#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/dcache.h>
#include <linux/percpu.h>
#include <linux/ptrace.h>
+#include <linux/reboot.h>
#include <linux/vmstat.h>
+#include <linux/device.h>
#include <linux/vmalloc.h>
#include <linux/hardirq.h>
#include <linux/rculist.h>
#include <linux/kernel_stat.h>
#include <linux/perf_event.h>
#include <linux/ftrace_event.h>
+#include <linux/hw_breakpoint.h>
#include <asm/irq_regs.h>
-/*
- * Each CPU has a list of per CPU events:
- */
-static DEFINE_PER_CPU(struct perf_cpu_context, perf_cpu_context);
+enum event_type_t {
+ EVENT_FLEXIBLE = 0x1,
+ EVENT_PINNED = 0x2,
+ EVENT_ALL = EVENT_FLEXIBLE | EVENT_PINNED,
+};
-static atomic_t nr_events __read_mostly;
+atomic_t perf_task_events __read_mostly;
static atomic_t nr_mmap_events __read_mostly;
static atomic_t nr_comm_events __read_mostly;
static atomic_t nr_task_events __read_mostly;
+static LIST_HEAD(pmus);
+static DEFINE_MUTEX(pmus_lock);
+static struct srcu_struct pmus_srcu;
+
/*
* perf event paranoia level:
* -1 - not paranoid at all
*/
int sysctl_perf_event_paranoid __read_mostly = 1;
-int sysctl_perf_event_mlock __read_mostly = 512; /* 'free' kb per user */
+/* 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
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);
+
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();
+}
+
void perf_pmu_disable(struct pmu *pmu)
{
int *count = this_cpu_ptr(pmu->pmu_disable_count);
pmu->pmu_enable(pmu);
}
-static void perf_pmu_rotate_start(void)
-{
- struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
-
- if (hrtimer_active(&cpuctx->timer))
- return;
-
- __hrtimer_start_range_ns(&cpuctx->timer,
- ns_to_ktime(cpuctx->timer_interval), 0,
- HRTIMER_MODE_REL_PINNED, 0);
-}
+static DEFINE_PER_CPU(struct list_head, rotation_list);
-static void perf_pmu_rotate_stop(void)
+/*
+ * perf_pmu_rotate_start() and perf_rotate_context() are fully serialized
+ * because they're strictly cpu affine and rotate_start is called with IRQs
+ * disabled, while rotate_context is called from IRQ context.
+ */
+static void perf_pmu_rotate_start(struct pmu *pmu)
{
- struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
+ struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
+ struct list_head *head = &__get_cpu_var(rotation_list);
- hrtimer_cancel(&cpuctx->timer);
+ WARN_ON(!irqs_disabled());
+
+ if (list_empty(&cpuctx->rotation_list))
+ list_add(&cpuctx->rotation_list, head);
}
static void get_ctx(struct perf_event_context *ctx)
}
}
+static u32 perf_event_pid(struct perf_event *event, struct task_struct *p)
+{
+ /*
+ * only top level events have the pid namespace they were created in
+ */
+ if (event->parent)
+ event = event->parent;
+
+ return task_tgid_nr_ns(p, event->ns);
+}
+
+static u32 perf_event_tid(struct perf_event *event, struct task_struct *p)
+{
+ /*
+ * only top level events have the pid namespace they were created in
+ */
+ if (event->parent)
+ event = event->parent;
+
+ return task_pid_nr_ns(p, event->ns);
+}
+
/*
* If we inherit events we want to return the parent event id
* to userspace.
* the context could get moved to another task.
*/
static struct perf_event_context *
-perf_lock_task_context(struct task_struct *task, unsigned long *flags)
+perf_lock_task_context(struct task_struct *task, int ctxn, unsigned long *flags)
{
struct perf_event_context *ctx;
rcu_read_lock();
retry:
- ctx = rcu_dereference(task->perf_event_ctxp);
+ ctx = rcu_dereference(task->perf_event_ctxp[ctxn]);
if (ctx) {
/*
* If this context is a clone of another, it might
* can't get swapped on us any more.
*/
raw_spin_lock_irqsave(&ctx->lock, *flags);
- if (ctx != rcu_dereference(task->perf_event_ctxp)) {
+ if (ctx != rcu_dereference(task->perf_event_ctxp[ctxn])) {
raw_spin_unlock_irqrestore(&ctx->lock, *flags);
goto retry;
}
* 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)
+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, &flags);
+ ctx = perf_lock_task_context(task, ctxn, &flags);
if (ctx) {
++ctx->pin_count;
raw_spin_unlock_irqrestore(&ctx->lock, flags);
put_ctx(ctx);
}
-static inline u64 perf_clock(void)
-{
- return local_clock();
-}
-
/*
* Update the record of the current time in a context.
*/
ctx->timestamp = now;
}
+static u64 perf_event_time(struct perf_event *event)
+{
+ struct perf_event_context *ctx = event->ctx;
+ return ctx ? ctx->time : 0;
+}
+
/*
* Update the total_time_enabled and total_time_running fields for a event.
*/
return;
if (ctx->is_active)
- run_end = ctx->time;
+ run_end = perf_event_time(event);
else
run_end = event->tstamp_stopped;
if (event->state == PERF_EVENT_STATE_INACTIVE)
run_end = event->tstamp_stopped;
else
- run_end = ctx->time;
+ run_end = perf_event_time(event);
event->total_time_running = run_end - event->tstamp_running;
}
list_add_rcu(&event->event_entry, &ctx->event_list);
if (!ctx->nr_events)
- perf_pmu_rotate_start();
+ perf_pmu_rotate_start(ctx->pmu);
ctx->nr_events++;
if (event->attr.inherit_stat)
ctx->nr_stat++;
}
+/*
+ * Called at perf_event creation and when events are attached/detached from a
+ * group.
+ */
+static void perf_event__read_size(struct perf_event *event)
+{
+ int entry = sizeof(u64); /* value */
+ int size = 0;
+ int nr = 1;
+
+ if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
+ size += sizeof(u64);
+
+ if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
+ size += sizeof(u64);
+
+ if (event->attr.read_format & PERF_FORMAT_ID)
+ entry += sizeof(u64);
+
+ if (event->attr.read_format & PERF_FORMAT_GROUP) {
+ nr += event->group_leader->nr_siblings;
+ size += sizeof(u64);
+ }
+
+ size += entry * nr;
+ event->read_size = size;
+}
+
+static void perf_event__header_size(struct perf_event *event)
+{
+ struct perf_sample_data *data;
+ u64 sample_type = event->attr.sample_type;
+ u16 size = 0;
+
+ perf_event__read_size(event);
+
+ 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_READ)
+ size += event->read_size;
+
+ event->header_size = size;
+}
+
+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_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;
+}
+
static void perf_group_attach(struct perf_event *event)
{
- struct perf_event *group_leader = event->group_leader;
+ struct perf_event *group_leader = event->group_leader, *pos;
+
+ /*
+ * We can have double attach due to group movement in perf_event_open.
+ */
+ if (event->attach_state & PERF_ATTACH_GROUP)
+ return;
- WARN_ON_ONCE(event->attach_state & PERF_ATTACH_GROUP);
event->attach_state |= PERF_ATTACH_GROUP;
if (group_leader == event)
list_add_tail(&event->group_entry, &group_leader->sibling_list);
group_leader->nr_siblings++;
+
+ perf_event__header_size(group_leader);
+
+ list_for_each_entry(pos, &group_leader->sibling_list, group_entry)
+ perf_event__header_size(pos);
}
/*
if (event->group_leader != event) {
list_del_init(&event->group_entry);
event->group_leader->nr_siblings--;
- return;
+ goto out;
}
if (!list_empty(&event->group_entry))
/* Inherit group flags from the previous leader */
sibling->group_flags = event->group_flags;
}
+
+out:
+ perf_event__header_size(event->group_leader);
+
+ list_for_each_entry(tmp, &event->group_leader->sibling_list, group_entry)
+ perf_event__header_size(tmp);
}
static inline int
struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx)
{
+ u64 tstamp = perf_event_time(event);
u64 delta;
/*
* An event which could not be activated because of
&& !event_filter_match(event)) {
delta = ctx->time - event->tstamp_stopped;
event->tstamp_running += delta;
- event->tstamp_stopped = ctx->time;
+ event->tstamp_stopped = tstamp;
}
if (event->state != PERF_EVENT_STATE_ACTIVE)
event->pending_disable = 0;
event->state = PERF_EVENT_STATE_OFF;
}
- event->tstamp_stopped = ctx->time;
+ event->tstamp_stopped = tstamp;
event->pmu->del(event, 0);
event->oncpu = -1;
cpuctx->exclusive = 0;
}
+static inline struct perf_cpu_context *
+__get_cpu_context(struct perf_event_context *ctx)
+{
+ return this_cpu_ptr(ctx->pmu->pmu_cpu_context);
+}
+
/*
* Cross CPU call to remove a performance event
*
*/
static void __perf_event_remove_from_context(void *info)
{
- struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
struct perf_event *event = info;
struct perf_event_context *ctx = event->ctx;
+ struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
/*
* If this is a task context, we need to check whether it is
static void __perf_event_disable(void *info)
{
struct perf_event *event = info;
- struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
struct perf_event_context *ctx = event->ctx;
+ struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
/*
* If this is a per-task event, need to check whether this
raw_spin_unlock_irq(&ctx->lock);
}
+#define MAX_INTERRUPTS (~0ULL)
+
+static void perf_log_throttle(struct perf_event *event, int enable);
+
static int
event_sched_in(struct perf_event *event,
struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx)
{
+ u64 tstamp = perf_event_time(event);
+
if (event->state <= PERF_EVENT_STATE_OFF)
return 0;
event->state = PERF_EVENT_STATE_ACTIVE;
event->oncpu = smp_processor_id();
+
+ /*
+ * 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;
+ }
+
/*
* The new state must be visible before we turn it on in the hardware:
*/
return -EAGAIN;
}
- event->tstamp_running += ctx->time - event->tstamp_stopped;
+ event->tstamp_running += tstamp - event->tstamp_stopped;
+
+ event->shadow_ctx_time = tstamp - ctx->timestamp;
if (!is_software_event(event))
cpuctx->active_oncpu++;
{
struct perf_event *event, *partial_group = NULL;
struct pmu *pmu = group_event->pmu;
+ u64 now = ctx->time;
+ bool simulate = false;
if (group_event->state == PERF_EVENT_STATE_OFF)
return 0;
/*
* 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,
+ * tstamp_stopped will be updated.
+ *
+ * The failed events and the remaining siblings need to have
+ * their timings updated as if they had gone thru event_sched_in()
+ * and event_sched_out(). This is required to get consistent timings
+ * across the group. This also takes care of the case where the group
+ * could never be scheduled by ensuring tstamp_stopped is set to mark
+ * the time the event was actually stopped, such that time delta
+ * calculation in update_event_times() is correct.
