#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:
*/
-DEFINE_PER_CPU(struct perf_cpu_context, perf_cpu_context);
+static DEFINE_PER_CPU(struct perf_cpu_context, perf_cpu_context);
int perf_max_events __read_mostly = 1;
static int perf_reserved_percpu __read_mostly;
* if so. If we locked the right context, then it
* can't get swapped on us any more.
*/
- spin_lock_irqsave(&ctx->lock, *flags);
+ raw_spin_lock_irqsave(&ctx->lock, *flags);
if (ctx != rcu_dereference(task->perf_event_ctxp)) {
- spin_unlock_irqrestore(&ctx->lock, *flags);
+ raw_spin_unlock_irqrestore(&ctx->lock, *flags);
goto retry;
}
if (!atomic_inc_not_zero(&ctx->refcount)) {
- spin_unlock_irqrestore(&ctx->lock, *flags);
+ raw_spin_unlock_irqrestore(&ctx->lock, *flags);
ctx = NULL;
}
}
ctx = perf_lock_task_context(task, &flags);
if (ctx) {
++ctx->pin_count;
- spin_unlock_irqrestore(&ctx->lock, flags);
+ raw_spin_unlock_irqrestore(&ctx->lock, flags);
}
return ctx;
}
{
unsigned long flags;
- spin_lock_irqsave(&ctx->lock, flags);
+ raw_spin_lock_irqsave(&ctx->lock, flags);
--ctx->pin_count;
- spin_unlock_irqrestore(&ctx->lock, flags);
+ raw_spin_unlock_irqrestore(&ctx->lock, flags);
put_ctx(ctx);
}
+static inline u64 perf_clock(void)
+{
+ return cpu_clock(smp_processor_id());
+}
+
+/*
+ * 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;
+}
+
+/*
+ * Update the total_time_enabled and total_time_running fields for a event.
+ */
+static void update_event_times(struct perf_event *event)
+{
+ struct perf_event_context *ctx = event->ctx;
+ u64 run_end;
+
+ if (event->state < PERF_EVENT_STATE_INACTIVE ||
+ event->group_leader->state < PERF_EVENT_STATE_INACTIVE)
+ return;
+
+ if (ctx->is_active)
+ run_end = ctx->time;
+ else
+ run_end = event->tstamp_stopped;
+
+ event->total_time_enabled = run_end - event->tstamp_enabled;
+
+ if (event->state == PERF_EVENT_STATE_INACTIVE)
+ run_end = event->tstamp_stopped;
+ else
+ run_end = ctx->time;
+
+ event->total_time_running = run_end - event->tstamp_running;
+}
+
/*
* Add a event from the lists for its context.
* Must be called with ctx->mutex and ctx->lock held.
if (event->group_leader != event)
event->group_leader->nr_siblings--;
+ update_event_times(event);
+
+ /*
+ * 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)
+ event->state = PERF_EVENT_STATE_OFF;
+
/*
* If this was a group event with sibling events then
* upgrade the siblings to singleton events by adding them
if (ctx->task && cpuctx->task_ctx != ctx)
return;
- spin_lock(&ctx->lock);
+ raw_spin_lock(&ctx->lock);
/*
* Protect the list operation against NMI by disabling the
* events on a global level.
}
perf_enable();
- spin_unlock(&ctx->lock);
+ raw_spin_unlock(&ctx->lock);
}
if (!task) {
/*
* Per cpu events are removed via an smp call and
- * the removal is always sucessful.
+ * the removal is always successful.
*/
smp_call_function_single(event->cpu,
__perf_event_remove_from_context,
task_oncpu_function_call(task, __perf_event_remove_from_context,
event);
- spin_lock_irq(&ctx->lock);
+ raw_spin_lock_irq(&ctx->lock);
/*
* If the context is active we need to retry the smp call.
*/
if (ctx->nr_active && !list_empty(&event->group_entry)) {
- spin_unlock_irq(&ctx->lock);
+ raw_spin_unlock_irq(&ctx->lock);
goto retry;
}
* can remove the event safely, if the call above did not
* succeed.
*/
- if (!list_empty(&event->group_entry)) {
+ if (!list_empty(&event->group_entry))
list_del_event(event, ctx);
- }
- spin_unlock_irq(&ctx->lock);
-}
-
-static inline u64 perf_clock(void)
-{
- return cpu_clock(smp_processor_id());
-}
-
-/*
- * 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;
-}
-
-/*
- * Update the total_time_enabled and total_time_running fields for a event.
