EVENT_PINNED = 0x2,
EVENT_TIME = 0x4,
/* see ctx_resched() for details */
EVENT_CPU = 0x8,
EVENT_ALL = EVENT_FLEXIBLE | EVENT_PINNED,
};
/*
* perf_sched_events : >0 events exist
* perf_cgroup_events: >0 per-cpu cgroup events exist on this cpu
*/
static void perf_sched_delayed(struct work_struct *work);
DEFINE_STATIC_KEY_FALSE(perf_sched_events);
static DECLARE_DELAYED_WORK(perf_sched_work, perf_sched_delayed);
static DEFINE_MUTEX(perf_sched_mutex);
static atomic_t perf_sched_count;
static DEFINE_PER_CPU(atomic_t, perf_cgroup_events);
static DEFINE_PER_CPU(int, perf_sched_cb_usages);
static DEFINE_PER_CPU(struct pmu_event_list, pmu_sb_events);
static atomic_t nr_mmap_events __read_mostly;
static atomic_t nr_comm_events __read_mostly;
static atomic_t nr_namespaces_events __read_mostly;
static atomic_t nr_task_events __read_mostly;
static atomic_t nr_freq_events __read_mostly;
static atomic_t nr_switch_events __read_mostly;
static atomic_t nr_ksymbol_events __read_mostly;
static atomic_t nr_bpf_events __read_mostly;
static LIST_HEAD(pmus);
static DEFINE_MUTEX(pmus_lock);
static struct srcu_struct pmus_srcu;
static cpumask_var_t perf_online_mask;
/*
* perf event paranoia level:
* -1 - not paranoid at all
* 0 - disallow raw tracepoint access for unpriv
* 1 - disallow cpu events for unpriv
* 2 - disallow kernel profiling for unpriv
*/
int sysctl_perf_event_paranoid __read_mostly = 2;
/* Minimum for 512 kiB + 1 user control page */
int sysctl_perf_event_mlock __read_mostly = 512 + (PAGE_SIZE / 1024); /* 'free' kiB per user */
/*
* max perf event sample rate
*/
#define DEFAULT_MAX_SAMPLE_RATE 100000
#define DEFAULT_SAMPLE_PERIOD_NS (NSEC_PER_SEC / DEFAULT_MAX_SAMPLE_RATE)
#define DEFAULT_CPU_TIME_MAX_PERCENT 25
int sysctl_perf_event_sample_rate __read_mostly = DEFAULT_MAX_SAMPLE_RATE;
static int max_samples_per_tick __read_mostly = DIV_ROUND_UP(DEFAULT_MAX_SAMPLE_RATE, HZ);
static int perf_sample_period_ns __read_mostly = DEFAULT_SAMPLE_PERIOD_NS;
static int perf_sample_allowed_ns __read_mostly =
DEFAULT_SAMPLE_PERIOD_NS * DEFAULT_CPU_TIME_MAX_PERCENT / 100;
static void update_perf_cpu_limits(void)
{
u64 tmp = perf_sample_period_ns;
tmp *= sysctl_perf_cpu_time_max_percent;
tmp = div_u64(tmp, 100);
if (!tmp)
tmp = 1;
WRITE_ONCE(perf_sample_allowed_ns, tmp);
}
static bool perf_rotate_context(struct perf_cpu_context *cpuctx);
int perf_proc_update_handler(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp,
loff_t *ppos)
{
int ret;
int perf_cpu = sysctl_perf_cpu_time_max_percent;
/*
* If throttling is disabled don't allow the write:
*/
if (write && (perf_cpu == 100 || perf_cpu == 0))
return -EINVAL;
ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
if (ret || !write)
return ret;
max_samples_per_tick = DIV_ROUND_UP(sysctl_perf_event_sample_rate, HZ);
perf_sample_period_ns = NSEC_PER_SEC / sysctl_perf_event_sample_rate;
update_perf_cpu_limits();
return 0;
}
int sysctl_perf_cpu_time_max_percent __read_mostly = DEFAULT_CPU_TIME_MAX_PERCENT;
int perf_cpu_time_max_percent_handler(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp,
loff_t *ppos)
{
int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
if (ret || !write)
return ret;
if (sysctl_perf_cpu_time_max_percent == 100 ||
sysctl_perf_cpu_time_max_percent == 0) {
printk(KERN_WARNING
"perf: Dynamic interrupt throttling disabled, can hang your system!\n");
WRITE_ONCE(perf_sample_allowed_ns, 0);
} else {
update_perf_cpu_limits();
}
return 0;
}
/*
* perf samples are done in some very critical code paths (NMIs).
* If they take too much CPU time, the system can lock up and not
* get any real work done. This will drop the sample rate when
* we detect that events are taking too long.
*/
#define NR_ACCUMULATED_SAMPLES 128
static DEFINE_PER_CPU(u64, running_sample_length);
static u64 __report_avg;
static u64 __report_allowed;
static void perf_duration_warn(struct irq_work *w)
{
printk_ratelimited(KERN_INFO
"perf: interrupt took too long (%lld > %lld), lowering "
"kernel.perf_event_max_sample_rate to %d\n",
__report_avg, __report_allowed,
sysctl_perf_event_sample_rate);
}
static DEFINE_IRQ_WORK(perf_duration_work, perf_duration_warn);
void perf_sample_event_took(u64 sample_len_ns)
{
u64 max_len = READ_ONCE(perf_sample_allowed_ns);
u64 running_len;
u64 avg_len;
u32 max;
if (max_len == 0)
return;
/* Decay the counter by 1 average sample. */
running_len = __this_cpu_read(running_sample_length);
running_len -= running_len/NR_ACCUMULATED_SAMPLES;
running_len += sample_len_ns;
__this_cpu_write(running_sample_length, running_len);
/*
* Note: this will be biased artifically low until we have
* seen NR_ACCUMULATED_SAMPLES. Doing it this way keeps us
* from having to maintain a count.
*/
avg_len = running_len/NR_ACCUMULATED_SAMPLES;
if (avg_len <= max_len)
return;
__report_avg = avg_len;
__report_allowed = max_len;
/*
* Compute a throttle threshold 25% below the current duration.
*/
avg_len += avg_len / 4;
max = (TICK_NSEC / 100) * sysctl_perf_cpu_time_max_percent;
if (avg_len < max)
max /= (u32)avg_len;
else
max = 1;
WRITE_ONCE(perf_sample_allowed_ns, avg_len);
WRITE_ONCE(max_samples_per_tick, max);
sysctl_perf_event_sample_rate = max * HZ;
perf_sample_period_ns = NSEC_PER_SEC / sysctl_perf_event_sample_rate;
if (!irq_work_queue(&perf_duration_work)) {
early_printk("perf: interrupt took too long (%lld > %lld), lowering "
"kernel.perf_event_max_sample_rate to %d\n",
__report_avg, __report_allowed,
sysctl_perf_event_sample_rate);
}
}
static atomic64_t perf_event_id;
static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx,
enum event_type_t event_type);
static void cpu_ctx_sched_in(struct perf_cpu_context *cpuctx,
enum event_type_t event_type,
struct task_struct *task);
static void update_context_time(struct perf_event_context *ctx);
static u64 perf_event_time(struct perf_event *event);
void __weak perf_event_print_debug(void) { }
extern __weak const char *perf_pmu_name(void)
{
return "pmu";
}
static inline u64 perf_clock(void)
{
return local_clock();
}
static inline u64 perf_event_clock(struct perf_event *event)
{
return event->clock();
}
/*
* State based event timekeeping...
*
* The basic idea is to use event->state to determine which (if any) time
* fields to increment with the current delta. This means we only need to
* update timestamps when we change state or when they are explicitly requested
* (read).
*
* Event groups make things a little more complicated, but not terribly so. The
* rules for a group are that if the group leader is OFF the entire group is
* OFF, irrespecive of what the group member states are. This results in
* __perf_effective_state().
*
* A futher ramification is that when a group leader flips between OFF and
* !OFF, we need to update all group member times.
*
*
* NOTE: perf_event_time() is based on the (cgroup) context time, and thus we
* need to make sure the relevant context time is updated before we try and
* update our timestamps.
*/
static __always_inline enum perf_event_state
__perf_effective_state(struct perf_event *event)
{
struct perf_event *leader = event->group_leader;
if (leader->state <= PERF_EVENT_STATE_OFF)
return leader->state;
return event->state;
}
static __always_inline void
__perf_update_times(struct perf_event *event, u64 now, u64 *enabled, u64 *running)
{
enum perf_event_state state = __perf_effective_state(event);
u64 delta = now - event->tstamp;
*enabled = event->total_time_enabled;
if (state >= PERF_EVENT_STATE_INACTIVE)
*enabled += delta;
*running = event->total_time_running;
if (state >= PERF_EVENT_STATE_ACTIVE)
*running += delta;
}
static void perf_event_update_time(struct perf_event *event)
{
u64 now = perf_event_time(event);
__perf_update_times(event, now, &event->total_time_enabled,
&event->total_time_running);
event->tstamp = now;
}
static void perf_event_update_sibling_time(struct perf_event *leader)
{
struct perf_event *sibling;
for_each_sibling_event(sibling, leader)
perf_event_update_time(sibling);
}
static void
perf_event_set_state(struct perf_event *event, enum perf_event_state state)
{
if (event->state == state)
return;
perf_event_update_time(event);
/*
* If a group leader gets enabled/disabled all its siblings
* are affected too.
*/
if ((event->state < 0) ^ (state < 0))
perf_event_update_sibling_time(event);
WRITE_ONCE(event->state, state);
}
#ifdef CONFIG_CGROUP_PERF
static inline bool
perf_cgroup_match(struct perf_event *event)
{
struct perf_event_context *ctx = event->ctx;
struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
/* @event doesn't care about cgroup */
if (!event->cgrp)
return true;
/* wants specific cgroup scope but @cpuctx isn't associated with any */
if (!cpuctx->cgrp)
return false;
/*
* Cgroup scoping is recursive. An event enabled for a cgroup is
* also enabled for all its descendant cgroups. If @cpuctx's
* cgroup is a descendant of @event's (the test covers identity
* case), it's a match.