*/
list_for_each_entry(event, &group_event->sibling_list, group_entry) {
if (event == partial_group)
- break;
- event_sched_out(event, cpuctx, ctx);
+ simulate = true;
+
+ if (simulate) {
+ event->tstamp_running += now - event->tstamp_stopped;
+ event->tstamp_stopped = now;
+ } else {
+ event_sched_out(event, cpuctx, ctx);
+ }
}
event_sched_out(group_event, cpuctx, ctx);
static void add_event_to_ctx(struct perf_event *event,
struct perf_event_context *ctx)
{
+ u64 tstamp = perf_event_time(event);
+
list_add_event(event, ctx);
perf_group_attach(event);
- event->tstamp_enabled = ctx->time;
- event->tstamp_running = ctx->time;
- event->tstamp_stopped = ctx->time;
+ event->tstamp_enabled = tstamp;
+ event->tstamp_running = tstamp;
+ event->tstamp_stopped = tstamp;
}
/*
*/
static void __perf_install_in_context(void *info)
{
- struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
struct perf_event *event = info;
struct perf_event_context *ctx = event->ctx;
struct perf_event *leader = event->group_leader;
+ struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
int err;
/*
add_event_to_ctx(event, ctx);
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
goto unlock;
/*
struct perf_event_context *ctx)
{
struct perf_event *sub;
+ u64 tstamp = perf_event_time(event);
event->state = PERF_EVENT_STATE_INACTIVE;
- event->tstamp_enabled = ctx->time - event->total_time_enabled;
+ event->tstamp_enabled = tstamp - event->total_time_enabled;
list_for_each_entry(sub, &event->sibling_list, group_entry) {
- if (sub->state >= PERF_EVENT_STATE_INACTIVE) {
- sub->tstamp_enabled =
- ctx->time - sub->total_time_enabled;
- }
+ if (sub->state >= PERF_EVENT_STATE_INACTIVE)
+ sub->tstamp_enabled = tstamp - sub->total_time_enabled;
}
}
static void __perf_event_enable(void *info)
{
struct perf_event *event = info;
- struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
struct perf_event_context *ctx = event->ctx;
struct perf_event *leader = event->group_leader;
+ struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
int err;
/*
goto unlock;
__perf_event_mark_enabled(event, ctx);
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
goto unlock;
/*
/*
* not supported on inherited events
*/
- if (event->attr.inherit)
+ if (event->attr.inherit || !is_sampling_event(event))
return -EINVAL;
atomic_add(refresh, &event->event_limit);
return 0;
}
-enum event_type_t {
- EVENT_FLEXIBLE = 0x1,
- EVENT_PINNED = 0x2,
- EVENT_ALL = EVENT_FLEXIBLE | EVENT_PINNED,
-};
-
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;
raw_spin_lock(&ctx->lock);
+ perf_pmu_disable(ctx->pmu);
ctx->is_active = 0;
if (likely(!ctx->nr_events))
goto out;
group_sched_out(event, cpuctx, ctx);
}
out:
+ perf_pmu_enable(ctx->pmu);
raw_spin_unlock(&ctx->lock);
}
}
}
-/*
- * 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)
+void perf_event_context_sched_out(struct task_struct *task, int ctxn,
+ struct task_struct *next)
{
- struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
- struct perf_event_context *ctx = task->perf_event_ctxp;
+ struct perf_event_context *ctx = task->perf_event_ctxp[ctxn];
struct perf_event_context *next_ctx;
struct perf_event_context *parent;
+ struct perf_cpu_context *cpuctx;
int do_switch = 1;
- perf_sw_event(PERF_COUNT_SW_CONTEXT_SWITCHES, 1, 1, NULL, 0);
+ if (likely(!ctx))
+ return;
- if (likely(!ctx || !cpuctx->task_ctx))
+ cpuctx = __get_cpu_context(ctx);
+ if (!cpuctx->task_ctx)
return;
rcu_read_lock();
parent = rcu_dereference(ctx->parent_ctx);
- next_ctx = next->perf_event_ctxp;
+ next_ctx = next->perf_event_ctxp[ctxn];
if (parent && next_ctx &&
rcu_dereference(next_ctx->parent_ctx) == parent) {
/*
* XXX do we need a memory barrier of sorts
* wrt to rcu_dereference() of perf_event_ctxp
*/
- task->perf_event_ctxp = next_ctx;
- next->perf_event_ctxp = ctx;
+ task->perf_event_ctxp[ctxn] = next_ctx;
+ next->perf_event_ctxp[ctxn] = ctx;
ctx->task = next;
next_ctx->task = task;
do_switch = 0;
}
}
+#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;
+
+ for_each_task_context_nr(ctxn)
+ perf_event_context_sched_out(task, ctxn, next);
+}
+
static void task_ctx_sched_out(struct perf_event_context *ctx,
enum event_type_t event_type)
{
- struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
+ struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
if (!cpuctx->task_ctx)
return;
/*
* Called with IRQs disabled
*/
-static void __perf_event_task_sched_out(struct perf_event_context *ctx)
-{
- task_ctx_sched_out(ctx, EVENT_ALL);
-}
-
-/*
- * Called with IRQs disabled
- */
static void cpu_ctx_sched_out(struct perf_cpu_context *cpuctx,
enum event_type_t event_type)
{
list_for_each_entry(event, &ctx->pinned_groups, group_entry) {
if (event->state <= PERF_EVENT_STATE_OFF)
continue;
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
continue;
if (group_can_go_on(event, cpuctx, 1))
* Listen to the 'cpu' scheduling filter constraint
* of events:
*/
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
continue;
if (group_can_go_on(event, cpuctx, can_add_hw)) {
ctx_sched_in(ctx, cpuctx, event_type);
}
-static void task_ctx_sched_in(struct task_struct *task,
+static void task_ctx_sched_in(struct perf_event_context *ctx,
enum event_type_t event_type)
{
- struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
- struct perf_event_context *ctx = task->perf_event_ctxp;
+ struct perf_cpu_context *cpuctx;
- if (likely(!ctx))
- return;
+ cpuctx = __get_cpu_context(ctx);
if (cpuctx->task_ctx == ctx)
return;
+
ctx_sched_in(ctx, cpuctx, event_type);
cpuctx->task_ctx = ctx;
}
-/*
- * Called from scheduler to add the events of the current task
- * with interrupts disabled.
- *
- * We restore the event value and then enable it.
- *
- * This does not protect us against NMI, but enable()
- * sets the enabled bit in the control field of event _before_
- * accessing the event control register. If a NMI hits, then it will
- * keep the event running.
- */
-void perf_event_task_sched_in(struct task_struct *task)
-{
- struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
- struct perf_event_context *ctx = task->perf_event_ctxp;
- if (likely(!ctx))
- return;
+void perf_event_context_sched_in(struct perf_event_context *ctx)
+{
+ struct perf_cpu_context *cpuctx;
+ cpuctx = __get_cpu_context(ctx);
if (cpuctx->task_ctx == ctx)
return;
+ perf_pmu_disable(ctx->pmu);
/*
* We want to keep the following priority order:
* cpu pinned (that don't need to move), task pinned,
* Since these rotations are per-cpu, we need to ensure the
* cpu-context we got scheduled on is actually rotating.
*/
- perf_pmu_rotate_start();
+ perf_pmu_rotate_start(ctx->pmu);
+ perf_pmu_enable(ctx->pmu);
}
-#define MAX_INTERRUPTS (~0ULL)
+/*
+ * Called from scheduler to add the events of the current task
+ * with interrupts disabled.
+ *
+ * We restore the event value and then enable it.
+ *
+ * This does not protect us against NMI, but enable()
+ * sets the enabled bit in the control field of event _before_
+ * accessing the event control register. If a NMI hits, then it will
+ * keep the event running.
+ */
+void __perf_event_task_sched_in(struct task_struct *task)
+{
+ struct perf_event_context *ctx;
+ int ctxn;
-static void perf_log_throttle(struct perf_event *event, int enable);
+ for_each_task_context_nr(ctxn) {
+ ctx = task->perf_event_ctxp[ctxn];
+ if (likely(!ctx))
+ continue;
+
+ perf_event_context_sched_in(ctx);
+ }
+}
static u64 perf_calculate_period(struct perf_event *event, u64 nsec, u64 count)
{
if (event->state != PERF_EVENT_STATE_ACTIVE)
continue;
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
continue;
hwc = &event->hw;
{
raw_spin_lock(&ctx->lock);
- /* Rotate the first entry last of non-pinned groups */
- list_rotate_left(&ctx->flexible_groups);
+ /*
+ * Rotate the first entry last of non-pinned groups. Rotation might be
+ * disabled by the inheritance code.
+ */
+ if (!ctx->rotate_disable)
+ list_rotate_left(&ctx->flexible_groups);
raw_spin_unlock(&ctx->lock);
}
/*
- * Cannot race with ->pmu_rotate_start() because this is ran from hardirq
- * context, and ->pmu_rotate_start() is called with irqs disabled (both are
- * cpu affine, so there are no SMP races).
+ * perf_pmu_rotate_start() and perf_rotate_context() are fully serialized
+ * because they're strictly cpu affine and rotate_start is called with IRQs
+ * disabled, while rotate_context is called from IRQ context.
*/
-static enum hrtimer_restart perf_event_context_tick(struct hrtimer *timer)
+static void perf_rotate_context(struct perf_cpu_context *cpuctx)
{
- enum hrtimer_restart restart = HRTIMER_NORESTART;
- struct perf_cpu_context *cpuctx;
- struct perf_event_context *ctx;
- int rotate = 0;
-
- cpuctx = container_of(timer, struct perf_cpu_context, timer);
+ u64 interval = (u64)cpuctx->jiffies_interval * TICK_NSEC;
+ struct perf_event_context *ctx = NULL;
+ int rotate = 0, remove = 1;
if (cpuctx->ctx.nr_events) {
- restart = HRTIMER_RESTART;
+ remove = 0;
if (cpuctx->ctx.nr_events != cpuctx->ctx.nr_active)
rotate = 1;
}
- ctx = current->perf_event_ctxp;
+ ctx = cpuctx->task_ctx;
if (ctx && ctx->nr_events) {
- restart = HRTIMER_RESTART;
+ remove = 0;
if (ctx->nr_events != ctx->nr_active)
rotate = 1;
}
- perf_ctx_adjust_freq(&cpuctx->ctx, cpuctx->timer_interval);
+ perf_pmu_disable(cpuctx->ctx.pmu);
+ perf_ctx_adjust_freq(&cpuctx->ctx, interval);
if (ctx)
- perf_ctx_adjust_freq(ctx, cpuctx->timer_interval);
+ perf_ctx_adjust_freq(ctx, interval);
if (!rotate)
goto done;
cpu_ctx_sched_in(cpuctx, EVENT_FLEXIBLE);
if (ctx)
- task_ctx_sched_in(current, EVENT_FLEXIBLE);
+ task_ctx_sched_in(ctx, EVENT_FLEXIBLE);
done:
- hrtimer_forward_now(timer, ns_to_ktime(cpuctx->timer_interval));
+ if (remove)
+ list_del_init(&cpuctx->rotation_list);
- return restart;
+ perf_pmu_enable(cpuctx->ctx.pmu);
+}
+
+void perf_event_task_tick(void)
+{
+ struct list_head *head = &__get_cpu_var(rotation_list);
+ struct perf_cpu_context *cpuctx, *tmp;
+
+ WARN_ON(!irqs_disabled());
+
+ list_for_each_entry_safe(cpuctx, tmp, head, rotation_list) {
+ if (cpuctx->jiffies_interval == 1 ||
+ !(jiffies % cpuctx->jiffies_interval))
+ perf_rotate_context(cpuctx);
+ }
}
static int event_enable_on_exec(struct perf_event *event,
* Enable all of a task's events that have been marked enable-on-exec.
* This expects task == current.
*/
-static void perf_event_enable_on_exec(struct task_struct *task)
+static void perf_event_enable_on_exec(struct perf_event_context *ctx)
{
- struct perf_event_context *ctx;
struct perf_event *event;
unsigned long flags;
int enabled = 0;
int ret;
local_irq_save(flags);
- ctx = task->perf_event_ctxp;
if (!ctx || !ctx->nr_events)
goto out;
- __perf_event_task_sched_out(ctx);
+ task_ctx_sched_out(ctx, EVENT_ALL);
raw_spin_lock(&ctx->lock);
raw_spin_unlock(&ctx->lock);
- perf_event_task_sched_in(task);
+ perf_event_context_sched_in(ctx);
out:
local_irq_restore(flags);
}
*/
static void __perf_event_read(void *info)
{
- struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
struct perf_event *event = info;
struct perf_event_context *ctx = event->ctx;
+ struct perf_cpu_context *cpuctx = __get_cpu_context(ctx);
/*
* If this is a task context, we need to check whether it is
return;
raw_spin_lock(&ctx->lock);
- update_context_time(ctx);
+ if (ctx->is_active)
+ update_context_time(ctx);
update_event_times(event);
+ if (event->state == PERF_EVENT_STATE_ACTIVE)
+ event->pmu->read(event);
raw_spin_unlock(&ctx->lock);
-
- event->pmu->read(event);
}
static inline u64 perf_event_count(struct perf_event *event)
unsigned long flags;
raw_spin_lock_irqsave(&ctx->lock, flags);
- update_context_time(ctx);
+ /*
+ * may read while context is not active
+ * (e.g., thread is blocked), in that case
+ * we cannot update context time
+ */
+ if (ctx->is_active)
+ update_context_time(ctx);
update_event_times(event);
raw_spin_unlock_irqrestore(&ctx->lock, flags);
}
* accessed from NMI. Use a temporary manual per cpu allocation
* until that gets sorted out.