- */
-static void update_event_times(struct perf_event *event)
-{
- struct perf_event_context *ctx = event->ctx;
- u64 run_end;
-
- if (event->state < PERF_EVENT_STATE_INACTIVE ||
- event->group_leader->state < PERF_EVENT_STATE_INACTIVE)
- return;
-
- event->total_time_enabled = ctx->time - event->tstamp_enabled;
-
- if (event->state == PERF_EVENT_STATE_INACTIVE)
- run_end = event->tstamp_stopped;
- else
- run_end = ctx->time;
-
- event->total_time_running = run_end - event->tstamp_running;
+ raw_spin_unlock_irq(&ctx->lock);
}
/*
if (ctx->task && cpuctx->task_ctx != ctx)
return;
- spin_lock(&ctx->lock);
+ raw_spin_lock(&ctx->lock);
/*
* If the event is on, turn it off.
event->state = PERF_EVENT_STATE_OFF;
}
- spin_unlock(&ctx->lock);
+ raw_spin_unlock(&ctx->lock);
}
/*
* 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)
+void perf_event_disable(struct perf_event *event)
{
struct perf_event_context *ctx = event->ctx;
struct task_struct *task = ctx->task;
retry:
task_oncpu_function_call(task, __perf_event_disable, event);
- spin_lock_irq(&ctx->lock);
+ raw_spin_lock_irq(&ctx->lock);
/*
* If the event is still active, we need to retry the cross-call.
*/
if (event->state == PERF_EVENT_STATE_ACTIVE) {
- spin_unlock_irq(&ctx->lock);
+ raw_spin_unlock_irq(&ctx->lock);
goto retry;
}
event->state = PERF_EVENT_STATE_OFF;
}
- spin_unlock_irq(&ctx->lock);
+ raw_spin_unlock_irq(&ctx->lock);
}
static int
cpuctx->task_ctx = ctx;
}
- spin_lock(&ctx->lock);
+ raw_spin_lock(&ctx->lock);
ctx->is_active = 1;
update_context_time(ctx);
add_event_to_ctx(event, ctx);
+ if (event->cpu != -1 && event->cpu != smp_processor_id())
+ goto unlock;
+
/*
* Don't put the event on if it is disabled or if
* it is in a group and the group isn't on.
unlock:
perf_enable();
- spin_unlock(&ctx->lock);
+ raw_spin_unlock(&ctx->lock);
}
/*
if (!task) {
/*
* Per cpu events are installed via an smp call and
- * the install is always sucessful.
+ * the install is always successful.
*/
smp_call_function_single(cpu, __perf_install_in_context,
event, 1);
task_oncpu_function_call(task, __perf_install_in_context,
event);
- spin_lock_irq(&ctx->lock);
+ raw_spin_lock_irq(&ctx->lock);
/*
* we need to retry the smp call.
*/
if (ctx->is_active && list_empty(&event->group_entry)) {
- spin_unlock_irq(&ctx->lock);
+ raw_spin_unlock_irq(&ctx->lock);
goto retry;
}
*/
if (list_empty(&event->group_entry))
add_event_to_ctx(event, ctx);
- spin_unlock_irq(&ctx->lock);
+ raw_spin_unlock_irq(&ctx->lock);
}
/*
cpuctx->task_ctx = ctx;
}
- spin_lock(&ctx->lock);
+ raw_spin_lock(&ctx->lock);
ctx->is_active = 1;
update_context_time(ctx);
goto unlock;
__perf_event_mark_enabled(event, ctx);
+ if (event->cpu != -1 && event->cpu != smp_processor_id())
+ goto unlock;
+
/*
* If the event is in a group and isn't the group leader,
* then don't put it on unless the group is on.