*/
return cgroup_is_descendant(cpuctx->cgrp->css.cgroup,
event->cgrp->css.cgroup);
}
static inline void perf_detach_cgroup(struct perf_event *event)
{
css_put(&event->cgrp->css);
event->cgrp = NULL;
}
static inline int is_cgroup_event(struct perf_event *event)
{
return event->cgrp != NULL;
}
static inline u64 perf_cgroup_event_time(struct perf_event *event)
{
struct perf_cgroup_info *t;
t = per_cpu_ptr(event->cgrp->info, event->cpu);
return t->time;
}
static inline void __update_cgrp_time(struct perf_cgroup *cgrp)
{
struct perf_cgroup_info *info;
u64 now;
now = perf_clock();
info = this_cpu_ptr(cgrp->info);
info->time += now - info->timestamp;
info->timestamp = now;
}
static inline void update_cgrp_time_from_cpuctx(struct perf_cpu_context *cpuctx)
{
struct perf_cgroup *cgrp = cpuctx->cgrp;
struct cgroup_subsys_state *css;
if (cgrp) {
for (css = &cgrp->css; css; css = css->parent) {
cgrp = container_of(css, struct perf_cgroup, css);
__update_cgrp_time(cgrp);
}
}
}
static inline void update_cgrp_time_from_event(struct perf_event *event)
{
struct perf_cgroup *cgrp;
/*
* ensure we access cgroup data only when needed and
* when we know the cgroup is pinned (css_get)
*/
if (!is_cgroup_event(event))
return;
cgrp = perf_cgroup_from_task(current, event->ctx);
/*
* Do not update time when cgroup is not active
*/
if (cgroup_is_descendant(cgrp->css.cgroup, event->cgrp->css.cgroup))
__update_cgrp_time(event->cgrp);
}
static inline void
perf_cgroup_set_timestamp(struct task_struct *task,
struct perf_event_context *ctx)
{
struct perf_cgroup *cgrp;
struct perf_cgroup_info *info;
struct cgroup_subsys_state *css;
/*
* ctx->lock held by caller
* ensure we do not access cgroup data
* unless we have the cgroup pinned (css_get)
*/
if (!task || !ctx->nr_cgroups)
return;
cgrp = perf_cgroup_from_task(task, ctx);
for (css = &cgrp->css; css; css = css->parent) {
cgrp = container_of(css, struct perf_cgroup, css);
info = this_cpu_ptr(cgrp->info);
info->timestamp = ctx->timestamp;
}
}
static DEFINE_PER_CPU(struct list_head, cgrp_cpuctx_list);
#define PERF_CGROUP_SWOUT 0x1 /* cgroup switch out every event */
#define PERF_CGROUP_SWIN 0x2 /* cgroup switch in events based on task */
/*
* reschedule events based on the cgroup constraint of task.
*
* mode SWOUT : schedule out everything
* mode SWIN : schedule in based on cgroup for next
*/
static void perf_cgroup_switch(struct task_struct *task, int mode)
{
struct perf_cpu_context *cpuctx, *tmp;
struct list_head *list;
unsigned long flags;
/*
* Disable interrupts and preemption to avoid this CPU's
* cgrp_cpuctx_entry to change under us.
*/
local_irq_save(flags);
list = this_cpu_ptr(&cgrp_cpuctx_list);
list_for_each_entry_safe(cpuctx, tmp, list, cgrp_cpuctx_entry) {
WARN_ON_ONCE(cpuctx->ctx.nr_cgroups == 0);
perf_ctx_lock(cpuctx, cpuctx->task_ctx);
perf_pmu_disable(cpuctx->ctx.pmu);
if (mode & PERF_CGROUP_SWOUT) {
cpu_ctx_sched_out(cpuctx, EVENT_ALL);
/*
* must not be done before ctxswout due
* to event_filter_match() in event_sched_out()
*/
cpuctx->cgrp = NULL;
}
if (mode & PERF_CGROUP_SWIN) {
WARN_ON_ONCE(cpuctx->cgrp);
/*
* set cgrp before ctxsw in to allow
* event_filter_match() to not have to pass
* task around
* we pass the cpuctx->ctx to perf_cgroup_from_task()
* because cgorup events are only per-cpu
*/
cpuctx->cgrp = perf_cgroup_from_task(task,
&cpuctx->ctx);
cpu_ctx_sched_in(cpuctx, EVENT_ALL, task);
}
perf_pmu_enable(cpuctx->ctx.pmu);
perf_ctx_unlock(cpuctx, cpuctx->task_ctx);
}
local_irq_restore(flags);
}
static inline void perf_cgroup_sched_out(struct task_struct *task,
struct task_struct *next)
{
struct perf_cgroup *cgrp1;
struct perf_cgroup *cgrp2 = NULL;
rcu_read_lock();
/*
* we come here when we know perf_cgroup_events > 0
* we do not need to pass the ctx here because we know
* we are holding the rcu lock
*/
cgrp1 = perf_cgroup_from_task(task, NULL);
cgrp2 = perf_cgroup_from_task(next, NULL);
/*
* only schedule out current cgroup events if we know
* that we are switching to a different cgroup. Otherwise,
* do no touch the cgroup events.
*/
if (cgrp1 != cgrp2)
perf_cgroup_switch(task, PERF_CGROUP_SWOUT);
rcu_read_unlock();
}
static inline void perf_cgroup_sched_in(struct task_struct *prev,
struct task_struct *task)
{
struct perf_cgroup *cgrp1;
struct perf_cgroup *cgrp2 = NULL;
rcu_read_lock();
/*
* we come here when we know perf_cgroup_events > 0
* we do not need to pass the ctx here because we know
* we are holding the rcu lock
*/
cgrp1 = perf_cgroup_from_task(task, NULL);
cgrp2 = perf_cgroup_from_task(prev, NULL);
/*
* only need to schedule in cgroup events if we are changing
* cgroup during ctxsw. Cgroup events were not scheduled
* out of ctxsw out if that was not the case.
*/
if (cgrp1 != cgrp2)
perf_cgroup_switch(task, PERF_CGROUP_SWIN);
rcu_read_unlock();
}
static inline int perf_cgroup_connect(int fd, struct perf_event *event,
struct perf_event_attr *attr,
struct perf_event *group_leader)
{
struct perf_cgroup *cgrp;
struct cgroup_subsys_state *css;
struct fd f = fdget(fd);
int ret = 0;
if (!f.file)
return -EBADF;
css = css_tryget_online_from_dir(f.file->f_path.dentry,
&perf_event_cgrp_subsys);
if (IS_ERR(css)) {
ret = PTR_ERR(css);
goto out;
}
cgrp = container_of(css, struct perf_cgroup, css);
event->cgrp = cgrp;
/*
* all events in a group must monitor
* the same cgroup because a task belongs
* to only one perf cgroup at a time
*/
if (group_leader && group_leader->cgrp != cgrp) {
perf_detach_cgroup(event);
ret = -EINVAL;
}
out:
fdput(f);
return ret;
}
static inline void
perf_cgroup_set_shadow_time(struct perf_event *event, u64 now)
{
struct perf_cgroup_info *t;
t = per_cpu_ptr(event->cgrp->info, event->cpu);
event->shadow_ctx_time = now - t->timestamp;
}
/*
* Update cpuctx->cgrp so that it is set when first cgroup event is added and
* cleared when last cgroup event is removed.
*/
static inline void
list_update_cgroup_event(struct perf_event *event,
struct perf_event_context *ctx, bool add)
{
struct perf_cpu_context *cpuctx;
struct list_head *cpuctx_entry;
if (!is_cgroup_event(event))
return;
/*
* Because cgroup events are always per-cpu events,
* this will always be called from the right CPU.
*/
cpuctx = __get_cpu_context(ctx);
/*
* Since setting cpuctx->cgrp is conditional on the current @cgrp
* matching the event's cgroup, we must do this for every new event,
* because if the first would mismatch, the second would not try again
* and we would leave cpuctx->cgrp unset.
*/
if (add && !cpuctx->cgrp) {
struct perf_cgroup *cgrp = perf_cgroup_from_task(current, ctx);
if (cgroup_is_descendant(cgrp->css.cgroup, event->cgrp->css.cgroup))
cpuctx->cgrp = cgrp;
}
if (add && ctx->nr_cgroups++)
return;
else if (!add && --ctx->nr_cgroups)
return;
/* no cgroup running */
if (!add)
cpuctx->cgrp = NULL;
cpuctx_entry = &cpuctx->cgrp_cpuctx_entry;
if (add)
list_add(cpuctx_entry, this_cpu_ptr(&cgrp_cpuctx_list));
else
list_del(cpuctx_entry);
}
#else /* !CONFIG_CGROUP_PERF */
static inline bool
perf_cgroup_match(struct perf_event *event)
{
return true;
}
static inline void perf_detach_cgroup(struct perf_event *event)
{}
static inline int is_cgroup_event(struct perf_event *event)
{
return 0;
}
static inline void update_cgrp_time_from_event(struct perf_event *event)
{
}
static inline void update_cgrp_time_from_cpuctx(struct perf_cpu_context *cpuctx)
{
}
static inline void perf_cgroup_sched_out(struct task_struct *task,
struct task_struct *next)
{
}
static inline void perf_cgroup_sched_in(struct task_struct *prev,
struct task_struct *task)
{
}
static inline int perf_cgroup_connect(pid_t pid, struct perf_event *event,
struct perf_event_attr *attr,
struct perf_event *group_leader)
{
return -EINVAL;
}
static inline void
perf_cgroup_set_timestamp(struct task_struct *task,
struct perf_event_context *ctx)
{
}
static inline void
perf_cgroup_switch(struct task_struct *task, struct task_struct *next)
{
}
static inline void
perf_cgroup_set_shadow_time(struct perf_event *event, u64 now)
{
}
static inline u64 perf_cgroup_event_time(struct perf_event *event)
{
return 0;
}
static inline void
list_update_cgroup_event(struct perf_event *event,
struct perf_event_context *ctx, bool add)
{
}
#endif
/*
* set default to be dependent on timer tick just
* like original code
*/
#define PERF_CPU_HRTIMER (1000 / HZ)
/*
* function must be called with interrupts disabled
*/
static enum hrtimer_restart perf_mux_hrtimer_handler(struct hrtimer *hr)
{
struct perf_cpu_context *cpuctx;
bool rotations;
lockdep_assert_irqs_disabled();
cpuctx = container_of(hr, struct perf_cpu_context, hrtimer);
rotations = perf_rotate_context(cpuctx);
raw_spin_lock(&cpuctx->hrtimer_lock);
if (rotations)
hrtimer_forward_now(hr, cpuctx->hrtimer_interval);
else
cpuctx->hrtimer_active = 0;
raw_spin_unlock(&cpuctx->hrtimer_lock);
return rotations ? HRTIMER_RESTART : HRTIMER_NORESTART;
}
static void __perf_mux_hrtimer_init(struct perf_cpu_context *cpuctx, int cpu)
{
struct hrtimer *timer = &cpuctx->hrtimer;
struct pmu *pmu = cpuctx->ctx.pmu;
u64 interval;
/* no multiplexing needed for SW PMU */
if (pmu->task_ctx_nr == perf_sw_context)
return;
/*
* check default is sane, if not set then force to
* default interval (1/tick)
*/
interval = pmu->hrtimer_interval_ms;
if (interval < 1)
interval = pmu->hrtimer_interval_ms = PERF_CPU_HRTIMER;
cpuctx->hrtimer_interval = ns_to_ktime(NSEC_PER_MSEC * interval);
raw_spin_lock_init(&cpuctx->hrtimer_lock);
hrtimer_init(timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS_PINNED_HARD);
timer->function = perf_mux_hrtimer_handler;
}
static int perf_mux_hrtimer_restart(struct perf_cpu_context *cpuctx)
{
struct hrtimer *timer = &cpuctx->hrtimer;
struct pmu *pmu = cpuctx->ctx.pmu;
unsigned long flags;
/* not for SW PMU */
if (pmu->task_ctx_nr == perf_sw_context)
return 0;
raw_spin_lock_irqsave(&cpuctx->hrtimer_lock, flags);
if (!cpuctx->hrtimer_active) {
cpuctx->hrtimer_active = 1;
hrtimer_forward_now(timer, cpuctx->hrtimer_interval);
hrtimer_start_expires(timer, HRTIMER_MODE_ABS_PINNED_HARD);
}
raw_spin_unlock_irqrestore(&cpuctx->hrtimer_lock, flags);
return 0;
}
static int perf_mux_hrtimer_restart_ipi(void *arg)
{
return perf_mux_hrtimer_restart(arg);
}
void perf_pmu_disable(struct pmu *pmu)
{
int *count = this_cpu_ptr(pmu->pmu_disable_count);
if (!(*count)++)
pmu->pmu_disable(pmu);
}
void perf_pmu_enable(struct pmu *pmu)
{
int *count = this_cpu_ptr(pmu->pmu_disable_count);
if (!--(*count))
pmu->pmu_enable(pmu);
}
static DEFINE_PER_CPU(struct list_head, active_ctx_list);
/*
* perf_event_ctx_activate(), perf_event_ctx_deactivate(), and
* perf_event_task_tick() are fully serialized because they're strictly cpu
* affine and perf_event_ctx{activate,deactivate} are called with IRQs
* disabled, while perf_event_task_tick is called from IRQ context.