*/
- size = sizeof(*entries) + sizeof(struct perf_callchain_entry *) *
- num_possible_cpus();
+ size = offsetof(struct callchain_cpus_entries, cpu_entries[nr_cpu_ids]);
entries = kzalloc(size, GFP_KERNEL);
if (!entries)
/*
* Initialize the perf_event context in a task_struct:
*/
-static void
-__perf_event_init_context(struct perf_event_context *ctx,
- struct task_struct *task)
+static void __perf_event_init_context(struct perf_event_context *ctx)
{
raw_spin_lock_init(&ctx->lock);
mutex_init(&ctx->mutex);
INIT_LIST_HEAD(&ctx->flexible_groups);
INIT_LIST_HEAD(&ctx->event_list);
atomic_set(&ctx->refcount, 1);
- ctx->task = task;
}
-static struct perf_event_context *find_get_context(pid_t pid, int cpu)
+static struct perf_event_context *
+alloc_perf_context(struct pmu *pmu, struct task_struct *task)
{
struct perf_event_context *ctx;
- struct perf_cpu_context *cpuctx;
+
+ ctx = kzalloc(sizeof(struct perf_event_context), GFP_KERNEL);
+ if (!ctx)
+ return NULL;
+
+ __perf_event_init_context(ctx);
+ if (task) {
+ ctx->task = task;
+ get_task_struct(task);
+ }
+ ctx->pmu = pmu;
+
+ return ctx;
+}
+
+static struct task_struct *
+find_lively_task_by_vpid(pid_t vpid)
+{
struct task_struct *task;
- unsigned long flags;
int err;
- if (pid == -1 && cpu != -1) {
+ rcu_read_lock();
+ if (!vpid)
+ task = current;
+ else
+ task = find_task_by_vpid(vpid);
+ if (task)
+ get_task_struct(task);
+ rcu_read_unlock();
+
+ if (!task)
+ return ERR_PTR(-ESRCH);
+
+ /* Reuse ptrace permission checks for now. */
+ err = -EACCES;
+ if (!ptrace_may_access(task, PTRACE_MODE_READ))
+ goto errout;
+
+ return task;
+errout:
+ put_task_struct(task);
+ return ERR_PTR(err);
+
+}
+
+static struct perf_event_context *
+find_get_context(struct pmu *pmu, struct task_struct *task, int cpu)
+{
+ struct perf_event_context *ctx;
+ struct perf_cpu_context *cpuctx;
+ unsigned long flags;
+ int ctxn, err;
+
+ if (!task) {
/* Must be root to operate on a CPU event: */
if (perf_paranoid_cpu() && !capable(CAP_SYS_ADMIN))
return ERR_PTR(-EACCES);
- if (cpu < 0 || cpu >= nr_cpumask_bits)
- return ERR_PTR(-EINVAL);
-
/*
* We could be clever and allow to attach a event to an
* offline CPU and activate it when the CPU comes up, but
if (!cpu_online(cpu))
return ERR_PTR(-ENODEV);
- cpuctx = &per_cpu(perf_cpu_context, cpu);
+ cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu);
ctx = &cpuctx->ctx;
get_ctx(ctx);
return ctx;
}
- rcu_read_lock();
- if (!pid)
- task = current;
- else
- task = find_task_by_vpid(pid);
- if (task)
- get_task_struct(task);
- rcu_read_unlock();
-
- if (!task)
- return ERR_PTR(-ESRCH);
-
- /*
- * Can't attach events to a dying task.
- */
- err = -ESRCH;
- if (task->flags & PF_EXITING)
- goto errout;
-
- /* Reuse ptrace permission checks for now. */
- err = -EACCES;
- if (!ptrace_may_access(task, PTRACE_MODE_READ))
+ err = -EINVAL;
+ ctxn = pmu->task_ctx_nr;
+ if (ctxn < 0)
goto errout;
retry:
- ctx = perf_lock_task_context(task, &flags);
+ ctx = perf_lock_task_context(task, ctxn, &flags);
if (ctx) {
unclone_ctx(ctx);
raw_spin_unlock_irqrestore(&ctx->lock, flags);
}
if (!ctx) {
- ctx = kzalloc(sizeof(struct perf_event_context), GFP_KERNEL);
+ ctx = alloc_perf_context(pmu, task);
err = -ENOMEM;
if (!ctx)
goto errout;
- __perf_event_init_context(ctx, task);
+
get_ctx(ctx);
- if (cmpxchg(&task->perf_event_ctxp, NULL, ctx)) {
- /*
- * We raced with some other task; use
- * the context they set.
- */
+
+ err = 0;
+ mutex_lock(&task->perf_event_mutex);
+ /*
+ * If it has already passed perf_event_exit_task().
+ * we must see PF_EXITING, it takes this mutex too.
+ */
+ if (task->flags & PF_EXITING)
+ err = -ESRCH;
+ else if (task->perf_event_ctxp[ctxn])
+ err = -EAGAIN;
+ else
+ rcu_assign_pointer(task->perf_event_ctxp[ctxn], ctx);
+ mutex_unlock(&task->perf_event_mutex);
+
+ if (unlikely(err)) {
+ put_task_struct(task);
kfree(ctx);
- goto retry;
+
+ if (err == -EAGAIN)
+ goto retry;
+ goto errout;
}
- get_task_struct(task);
}
- put_task_struct(task);
return ctx;
errout:
- put_task_struct(task);
return ERR_PTR(err);
}
kfree(event);
}
-static void perf_pending_sync(struct perf_event *event);
static void perf_buffer_put(struct perf_buffer *buffer);
static void free_event(struct perf_event *event)
{
- perf_pending_sync(event);
+ irq_work_sync(&event->pending);
if (!event->parent) {
- atomic_dec(&nr_events);
+ if (event->attach_state & PERF_ATTACH_TASK)
+ jump_label_dec(&perf_task_events);
if (event->attr.mmap || event->attr.mmap_data)
atomic_dec(&nr_mmap_events);
if (event->attr.comm)
if (event->destroy)
event->destroy(event);
- put_ctx(event->ctx);
+ if (event->ctx)
+ put_ctx(event->ctx);
+
call_rcu(&event->rcu_head, free_event_rcu);
}
raw_spin_unlock_irq(&ctx->lock);
mutex_unlock(&ctx->mutex);
- mutex_lock(&event->owner->perf_event_mutex);
- list_del_init(&event->owner_entry);
- mutex_unlock(&event->owner->perf_event_mutex);
- put_task_struct(event->owner);
-
free_event(event);
return 0;
static int perf_release(struct inode *inode, struct file *file)
{
struct perf_event *event = file->private_data;
+ struct task_struct *owner;
file->private_data = NULL;
- return perf_event_release_kernel(event);
-}
-
-static int perf_event_read_size(struct perf_event *event)
-{
- int entry = sizeof(u64); /* value */
- int size = 0;
- int nr = 1;
-
- if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
- size += sizeof(u64);
-
- if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
- size += sizeof(u64);
-
- if (event->attr.read_format & PERF_FORMAT_ID)
- entry += sizeof(u64);
-
- if (event->attr.read_format & PERF_FORMAT_GROUP) {
- nr += event->group_leader->nr_siblings;
- size += sizeof(u64);
+ rcu_read_lock();
+ owner = ACCESS_ONCE(event->owner);
+ /*
+ * Matches the smp_wmb() in perf_event_exit_task(). If we observe
+ * !owner it means the list deletion is complete and we can indeed
+ * free this event, otherwise we need to serialize on
+ * owner->perf_event_mutex.
+ */
+ smp_read_barrier_depends();
+ if (owner) {
+ /*
+ * Since delayed_put_task_struct() also drops the last
+ * task reference we can safely take a new reference
+ * while holding the rcu_read_lock().
+ */
+ get_task_struct(owner);
}
+ rcu_read_unlock();
- size += entry * nr;
+ if (owner) {
+ mutex_lock(&owner->perf_event_mutex);
+ /*
+ * We have to re-check the event->owner field, if it is cleared
+ * we raced with perf_event_exit_task(), acquiring the mutex
+ * ensured they're done, and we can proceed with freeing the
+ * event.
+ */
+ if (event->owner)
+ list_del_init(&event->owner_entry);
+ mutex_unlock(&owner->perf_event_mutex);
+ put_task_struct(owner);
+ }
- return size;
+ return perf_event_release_kernel(event);
}
u64 perf_event_read_value(struct perf_event *event, u64 *enabled, u64 *running)
if (event->state == PERF_EVENT_STATE_ERROR)
return 0;
- if (count < perf_event_read_size(event))
+ if (count < event->read_size)
return -ENOSPC;
WARN_ON_ONCE(event->ctx->parent_ctx);
static int perf_event_period(struct perf_event *event, u64 __user *arg)
{
struct perf_event_context *ctx = event->ctx;
- unsigned long size;
int ret = 0;
u64 value;
- if (!event->attr.sample_period)
+ if (!is_sampling_event(event))
return -EINVAL;
- size = copy_from_user(&value, arg, sizeof(value));
- if (size != sizeof(value))
+ if (copy_from_user(&value, arg, sizeof(value)))
return -EFAULT;
if (!value)
}
}
-/*
- * Pending wakeups
- *
- * Handle the case where we need to wakeup up from NMI (or rq->lock) context.
- *
- * The NMI bit means we cannot possibly take locks. Therefore, maintain a
- * single linked list and use cmpxchg() to add entries lockless.
- */
-
-static void perf_pending_event(struct perf_pending_entry *entry)
+static void perf_pending_event(struct irq_work *entry)
{
struct perf_event *event = container_of(entry,
struct perf_event, pending);
}
}
-#define PENDING_TAIL ((struct perf_pending_entry *)-1UL)
-
-static DEFINE_PER_CPU(struct perf_pending_entry *, perf_pending_head) = {
- PENDING_TAIL,
-};
-
-static void perf_pending_queue(struct perf_pending_entry *entry,
- void (*func)(struct perf_pending_entry *))
-{
- struct perf_pending_entry **head;
-
- if (cmpxchg(&entry->next, NULL, PENDING_TAIL) != NULL)
- return;
-
- entry->func = func;
-
- head = &get_cpu_var(perf_pending_head);
-
- do {
- entry->next = *head;
- } while (cmpxchg(head, entry->next, entry) != entry->next);
-
- set_perf_event_pending();
-
- put_cpu_var(perf_pending_head);
-}
-
-static int __perf_pending_run(void)
-{
- struct perf_pending_entry *list;
- int nr = 0;
-
- list = xchg(&__get_cpu_var(perf_pending_head), PENDING_TAIL);
- while (list != PENDING_TAIL) {
- void (*func)(struct perf_pending_entry *);
- struct perf_pending_entry *entry = list;
-
- list = list->next;
-
- func = entry->func;
- entry->next = NULL;
- /*
- * Ensure we observe the unqueue before we issue the wakeup,
- * so that we won't be waiting forever.
- * -- see perf_not_pending().
- */
- smp_wmb();
-
- func(entry);
- nr++;
- }
-
- return nr;
-}
-
-static inline int perf_not_pending(struct perf_event *event)
-{
- /*
- * If we flush on whatever cpu we run, there is a chance we don't
- * need to wait.
- */
- get_cpu();
- __perf_pending_run();
- put_cpu();
-
- /*
- * Ensure we see the proper queue state before going to sleep
- * so that we do not miss the wakeup. -- see perf_pending_handle()
- */
- smp_rmb();
- return event->pending.next == NULL;
-}
-
-static void perf_pending_sync(struct perf_event *event)
-{
- wait_event(event->waitq, perf_not_pending(event));
-}
-
-void perf_event_do_pending(void)
-{
- __perf_pending_run();
-}
-
/*
* We assume there is only KVM supporting the callbacks.