}
unlock:
- spin_unlock(&ctx->lock);
+ raw_spin_unlock(&ctx->lock);
}
/*
* 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)
+void perf_event_enable(struct perf_event *event)
{
struct perf_event_context *ctx = event->ctx;
struct task_struct *task = ctx->task;
return;
}
- spin_lock_irq(&ctx->lock);
+ raw_spin_lock_irq(&ctx->lock);
if (event->state >= PERF_EVENT_STATE_INACTIVE)
goto out;
event->state = PERF_EVENT_STATE_OFF;
retry:
- spin_unlock_irq(&ctx->lock);
+ raw_spin_unlock_irq(&ctx->lock);
task_oncpu_function_call(task, __perf_event_enable, event);
- spin_lock_irq(&ctx->lock);
+ raw_spin_lock_irq(&ctx->lock);
/*
* If the context is active and the event is still off,
__perf_event_mark_enabled(event, ctx);
out:
- spin_unlock_irq(&ctx->lock);
+ raw_spin_unlock_irq(&ctx->lock);
}
static int perf_event_refresh(struct perf_event *event, int refresh)
{
struct perf_event *event;
- spin_lock(&ctx->lock);
+ raw_spin_lock(&ctx->lock);
ctx->is_active = 0;
if (likely(!ctx->nr_events))
goto out;
update_context_time(ctx);
perf_disable();
- if (ctx->nr_active)
+ if (ctx->nr_active) {
list_for_each_entry(event, &ctx->group_list, group_entry)
group_sched_out(event, cpuctx, ctx);
-
+ }
perf_enable();
out:
- spin_unlock(&ctx->lock);
+ raw_spin_unlock(&ctx->lock);
}
/*
&& !ctx1->pin_count && !ctx2->pin_count;
}
-static void __perf_event_read(void *event);
-
static void __perf_event_sync_stat(struct perf_event *event,
struct perf_event *next_event)
{
*/
switch (event->state) {
case PERF_EVENT_STATE_ACTIVE:
- __perf_event_read(event);
- break;
+ event->pmu->read(event);
+ /* fall-through */
case PERF_EVENT_STATE_INACTIVE:
update_event_times(event);
if (!ctx->nr_stat)
return;
+ update_context_time(ctx);
+
event = list_first_entry(&ctx->event_list,
struct perf_event, event_entry);
if (likely(!ctx || !cpuctx->task_ctx))
return;
- update_context_time(ctx);
-
rcu_read_lock();
parent = rcu_dereference(ctx->parent_ctx);
next_ctx = next->perf_event_ctxp;
* order we take the locks because no other cpu could
* be trying to lock both of these tasks.
*/
- spin_lock(&ctx->lock);
- spin_lock_nested(&next_ctx->lock, SINGLE_DEPTH_NESTING);
+ raw_spin_lock(&ctx->lock);
+ raw_spin_lock_nested(&next_ctx->lock, SINGLE_DEPTH_NESTING);
if (context_equiv(ctx, next_ctx)) {
/*
* XXX do we need a memory barrier of sorts
perf_event_sync_stat(ctx, next_ctx);
}
- spin_unlock(&next_ctx->lock);
- spin_unlock(&ctx->lock);
+ raw_spin_unlock(&next_ctx->lock);
+ raw_spin_unlock(&ctx->lock);
}
rcu_read_unlock();
struct perf_event *event;
int can_add_hw = 1;
- spin_lock(&ctx->lock);
+ raw_spin_lock(&ctx->lock);
ctx->is_active = 1;
if (likely(!ctx->nr_events))
goto out;
}
perf_enable();
out:
- spin_unlock(&ctx->lock);
+ raw_spin_unlock(&ctx->lock);
}
/*
struct hw_perf_event *hwc;
u64 interrupts, freq;
- spin_lock(&ctx->lock);
- list_for_each_entry(event, &ctx->group_list, group_entry) {
+ raw_spin_lock(&ctx->lock);
+ list_for_each_entry_rcu(event, &ctx->event_list, event_entry) {
if (event->state != PERF_EVENT_STATE_ACTIVE)
continue;
+ if (event->cpu != -1 && event->cpu != smp_processor_id())
+ continue;
+
hwc = &event->hw;
interrupts = hwc->interrupts;
perf_enable();
}
}
- spin_unlock(&ctx->lock);
+ raw_spin_unlock(&ctx->lock);
}
/*
if (!ctx->nr_events)
return;
- spin_lock(&ctx->lock);