*/
static void perf_event_ctx_activate(struct perf_event_context *ctx)
{
struct list_head *head = this_cpu_ptr(&active_ctx_list);
lockdep_assert_irqs_disabled();
WARN_ON(!list_empty(&ctx->active_ctx_list));
list_add(&ctx->active_ctx_list, head);
}
static void perf_event_ctx_deactivate(struct perf_event_context *ctx)
{
lockdep_assert_irqs_disabled();
WARN_ON(list_empty(&ctx->active_ctx_list));
list_del_init(&ctx->active_ctx_list);
}
static void get_ctx(struct perf_event_context *ctx)
{
refcount_inc(&ctx->refcount);
}
static void free_ctx(struct rcu_head *head)
{
struct perf_event_context *ctx;
ctx = container_of(head, struct perf_event_context, rcu_head);
kfree(ctx->task_ctx_data);
kfree(ctx);
}
static void put_ctx(struct perf_event_context *ctx)
{
if (refcount_dec_and_test(&ctx->refcount)) {
if (ctx->parent_ctx)
put_ctx(ctx->parent_ctx);
if (ctx->task && ctx->task != TASK_TOMBSTONE)
put_task_struct(ctx->task);
call_rcu(&ctx->rcu_head, free_ctx);
}
}
/*
* Because of perf_event::ctx migration in sys_perf_event_open::move_group and
* perf_pmu_migrate_context() we need some magic.
*
* Those places that change perf_event::ctx will hold both
* perf_event_ctx::mutex of the 'old' and 'new' ctx value.
*
* Lock ordering is by mutex address. There are two other sites where
* perf_event_context::mutex nests and those are:
*
* - perf_event_exit_task_context() [ child , 0 ]
* perf_event_exit_event()
* put_event() [ parent, 1 ]
*
* - perf_event_init_context() [ parent, 0 ]
* inherit_task_group()
* inherit_group()
* inherit_event()
* perf_event_alloc()
* perf_init_event()
* perf_try_init_event() [ child , 1 ]
*
* While it appears there is an obvious deadlock here -- the parent and child
* nesting levels are inverted between the two. This is in fact safe because
* life-time rules separate them. That is an exiting task cannot fork, and a
* spawning task cannot (yet) exit.
*
* But remember that that these are parent<->child context relations, and
* migration does not affect children, therefore these two orderings should not
* interact.
*
* The change in perf_event::ctx does not affect children (as claimed above)
* because the sys_perf_event_open() case will install a new event and break
* the ctx parent<->child relation, and perf_pmu_migrate_context() is only
* concerned with cpuctx and that doesn't have children.
*
* The places that change perf_event::ctx will issue:
*
* perf_remove_from_context();
* synchronize_rcu();
* perf_install_in_context();
*
* to affect the change. The remove_from_context() + synchronize_rcu() should
* quiesce the event, after which we can install it in the new location. This
* means that only external vectors (perf_fops, prctl) can perturb the event
* while in transit. Therefore all such accessors should also acquire
* perf_event_context::mutex to serialize against this.
*
* However; because event->ctx can change while we're waiting to acquire
* ctx->mutex we must be careful and use the below perf_event_ctx_lock()
* function.
*
* Lock order:
* exec_update_lock
* task_struct::perf_event_mutex
* perf_event_context::mutex
* perf_event::child_mutex;
* perf_event_context::lock
* perf_event::mmap_mutex
* mmap_sem
* perf_addr_filters_head::lock
*
* cpu_hotplug_lock
* pmus_lock
* cpuctx->mutex / perf_event_context::mutex
*/
static struct perf_event_context *
perf_event_ctx_lock_nested(struct perf_event *event, int nesting)
{
struct perf_event_context *ctx;
again:
rcu_read_lock();
ctx = READ_ONCE(event->ctx);
if (!refcount_inc_not_zero(&ctx->refcount)) {
rcu_read_unlock();
goto again;
}
rcu_read_unlock();
mutex_lock_nested(&ctx->mutex, nesting);
if (event->ctx != ctx) {
mutex_unlock(&ctx->mutex);
put_ctx(ctx);
goto again;
}
return ctx;
}
static inline struct perf_event_context *
perf_event_ctx_lock(struct perf_event *event)
{
return perf_event_ctx_lock_nested(event, 0);
}
static void perf_event_ctx_unlock(struct perf_event *event,
struct perf_event_context *ctx)
{
mutex_unlock(&ctx->mutex);
put_ctx(ctx);
}
/*
* This must be done under the ctx->lock, such as to serialize against
* context_equiv(), therefore we cannot call put_ctx() since that might end up
* calling scheduler related locks and ctx->lock nests inside those.
*/
static __must_check struct perf_event_context *
unclone_ctx(struct perf_event_context *ctx)
{
struct perf_event_context *parent_ctx = ctx->parent_ctx;
lockdep_assert_held(&ctx->lock);
if (parent_ctx)
ctx->parent_ctx = NULL;
ctx->generation++;
return parent_ctx;
}
static u32 perf_event_pid_type(struct perf_event *event, struct task_struct *p,
enum pid_type type)
{
u32 nr;
/*
* only top level events have the pid namespace they were created in
*/
if (event->parent)
event = event->parent;
nr = __task_pid_nr_ns(p, type, event->ns);
/* avoid -1 if it is idle thread or runs in another ns */
if (!nr && !pid_alive(p))
nr = -1;
return nr;
}
static u32 perf_event_pid(struct perf_event *event, struct task_struct *p)
{
return perf_event_pid_type(event, p, PIDTYPE_TGID);
}
static u32 perf_event_tid(struct perf_event *event, struct task_struct *p)
{
return perf_event_pid_type(event, p, PIDTYPE_PID);
}
/*
* If we inherit events we want to return the parent event id
* to userspace.
*/
static u64 primary_event_id(struct perf_event *event)
{
u64 id = event->id;
if (event->parent)
id = event->parent->id;
return id;
}
/*
* Get the perf_event_context for a task and lock it.
*
* This has to cope with with the fact that until it is locked,
* the context could get moved to another task.
*/
static struct perf_event_context *
perf_lock_task_context(struct task_struct *task, int ctxn, unsigned long *flags)
{
struct perf_event_context *ctx;
retry:
/*
* One of the few rules of preemptible RCU is that one cannot do
* rcu_read_unlock() while holding a scheduler (or nested) lock when
* part of the read side critical section was irqs-enabled -- see
* rcu_read_unlock_special().
*
* Since ctx->lock nests under rq->lock we must ensure the entire read
* side critical section has interrupts disabled.
*/
local_irq_save(*flags);
rcu_read_lock();
ctx = rcu_dereference(task->perf_event_ctxp[ctxn]);
if (ctx) {
/*
* If this context is a clone of another, it might
* get swapped for another underneath us by
* perf_event_task_sched_out, though the
* rcu_read_lock() protects us from any context
* getting freed. Lock the context and check if it
* got swapped before we could get the lock, and retry
* if so. If we locked the right context, then it
* can't get swapped on us any more.
*/
raw_spin_lock(&ctx->lock);
if (ctx != rcu_dereference(task->perf_event_ctxp[ctxn])) {
raw_spin_unlock(&ctx->lock);
rcu_read_unlock();
local_irq_restore(*flags);
goto retry;
}
if (ctx->task == TASK_TOMBSTONE ||
!refcount_inc_not_zero(&ctx->refcount)) {
raw_spin_unlock(&ctx->lock);
ctx = NULL;
} else {
WARN_ON_ONCE(ctx->task != task);
}
}
rcu_read_unlock();
if (!ctx)
local_irq_restore(*flags);
return ctx;
}
/*
* Get the context for a task and increment its pin_count so it
* can't get swapped to another task. This also increments its
* reference count so that the context can't get freed.
*/
static struct perf_event_context *
perf_pin_task_context(struct task_struct *task, int ctxn)
{
struct perf_event_context *ctx;
unsigned long flags;
ctx = perf_lock_task_context(task, ctxn, &flags);
if (ctx) {
++ctx->pin_count;
raw_spin_unlock_irqrestore(&ctx->lock, flags);
}
return ctx;
}
static void perf_unpin_context(struct perf_event_context *ctx)
{
unsigned long flags;
raw_spin_lock_irqsave(&ctx->lock, flags);
--ctx->pin_count;
raw_spin_unlock_irqrestore(&ctx->lock, flags);
}
/*
* Update the record of the current time in a context.