* Later on, we might change it to a list if there is
if (handle->nmi) {
handle->event->pending_wakeup = 1;
- perf_pending_queue(&handle->event->pending,
- perf_pending_event);
+ irq_work_queue(&handle->event->pending);
} else
perf_event_wakeup(handle->event);
}
} while (len);
}
+static void __perf_event_header__init_id(struct perf_event_header *header,
+ struct perf_sample_data *data,
+ struct perf_event *event)
+{
+ u64 sample_type = event->attr.sample_type;
+
+ data->type = sample_type;
+ header->size += event->id_header_size;
+
+ if (sample_type & PERF_SAMPLE_TID) {
+ /* namespace issues */
+ data->tid_entry.pid = perf_event_pid(event, current);
+ data->tid_entry.tid = perf_event_tid(event, current);
+ }
+
+ if (sample_type & PERF_SAMPLE_TIME)
+ data->time = perf_clock();
+
+ if (sample_type & PERF_SAMPLE_ID)
+ data->id = primary_event_id(event);
+
+ if (sample_type & PERF_SAMPLE_STREAM_ID)
+ data->stream_id = event->id;
+
+ if (sample_type & PERF_SAMPLE_CPU) {
+ data->cpu_entry.cpu = raw_smp_processor_id();
+ data->cpu_entry.reserved = 0;
+ }
+}
+
+static void perf_event_header__init_id(struct perf_event_header *header,
+ struct perf_sample_data *data,
+ struct perf_event *event)
+{
+ if (event->attr.sample_id_all)
+ __perf_event_header__init_id(header, data, event);
+}
+
+static void __perf_event__output_id_sample(struct perf_output_handle *handle,
+ struct perf_sample_data *data)
+{
+ u64 sample_type = data->type;
+
+ if (sample_type & PERF_SAMPLE_TID)
+ perf_output_put(handle, data->tid_entry);
+
+ if (sample_type & PERF_SAMPLE_TIME)
+ perf_output_put(handle, data->time);
+
+ if (sample_type & PERF_SAMPLE_ID)
+ perf_output_put(handle, data->id);
+
+ if (sample_type & PERF_SAMPLE_STREAM_ID)
+ perf_output_put(handle, data->stream_id);
+
+ if (sample_type & PERF_SAMPLE_CPU)
+ perf_output_put(handle, data->cpu_entry);
+}
+
+static void perf_event__output_id_sample(struct perf_event *event,
+ struct perf_output_handle *handle,
+ struct perf_sample_data *sample)
+{
+ if (event->attr.sample_id_all)
+ __perf_event__output_id_sample(handle, sample);
+}
+
int perf_output_begin(struct perf_output_handle *handle,
struct perf_event *event, unsigned int size,
int nmi, int sample)
struct perf_buffer *buffer;
unsigned long tail, offset, head;
int have_lost;
+ struct perf_sample_data sample_data;
struct {
struct perf_event_header header;
u64 id;
goto out;
have_lost = local_read(&buffer->lost);
- if (have_lost)
- size += sizeof(lost_event);
+ if (have_lost) {
+ lost_event.header.size = sizeof(lost_event);
+ perf_event_header__init_id(&lost_event.header, &sample_data,
+ event);
+ size += lost_event.header.size;
+ }
perf_output_get_handle(handle);
if (have_lost) {
lost_event.header.type = PERF_RECORD_LOST;
lost_event.header.misc = 0;
- lost_event.header.size = sizeof(lost_event);
lost_event.id = event->id;
lost_event.lost = local_xchg(&buffer->lost, 0);
perf_output_put(handle, lost_event);
+ perf_event__output_id_sample(event, handle, &sample_data);
}
return 0;
rcu_read_unlock();
}
-static u32 perf_event_pid(struct perf_event *event, struct task_struct *p)
-{
- /*
- * only top level events have the pid namespace they were created in
- */
- if (event->parent)
- event = event->parent;
-
- return task_tgid_nr_ns(p, event->ns);
-}
-
-static u32 perf_event_tid(struct perf_event *event, struct task_struct *p)
-{
- /*
- * only top level events have the pid namespace they were created in
- */
- if (event->parent)
- event = event->parent;
-
- return task_pid_nr_ns(p, event->ns);
-}
-
static void perf_output_read_one(struct perf_output_handle *handle,
- struct perf_event *event)
+ struct perf_event *event,
+ u64 enabled, u64 running)
{
u64 read_format = event->attr.read_format;
u64 values[4];
values[n++] = perf_event_count(event);
if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) {
- values[n++] = event->total_time_enabled +
+ values[n++] = enabled +
atomic64_read(&event->child_total_time_enabled);
}
if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) {
- values[n++] = event->total_time_running +
+ values[n++] = running +
atomic64_read(&event->child_total_time_running);
}
if (read_format & PERF_FORMAT_ID)
* XXX PERF_FORMAT_GROUP vs inherited events seems difficult.
*/
static void perf_output_read_group(struct perf_output_handle *handle,
- struct perf_event *event)
+ struct perf_event *event,
+ u64 enabled, u64 running)
{
struct perf_event *leader = event->group_leader, *sub;
u64 read_format = event->attr.read_format;
values[n++] = 1 + leader->nr_siblings;
if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
- values[n++] = leader->total_time_enabled;
+ values[n++] = enabled;
if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
- values[n++] = leader->total_time_running;
+ values[n++] = running;
if (leader != event)
leader->pmu->read(leader);
}
}
+#define PERF_FORMAT_TOTAL_TIMES (PERF_FORMAT_TOTAL_TIME_ENABLED|\
+ PERF_FORMAT_TOTAL_TIME_RUNNING)
+
static void perf_output_read(struct perf_output_handle *handle,
struct perf_event *event)
{
+ u64 enabled = 0, running = 0, now, ctx_time;
+ u64 read_format = event->attr.read_format;
+
+ /*
+ * compute total_time_enabled, total_time_running
+ * based on snapshot values taken when the event
+ * was last scheduled in.
+ *
+ * we cannot simply called update_context_time()
+ * because of locking issue as we are called in
+ * NMI context
+ */
+ if (read_format & PERF_FORMAT_TOTAL_TIMES) {
+ now = perf_clock();
+ ctx_time = event->shadow_ctx_time + now;
+ enabled = ctx_time - event->tstamp_enabled;
+ running = ctx_time - event->tstamp_running;
+ }
+
if (event->attr.read_format & PERF_FORMAT_GROUP)
- perf_output_read_group(handle, event);
+ perf_output_read_group(handle, event, enabled, running);
else
- perf_output_read_one(handle, event);
+ perf_output_read_one(handle, event, enabled, running);
}
void perf_output_sample(struct perf_output_handle *handle,
{
u64 sample_type = event->attr.sample_type;
- data->type = sample_type;
-
header->type = PERF_RECORD_SAMPLE;
- header->size = sizeof(*header);
+ header->size = sizeof(*header) + event->header_size;
header->misc = 0;
header->misc |= perf_misc_flags(regs);
- if (sample_type & PERF_SAMPLE_IP) {
+ __perf_event_header__init_id(header, data, event);
+
+ if (sample_type & PERF_SAMPLE_IP)
data->ip = perf_instruction_pointer(regs);
- header->size += sizeof(data->ip);
- }
+ if (sample_type & PERF_SAMPLE_CALLCHAIN) {
+ int size = 1;
- if (sample_type & PERF_SAMPLE_TID) {
- /* namespace issues */
- data->tid_entry.pid = perf_event_pid(event, current);
- data->tid_entry.tid = perf_event_tid(event, current);
-
- header->size += sizeof(data->tid_entry);
- }
-
- if (sample_type & PERF_SAMPLE_TIME) {
- data->time = perf_clock();
-
- header->size += sizeof(data->time);
- }
-
- if (sample_type & PERF_SAMPLE_ADDR)
- header->size += sizeof(data->addr);
-
- if (sample_type & PERF_SAMPLE_ID) {
- data->id = primary_event_id(event);
-
- header->size += sizeof(data->id);
- }
-
- if (sample_type & PERF_SAMPLE_STREAM_ID) {
- data->stream_id = event->id;
-
- header->size += sizeof(data->stream_id);
- }
-
- if (sample_type & PERF_SAMPLE_CPU) {
- data->cpu_entry.cpu = raw_smp_processor_id();
- data->cpu_entry.reserved = 0;
-
- header->size += sizeof(data->cpu_entry);
- }
-
- if (sample_type & PERF_SAMPLE_PERIOD)
- header->size += sizeof(data->period);
-
- if (sample_type & PERF_SAMPLE_READ)
- header->size += perf_event_read_size(event);
-
- if (sample_type & PERF_SAMPLE_CALLCHAIN) {
- int size = 1;
-
- data->callchain = perf_callchain(regs);
+ data->callchain = perf_callchain(regs);
if (data->callchain)
size += data->callchain->nr;
struct task_struct *task)
{
struct perf_output_handle handle;
+ struct perf_sample_data sample;
struct perf_read_event read_event = {
.header = {
.type = PERF_RECORD_READ,
.misc = 0,
- .size = sizeof(read_event) + perf_event_read_size(event),
+ .size = sizeof(read_event) + event->read_size,
},
.pid = perf_event_pid(event, task),
.tid = perf_event_tid(event, task),
};
int ret;
+ perf_event_header__init_id(&read_event.header, &sample, event);
ret = perf_output_begin(&handle, event, read_event.header.size, 0, 0);
if (ret)
return;
perf_output_put(&handle, read_event);
perf_output_read(&handle, event);
+ perf_event__output_id_sample(event, &handle, &sample);
perf_output_end(&handle);
}
struct perf_task_event *task_event)
{
struct perf_output_handle handle;
+ struct perf_sample_data sample;
struct task_struct *task = task_event->task;
- int size, ret;
+ int ret, size = task_event->event_id.header.size;
- size = task_event->event_id.header.size;
- ret = perf_output_begin(&handle, event, size, 0, 0);
+ perf_event_header__init_id(&task_event->event_id.header, &sample, event);
+ ret = perf_output_begin(&handle, event,
+ task_event->event_id.header.size, 0, 0);
if (ret)
- return;
+ goto out;
task_event->event_id.pid = perf_event_pid(event, task);
task_event->event_id.ppid = perf_event_pid(event, current);
perf_output_put(&handle, task_event->event_id);
+ perf_event__output_id_sample(event, &handle, &sample);
+
perf_output_end(&handle);
+out:
+ task_event->event_id.header.size = size;
}
static int perf_event_task_match(struct perf_event *event)
if (event->state < PERF_EVENT_STATE_INACTIVE)
return 0;
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
return 0;
if (event->attr.comm || event->attr.mmap ||
static void perf_event_task_event(struct perf_task_event *task_event)
{
struct perf_cpu_context *cpuctx;
- struct perf_event_context *ctx = task_event->task_ctx;
+ struct perf_event_context *ctx;
+ struct pmu *pmu;
+ int ctxn;
rcu_read_lock();
- cpuctx = &get_cpu_var(perf_cpu_context);
- perf_event_task_ctx(&cpuctx->ctx, task_event);
- if (!ctx)
- ctx = rcu_dereference(current->perf_event_ctxp);
- if (ctx)
- perf_event_task_ctx(ctx, task_event);
- put_cpu_var(perf_cpu_context);
+ list_for_each_entry_rcu(pmu, &pmus, entry) {
+ cpuctx = get_cpu_ptr(pmu->pmu_cpu_context);
+ if (cpuctx->active_pmu != pmu)
+ goto next;