+ raw_spin_lock(&ctx->lock);
/*
* Rotate the first entry last (works just fine for group events too):
*/
}
perf_enable();
- spin_unlock(&ctx->lock);
+ raw_spin_unlock(&ctx->lock);
}
void perf_event_task_tick(struct task_struct *curr, int cpu)
__perf_event_task_sched_out(ctx);
- spin_lock(&ctx->lock);
+ raw_spin_lock(&ctx->lock);
list_for_each_entry(event, &ctx->group_list, group_entry) {
if (!event->attr.enable_on_exec)
if (enabled)
unclone_ctx(ctx);
- spin_unlock(&ctx->lock);
+ raw_spin_unlock(&ctx->lock);
perf_event_task_sched_in(task, smp_processor_id());
out:
struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
struct perf_event *event = info;
struct perf_event_context *ctx = event->ctx;
- unsigned long flags;
/*
* If this is a task context, we need to check whether it is
if (ctx->task && cpuctx->task_ctx != ctx)
return;
- local_irq_save(flags);
- if (ctx->is_active)
- update_context_time(ctx);
- event->pmu->read(event);
+ raw_spin_lock(&ctx->lock);
+ update_context_time(ctx);
update_event_times(event);
- local_irq_restore(flags);
+ raw_spin_unlock(&ctx->lock);
+
+ event->pmu->read(event);
}
static u64 perf_event_read(struct perf_event *event)
smp_call_function_single(event->oncpu,
__perf_event_read, event, 1);
} else if (event->state == PERF_EVENT_STATE_INACTIVE) {
+ struct perf_event_context *ctx = event->ctx;
+ unsigned long flags;
+
+ raw_spin_lock_irqsave(&ctx->lock, flags);
+ update_context_time(ctx);
update_event_times(event);
+ raw_spin_unlock_irqrestore(&ctx->lock, flags);
}
return atomic64_read(&event->count);
__perf_event_init_context(struct perf_event_context *ctx,
struct task_struct *task)
{
- memset(ctx, 0, sizeof(*ctx));
- spin_lock_init(&ctx->lock);
+ raw_spin_lock_init(&ctx->lock);
mutex_init(&ctx->mutex);
INIT_LIST_HEAD(&ctx->group_list);
INIT_LIST_HEAD(&ctx->event_list);
unsigned long flags;
int err;
- /*
- * If cpu is not a wildcard then this is a percpu event:
- */
- if (cpu != -1) {
+ if (pid == -1 && cpu != -1) {
/* 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 > num_possible_cpus())
+ if (cpu < 0 || cpu >= nr_cpumask_bits)
return ERR_PTR(-EINVAL);
/*
* offline CPU and activate it when the CPU comes up, but
* that's for later.
*/
- if (!cpu_isset(cpu, cpu_online_map))
+ if (!cpu_online(cpu))
return ERR_PTR(-ENODEV);
cpuctx = &per_cpu(perf_cpu_context, cpu);
ctx = perf_lock_task_context(task, &flags);
if (ctx) {
unclone_ctx(ctx);
- spin_unlock_irqrestore(&ctx->lock, flags);
+ raw_spin_unlock_irqrestore(&ctx->lock, flags);
}
if (!ctx) {
- ctx = kmalloc(sizeof(struct perf_event_context), GFP_KERNEL);
+ ctx = kzalloc(sizeof(struct perf_event_context), GFP_KERNEL);
err = -ENOMEM;
if (!ctx)
goto errout;
call_rcu(&event->rcu_head, free_event_rcu);
}
-/*
- * Called when the last reference to the file is gone.
- */
-static int perf_release(struct inode *inode, struct file *file)
+int perf_event_release_kernel(struct perf_event *event)
{
- struct perf_event *event = file->private_data;
struct perf_event_context *ctx = event->ctx;
- file->private_data = NULL;
-
WARN_ON_ONCE(ctx->parent_ctx);
mutex_lock(&ctx->mutex);
perf_event_remove_from_context(event);
return 0;
}
+EXPORT_SYMBOL_GPL(perf_event_release_kernel);
-int perf_event_release_kernel(struct perf_event *event)
+/*
+ * Called when the last reference to the file is gone.