*/
static void update_context_time(struct perf_event_context *ctx)
{
u64 now = perf_clock();
ctx->time += now - ctx->timestamp;
ctx->timestamp = now;
}
static u64 perf_event_time(struct perf_event *event)
{
struct perf_event_context *ctx = event->ctx;
if (is_cgroup_event(event))
return perf_cgroup_event_time(event);
return ctx ? ctx->time : 0;
}
static enum event_type_t get_event_type(struct perf_event *event)
{
struct perf_event_context *ctx = event->ctx;
enum event_type_t event_type;
lockdep_assert_held(&ctx->lock);
/*
* It's 'group type', really, because if our group leader is
* pinned, so are we.
*/
if (event->group_leader != event)
event = event->group_leader;
event_type = event->attr.pinned ? EVENT_PINNED : EVENT_FLEXIBLE;
if (!ctx->task)
event_type |= EVENT_CPU;
return event_type;
}
/*
* Helper function to initialize event group nodes.
*/
static void init_event_group(struct perf_event *event)
{
RB_CLEAR_NODE(&event->group_node);
event->group_index = 0;
}
/*
* Extract pinned or flexible groups from the context
* based on event attrs bits.
*/
static struct perf_event_groups *
get_event_groups(struct perf_event *event, struct perf_event_context *ctx)
{
if (event->attr.pinned)
return &ctx->pinned_groups;
else
return &ctx->flexible_groups;
}
/*
* Helper function to initializes perf_event_group trees.
*/
static void perf_event_groups_init(struct perf_event_groups *groups)
{
groups->tree = RB_ROOT;
groups->index = 0;
}
/*
* Compare function for event groups;
*
* Implements complex key that first sorts by CPU and then by virtual index
* which provides ordering when rotating groups for the same CPU.
*/
static bool
perf_event_groups_less(struct perf_event *left, struct perf_event *right)
{
if (left->cpu < right->cpu)
return true;
if (left->cpu > right->cpu)
return false;
if (left->group_index < right->group_index)
return true;
if (left->group_index > right->group_index)
return false;
return false;
}
/*
* Insert @event into @groups' tree; using {@event->cpu, ++@groups->index} for
* key (see perf_event_groups_less). This places it last inside the CPU
* subtree.
*/
static void
perf_event_groups_insert(struct perf_event_groups *groups,
struct perf_event *event)
{
struct perf_event *node_event;
struct rb_node *parent;
struct rb_node **node;
event->group_index = ++groups->index;
node = &groups->tree.rb_node;
parent = *node;
while (*node) {
parent = *node;
node_event = container_of(*node, struct perf_event, group_node);
if (perf_event_groups_less(event, node_event))
node = &parent->rb_left;
else
node = &parent->rb_right;
}
rb_link_node(&event->group_node, parent, node);
rb_insert_color(&event->group_node, &groups->tree);
}
/*
* Helper function to insert event into the pinned or flexible groups.
*/
static void
add_event_to_groups(struct perf_event *event, struct perf_event_context *ctx)
{
struct perf_event_groups *groups;
groups = get_event_groups(event, ctx);
perf_event_groups_insert(groups, event);
}
/*
* Delete a group from a tree.
*/
static void
perf_event_groups_delete(struct perf_event_groups *groups,
struct perf_event *event)
{
WARN_ON_ONCE(RB_EMPTY_NODE(&event->group_node) ||
RB_EMPTY_ROOT(&groups->tree));
rb_erase(&event->group_node, &groups->tree);
init_event_group(event);
}
/*
* Helper function to delete event from its groups.
*/
static void
del_event_from_groups(struct perf_event *event, struct perf_event_context *ctx)
{
struct perf_event_groups *groups;
groups = get_event_groups(event, ctx);
perf_event_groups_delete(groups, event);
}
/*
* Get the leftmost event in the @cpu subtree.
*/
static struct perf_event *
perf_event_groups_first(struct perf_event_groups *groups, int cpu)
{
struct perf_event *node_event = NULL, *match = NULL;
struct rb_node *node = groups->tree.rb_node;
while (node) {
node_event = container_of(node, struct perf_event, group_node);
if (cpu < node_event->cpu) {
node = node->rb_left;
} else if (cpu > node_event->cpu) {
node = node->rb_right;
} else {
match = node_event;
node = node->rb_left;
}
}
return match;
}
/*
* Like rb_entry_next_safe() for the @cpu subtree.
*/
static struct perf_event *
perf_event_groups_next(struct perf_event *event)
{
struct perf_event *next;
next = rb_entry_safe(rb_next(&event->group_node), typeof(*event), group_node);
if (next && next->cpu == event->cpu)
return next;
return NULL;
}
/*
* Iterate through the whole groups tree.
*/
#define perf_event_groups_for_each(event, groups) \
for (event = rb_entry_safe(rb_first(&((groups)->tree)), \
typeof(*event), group_node); event; \
event = rb_entry_safe(rb_next(&event->group_node), \
typeof(*event), group_node))
/*
* Add an event from the lists for its context.
* Must be called with ctx->mutex and ctx->lock held.
*/
static void
list_add_event(struct perf_event *event, struct perf_event_context *ctx)
{
lockdep_assert_held(&ctx->lock);
WARN_ON_ONCE(event->attach_state & PERF_ATTACH_CONTEXT);
event->attach_state |= PERF_ATTACH_CONTEXT;
event->tstamp = perf_event_time(event);
/*
* If we're a stand alone event or group leader, we go to the context
* list, group events are kept attached to the group so that
* perf_group_detach can, at all times, locate all siblings.
*/
if (event->group_leader == event) {
event->group_caps = event->event_caps;
add_event_to_groups(event, ctx);
}
list_update_cgroup_event(event, ctx, true);
list_add_rcu(&event->event_entry, &ctx->event_list);
ctx->nr_events++;
if (event->attr.inherit_stat)
ctx->nr_stat++;
ctx->generation++;
}
/*
* Initialize event state based on the perf_event_attr::disabled.
*/
static inline void perf_event__state_init(struct perf_event *event)
{
event->state = event->attr.disabled ? PERF_EVENT_STATE_OFF :
PERF_EVENT_STATE_INACTIVE;
}
static int __perf_event_read_size(u64 read_format, int nr_siblings)
{
int entry = sizeof(u64); /* value */
int size = 0;
int nr = 1;
if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
size += sizeof(u64);
if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
size += sizeof(u64);
if (read_format & PERF_FORMAT_ID)
entry += sizeof(u64);
if (read_format & PERF_FORMAT_GROUP) {
nr += nr_siblings;
size += sizeof(u64);
}
/*
* Since perf_event_validate_size() limits this to 16k and inhibits
* adding more siblings, this will never overflow.
*/
return size + nr * entry;
}
static void __perf_event_header_size(struct perf_event *event, u64 sample_type)
{
struct perf_sample_data *data;
u16 size = 0;
if (sample_type & PERF_SAMPLE_IP)
size += sizeof(data->ip);
if (sample_type & PERF_SAMPLE_ADDR)
size += sizeof(data->addr);
if (sample_type & PERF_SAMPLE_PERIOD)
size += sizeof(data->period);
if (sample_type & PERF_SAMPLE_WEIGHT)
size += sizeof(data->weight);
if (sample_type & PERF_SAMPLE_READ)
size += event->read_size;
if (sample_type & PERF_SAMPLE_DATA_SRC)
size += sizeof(data->data_src.val);
if (sample_type & PERF_SAMPLE_TRANSACTION)
size += sizeof(data->txn);
if (sample_type & PERF_SAMPLE_PHYS_ADDR)
size += sizeof(data->phys_addr);
event->header_size = size;
}
/*
* Called at perf_event creation and when events are attached/detached from a
* group.
*/
static void perf_event__header_size(struct perf_event *event)
{
event->read_size =
__perf_event_read_size(event->attr.read_format,
event->group_leader->nr_siblings);
__perf_event_header_size(event, event->attr.sample_type);
}
static void perf_event__id_header_size(struct perf_event *event)
{
struct perf_sample_data *data;
u64 sample_type = event->attr.sample_type;
u16 size = 0;
if (sample_type & PERF_SAMPLE_TID)
size += sizeof(data->tid_entry);
if (sample_type & PERF_SAMPLE_TIME)
size += sizeof(data->time);
if (sample_type & PERF_SAMPLE_IDENTIFIER)
size += sizeof(data->id);
if (sample_type & PERF_SAMPLE_ID)
size += sizeof(data->id);
if (sample_type & PERF_SAMPLE_STREAM_ID)
size += sizeof(data->stream_id);
if (sample_type & PERF_SAMPLE_CPU)
size += sizeof(data->cpu_entry);
event->id_header_size = size;
}
/*
* Check that adding an event to the group does not result in anybody
* overflowing the 64k event limit imposed by the output buffer.
*
* Specifically, check that the read_size for the event does not exceed 16k,
* read_size being the one term that grows with groups size. Since read_size
* depends on per-event read_format, also (re)check the existing events.
*
* This leaves 48k for the constant size fields and things like callchains,
* branch stacks and register sets.
*/
static bool perf_event_validate_size(struct perf_event *event)
{
struct perf_event *sibling, *group_leader = event->group_leader;
if (__perf_event_read_size(event->attr.read_format,
group_leader->nr_siblings + 1) > 16*1024)
return false;
if (__perf_event_read_size(group_leader->attr.read_format,
group_leader->nr_siblings + 1) > 16*1024)
return false;
for_each_sibling_event(sibling, group_leader) {
if (__perf_event_read_size(sibling->attr.read_format,
group_leader->nr_siblings + 1) > 16*1024)
return false;
}
return true;
}
static void perf_group_attach(struct perf_event *event)
{
struct perf_event *group_leader = event->group_leader, *pos;
lockdep_assert_held(&event->ctx->lock);
/*
* We can have double attach due to group movement in perf_event_open.
*/
if (event->attach_state & PERF_ATTACH_GROUP)
return;
event->attach_state |= PERF_ATTACH_GROUP;
if (group_leader == event)
return;
WARN_ON_ONCE(group_leader->ctx != event->ctx);
group_leader->group_caps &= event->event_caps;
list_add_tail(&event->sibling_list, &group_leader->sibling_list);
group_leader->nr_siblings++;
group_leader->group_generation++;
perf_event__header_size(group_leader);
for_each_sibling_event(pos, group_leader)
perf_event__header_size(pos);
}
/*
* Remove an event from the lists for its context.