+ perf_event_task_ctx(&cpuctx->ctx, task_event);
+
+ ctx = task_event->task_ctx;
+ if (!ctx) {
+ ctxn = pmu->task_ctx_nr;
+ if (ctxn < 0)
+ goto next;
+ ctx = rcu_dereference(current->perf_event_ctxp[ctxn]);
+ }
+ if (ctx)
+ perf_event_task_ctx(ctx, task_event);
+next:
+ put_cpu_ptr(pmu->pmu_cpu_context);
+ }
rcu_read_unlock();
}
struct perf_comm_event *comm_event)
{
struct perf_output_handle handle;
+ struct perf_sample_data sample;
int size = comm_event->event_id.header.size;
- int ret = perf_output_begin(&handle, event, size, 0, 0);
+ int ret;
+
+ perf_event_header__init_id(&comm_event->event_id.header, &sample, event);
+ ret = perf_output_begin(&handle, event,
+ comm_event->event_id.header.size, 0, 0);
if (ret)
- return;
+ goto out;
comm_event->event_id.pid = perf_event_pid(event, comm_event->task);
comm_event->event_id.tid = perf_event_tid(event, comm_event->task);
perf_output_put(&handle, comm_event->event_id);
perf_output_copy(&handle, comm_event->comm,
comm_event->comm_size);
+
+ perf_event__output_id_sample(event, &handle, &sample);
+
perf_output_end(&handle);
+out:
+ comm_event->event_id.header.size = size;
}
static int perf_event_comm_match(struct perf_event *event)
if (event->state < PERF_EVENT_STATE_INACTIVE)
return 0;
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
return 0;
if (event->attr.comm)
{
struct perf_cpu_context *cpuctx;
struct perf_event_context *ctx;
- unsigned int size;
char comm[TASK_COMM_LEN];
+ unsigned int size;
+ struct pmu *pmu;
+ int ctxn;
memset(comm, 0, sizeof(comm));
strlcpy(comm, comm_event->task->comm, sizeof(comm));
comm_event->comm_size = size;
comm_event->event_id.header.size = sizeof(comm_event->event_id) + size;
-
rcu_read_lock();
- cpuctx = &get_cpu_var(perf_cpu_context);
- perf_event_comm_ctx(&cpuctx->ctx, comm_event);
- ctx = rcu_dereference(current->perf_event_ctxp);
- if (ctx)
- perf_event_comm_ctx(ctx, comm_event);
- put_cpu_var(perf_cpu_context);
+ list_for_each_entry_rcu(pmu, &pmus, entry) {
+ cpuctx = get_cpu_ptr(pmu->pmu_cpu_context);
+ if (cpuctx->active_pmu != pmu)
+ goto next;
+ perf_event_comm_ctx(&cpuctx->ctx, comm_event);
+
+ ctxn = pmu->task_ctx_nr;
+ if (ctxn < 0)
+ goto next;
+
+ ctx = rcu_dereference(current->perf_event_ctxp[ctxn]);
+ if (ctx)
+ perf_event_comm_ctx(ctx, comm_event);
+next:
+ put_cpu_ptr(pmu->pmu_cpu_context);
+ }
rcu_read_unlock();
}
void perf_event_comm(struct task_struct *task)
{
struct perf_comm_event comm_event;
+ struct perf_event_context *ctx;
+ int ctxn;
+
+ for_each_task_context_nr(ctxn) {
+ ctx = task->perf_event_ctxp[ctxn];
+ if (!ctx)
+ continue;
- if (task->perf_event_ctxp)
- perf_event_enable_on_exec(task);
+ perf_event_enable_on_exec(ctx);
+ }
if (!atomic_read(&nr_comm_events))
return;
struct perf_mmap_event *mmap_event)
{
struct perf_output_handle handle;
+ struct perf_sample_data sample;
int size = mmap_event->event_id.header.size;
- int ret = perf_output_begin(&handle, event, size, 0, 0);
+ int ret;
+ perf_event_header__init_id(&mmap_event->event_id.header, &sample, event);
+ ret = perf_output_begin(&handle, event,
+ mmap_event->event_id.header.size, 0, 0);
if (ret)
- return;
+ goto out;
mmap_event->event_id.pid = perf_event_pid(event, current);
mmap_event->event_id.tid = perf_event_tid(event, current);
perf_output_put(&handle, mmap_event->event_id);
perf_output_copy(&handle, mmap_event->file_name,
mmap_event->file_size);
+
+ perf_event__output_id_sample(event, &handle, &sample);
+
perf_output_end(&handle);
+out:
+ mmap_event->event_id.header.size = size;
}
static int perf_event_mmap_match(struct perf_event *event,
if (event->state < PERF_EVENT_STATE_INACTIVE)
return 0;
- if (event->cpu != -1 && event->cpu != smp_processor_id())
+ if (!event_filter_match(event))
return 0;
if ((!executable && event->attr.mmap_data) ||
char tmp[16];
char *buf = NULL;
const char *name;
+ struct pmu *pmu;
+ int ctxn;
memset(tmp, 0, sizeof(tmp));
mmap_event->event_id.header.size = sizeof(mmap_event->event_id) + size;
rcu_read_lock();
- cpuctx = &get_cpu_var(perf_cpu_context);
- perf_event_mmap_ctx(&cpuctx->ctx, mmap_event, vma->vm_flags & VM_EXEC);
- ctx = rcu_dereference(current->perf_event_ctxp);
- if (ctx)
- perf_event_mmap_ctx(ctx, mmap_event, vma->vm_flags & VM_EXEC);
- put_cpu_var(perf_cpu_context);
+ list_for_each_entry_rcu(pmu, &pmus, entry) {
+ cpuctx = get_cpu_ptr(pmu->pmu_cpu_context);
+ if (cpuctx->active_pmu != pmu)
+ goto next;
+ perf_event_mmap_ctx(&cpuctx->ctx, mmap_event,
+ vma->vm_flags & VM_EXEC);
+
+ ctxn = pmu->task_ctx_nr;
+ if (ctxn < 0)
+ goto next;
+
+ ctx = rcu_dereference(current->perf_event_ctxp[ctxn]);
+ if (ctx) {
+ perf_event_mmap_ctx(ctx, mmap_event,
+ vma->vm_flags & VM_EXEC);
+ }
+next:
+ put_cpu_ptr(pmu->pmu_cpu_context);
+ }
rcu_read_unlock();
kfree(buf);
static void perf_log_throttle(struct perf_event *event, int enable)
{
struct perf_output_handle handle;
+ struct perf_sample_data sample;
int ret;
struct {
if (enable)
throttle_event.header.type = PERF_RECORD_UNTHROTTLE;
- ret = perf_output_begin(&handle, event, sizeof(throttle_event), 1, 0);
+ perf_event_header__init_id(&throttle_event.header, &sample, event);
+
+ ret = perf_output_begin(&handle, event,
+ throttle_event.header.size, 1, 0);
if (ret)
return;
perf_output_put(&handle, throttle_event);
+ perf_event__output_id_sample(event, &handle, &sample);
perf_output_end(&handle);
}
struct hw_perf_event *hwc = &event->hw;
int ret = 0;
+ /*
+ * Non-sampling counters might still use the PMI to fold short
+ * hardware counters, ignore those.
+ */
+ if (unlikely(!is_sampling_event(event)))
+ return 0;
+
if (!throttle) {
hwc->interrupts++;
} else {
if (events && atomic_dec_and_test(&event->event_limit)) {
ret = 1;
event->pending_kill = POLL_HUP;
- if (nmi) {
- event->pending_disable = 1;
- perf_pending_queue(&event->pending,
- perf_pending_event);
- } else
- perf_event_disable(event);
+ event->pending_disable = 1;
+ irq_work_queue(&event->pending);
}
if (event->overflow_handler)
if (!regs)
return;
- if (!hwc->sample_period)
+ if (!is_sampling_event(event))
return;
if (nr == 1 && hwc->sample_period == 1 && !event->attr.freq)
struct pt_regs *regs)
{
if (event->hw.state & PERF_HES_STOPPED)
- return 0;
+ return 1;
if (regs) {
if (event->attr.exclude_user && user_mode(regs))
}
EXPORT_SYMBOL_GPL(perf_swevent_get_recursion_context);
-void inline perf_swevent_put_recursion_context(int rctx)
+inline void perf_swevent_put_recursion_context(int rctx)
{
struct swevent_htable *swhash = &__get_cpu_var(swevent_htable);
struct hw_perf_event *hwc = &event->hw;
struct hlist_head *head;
- if (hwc->sample_period) {
+ if (is_sampling_event(event)) {
hwc->last_period = hwc->sample_period;
perf_swevent_set_period(event);
}
WARN_ON(event->parent);
- atomic_dec(&perf_swevent_enabled[event_id]);
+ jump_label_dec(&perf_swevent_enabled[event_id]);
swevent_hlist_put(event);
}
break;
}
- if (event_id > PERF_COUNT_SW_MAX)
+ if (event_id >= PERF_COUNT_SW_MAX)
return -ENOENT;
if (!event->parent) {
if (err)
return err;
- atomic_inc(&perf_swevent_enabled[event_id]);
+ jump_label_inc(&perf_swevent_enabled[event_id]);
event->destroy = sw_perf_event_destroy;
}
}
static struct pmu perf_swevent = {
+ .task_ctx_nr = perf_sw_context,
+
.event_init = perf_swevent_init,
.add = perf_swevent_add,
.del = perf_swevent_del,
struct perf_sample_data *data,
struct pt_regs *regs)
{
+ if (event->hw.state & PERF_HES_STOPPED)
+ return 0;
/*
* All tracepoints are from kernel-space.
*/
if (event->attr.type != PERF_TYPE_TRACEPOINT)
return -ENOENT;
- /*
- * Raw tracepoint data is a severe data leak, only allow root to
- * have these.
- */
- if ((event->attr.sample_type & PERF_SAMPLE_RAW) &&
- perf_paranoid_tracepoint_raw() &&
- !capable(CAP_SYS_ADMIN))
- return -EPERM;
-
err = perf_trace_init(event);
if (err)
return err;
}
static struct pmu perf_tracepoint = {
+ .task_ctx_nr = perf_sw_context,
+
.event_init = perf_tp_event_init,
.add = perf_trace_add,
.del = perf_trace_del,
static inline void perf_tp_register(void)
{
- perf_pmu_register(&perf_tracepoint);
+ perf_pmu_register(&perf_tracepoint, "tracepoint", PERF_TYPE_TRACEPOINT);
}
static int perf_event_set_filter(struct perf_event *event, void __user *arg)
static void perf_swevent_start_hrtimer(struct perf_event *event)
{
struct hw_perf_event *hwc = &event->hw;
+ s64 period;
+
+ if (!is_sampling_event(event))
+ return;
hrtimer_init(&hwc->hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
hwc->hrtimer.function = perf_swevent_hrtimer;
- if (hwc->sample_period) {
- s64 period = local64_read(&hwc->period_left);
- if (period) {
- if (period < 0)
- period = 10000;
+ period = local64_read(&hwc->period_left);
+ if (period) {
+ if (period < 0)
+ period = 10000;
- local64_set(&hwc->period_left, 0);
- } else {
- period = max_t(u64, 10000, hwc->sample_period);
- }
- __hrtimer_start_range_ns(&hwc->hrtimer,
+ local64_set(&hwc->period_left, 0);
+ } else {
+ period = max_t(u64, 10000, hwc->sample_period);
+ }
+ __hrtimer_start_range_ns(&hwc->hrtimer,
ns_to_ktime(period), 0,
HRTIMER_MODE_REL_PINNED, 0);
- }
}
static void perf_swevent_cancel_hrtimer(struct perf_event *event)
{
struct hw_perf_event *hwc = &event->hw;
- if (hwc->sample_period) {
+ if (is_sampling_event(event)) {
ktime_t remaining = hrtimer_get_remaining(&hwc->hrtimer);
local64_set(&hwc->period_left, ktime_to_ns(remaining));
}
static struct pmu perf_cpu_clock = {
+ .task_ctx_nr = perf_sw_context,
+
.event_init = cpu_clock_event_init,
.add = cpu_clock_event_add,
.del = cpu_clock_event_del,
}
static struct pmu perf_task_clock = {
+ .task_ctx_nr = perf_sw_context,
+
.event_init = task_clock_event_init,
.add = task_clock_event_add,
.del = task_clock_event_del,
.read = task_clock_event_read,
};
-static LIST_HEAD(pmus);
-static DEFINE_MUTEX(pmus_lock);
-static struct srcu_struct pmus_srcu;
-
static void perf_pmu_nop_void(struct pmu *pmu)
{
}
perf_pmu_enable(pmu);
}
-int perf_pmu_register(struct pmu *pmu)
+/*
+ * Ensures all contexts with the same task_ctx_nr have the same
+ * pmu_cpu_context too.