+ */
+static int perf_release(struct inode *inode, struct file *file)
{
- struct perf_event_context *ctx = event->ctx;
-
- WARN_ON_ONCE(ctx->parent_ctx);
- mutex_lock(&ctx->mutex);
- perf_event_remove_from_context(event);
- 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);
+ struct perf_event *event = file->private_data;
- free_event(event);
+ file->private_data = NULL;
- return 0;
+ return perf_event_release_kernel(event);
}
-EXPORT_SYMBOL_GPL(perf_event_release_kernel);
static int perf_event_read_size(struct perf_event *event)
{
return size;
}
-u64 perf_event_read_value(struct perf_event *event)
+u64 perf_event_read_value(struct perf_event *event, u64 *enabled, u64 *running)
{
struct perf_event *child;
u64 total = 0;
+ *enabled = 0;
+ *running = 0;
+
+ mutex_lock(&event->child_mutex);
total += perf_event_read(event);
- list_for_each_entry(child, &event->child_list, child_list)
+ *enabled += event->total_time_enabled +
+ atomic64_read(&event->child_total_time_enabled);
+ *running += event->total_time_running +
+ atomic64_read(&event->child_total_time_running);
+
+ list_for_each_entry(child, &event->child_list, child_list) {
total += perf_event_read(child);
+ *enabled += child->total_time_enabled;
+ *running += child->total_time_running;
+ }
+ mutex_unlock(&event->child_mutex);
return total;
}
EXPORT_SYMBOL_GPL(perf_event_read_value);
-static int perf_event_read_entry(struct perf_event *event,
- u64 read_format, char __user *buf)
-{
- int n = 0, count = 0;
- u64 values[2];
-
- values[n++] = perf_event_read_value(event);
- if (read_format & PERF_FORMAT_ID)
- values[n++] = primary_event_id(event);
-
- count = n * sizeof(u64);
-
- if (copy_to_user(buf, values, count))
- return -EFAULT;
-
- return count;
-}
-
static int perf_event_read_group(struct perf_event *event,
u64 read_format, char __user *buf)
{
struct perf_event *leader = event->group_leader, *sub;
- int n = 0, size = 0, err = -EFAULT;
- u64 values[3];
+ int n = 0, size = 0, ret = -EFAULT;
+ struct perf_event_context *ctx = leader->ctx;
+ u64 values[5];
+ u64 count, enabled, running;
+
+ mutex_lock(&ctx->mutex);
+ count = perf_event_read_value(leader, &enabled, &running);
values[n++] = 1 + leader->nr_siblings;
- if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) {
- values[n++] = leader->total_time_enabled +
- atomic64_read(&leader->child_total_time_enabled);
- }
- if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) {
- values[n++] = leader->total_time_running +
- atomic64_read(&leader->child_total_time_running);
- }
+ if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
+ values[n++] = enabled;
+ if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
+ values[n++] = running;
+ values[n++] = count;
+ if (read_format & PERF_FORMAT_ID)
+ values[n++] = primary_event_id(leader);
size = n * sizeof(u64);
if (copy_to_user(buf, values, size))
- return -EFAULT;
-
- err = perf_event_read_entry(leader, read_format, buf + size);
- if (err < 0)
- return err;
+ goto unlock;
- size += err;
+ ret = size;
list_for_each_entry(sub, &leader->sibling_list, group_entry) {
- err = perf_event_read_entry(sub, read_format,
- buf + size);
- if (err < 0)
- return err;
+ n = 0;
+
+ values[n++] = perf_event_read_value(sub, &enabled, &running);
+ if (read_format & PERF_FORMAT_ID)
+ values[n++] = primary_event_id(sub);
+
+ size = n * sizeof(u64);
- size += err;
+ if (copy_to_user(buf + ret, values, size)) {
+ ret = -EFAULT;
+ goto unlock;
+ }
+
+ ret += size;
}
+unlock:
+ mutex_unlock(&ctx->mutex);
- return size;
+ return ret;
}
static int perf_event_read_one(struct perf_event *event,
u64 read_format, char __user *buf)
{
+ u64 enabled, running;
u64 values[4];
int n = 0;
- values[n++] = perf_event_read_value(event);
- if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) {
- values[n++] = event->total_time_enabled +
- atomic64_read(&event->child_total_time_enabled);
- }
- if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) {
- values[n++] = event->total_time_running +