* Must be called with ctx->mutex and ctx->lock held.
*/
static void
list_del_event(struct perf_event *event, struct perf_event_context *ctx)
{
WARN_ON_ONCE(event->ctx != ctx);
lockdep_assert_held(&ctx->lock);
/*
* We can have double detach due to exit/hot-unplug + close.
*/
if (!(event->attach_state & PERF_ATTACH_CONTEXT))
return;
event->attach_state &= ~PERF_ATTACH_CONTEXT;
list_update_cgroup_event(event, ctx, false);
ctx->nr_events--;
if (event->attr.inherit_stat)
ctx->nr_stat--;
list_del_rcu(&event->event_entry);
if (event->group_leader == event)
del_event_from_groups(event, ctx);
/*
* If event was in error state, then keep it
* that way, otherwise bogus counts will be
* returned on read(). The only way to get out
* of error state is by explicit re-enabling
* of the event
*/
if (event->state > PERF_EVENT_STATE_OFF)
perf_event_set_state(event, PERF_EVENT_STATE_OFF);
ctx->generation++;
}
static int
perf_aux_output_match(struct perf_event *event, struct perf_event *aux_event)
{
if (!has_aux(aux_event))
return 0;
if (!event->pmu->aux_output_match)
return 0;
return event->pmu->aux_output_match(aux_event);
}
static void put_event(struct perf_event *event);
static void event_sched_out(struct perf_event *event,
struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx);
static void perf_put_aux_event(struct perf_event *event)
{
struct perf_event_context *ctx = event->ctx;
struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
struct perf_event *iter;
/*
* If event uses aux_event tear down the link
*/
if (event->aux_event) {
iter = event->aux_event;
event->aux_event = NULL;
put_event(iter);
return;
}
/*
* If the event is an aux_event, tear down all links to
* it from other events.
*/
for_each_sibling_event(iter, event->group_leader) {
if (iter->aux_event != event)
continue;
iter->aux_event = NULL;
put_event(event);
/*
* If it's ACTIVE, schedule it out and put it into ERROR
* state so that we don't try to schedule it again. Note
* that perf_event_enable() will clear the ERROR status.
*/
event_sched_out(iter, cpuctx, ctx);
perf_event_set_state(event, PERF_EVENT_STATE_ERROR);
}
}
static int perf_get_aux_event(struct perf_event *event,
struct perf_event *group_leader)
{
/*
* Our group leader must be an aux event if we want to be
* an aux_output. This way, the aux event will precede its
* aux_output events in the group, and therefore will always
* schedule first.
*/
if (!group_leader)
return 0;
if (!perf_aux_output_match(event, group_leader))
return 0;
if (!atomic_long_inc_not_zero(&group_leader->refcount))
return 0;
/*
* Link aux_outputs to their aux event; this is undone in
* perf_group_detach() by perf_put_aux_event(). When the
* group in torn down, the aux_output events loose their
* link to the aux_event and can't schedule any more.
*/
event->aux_event = group_leader;
return 1;
}
static void perf_group_detach(struct perf_event *event)
{
struct perf_event *sibling, *tmp;
struct perf_event_context *ctx = event->ctx;
lockdep_assert_held(&ctx->lock);
/*
* We can have double detach due to exit/hot-unplug + close.
*/
if (!(event->attach_state & PERF_ATTACH_GROUP))
return;
event->attach_state &= ~PERF_ATTACH_GROUP;
perf_put_aux_event(event);
/*
* If this is a sibling, remove it from its group.
*/
if (event->group_leader != event) {
list_del_init(&event->sibling_list);
event->group_leader->nr_siblings--;
event->group_leader->group_generation++;
goto out;
}
/*
* If this was a group event with sibling events then
* upgrade the siblings to singleton events by adding them
* to whatever list we are on.
*/
list_for_each_entry_safe(sibling, tmp, &event->sibling_list, sibling_list) {
sibling->group_leader = sibling;
list_del_init(&sibling->sibling_list);
/* Inherit group flags from the previous leader */
sibling->group_caps = event->group_caps;
if (!RB_EMPTY_NODE(&event->group_node)) {
add_event_to_groups(sibling, event->ctx);
if (sibling->state == PERF_EVENT_STATE_ACTIVE) {
struct list_head *list = sibling->attr.pinned ?
&ctx->pinned_active : &ctx->flexible_active;
list_add_tail(&sibling->active_list, list);
}
}
WARN_ON_ONCE(sibling->ctx != event->ctx);
}
out:
perf_event__header_size(event->group_leader);
for_each_sibling_event(tmp, event->group_leader)
perf_event__header_size(tmp);
}
static bool is_orphaned_event(struct perf_event *event)
{
return event->state == PERF_EVENT_STATE_DEAD;
}
static inline int __pmu_filter_match(struct perf_event *event)
{
struct pmu *pmu = event->pmu;
return pmu->filter_match ? pmu->filter_match(event) : 1;
}
/*
* Check whether we should attempt to schedule an event group based on
* PMU-specific filtering. An event group can consist of HW and SW events,
* potentially with a SW leader, so we must check all the filters, to
* determine whether a group is schedulable:
*/
static inline int pmu_filter_match(struct perf_event *event)
{
struct perf_event *sibling;
if (!__pmu_filter_match(event))
return 0;
for_each_sibling_event(sibling, event) {
if (!__pmu_filter_match(sibling))
return 0;
}
return 1;
}
static inline int
event_filter_match(struct perf_event *event)
{
return (event->cpu == -1 || event->cpu == smp_processor_id()) &&
perf_cgroup_match(event) && pmu_filter_match(event);
}
static void
event_sched_out(struct perf_event *event,
struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx)
{
enum perf_event_state state = PERF_EVENT_STATE_INACTIVE;
WARN_ON_ONCE(event->ctx != ctx);
lockdep_assert_held(&ctx->lock);
if (event->state != PERF_EVENT_STATE_ACTIVE)
return;
/*
* Asymmetry; we only schedule events _IN_ through ctx_sched_in(), but
* we can schedule events _OUT_ individually through things like
* __perf_remove_from_context().
*/
list_del_init(&event->active_list);
perf_pmu_disable(event->pmu);
event->pmu->del(event, 0);
event->oncpu = -1;
if (READ_ONCE(event->pending_disable) >= 0) {
WRITE_ONCE(event->pending_disable, -1);
state = PERF_EVENT_STATE_OFF;
}
perf_event_set_state(event, state);
if (!is_software_event(event))
cpuctx->active_oncpu--;
if (!--ctx->nr_active)
perf_event_ctx_deactivate(ctx);
if (event->attr.freq && event->attr.sample_freq)
ctx->nr_freq--;
if (event->attr.exclusive || !cpuctx->active_oncpu)
cpuctx->exclusive = 0;
perf_pmu_enable(event->pmu);
}
static void
group_sched_out(struct perf_event *group_event,
struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx)
{
struct perf_event *event;
if (group_event->state != PERF_EVENT_STATE_ACTIVE)
return;
perf_pmu_disable(ctx->pmu);
event_sched_out(group_event, cpuctx, ctx);
/*
* Schedule out siblings (if any):
*/
for_each_sibling_event(event, group_event)
event_sched_out(event, cpuctx, ctx);
perf_pmu_enable(ctx->pmu);
if (group_event->attr.exclusive)
cpuctx->exclusive = 0;
}
#define DETACH_GROUP 0x01UL
/*
* Cross CPU call to remove a performance event
*
* We disable the event on the hardware level first. After that we
* remove it from the context list.
*/
static void
__perf_remove_from_context(struct perf_event *event,
struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx,
void *info)
{
unsigned long flags = (unsigned long)info;
if (ctx->is_active & EVENT_TIME) {
update_context_time(ctx);
update_cgrp_time_from_cpuctx(cpuctx);
}
event_sched_out(event, cpuctx, ctx);
if (flags & DETACH_GROUP)
perf_group_detach(event);
list_del_event(event, ctx);
if (!ctx->nr_events && ctx->is_active) {
ctx->is_active = 0;
ctx->rotate_necessary = 0;
if (ctx->task) {
WARN_ON_ONCE(cpuctx->task_ctx != ctx);
cpuctx->task_ctx = NULL;
}
}
}
/*
* Remove the event from a task's (or a CPU's) list of events.
*
* If event->ctx is a cloned context, callers must make sure that
* every task struct that event->ctx->task could possibly point to
* remains valid. This is OK when called from perf_release since
* that only calls us on the top-level context, which can't be a clone.
* When called from perf_event_exit_task, it's OK because the
* context has been detached from its task.
*/
static void perf_remove_from_context(struct perf_event *event, unsigned long flags)
{
struct perf_event_context *ctx = event->ctx;
lockdep_assert_held(&ctx->mutex);
event_function_call(event, __perf_remove_from_context, (void *)flags);
/*
* The above event_function_call() can NO-OP when it hits
* TASK_TOMBSTONE. In that case we must already have been detached
* from the context (by perf_event_exit_event()) but the grouping
* might still be in-tact.
*/
WARN_ON_ONCE(event->attach_state & PERF_ATTACH_CONTEXT);
if ((flags & DETACH_GROUP) &&
(event->attach_state & PERF_ATTACH_GROUP)) {
/*
* Since in that case we cannot possibly be scheduled, simply
* detach now.
*/
raw_spin_lock_irq(&ctx->lock);
perf_group_detach(event);
raw_spin_unlock_irq(&ctx->lock);
}
}
/*
* Cross CPU call to disable a performance event
*/
static void __perf_event_disable(struct perf_event *event,
struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx,
void *info)
{
if (event->state < PERF_EVENT_STATE_INACTIVE)
return;
if (ctx->is_active & EVENT_TIME) {
update_context_time(ctx);
update_cgrp_time_from_event(event);
}
if (event == event->group_leader)
group_sched_out(event, cpuctx, ctx);
else
event_sched_out(event, cpuctx, ctx);
perf_event_set_state(event, PERF_EVENT_STATE_OFF);
}
/*
* Disable an event.
*
* If event->ctx is a cloned context, callers must make sure that
* every task struct that event->ctx->task could possibly point to
* remains valid. This condition is satisfied when called through
* perf_event_for_each_child or perf_event_for_each because they
* hold the top-level event's child_mutex, so any descendant that
* goes to exit will block in perf_event_exit_event().