+ */
+static void *find_pmu_context(int ctxn)
{
- int ret;
+ struct pmu *pmu;
+
+ if (ctxn < 0)
+ return NULL;
+
+ list_for_each_entry(pmu, &pmus, entry) {
+ if (pmu->task_ctx_nr == ctxn)
+ return pmu->pmu_cpu_context;
+ }
+
+ return NULL;
+}
+
+static void update_pmu_context(struct pmu *pmu, struct pmu *old_pmu)
+{
+ int cpu;
+
+ for_each_possible_cpu(cpu) {
+ struct perf_cpu_context *cpuctx;
+
+ cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu);
+
+ if (cpuctx->active_pmu == old_pmu)
+ cpuctx->active_pmu = pmu;
+ }
+}
+
+static void free_pmu_context(struct pmu *pmu)
+{
+ struct pmu *i;
+
+ mutex_lock(&pmus_lock);
+ /*
+ * Like a real lame refcount.
+ */
+ list_for_each_entry(i, &pmus, entry) {
+ if (i->pmu_cpu_context == pmu->pmu_cpu_context) {
+ update_pmu_context(i, pmu);
+ goto out;
+ }
+ }
+
+ free_percpu(pmu->pmu_cpu_context);
+out:
+ mutex_unlock(&pmus_lock);
+}
+static struct idr pmu_idr;
+
+static ssize_t
+type_show(struct device *dev, struct device_attribute *attr, char *page)
+{
+ struct pmu *pmu = dev_get_drvdata(dev);
+
+ return snprintf(page, PAGE_SIZE-1, "%d\n", pmu->type);
+}
+
+static struct device_attribute pmu_dev_attrs[] = {
+ __ATTR_RO(type),
+ __ATTR_NULL,
+};
+
+static int pmu_bus_running;
+static struct bus_type pmu_bus = {
+ .name = "event_source",
+ .dev_attrs = pmu_dev_attrs,
+};
+
+static void pmu_dev_release(struct device *dev)
+{
+ kfree(dev);
+}
+
+static int pmu_dev_alloc(struct pmu *pmu)
+{
+ int ret = -ENOMEM;
+
+ pmu->dev = kzalloc(sizeof(struct device), GFP_KERNEL);
+ if (!pmu->dev)
+ goto out;
+
+ device_initialize(pmu->dev);
+ ret = dev_set_name(pmu->dev, "%s", pmu->name);
+ if (ret)
+ goto free_dev;
+
+ dev_set_drvdata(pmu->dev, pmu);
+ pmu->dev->bus = &pmu_bus;
+ pmu->dev->release = pmu_dev_release;
+ ret = device_add(pmu->dev);
+ if (ret)
+ goto free_dev;
+
+out:
+ return ret;
+
+free_dev:
+ put_device(pmu->dev);
+ goto out;
+}
+
+static struct lock_class_key cpuctx_mutex;
+
+int perf_pmu_register(struct pmu *pmu, char *name, int type)
+{
+ int cpu, ret;
mutex_lock(&pmus_lock);
ret = -ENOMEM;
if (!pmu->pmu_disable_count)
goto unlock;
+ pmu->type = -1;
+ if (!name)
+ goto skip_type;
+ pmu->name = name;
+
+ if (type < 0) {
+ int err = idr_pre_get(&pmu_idr, GFP_KERNEL);
+ if (!err)
+ goto free_pdc;
+
+ err = idr_get_new_above(&pmu_idr, pmu, PERF_TYPE_MAX, &type);
+ if (err) {
+ ret = err;
+ goto free_pdc;
+ }
+ }
+ pmu->type = type;
+
+ if (pmu_bus_running) {
+ ret = pmu_dev_alloc(pmu);
+ if (ret)
+ goto free_idr;
+ }
+
+skip_type:
+ pmu->pmu_cpu_context = find_pmu_context(pmu->task_ctx_nr);
+ if (pmu->pmu_cpu_context)
+ goto got_cpu_context;
+
+ pmu->pmu_cpu_context = alloc_percpu(struct perf_cpu_context);
+ if (!pmu->pmu_cpu_context)
+ goto free_dev;
+
+ for_each_possible_cpu(cpu) {
+ struct perf_cpu_context *cpuctx;
+
+ cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu);
+ __perf_event_init_context(&cpuctx->ctx);
+ lockdep_set_class(&cpuctx->ctx.mutex, &cpuctx_mutex);
+ cpuctx->ctx.type = cpu_context;
+ cpuctx->ctx.pmu = pmu;
+ cpuctx->jiffies_interval = 1;
+ INIT_LIST_HEAD(&cpuctx->rotation_list);
+ cpuctx->active_pmu = pmu;
+ }
+
+got_cpu_context:
if (!pmu->start_txn) {
if (pmu->pmu_enable) {
/*
mutex_unlock(&pmus_lock);
return ret;
+
+free_dev:
+ device_del(pmu->dev);
+ put_device(pmu->dev);
+
+free_idr:
+ if (pmu->type >= PERF_TYPE_MAX)
+ idr_remove(&pmu_idr, pmu->type);
+
+free_pdc:
+ free_percpu(pmu->pmu_disable_count);
+ goto unlock;
}
void perf_pmu_unregister(struct pmu *pmu)
list_del_rcu(&pmu->entry);
mutex_unlock(&pmus_lock);
+ /*
+ * We dereference the pmu list under both SRCU and regular RCU, so
+ * synchronize against both of those.
+ */
synchronize_srcu(&pmus_srcu);
+ synchronize_rcu();
free_percpu(pmu->pmu_disable_count);
+ if (pmu->type >= PERF_TYPE_MAX)
+ idr_remove(&pmu_idr, pmu->type);
+ device_del(pmu->dev);
+ put_device(pmu->dev);
+ free_pmu_context(pmu);
}
struct pmu *perf_init_event(struct perf_event *event)
int idx;
idx = srcu_read_lock(&pmus_srcu);
+
+ rcu_read_lock();
+ pmu = idr_find(&pmu_idr, event->attr.type);
+ rcu_read_unlock();
+ if (pmu)
+ goto unlock;
+
list_for_each_entry_rcu(pmu, &pmus, entry) {
int ret = pmu->event_init(event);
if (!ret)
- break;
+ goto unlock;
+
if (ret != -ENOENT) {
pmu = ERR_PTR(ret);
- break;
+ goto unlock;
}
}
+ pmu = ERR_PTR(-ENOENT);
+unlock:
srcu_read_unlock(&pmus_srcu, idx);
return pmu;
*/
static struct perf_event *
perf_event_alloc(struct perf_event_attr *attr, int cpu,
- struct perf_event *group_leader,
- struct perf_event *parent_event,
- perf_overflow_handler_t overflow_handler)
+ struct task_struct *task,
+ struct perf_event *group_leader,
+ struct perf_event *parent_event,
+ perf_overflow_handler_t overflow_handler)
{
struct pmu *pmu;
struct perf_event *event;
struct hw_perf_event *hwc;
long err;
+ if ((unsigned)cpu >= nr_cpu_ids) {
+ if (!task || cpu != -1)
+ return ERR_PTR(-EINVAL);
+ }
+
event = kzalloc(sizeof(*event), GFP_KERNEL);
if (!event)
return ERR_PTR(-ENOMEM);
INIT_LIST_HEAD(&event->event_entry);
INIT_LIST_HEAD(&event->sibling_list);
init_waitqueue_head(&event->waitq);
+ init_irq_work(&event->pending, perf_pending_event);
mutex_init(&event->mmap_mutex);
event->state = PERF_EVENT_STATE_INACTIVE;
+ if (task) {
+ event->attach_state = PERF_ATTACH_TASK;
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+ /*
+ * hw_breakpoint is a bit difficult here..
+ */
+ if (attr->type == PERF_TYPE_BREAKPOINT)
+ event->hw.bp_target = task;
+#endif
+ }
+
if (!overflow_handler && parent_event)
overflow_handler = parent_event->overflow_handler;
-
+
event->overflow_handler = overflow_handler;
if (attr->disabled)
event->pmu = pmu;
if (!event->parent) {
- atomic_inc(&nr_events);
+ if (event->attach_state & PERF_ATTACH_TASK)
+ jump_label_inc(&perf_task_events);
if (event->attr.mmap || event->attr.mmap_data)
atomic_inc(&nr_mmap_events);
if (event->attr.comm)
struct perf_event_attr __user *, attr_uptr,
pid_t, pid, int, cpu, int, group_fd, unsigned long, flags)
{
- struct perf_event *event, *group_leader = NULL, *output_event = NULL;
+ struct perf_event *group_leader = NULL, *output_event = NULL;
+ struct perf_event *event, *sibling;
struct perf_event_attr attr;
struct perf_event_context *ctx;
struct file *event_file = NULL;
struct file *group_file = NULL;
+ struct task_struct *task = NULL;
+ struct pmu *pmu;
int event_fd;
+ int move_group = 0;
int fput_needed = 0;
int err;
if (event_fd < 0)
return event_fd;
- event = perf_event_alloc(&attr, cpu, group_leader, NULL, NULL);
+ if (group_fd != -1) {
+ group_leader = perf_fget_light(group_fd, &fput_needed);
+ if (IS_ERR(group_leader)) {
+ err = PTR_ERR(group_leader);
+ goto err_fd;
+ }
+ group_file = group_leader->filp;
+ if (flags & PERF_FLAG_FD_OUTPUT)
+ output_event = group_leader;
+ if (flags & PERF_FLAG_FD_NO_GROUP)
+ group_leader = NULL;
+ }
+
+ if (pid != -1) {
+ task = find_lively_task_by_vpid(pid);
+ if (IS_ERR(task)) {
+ err = PTR_ERR(task);
+ goto err_group_fd;
+ }
+ }
+
+ event = perf_event_alloc(&attr, cpu, task, group_leader, NULL, NULL);
if (IS_ERR(event)) {
err = PTR_ERR(event);
- goto err_fd;
+ goto err_task;
+ }
+
+ /*
+ * Special case software events and allow them to be part of
+ * any hardware group.
+ */
+ pmu = event->pmu;
+
+ if (group_leader &&
+ (is_software_event(event) != is_software_event(group_leader))) {
+ if (is_software_event(event)) {
+ /*
+ * If event and group_leader are not both a software
+ * event, and event is, then group leader is not.
+ *
+ * Allow the addition of software events to !software
+ * groups, this is safe because software events never
+ * fail to schedule.
+ */
+ pmu = group_leader->pmu;
+ } else if (is_software_event(group_leader) &&
+ (group_leader->group_flags & PERF_GROUP_SOFTWARE)) {
+ /*
+ * In case the group is a pure software group, and we
+ * try to add a hardware event, move the whole group to
+ * the hardware context.
+ */
+ move_group = 1;
+ }
}
/*
* Get the target context (task or percpu):
*/
- ctx = find_get_context(pid, cpu);
+ ctx = find_get_context(pmu, task, cpu);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
goto err_alloc;
}
- if (group_fd != -1) {
- group_leader = perf_fget_light(group_fd, &fput_needed);
- if (IS_ERR(group_leader)) {
- err = PTR_ERR(group_leader);
- goto err_context;
- }
- group_file = group_leader->filp;
- if (flags & PERF_FLAG_FD_OUTPUT)
- output_event = group_leader;
- if (flags & PERF_FLAG_FD_NO_GROUP)
- group_leader = NULL;
+ if (task) {
+ put_task_struct(task);
+ task = NULL;
}
/*
* Do not allow to attach to a group in a different
* task or CPU context:
*/
- if (group_leader->ctx != ctx)
- goto err_context;
+ if (move_group) {
+ if (group_leader->ctx->type != ctx->type)
+ goto err_context;
+ } else {
+ if (group_leader->ctx != ctx)
+ goto err_context;
+ }
+
/*
* Only a group leader can be exclusive or pinned
*/
goto err_context;
}
+ if (move_group) {
+ struct perf_event_context *gctx = group_leader->ctx;
+
+ mutex_lock(&gctx->mutex);
+ perf_event_remove_from_context(group_leader);
+ list_for_each_entry(sibling, &group_leader->sibling_list,
+ group_entry) {
+ perf_event_remove_from_context(sibling);
+ put_ctx(gctx);
+ }
+ mutex_unlock(&gctx->mutex);
+ put_ctx(gctx);
+ }
+
event->filp = event_file;
WARN_ON_ONCE(ctx->parent_ctx);
mutex_lock(&ctx->mutex);
+
+ if (move_group) {
+ perf_install_in_context(ctx, group_leader, cpu);
+ get_ctx(ctx);
+ list_for_each_entry(sibling, &group_leader->sibling_list,
+ group_entry) {
+ perf_install_in_context(ctx, sibling, cpu);
+ get_ctx(ctx);
+ }
+ }
+
perf_install_in_context(ctx, event, cpu);
++ctx->generation;
mutex_unlock(&ctx->mutex);
event->owner = current;
- get_task_struct(current);
+
mutex_lock(¤t->perf_event_mutex);
list_add_tail(&event->owner_entry, ¤t->perf_event_list);
mutex_unlock(¤t->perf_event_mutex);
/*
+ * Precalculate sample_data sizes
+ */
+ perf_event__header_size(event);
+ perf_event__id_header_size(event);
+
+ /*
* Drop the reference on the group_event after placing the
* new event on the sibling_list. This ensures destruction
* of the group leader will find the pointer to itself in
return event_fd;
err_context:
- fput_light(group_file, fput_needed);
put_ctx(ctx);
err_alloc:
free_event(event);
+err_task:
+ if (task)
+ put_task_struct(task);
+err_group_fd:
+ fput_light(group_file, fput_needed);
err_fd:
put_unused_fd(event_fd);
return err;
*
* @attr: attributes of the counter to create
* @cpu: cpu in which the counter is bound
- * @pid: task to profile
+ * @task: task to profile (NULL for percpu)
*/
struct perf_event *
perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu,
- pid_t pid,
+ struct task_struct *task,
perf_overflow_handler_t overflow_handler)
{
struct perf_event_context *ctx;
* Get the target context (task or percpu):
*/
- event = perf_event_alloc(attr, cpu, NULL, NULL, overflow_handler);
+ event = perf_event_alloc(attr, cpu, task, NULL, NULL, overflow_handler);
if (IS_ERR(event)) {
err = PTR_ERR(event);
goto err;
}
- ctx = find_get_context(pid, cpu);
+ ctx = find_get_context(event->pmu, task, cpu);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
goto err_free;
++ctx->generation;
mutex_unlock(&ctx->mutex);
- event->owner = current;
- get_task_struct(current);
- mutex_lock(¤t->perf_event_mutex);
- list_add_tail(&event->owner_entry, ¤t->perf_event_list);
- mutex_unlock(¤t->perf_event_mutex);
-
return event;
err_free:
}
EXPORT_SYMBOL_GPL(perf_event_create_kernel_counter);
-/*
- * inherit a event from parent task to child task:
- */
-static struct perf_event *
-inherit_event(struct perf_event *parent_event,
- struct task_struct *parent,
- struct perf_event_context *parent_ctx,
- struct task_struct *child,
- struct perf_event *group_leader,
- struct perf_event_context *child_ctx)
-{
- struct perf_event *child_event;
-
- /*
- * Instead of creating recursive hierarchies of events,
- * we link inherited events back to the original parent,
- * which has a filp for sure, which we use as the reference
- * count:
- */
- if (parent_event->parent)
- parent_event = parent_event->parent;
-
- child_event = perf_event_alloc(&parent_event->attr,
- parent_event->cpu,
- group_leader, parent_event,
- NULL);
- if (IS_ERR(child_event))
- return child_event;
- get_ctx(child_ctx);
-
- /*
- * Make the child state follow the state of the parent event,
- * not its attr.disabled bit. We hold the parent's mutex,
- * so we won't race with perf_event_{en, dis}able_family.
- */
- if (parent_event->state >= PERF_EVENT_STATE_INACTIVE)
- child_event->state = PERF_EVENT_STATE_INACTIVE;
- else
- child_event->state = PERF_EVENT_STATE_OFF;
-
- if (parent_event->attr.freq) {
- u64 sample_period = parent_event->hw.sample_period;
- struct hw_perf_event *hwc = &child_event->hw;
-
- hwc->sample_period = sample_period;
- hwc->last_period = sample_period;
-
- local64_set(&hwc->period_left, sample_period);
- }
-
- child_event->ctx = child_ctx;
- child_event->overflow_handler = parent_event->overflow_handler;
-
- /*
- * Link it up in the child's context:
- */
- add_event_to_ctx(child_event, child_ctx);
-
- /*
- * Get a reference to the parent filp - we will fput it
- * when the child event exits. This is safe to do because
- * we are in the parent and we know that the filp still
- * exists and has a nonzero count:
- */
- atomic_long_inc(&parent_event->filp->f_count);
-
- /*
- * Link this into the parent event's child list
- */
- WARN_ON_ONCE(parent_event->ctx->parent_ctx);
- mutex_lock(&parent_event->child_mutex);
- list_add_tail(&child_event->child_list, &parent_event->child_list);
- mutex_unlock(&parent_event->child_mutex);
-
- return child_event;
-}
-
-static int inherit_group(struct perf_event *parent_event,
- struct task_struct *parent,
- struct perf_event_context *parent_ctx,
- struct task_struct *child,
- struct perf_event_context *child_ctx)
-{
- struct perf_event *leader;
- struct perf_event *sub;
- struct perf_event *child_ctr;
-
- leader = inherit_event(parent_event, parent, parent_ctx,
- child, NULL, child_ctx);
- if (IS_ERR(leader))
- return PTR_ERR(leader);
- list_for_each_entry(sub, &parent_event->sibling_list, group_entry) {
- child_ctr = inherit_event(sub, parent, parent_ctx,
- child, leader, child_ctx);
- if (IS_ERR(child_ctr))
- return PTR_ERR(child_ctr);
- }
- return 0;
-}
-
static void sync_child_event(struct perf_event *child_event,
struct task_struct *child)
{
struct perf_event_context *child_ctx,
struct task_struct *child)
{
- struct perf_event *parent_event;
+ if (child_event->parent) {
+ raw_spin_lock_irq(&child_ctx->lock);
+ perf_group_detach(child_event);
+ raw_spin_unlock_irq(&child_ctx->lock);
+ }
perf_event_remove_from_context(child_event);
- parent_event = child_event->parent;
/*
- * It can happen that parent exits first, and has events
+ * It can happen that the parent exits first, and has events
* that are still around due to the child reference. These
- * events need to be zapped - but otherwise linger.
+ * events need to be zapped.
*/
- if (parent_event) {
+ if (child_event->parent) {
sync_child_event(child_event, child);
free_event(child_event);
}
}
-/*
- * When a child task exits, feed back event values to parent events.
- */
-void perf_event_exit_task(struct task_struct *child)
+static void perf_event_exit_task_context(struct task_struct *child, int ctxn)
{
struct perf_event *child_event, *tmp;
struct perf_event_context *child_ctx;
unsigned long flags;
- if (likely(!child->perf_event_ctxp)) {
+ if (likely(!child->perf_event_ctxp[ctxn])) {
perf_event_task(child, NULL, 0);
return;
}
* scheduled, so we are now safe from rescheduling changing
* our context.
*/
- child_ctx = child->perf_event_ctxp;
- __perf_event_task_sched_out(child_ctx);
+ child_ctx = rcu_dereference_raw(child->perf_event_ctxp[ctxn]);
+ task_ctx_sched_out(child_ctx, EVENT_ALL);
/*
* Take the context lock here so that if find_get_context is
* incremented the context's refcount before we do put_ctx below.
*/
raw_spin_lock(&child_ctx->lock);
- child->perf_event_ctxp = NULL;
+ child->perf_event_ctxp[ctxn] = NULL;
/*
* If this context is a clone; unclone it so it can't get
* swapped to another process while we're removing all
put_ctx(child_ctx);
}
+/*
+ * When a child task exits, feed back event values to parent events.
+ */
+void perf_event_exit_task(struct task_struct *child)
+{
+ struct perf_event *event, *tmp;
+ int ctxn;
+
+ mutex_lock(&child->perf_event_mutex);
+ list_for_each_entry_safe(event, tmp, &child->perf_event_list,
+ owner_entry) {
+ list_del_init(&event->owner_entry);
+
+ /*
+ * Ensure the list deletion is visible before we clear
+ * the owner, closes a race against perf_release() where
+ * we need to serialize on the owner->perf_event_mutex.
+ */
+ smp_wmb();
+ event->owner = NULL;
+ }
+ mutex_unlock(&child->perf_event_mutex);
+
+ for_each_task_context_nr(ctxn)
+ perf_event_exit_task_context(child, ctxn);
+}
+
static void perf_free_event(struct perf_event *event,
struct perf_event_context *ctx)
{
/*
* free an unexposed, unused context as created by inheritance by
- * init_task below, used by fork() in case of fail.
+ * perf_event_init_task below, used by fork() in case of fail.
*/
void perf_event_free_task(struct task_struct *task)
{
- struct perf_event_context *ctx = task->perf_event_ctxp;
+ struct perf_event_context *ctx;
struct perf_event *event, *tmp;
+ int ctxn;
- if (!ctx)
- return;
+ for_each_task_context_nr(ctxn) {
+ ctx = task->perf_event_ctxp[ctxn];
+ if (!ctx)
+ continue;
- mutex_lock(&ctx->mutex);
+ mutex_lock(&ctx->mutex);
again:
- list_for_each_entry_safe(event, tmp, &ctx->pinned_groups, group_entry)
- perf_free_event(event, ctx);
+ list_for_each_entry_safe(event, tmp, &ctx->pinned_groups,
+ group_entry)
+ perf_free_event(event, ctx);
- list_for_each_entry_safe(event, tmp, &ctx->flexible_groups,
- group_entry)
- perf_free_event(event, ctx);
+ list_for_each_entry_safe(event, tmp, &ctx->flexible_groups,
+ group_entry)
+ perf_free_event(event, ctx);
- if (!list_empty(&ctx->pinned_groups) ||
- !list_empty(&ctx->flexible_groups))
- goto again;
+ if (!list_empty(&ctx->pinned_groups) ||
+ !list_empty(&ctx->flexible_groups))
+ goto again;
- mutex_unlock(&ctx->mutex);
+ mutex_unlock(&ctx->mutex);
- put_ctx(ctx);
+ put_ctx(ctx);
+ }
+}
+
+void perf_event_delayed_put(struct task_struct *task)
+{
+ int ctxn;
+
+ for_each_task_context_nr(ctxn)
+ WARN_ON_ONCE(task->perf_event_ctxp[ctxn]);
+}
+
+/*
+ * inherit a event from parent task to child task:
+ */
+static struct perf_event *
+inherit_event(struct perf_event *parent_event,
+ struct task_struct *parent,
+ struct perf_event_context *parent_ctx,
+ struct task_struct *child,
+ struct perf_event *group_leader,
+ struct perf_event_context *child_ctx)
+{
+ struct perf_event *child_event;
+ unsigned long flags;
+
+ /*
+ * Instead of creating recursive hierarchies of events,
+ * we link inherited events back to the original parent,
+ * which has a filp for sure, which we use as the reference
+ * count:
+ */
+ if (parent_event->parent)
+ parent_event = parent_event->parent;
+
+ child_event = perf_event_alloc(&parent_event->attr,
+ parent_event->cpu,
+ child,
+ group_leader, parent_event,
+ NULL);
+ if (IS_ERR(child_event))
+ return child_event;
+ get_ctx(child_ctx);
+
+ /*
+ * Make the child state follow the state of the parent event,
+ * not its attr.disabled bit. We hold the parent's mutex,
+ * so we won't race with perf_event_{en, dis}able_family.