*
* When called from perf_pending_event it's OK because event->ctx
* is the current context on this CPU and preemption is disabled,
* hence we can't get into perf_event_task_sched_out for this context.
*/
static void _perf_event_disable(struct perf_event *event)
{
struct perf_event_context *ctx = event->ctx;
raw_spin_lock_irq(&ctx->lock);
if (event->state <= PERF_EVENT_STATE_OFF) {
raw_spin_unlock_irq(&ctx->lock);
return;
}
raw_spin_unlock_irq(&ctx->lock);
event_function_call(event, __perf_event_disable, NULL);
}
void perf_event_disable_local(struct perf_event *event)
{
event_function_local(event, __perf_event_disable, NULL);
}
/*
* Strictly speaking kernel users cannot create groups and therefore this
* interface does not need the perf_event_ctx_lock() magic.
*/
void perf_event_disable(struct perf_event *event)
{
struct perf_event_context *ctx;
ctx = perf_event_ctx_lock(event);
_perf_event_disable(event);
perf_event_ctx_unlock(event, ctx);
}
EXPORT_SYMBOL_GPL(perf_event_disable);
void perf_event_disable_inatomic(struct perf_event *event)
{
WRITE_ONCE(event->pending_disable, smp_processor_id());
/* can fail, see perf_pending_event_disable() */
irq_work_queue(&event->pending);
}
static void perf_set_shadow_time(struct perf_event *event,
struct perf_event_context *ctx)
{
/*
* use the correct time source for the time snapshot
*
* We could get by without this by leveraging the
* fact that to get to this function, the caller
* has most likely already called update_context_time()
* and update_cgrp_time_xx() and thus both timestamp
* are identical (or very close). Given that tstamp is,
* already adjusted for cgroup, we could say that:
* tstamp - ctx->timestamp
* is equivalent to
* tstamp - cgrp->timestamp.
*
* Then, in perf_output_read(), the calculation would
* work with no changes because:
* - event is guaranteed scheduled in
* - no scheduled out in between
* - thus the timestamp would be the same
*
* But this is a bit hairy.
*
* So instead, we have an explicit cgroup call to remain
* within the time time source all along. We believe it
* is cleaner and simpler to understand.
*/
if (is_cgroup_event(event))
perf_cgroup_set_shadow_time(event, event->tstamp);
else
event->shadow_ctx_time = event->tstamp - ctx->timestamp;
}
#define MAX_INTERRUPTS (~0ULL)
static void perf_log_throttle(struct perf_event *event, int enable);
static void perf_log_itrace_start(struct perf_event *event);
static int
event_sched_in(struct perf_event *event,
struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx)
{
int ret = 0;
lockdep_assert_held(&ctx->lock);
if (event->state <= PERF_EVENT_STATE_OFF)
return 0;
WRITE_ONCE(event->oncpu, smp_processor_id());
/*
* Order event::oncpu write to happen before the ACTIVE state is
* visible. This allows perf_event_{stop,read}() to observe the correct
* ->oncpu if it sees ACTIVE.
*/
smp_wmb();
perf_event_set_state(event, PERF_EVENT_STATE_ACTIVE);
/*
* Unthrottle events, since we scheduled we might have missed several
* ticks already, also for a heavily scheduling task there is little
* guarantee it'll get a tick in a timely manner.
*/
if (unlikely(event->hw.interrupts == MAX_INTERRUPTS)) {
perf_log_throttle(event, 1);
event->hw.interrupts = 0;
}
perf_pmu_disable(event->pmu);
perf_set_shadow_time(event, ctx);
perf_log_itrace_start(event);
if (event->pmu->add(event, PERF_EF_START)) {
perf_event_set_state(event, PERF_EVENT_STATE_INACTIVE);
event->oncpu = -1;
ret = -EAGAIN;
goto out;
}
if (!is_software_event(event))
cpuctx->active_oncpu++;
if (!ctx->nr_active++)
perf_event_ctx_activate(ctx);
if (event->attr.freq && event->attr.sample_freq)
ctx->nr_freq++;
if (event->attr.exclusive)
cpuctx->exclusive = 1;
out:
perf_pmu_enable(event->pmu);
return ret;
}
static int
group_sched_in(struct perf_event *group_event,
struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx)
{
struct perf_event *event, *partial_group = NULL;
struct pmu *pmu = ctx->pmu;
if (group_event->state == PERF_EVENT_STATE_OFF)
return 0;
pmu->start_txn(pmu, PERF_PMU_TXN_ADD);
if (event_sched_in(group_event, cpuctx, ctx)) {
pmu->cancel_txn(pmu);
perf_mux_hrtimer_restart(cpuctx);
return -EAGAIN;
}
/*
* Schedule in siblings as one group (if any):
*/
for_each_sibling_event(event, group_event) {
if (event_sched_in(event, cpuctx, ctx)) {
partial_group = event;
goto group_error;
}
}
if (!pmu->commit_txn(pmu))
return 0;
group_error:
/*
* Groups can be scheduled in as one unit only, so undo any
* partial group before returning:
* The events up to the failed event are scheduled out normally.
*/
for_each_sibling_event(event, group_event) {
if (event == partial_group)
break;
event_sched_out(event, cpuctx, ctx);
}
event_sched_out(group_event, cpuctx, ctx);
pmu->cancel_txn(pmu);
perf_mux_hrtimer_restart(cpuctx);
return -EAGAIN;
}
/*
* Work out whether we can put this event group on the CPU now.
*/
static int group_can_go_on(struct perf_event *event,
struct perf_cpu_context *cpuctx,
int can_add_hw)
{
/*
* Groups consisting entirely of software events can always go on.
*/
if (event->group_caps & PERF_EV_CAP_SOFTWARE)
return 1;
/*
* If an exclusive group is already on, no other hardware
* events can go on.
*/
if (cpuctx->exclusive)
return 0;
/*
* If this group is exclusive and there are already
* events on the CPU, it can't go on.
*/
if (event->attr.exclusive && cpuctx->active_oncpu)
return 0;
/*
* Otherwise, try to add it if all previous groups were able
* to go on.
*/
return can_add_hw;
}
static void add_event_to_ctx(struct perf_event *event,
struct perf_event_context *ctx)
{
list_add_event(event, ctx);
perf_group_attach(event);
}
static void ctx_sched_out(struct perf_event_context *ctx,
struct perf_cpu_context *cpuctx,
enum event_type_t event_type);
static void
ctx_sched_in(struct perf_event_context *ctx,
struct perf_cpu_context *cpuctx,
enum event_type_t event_type,
struct task_struct *task);
static void task_ctx_sched_out(struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx,
enum event_type_t event_type)
{
if (!cpuctx->task_ctx)
return;
if (WARN_ON_ONCE(ctx != cpuctx->task_ctx))
return;
ctx_sched_out(ctx, cpuctx, event_type);
}
static void perf_event_sched_in(struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx,
struct task_struct *task)
{
cpu_ctx_sched_in(cpuctx, EVENT_PINNED, task);
if (ctx)
ctx_sched_in(ctx, cpuctx, EVENT_PINNED, task);
cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE, task);
if (ctx)
ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE, task);
}
/*
* We want to maintain the following priority of scheduling:
* - CPU pinned (EVENT_CPU | EVENT_PINNED)
* - task pinned (EVENT_PINNED)
* - CPU flexible (EVENT_CPU | EVENT_FLEXIBLE)
* - task flexible (EVENT_FLEXIBLE).
*
* In order to avoid unscheduling and scheduling back in everything every
* time an event is added, only do it for the groups of equal priority and
* below.
*
* This can be called after a batch operation on task events, in which case
* event_type is a bit mask of the types of events involved. For CPU events,
* event_type is only either EVENT_PINNED or EVENT_FLEXIBLE.
*/
static void ctx_resched(struct perf_cpu_context *cpuctx,
struct perf_event_context *task_ctx,
enum event_type_t event_type)
{
enum event_type_t ctx_event_type;
bool cpu_event = !!(event_type & EVENT_CPU);
/*
* If pinned groups are involved, flexible groups also need to be
* scheduled out.
*/
if (event_type & EVENT_PINNED)
event_type |= EVENT_FLEXIBLE;
ctx_event_type = event_type & EVENT_ALL;
perf_pmu_disable(cpuctx->ctx.pmu);
if (task_ctx)
task_ctx_sched_out(cpuctx, task_ctx, event_type);
/*
* Decide which cpu ctx groups to schedule out based on the types
* of events that caused rescheduling:
* - EVENT_CPU: schedule out corresponding groups;
* - EVENT_PINNED task events: schedule out EVENT_FLEXIBLE groups;
* - otherwise, do nothing more.
*/
if (cpu_event)
cpu_ctx_sched_out(cpuctx, ctx_event_type);
else if (ctx_event_type & EVENT_PINNED)
cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE);
perf_event_sched_in(cpuctx, task_ctx, current);
perf_pmu_enable(cpuctx->ctx.pmu);
}
void perf_pmu_resched(struct pmu *pmu)
{
struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
struct perf_event_context *task_ctx = cpuctx->task_ctx;
perf_ctx_lock(cpuctx, task_ctx);
ctx_resched(cpuctx, task_ctx, EVENT_ALL|EVENT_CPU);
perf_ctx_unlock(cpuctx, task_ctx);
}
/*
* Cross CPU call to install and enable a performance event
*
* Very similar to remote_function() + event_function() but cannot assume that
* things like ctx->is_active and cpuctx->task_ctx are set.
*/
static int __perf_install_in_context(void *info)
{
struct perf_event *event = info;
struct perf_event_context *ctx = event->ctx;
struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
struct perf_event_context *task_ctx = cpuctx->task_ctx;
bool reprogram = true;
int ret = 0;
raw_spin_lock(&cpuctx->ctx.lock);
if (ctx->task) {
raw_spin_lock(&ctx->lock);
task_ctx = ctx;
reprogram = (ctx->task == current);
/*
* If the task is running, it must be running on this CPU,
* otherwise we cannot reprogram things.
*
* If its not running, we don't care, ctx->lock will
* serialize against it becoming runnable.
*/
if (task_curr(ctx->task) && !reprogram) {
ret = -ESRCH;
goto unlock;
}
WARN_ON_ONCE(reprogram && cpuctx->task_ctx && cpuctx->task_ctx != ctx);
} else if (task_ctx) {
raw_spin_lock(&task_ctx->lock);
}
#ifdef CONFIG_CGROUP_PERF
if (is_cgroup_event(event)) {
/*
* If the current cgroup doesn't match the event's
* cgroup, we should not try to schedule it.