+ */
+ if (parent_event->state >= PERF_EVENT_STATE_INACTIVE)
+ child_event->state = PERF_EVENT_STATE_INACTIVE;
+ else
+ child_event->state = PERF_EVENT_STATE_OFF;
+
+ if (parent_event->attr.freq) {
+ u64 sample_period = parent_event->hw.sample_period;
+ struct hw_perf_event *hwc = &child_event->hw;
+
+ hwc->sample_period = sample_period;
+ hwc->last_period = sample_period;
+
+ local64_set(&hwc->period_left, sample_period);
+ }
+
+ child_event->ctx = child_ctx;
+ child_event->overflow_handler = parent_event->overflow_handler;
+
+ /*
+ * Precalculate sample_data sizes
+ */
+ perf_event__header_size(child_event);
+ perf_event__id_header_size(child_event);
+
+ /*
+ * Link it up in the child's context:
+ */
+ raw_spin_lock_irqsave(&child_ctx->lock, flags);
+ add_event_to_ctx(child_event, child_ctx);
+ raw_spin_unlock_irqrestore(&child_ctx->lock, flags);
+
+ /*
+ * Get a reference to the parent filp - we will fput it
+ * when the child event exits. This is safe to do because
+ * we are in the parent and we know that the filp still
+ * exists and has a nonzero count:
+ */
+ atomic_long_inc(&parent_event->filp->f_count);
+
+ /*
+ * Link this into the parent event's child list
+ */
+ WARN_ON_ONCE(parent_event->ctx->parent_ctx);
+ mutex_lock(&parent_event->child_mutex);
+ list_add_tail(&child_event->child_list, &parent_event->child_list);
+ mutex_unlock(&parent_event->child_mutex);
+
+ return child_event;
+}
+
+static int inherit_group(struct perf_event *parent_event,
+ struct task_struct *parent,
+ struct perf_event_context *parent_ctx,
+ struct task_struct *child,
+ struct perf_event_context *child_ctx)
+{
+ struct perf_event *leader;
+ struct perf_event *sub;
+ struct perf_event *child_ctr;
+
+ leader = inherit_event(parent_event, parent, parent_ctx,
+ child, NULL, child_ctx);
+ if (IS_ERR(leader))
+ return PTR_ERR(leader);
+ list_for_each_entry(sub, &parent_event->sibling_list, group_entry) {
+ child_ctr = inherit_event(sub, parent, parent_ctx,
+ child, leader, child_ctx);
+ if (IS_ERR(child_ctr))
+ return PTR_ERR(child_ctr);
+ }
+ return 0;
}
static int
inherit_task_group(struct perf_event *event, struct task_struct *parent,
struct perf_event_context *parent_ctx,
- struct task_struct *child,
+ struct task_struct *child, int ctxn,
int *inherited_all)
{
int ret;
- struct perf_event_context *child_ctx = child->perf_event_ctxp;
+ struct perf_event_context *child_ctx;
if (!event->attr.inherit) {
*inherited_all = 0;
return 0;
}
+ child_ctx = child->perf_event_ctxp[ctxn];
if (!child_ctx) {
/*
* This is executed from the parent task context, so
* child.
*/
- child_ctx = kzalloc(sizeof(struct perf_event_context),
- GFP_KERNEL);
+ child_ctx = alloc_perf_context(event->pmu, child);
if (!child_ctx)
return -ENOMEM;
- __perf_event_init_context(child_ctx, child);
- child->perf_event_ctxp = child_ctx;
- get_task_struct(child);
+ child->perf_event_ctxp[ctxn] = child_ctx;
}
ret = inherit_group(event, parent, parent_ctx,
return ret;
}
-
/*
* Initialize the perf_event context in task_struct
*/
-int perf_event_init_task(struct task_struct *child)
+int perf_event_init_context(struct task_struct *child, int ctxn)
{
struct perf_event_context *child_ctx, *parent_ctx;
struct perf_event_context *cloned_ctx;
struct perf_event *event;
struct task_struct *parent = current;
int inherited_all = 1;
+ unsigned long flags;
int ret = 0;
- child->perf_event_ctxp = NULL;
-
- mutex_init(&child->perf_event_mutex);
- INIT_LIST_HEAD(&child->perf_event_list);
-
- if (likely(!parent->perf_event_ctxp))
+ if (likely(!parent->perf_event_ctxp[ctxn]))
return 0;
/*
* If the parent's context is a clone, pin it so it won't get
* swapped under us.
*/
- parent_ctx = perf_pin_task_context(parent);
+ parent_ctx = perf_pin_task_context(parent, ctxn);
/*
* No need to check if parent_ctx != NULL here; since we saw
* the list, not manipulating it:
*/
list_for_each_entry(event, &parent_ctx->pinned_groups, group_entry) {
- ret = inherit_task_group(event, parent, parent_ctx, child,
- &inherited_all);
+ ret = inherit_task_group(event, parent, parent_ctx,
+ child, ctxn, &inherited_all);
if (ret)
break;
}
+ /*
+ * We can't hold ctx->lock when iterating the ->flexible_group list due
+ * to allocations, but we need to prevent rotation because
+ * rotate_ctx() will change the list from interrupt context.
+ */
+ raw_spin_lock_irqsave(&parent_ctx->lock, flags);
+ parent_ctx->rotate_disable = 1;
+ raw_spin_unlock_irqrestore(&parent_ctx->lock, flags);
+
list_for_each_entry(event, &parent_ctx->flexible_groups, group_entry) {
- ret = inherit_task_group(event, parent, parent_ctx, child,
- &inherited_all);
+ ret = inherit_task_group(event, parent, parent_ctx,
+ child, ctxn, &inherited_all);
if (ret)
break;
}
- child_ctx = child->perf_event_ctxp;
+ raw_spin_lock_irqsave(&parent_ctx->lock, flags);
+ parent_ctx->rotate_disable = 0;
+
+ child_ctx = child->perf_event_ctxp[ctxn];
if (child_ctx && inherited_all) {
/*
* Mark the child context as a clone of the parent
* context, or of whatever the parent is a clone of.
- * Note that if the parent is a clone, it could get
- * uncloned at any point, but that doesn't matter
- * because the list of events and the generation
- * count can't have changed since we took the mutex.
+ *
+ * Note that if the parent is a clone, the holding of
+ * parent_ctx->lock avoids it from being uncloned.
*/
- cloned_ctx = rcu_dereference(parent_ctx->parent_ctx);
+ cloned_ctx = parent_ctx->parent_ctx;
if (cloned_ctx) {
child_ctx->parent_ctx = cloned_ctx;
child_ctx->parent_gen = parent_ctx->parent_gen;
get_ctx(child_ctx->parent_ctx);
}
+ raw_spin_unlock_irqrestore(&parent_ctx->lock, flags);
mutex_unlock(&parent_ctx->mutex);
perf_unpin_context(parent_ctx);
return ret;
}
+/*
+ * Initialize the perf_event context in task_struct
+ */
+int perf_event_init_task(struct task_struct *child)
+{
+ int ctxn, ret;
+
+ memset(child->perf_event_ctxp, 0, sizeof(child->perf_event_ctxp));
+ mutex_init(&child->perf_event_mutex);
+ INIT_LIST_HEAD(&child->perf_event_list);
+
+ for_each_task_context_nr(ctxn) {
+ ret = perf_event_init_context(child, ctxn);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
+}
+
static void __init perf_event_init_all_cpus(void)
{
- struct perf_cpu_context *cpuctx;
struct swevent_htable *swhash;
int cpu;
for_each_possible_cpu(cpu) {
swhash = &per_cpu(swevent_htable, cpu);
mutex_init(&swhash->hlist_mutex);
-
- cpuctx = &per_cpu(perf_cpu_context, cpu);
- __perf_event_init_context(&cpuctx->ctx, NULL);
- cpuctx->timer_interval = TICK_NSEC;
- hrtimer_init(&cpuctx->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
- cpuctx->timer.function = perf_event_context_tick;
+ INIT_LIST_HEAD(&per_cpu(rotation_list, cpu));
}
}
static void __cpuinit perf_event_init_cpu(int cpu)
{
- struct perf_cpu_context *cpuctx;
- struct swevent_htable *swhash;
-
- cpuctx = &per_cpu(perf_cpu_context, cpu);
-
- swhash = &per_cpu(swevent_htable, cpu);
+ struct swevent_htable *swhash = &per_cpu(swevent_htable, cpu);
mutex_lock(&swhash->hlist_mutex);
if (swhash->hlist_refcount > 0) {
mutex_unlock(&swhash->hlist_mutex);
}
-#ifdef CONFIG_HOTPLUG_CPU
-static void __perf_event_exit_cpu(void *info)
+#if defined CONFIG_HOTPLUG_CPU || defined CONFIG_KEXEC
+static void perf_pmu_rotate_stop(struct pmu *pmu)
{
- struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
- struct perf_event_context *ctx = &cpuctx->ctx;
+ struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
+
+ WARN_ON(!irqs_disabled());
+
+ list_del_init(&cpuctx->rotation_list);
+}
+
+static void __perf_event_exit_context(void *__info)
+{
+ struct perf_event_context *ctx = __info;
struct perf_event *event, *tmp;
- perf_pmu_rotate_stop();
+ perf_pmu_rotate_stop(ctx->pmu);
list_for_each_entry_safe(event, tmp, &ctx->pinned_groups, group_entry)
__perf_event_remove_from_context(event);
list_for_each_entry_safe(event, tmp, &ctx->flexible_groups, group_entry)
__perf_event_remove_from_context(event);
}
+
+static void perf_event_exit_cpu_context(int cpu)
+{
+ struct perf_event_context *ctx;
+ struct pmu *pmu;
+ int idx;
+
+ idx = srcu_read_lock(&pmus_srcu);
+ list_for_each_entry_rcu(pmu, &pmus, entry) {
+ ctx = &per_cpu_ptr(pmu->pmu_cpu_context, cpu)->ctx;
+
+ mutex_lock(&ctx->mutex);
+ smp_call_function_single(cpu, __perf_event_exit_context, ctx, 1);
+ mutex_unlock(&ctx->mutex);
+ }
+ srcu_read_unlock(&pmus_srcu, idx);
+}
+
static void perf_event_exit_cpu(int cpu)
{
- struct perf_cpu_context *cpuctx = &per_cpu(perf_cpu_context, cpu);
struct swevent_htable *swhash = &per_cpu(swevent_htable, cpu);
- struct perf_event_context *ctx = &cpuctx->ctx;
mutex_lock(&swhash->hlist_mutex);
swevent_hlist_release(swhash);
mutex_unlock(&swhash->hlist_mutex);
- mutex_lock(&ctx->mutex);
- smp_call_function_single(cpu, __perf_event_exit_cpu, NULL, 1);
- mutex_unlock(&ctx->mutex);
+ perf_event_exit_cpu_context(cpu);
}
#else
static inline void perf_event_exit_cpu(int cpu) { }
#endif
+static int
+perf_reboot(struct notifier_block *notifier, unsigned long val, void *v)
+{
+ int cpu;
+
+ for_each_online_cpu(cpu)
+ perf_event_exit_cpu(cpu);
+
+ return NOTIFY_OK;
+}
+
+/*
+ * Run the perf reboot notifier at the very last possible moment so that
+ * the generic watchdog code runs as long as possible.
+ */
+static struct notifier_block perf_reboot_notifier = {
+ .notifier_call = perf_reboot,
+ .priority = INT_MIN,
+};
+
static int __cpuinit
perf_cpu_notify(struct notifier_block *self, unsigned long action, void *hcpu)
{
void __init perf_event_init(void)
{
+ int ret;
+
+ idr_init(&pmu_idr);
+
perf_event_init_all_cpus();
init_srcu_struct(&pmus_srcu);
- perf_pmu_register(&perf_swevent);
- perf_pmu_register(&perf_cpu_clock);
- perf_pmu_register(&perf_task_clock);
+ perf_pmu_register(&perf_swevent, "software", PERF_TYPE_SOFTWARE);
+ perf_pmu_register(&perf_cpu_clock, NULL, -1);
+ perf_pmu_register(&perf_task_clock, NULL, -1);
perf_tp_register();
perf_cpu_notifier(perf_cpu_notify);
+ register_reboot_notifier(&perf_reboot_notifier);
+
+ ret = init_hw_breakpoint();
+ WARN(ret, "hw_breakpoint initialization failed with: %d", ret);
+}
+
+static int __init perf_event_sysfs_init(void)
+{
+ struct pmu *pmu;
+ int ret;
+
+ mutex_lock(&pmus_lock);
+
+ ret = bus_register(&pmu_bus);
+ if (ret)
+ goto unlock;
+
+ list_for_each_entry(pmu, &pmus, entry) {
+ if (!pmu->name || pmu->type < 0)
+ continue;
+
+ ret = pmu_dev_alloc(pmu);
+ WARN(ret, "Failed to register pmu: %s, reason %d\n", pmu->name, ret);
+ }
+ pmu_bus_running = 1;
+ ret = 0;
+
+unlock:
+ mutex_unlock(&pmus_lock);
+
+ return ret;
}
+device_initcall(perf_event_sysfs_init);