*/
struct perf_cgroup *cgrp = perf_cgroup_from_task(current, ctx);
reprogram = cgroup_is_descendant(cgrp->css.cgroup,
event->cgrp->css.cgroup);
}
#endif
if (reprogram) {
ctx_sched_out(ctx, cpuctx, EVENT_TIME);
add_event_to_ctx(event, ctx);
ctx_resched(cpuctx, task_ctx, get_event_type(event));
} else {
add_event_to_ctx(event, ctx);
}
unlock:
perf_ctx_unlock(cpuctx, task_ctx);
return ret;
}
static bool exclusive_event_installable(struct perf_event *event,
struct perf_event_context *ctx);
/*
* Attach a performance event to a context.
*
* Very similar to event_function_call, see comment there.
*/
static void
perf_install_in_context(struct perf_event_context *ctx,
struct perf_event *event,
int cpu)
{
struct task_struct *task = READ_ONCE(ctx->task);
lockdep_assert_held(&ctx->mutex);
WARN_ON_ONCE(!exclusive_event_installable(event, ctx));
if (event->cpu != -1)
event->cpu = cpu;
/*
* Ensures that if we can observe event->ctx, both the event and ctx
* will be 'complete'. See perf_iterate_sb_cpu().
*/
smp_store_release(&event->ctx, ctx);
if (!task) {
cpu_function_call(cpu, __perf_install_in_context, event);
return;
}
/*
* Should not happen, we validate the ctx is still alive before calling.
*/
if (WARN_ON_ONCE(task == TASK_TOMBSTONE))
return;
/*
* Installing events is tricky because we cannot rely on ctx->is_active
* to be set in case this is the nr_events 0 -> 1 transition.
*
* Instead we use task_curr(), which tells us if the task is running.
* However, since we use task_curr() outside of rq::lock, we can race
* against the actual state. This means the result can be wrong.
*
* If we get a false positive, we retry, this is harmless.
*
* If we get a false negative, things are complicated. If we are after
* perf_event_context_sched_in() ctx::lock will serialize us, and the
* value must be correct. If we're before, it doesn't matter since
* perf_event_context_sched_in() will program the counter.
*
* However, this hinges on the remote context switch having observed
* our task->perf_event_ctxp[] store, such that it will in fact take
* ctx::lock in perf_event_context_sched_in().
*
* We do this by task_function_call(), if the IPI fails to hit the task
* we know any future context switch of task must see the
* perf_event_ctpx[] store.
*/
/*
* This smp_mb() orders the task->perf_event_ctxp[] store with the
* task_cpu() load, such that if the IPI then does not find the task
* running, a future context switch of that task must observe the
* store.
*/
smp_mb();
again:
if (!task_function_call(task, __perf_install_in_context, event))
return;
raw_spin_lock_irq(&ctx->lock);
task = ctx->task;
if (WARN_ON_ONCE(task == TASK_TOMBSTONE)) {
/*
* Cannot happen because we already checked above (which also
* cannot happen), and we hold ctx->mutex, which serializes us
* against perf_event_exit_task_context().
*/
raw_spin_unlock_irq(&ctx->lock);
return;
}
/*
* If the task is not running, ctx->lock will avoid it becoming so,
* thus we can safely install the event.
*/
if (task_curr(task)) {
raw_spin_unlock_irq(&ctx->lock);
goto again;
}
add_event_to_ctx(event, ctx);
raw_spin_unlock_irq(&ctx->lock);
}
/*
* Cross CPU call to enable a performance event
*/
static void __perf_event_enable(struct perf_event *event,
struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx,
void *info)
{
struct perf_event *leader = event->group_leader;
struct perf_event_context *task_ctx;
if (event->state >= PERF_EVENT_STATE_INACTIVE ||
event->state <= PERF_EVENT_STATE_ERROR)
return;
if (ctx->is_active)
ctx_sched_out(ctx, cpuctx, EVENT_TIME);
perf_event_set_state(event, PERF_EVENT_STATE_INACTIVE);
if (!ctx->is_active)
return;
if (!event_filter_match(event)) {
ctx_sched_in(ctx, cpuctx, EVENT_TIME, current);
return;
}
/*
* If the event is in a group and isn't the group leader,
* then don't put it on unless the group is on.
*/
if (leader != event && leader->state != PERF_EVENT_STATE_ACTIVE) {
ctx_sched_in(ctx, cpuctx, EVENT_TIME, current);
return;
}
task_ctx = cpuctx->task_ctx;
if (ctx->task)
WARN_ON_ONCE(task_ctx != ctx);
ctx_resched(cpuctx, task_ctx, get_event_type(event));
}
/*
* Enable an event.
*
* If event->ctx is a cloned context, callers must make sure that
* every task struct that event->ctx->task could possibly point to
* remains valid. This condition is satisfied when called through
* perf_event_for_each_child or perf_event_for_each as described
* for perf_event_disable.
*/
static void _perf_event_enable(struct perf_event *event)
{
struct perf_event_context *ctx = event->ctx;
raw_spin_lock_irq(&ctx->lock);
if (event->state >= PERF_EVENT_STATE_INACTIVE ||
event->state < PERF_EVENT_STATE_ERROR) {
raw_spin_unlock_irq(&ctx->lock);
return;
}
/*
* If the event is in error state, clear that first.
*
* That way, if we see the event in error state below, we know that it
* has gone back into error state, as distinct from the task having
* been scheduled away before the cross-call arrived.
*/
if (event->state == PERF_EVENT_STATE_ERROR)
event->state = PERF_EVENT_STATE_OFF;
raw_spin_unlock_irq(&ctx->lock);
event_function_call(event, __perf_event_enable, NULL);
}
/*
* See perf_event_disable();
*/
void perf_event_enable(struct perf_event *event)
{
struct perf_event_context *ctx;
ctx = perf_event_ctx_lock(event);
_perf_event_enable(event);
perf_event_ctx_unlock(event, ctx);
}
EXPORT_SYMBOL_GPL(perf_event_enable);
struct stop_event_data {
struct perf_event *event;
unsigned int restart;
};
static int __perf_event_stop(void *info)
{
struct stop_event_data *sd = info;
struct perf_event *event = sd->event;
/* if it's already INACTIVE, do nothing */
if (READ_ONCE(event->state) != PERF_EVENT_STATE_ACTIVE)
return 0;
/* matches smp_wmb() in event_sched_in() */
smp_rmb();
/*
* There is a window with interrupts enabled before we get here,
* so we need to check again lest we try to stop another CPU's event.
*/
if (READ_ONCE(event->oncpu) != smp_processor_id())
return -EAGAIN;
event->pmu->stop(event, PERF_EF_UPDATE);
/*
* May race with the actual stop (through perf_pmu_output_stop()),
* but it is only used for events with AUX ring buffer, and such
* events will refuse to restart because of rb::aux_mmap_count==0,
* see comments in perf_aux_output_begin().
*
* Since this is happening on an event-local CPU, no trace is lost
* while restarting.
*/
if (sd->restart)
event->pmu->start(event, 0);
return 0;
}
static int perf_event_stop(struct perf_event *event, int restart)
{
struct stop_event_data sd = {
.event = event,
.restart = restart,
};
int ret = 0;
do {
if (READ_ONCE(event->state) != PERF_EVENT_STATE_ACTIVE)
return 0;
/* matches smp_wmb() in event_sched_in() */
smp_rmb();
/*
* We only want to restart ACTIVE events, so if the event goes
* inactive here (event->oncpu==-1), there's nothing more to do;
* fall through with ret==-ENXIO.
*/
ret = cpu_function_call(READ_ONCE(event->oncpu),
__perf_event_stop, &sd);
} while (ret == -EAGAIN);
return ret;
}
/*
* In order to contain the amount of racy and tricky in the address filter
* configuration management, it is a two part process:
*
* (p1) when userspace mappings change as a result of (1) or (2) or (3) below,
* we update the addresses of corresponding vmas in
* event::addr_filter_ranges array and bump the event::addr_filters_gen;
* (p2) when an event is scheduled in (pmu::add), it calls
* perf_event_addr_filters_sync() which calls pmu::addr_filters_sync()
* if the generation has changed since the previous call.
*
* If (p1) happens while the event is active, we restart it to force (p2).
*
* (1) perf_addr_filters_apply(): adjusting filters' offsets based on
* pre-existing mappings, called once when new filters arrive via SET_FILTER
* ioctl;
* (2) perf_addr_filters_adjust(): adjusting filters' offsets based on newly
* registered mapping, called for every new mmap(), with mm::mmap_sem down
* for reading;
* (3) perf_event_addr_filters_exec(): clearing filters' offsets in the process
* of exec.
*/
void perf_event_addr_filters_sync(struct perf_event *event)
{
struct perf_addr_filters_head *ifh = perf_event_addr_filters(event);
if (!has_addr_filter(event))
return;
raw_spin_lock(&ifh->lock);
if (event->addr_filters_gen != event->hw.addr_filters_gen) {
event->pmu->addr_filters_sync(event);
event->hw.addr_filters_gen = event->addr_filters_gen;
}
raw_spin_unlock(&ifh->lock);
}
EXPORT_SYMBOL_GPL(perf_event_addr_filters_sync);
static int _perf_event_refresh(struct perf_event *event, int refresh)
{
/*
* not supported on inherited events
*/
if (event->attr.inherit || !is_sampling_event(event))
return -EINVAL;
atomic_add(refresh, &event->event_limit);
_perf_event_enable(event);
return 0;
}
/*
* See perf_event_disable()
*/
int perf_event_refresh(struct perf_event *event, int refresh)
{
struct perf_event_context *ctx;
int ret;
ctx = perf_event_ctx_lock(event);
ret = _perf_event_refresh(event, refresh);
perf_event_ctx_unlock(event, ctx);
return ret;
}
EXPORT_SYMBOL_GPL(perf_event_refresh);
static int perf_event_modify_breakpoint(struct perf_event *bp,
struct perf_event_attr *attr)
{
int err;
_perf_event_disable(bp);
err = modify_user_hw_breakpoint_check(bp, attr, true);
if (!bp->attr.disabled)
_perf_event_enable(bp);
return err;
}
static int perf_event_modify_attr(struct perf_event *event,
struct perf_event_attr *attr)
{
if (event->attr.type != attr->type)
return -EINVAL;
switch (event->attr.type) {
case PERF_TYPE_BREAKPOINT:
return perf_event_modify_breakpoint(event, attr);
default:
/* Place holder for future additions. */
return -EOPNOTSUPP;
}
}
static void ctx_sched_out(struct perf_event_context *ctx,
struct perf_cpu_context *cpuctx,
enum event_type_t event_type)
{
struct perf_event *event, *tmp;
int is_active = ctx->is_active;
lockdep_assert_held(&ctx->lock);
if (likely(!ctx->nr_events)) {
/*
* See __perf_remove_from_context().
*/
WARN_ON_ONCE(ctx->is_active);
if (ctx->task)
WARN_ON_ONCE(cpuctx->task_ctx);
return;
}
ctx->is_active &= ~event_type;
if (!(ctx->is_active & EVENT_ALL))
ctx->is_active = 0;
if (ctx->task) {
WARN_ON_ONCE(cpuctx->task_ctx != ctx);
if (!ctx->is_active)
cpuctx->task_ctx = NULL;
}
/*
* Always update time if it was set; not only when it changes.
* Otherwise we can 'forget' to update time for any but the last
* context we sched out. For example:
*
* ctx_sched_out(.event_type = EVENT_FLEXIBLE)
* ctx_sched_out(.event_type = EVENT_PINNED)
*
* would only update time for the pinned events.
*/
if (is_active & EVENT_TIME) {
/* update (and stop) ctx time */
update_context_time(ctx);
update_cgrp_time_from_cpuctx(cpuctx);
}
is_active ^= ctx->is_active; /* changed bits */
if (!ctx->nr_active || !(is_active & EVENT_ALL))
return;
perf_pmu_disable(ctx->pmu);
if (is_active & EVENT_PINNED) {
list_for_each_entry_safe(event, tmp, &ctx->pinned_active, active_list)
group_sched_out(event, cpuctx, ctx);
}
if (is_active & EVENT_FLEXIBLE) {
list_for_each_entry_safe(event, tmp, &ctx->flexible_active, active_list)
group_sched_out(event, cpuctx, ctx);
/*
* Since we cleared EVENT_FLEXIBLE, also clear
* rotate_necessary, is will be reset by
* ctx_flexible_sched_in() when needed.
*/
ctx->rotate_necessary = 0;
}
perf_pmu_enable(ctx->pmu);
}
/*
* Test whether two contexts are equivalent, i.e. whether they have both been
* cloned from the same version of the same context.
*
* Equivalence is measured using a generation number in the context that is
* incremented on each modification to it; see unclone_ctx(), list_add_event()
* and list_del_event().
*/
static int context_equiv(struct perf_event_context *ctx1,
struct perf_event_context *ctx2)
{
lockdep_assert_held(&ctx1->lock);
lockdep_assert_held(&ctx2->lock);
/* Pinning disables the swap optimization */
if (ctx1->pin_count || ctx2->pin_count)
return 0;
/* If ctx1 is the parent of ctx2 */
if (ctx1 == ctx2->parent_ctx && ctx1->generation == ctx2->parent_gen)
return 1;
/* If ctx2 is the parent of ctx1 */
if (ctx1->parent_ctx == ctx2 && ctx1->parent_gen == ctx2->generation)
return 1;
/*
* If ctx1 and ctx2 have the same parent; we flatten the parent
* hierarchy, see perf_event_init_context().
*/
if (ctx1->parent_ctx && ctx1->parent_ctx == ctx2->parent_ctx &&
ctx1->parent_gen == ctx2->parent_gen)
return 1;
/* Unmatched */
return 0;
}
static void __perf_event_sync_stat(struct perf_event *event,
struct perf_event *next_event)
{
u64 value;
if (!event->attr.inherit_stat)
return;
/*
* Update the event value, we cannot use perf_event_read()
* because we're in the middle of a context switch and have IRQs
* disabled, which upsets smp_call_function_single(), however
* we know the event must be on the current CPU, therefore we
* don't need to use it.
*/
if (event->state == PERF_EVENT_STATE_ACTIVE)
event->pmu->read(event);
perf_event_update_time(event);
/*
* In order to keep per-task stats reliable we need to flip the event
* values when we flip the contexts.
*/
value = local64_read(&next_event->count);
value = local64_xchg(&event->count, value);
local64_set(&next_event->count, value);
swap(event->total_time_enabled, next_event->total_time_enabled);
swap(event->total_time_running, next_event->total_time_running);
/*
* Since we swizzled the values, update the user visible data too.
*/
perf_event_update_userpage(event);
perf_event_update_userpage(next_event);
}
static void perf_event_sync_stat(struct perf_event_context *ctx,
struct perf_event_context *next_ctx)
{
struct perf_event *event, *next_event;
if (!ctx->nr_stat)
return;
update_context_time(ctx);
event = list_first_entry(&ctx->event_list,
struct perf_event, event_entry);
next_event = list_first_entry(&next_ctx->event_list,
struct perf_event, event_entry);
while (&event->event_entry != &ctx->event_list &&
&next_event->event_entry != &next_ctx->event_list) {
__perf_event_sync_stat(event, next_event);
event = list_next_entry(event, event_entry);
next_event = list_next_entry(next_event, event_entry);
}
}
static void perf_event_context_sched_out(struct task_struct *task, int ctxn,
struct task_struct *next)
{
struct perf_event_context *ctx = task->perf_event_ctxp[ctxn];
struct perf_event_context *next_ctx;
struct perf_event_context *parent, *next_parent;
struct perf_cpu_context *cpuctx;
int do_switch = 1;
if (likely(!ctx))
return;
cpuctx = __get_cpu_context(ctx);
if (!cpuctx->task_ctx)
return;
rcu_read_lock();
next_ctx = next->perf_event_ctxp[ctxn];
if (!next_ctx)
goto unlock;
parent = rcu_dereference(ctx->parent_ctx);
next_parent = rcu_dereference(next_ctx->parent_ctx);
/* If neither context have a parent context; they cannot be clones. */
if (!parent && !next_parent)
goto unlock;
if (next_parent == ctx || next_ctx == parent || next_parent == parent) {
/*
* Looks like the two contexts are clones, so we might be
* able to optimize the context switch. We lock both
* contexts and check that they are clones under the
* lock (including re-checking that neither has been
* uncloned in the meantime). It doesn't matter which
* order we take the locks because no other cpu could
* be trying to lock both of these tasks.
*/
raw_spin_lock(&ctx->lock);
raw_spin_lock_nested(&next_ctx->lock, SINGLE_DEPTH_NESTING);
if (context_equiv(ctx, next_ctx)) {
WRITE_ONCE(ctx->task, next);
WRITE_ONCE(next_ctx->task, task);
swap(ctx->task_ctx_data, next_ctx->task_ctx_data);
/*
* RCU_INIT_POINTER here is safe because we've not
* modified the ctx and the above modification of
* ctx->task and ctx->task_ctx_data are immaterial
* since those values are always verified under
* ctx->lock which we're now holding.
*/
RCU_INIT_POINTER(task->perf_event_ctxp[ctxn], next_ctx);
RCU_INIT_POINTER(next->perf_event_ctxp[ctxn], ctx);
do_switch = 0;
perf_event_sync_stat(ctx, next_ctx);
}
raw_spin_unlock(&next_ctx->lock);
raw_spin_unlock(&ctx->lock);
}
unlock:
rcu_read_unlock();
if (do_switch) {
raw_spin_lock(&ctx->lock);
task_ctx_sched_out(cpuctx, ctx, EVENT_ALL);
raw_spin_unlock(&ctx->lock);
}
}
static DEFINE_PER_CPU(struct list_head, sched_cb_list);
void perf_sched_cb_dec(struct pmu *pmu)
{
struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
this_cpu_dec(perf_sched_cb_usages);
if (!--cpuctx->sched_cb_usage)
list_del(&cpuctx->sched_cb_entry);
}
void perf_sched_cb_inc(struct pmu *pmu)
{
struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
if (!cpuctx->sched_cb_usage++)
list_add(&cpuctx->sched_cb_entry, this_cpu_ptr(&sched_cb_list));
this_cpu_inc(perf_sched_cb_usages);
}
/*
* This function provides the context switch callback to the lower code
* layer. It is invoked ONLY when the context switch callback is enabled.
*
* This callback is relevant even to per-cpu events; for example multi event
* PEBS requires this to provide PID/TID information. This requires we flush
* all queued PEBS records before we context switch to a new task.
*/
static void perf_pmu_sched_task(struct task_struct *prev,
struct task_struct *next,
bool sched_in)
{
struct perf_cpu_context *cpuctx;
struct pmu *pmu;
if (prev == next)
return;
list_for_each_entry(cpuctx, this_cpu_ptr(&sched_cb_list), sched_cb_entry) {
pmu = cpuctx->ctx.pmu; /* software PMUs will not have sched_task */
if (WARN_ON_ONCE(!pmu->sched_task))
continue;
perf_ctx_lock(cpuctx, cpuctx->task_ctx);
perf_pmu_disable(pmu);
pmu->sched_task(cpuctx->task_ctx, sched_in);
perf_pmu_enable(pmu);
perf_ctx_unlock(cpuctx, cpuctx->task_ctx);
}
}
static void perf_event_switch(struct task_struct *task,
struct task_struct *next_prev, bool sched_in);
#define for_each_task_context_nr(ctxn) \
for ((ctxn) = 0; (ctxn) < perf_nr_task_contexts; (ctxn)++)
/*
* Called from scheduler to remove the events of the current task,
* with interrupts disabled.
*
* We stop each event and update the event value in event->count.
*
* This does not protect us against NMI, but disable()
* sets the disabled bit in the control field of event _before_
* accessing the event control register. If a NMI hits, then it will
* not restart the event.
*/
void __perf_event_task_sched_out(struct task_struct *task,
struct task_struct *next)
{
int ctxn;
if (__this_cpu_read(perf_sched_cb_usages))
perf_pmu_sched_task(task, next, false);
if (atomic_read(&nr_switch_events))
perf_event_switch(task, next, false);
for_each_task_context_nr(ctxn)
perf_event_context_sched_out(task, ctxn, next);
/*
* if cgroup events exist on this CPU, then we need
* to check if we have to switch out PMU state.
* cgroup event are system-wide mode only
*/
if (atomic_read(this_cpu_ptr(&perf_cgroup_events)))
perf_cgroup_sched_out(task, next);
}
/*
* Called with IRQs disabled
*/
static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx,
enum event_type_t event_type)
{
ctx_sched_out(&cpuctx->ctx, cpuctx, event_type);
}
static int visit_groups_merge(struct perf_event_groups *groups, int cpu,
int (*func)(struct perf_event *, void *), void *data)
{
struct perf_event **evt, *evt1, *evt2;
int ret;