- atomic64_read(&event->child_total_time_running);
- }
+ values[n++] = perf_event_read_value(event, &enabled, &running);
+ if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
+ values[n++] = enabled;
+ if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
+ values[n++] = running;
if (read_format & PERF_FORMAT_ID)
values[n++] = primary_event_id(event);
return -ENOSPC;
WARN_ON_ONCE(event->ctx->parent_ctx);
- mutex_lock(&event->child_mutex);
if (read_format & PERF_FORMAT_GROUP)
ret = perf_event_read_group(event, read_format, buf);
else
ret = perf_event_read_one(event, read_format, buf);
- mutex_unlock(&event->child_mutex);
return ret;
}
if (!value)
return -EINVAL;
- spin_lock_irq(&ctx->lock);
+ raw_spin_lock_irq(&ctx->lock);
if (event->attr.freq) {
if (value > sysctl_perf_event_sample_rate) {
ret = -EINVAL;
event->hw.sample_period = value;
}
unlock:
- spin_unlock_irq(&ctx->lock);
+ raw_spin_unlock_irq(&ctx->lock);
return ret;
}
perf_mmap_free_page((unsigned long)data->user_page);
for (i = 0; i < data->nr_pages; i++)
perf_mmap_free_page((unsigned long)data->data_pages[i]);
+ kfree(data);
}
#else
perf_mmap_unmark_page(base + (i * PAGE_SIZE));
vfree(base);
+ kfree(data);
}
static void perf_mmap_data_free(struct perf_mmap_data *data)
}
if (!data->watermark)
- data->watermark = max_t(long, PAGE_SIZE, max_size / 2);
+ data->watermark = max_size / 2;
rcu_assign_pointer(event->data, data);
data = container_of(rcu_head, struct perf_mmap_data, rcu_head);
perf_mmap_data_free(data);
- kfree(data);
}
static void perf_mmap_data_release(struct perf_event *event)
static void perf_output_lock(struct perf_output_handle *handle)
{
struct perf_mmap_data *data = handle->data;
- int cpu;
+ int cur, cpu = get_cpu();
handle->locked = 0;
- local_irq_save(handle->flags);
- cpu = smp_processor_id();
-
- if (in_nmi() && atomic_read(&data->lock) == cpu)
- return;
+ for (;;) {
+ cur = atomic_cmpxchg(&data->lock, -1, cpu);
+ if (cur == -1) {
+ handle->locked = 1;
+ break;
+ }
+ if (cur == cpu)
+ break;
- while (atomic_cmpxchg(&data->lock, -1, cpu) != -1)
cpu_relax();
-
- handle->locked = 1;
+ }
}
static void perf_output_unlock(struct perf_output_handle *handle)
if (atomic_xchg(&data->wakeup, 0))
perf_output_wakeup(handle);
out:
- local_irq_restore(handle->flags);
+ put_cpu();
}
void perf_output_copy(struct perf_output_handle *handle,
static int perf_event_task_match(struct perf_event *event)
{
+ if (event->state != PERF_EVENT_STATE_ACTIVE)
+ return 0;
+
+ if (event->cpu != -1 && event->cpu != smp_processor_id())
+ return 0;
+
if (event->attr.comm || event->attr.mmap || event->attr.task)
return 1;
{
struct perf_event *event;
- if (system_state != SYSTEM_RUNNING || list_empty(&ctx->event_list))
- return;
-
- rcu_read_lock();
list_for_each_entry_rcu(event, &ctx->event_list, event_entry) {
if (perf_event_task_match(event))
perf_event_task_output(event, task_event);
}
- rcu_read_unlock();
}
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;
+ rcu_read_lock();
cpuctx = &get_cpu_var(perf_cpu_context);
perf_event_task_ctx(&cpuctx->ctx, task_event);
- put_cpu_var(perf_cpu_context);
-
- rcu_read_lock();
if (!ctx)
ctx = rcu_dereference(task_event->task->perf_event_ctxp);
if (ctx)
perf_event_task_ctx(ctx, task_event);
+ put_cpu_var(perf_cpu_context);
rcu_read_unlock();
}
static int perf_event_comm_match(struct perf_event *event)
{
+ if (event->state != PERF_EVENT_STATE_ACTIVE)
+ return 0;
+
+ if (event->cpu != -1 && event->cpu != smp_processor_id())
+ return 0;
+
if (event->attr.comm)
return 1;
{
struct perf_event *event;
- if (system_state != SYSTEM_RUNNING || list_empty(&ctx->event_list))
- return;
-
- rcu_read_lock();
list_for_each_entry_rcu(event, &ctx->event_list, event_entry) {
if (perf_event_comm_match(event))
perf_event_comm_output(event, comm_event);
}
- rcu_read_unlock();
}
static void perf_event_comm_event(struct perf_comm_event *comm_event)
char comm[TASK_COMM_LEN];
memset(comm, 0, sizeof(comm));
- strncpy(comm, comm_event->task->comm, sizeof(comm));
+ strlcpy(comm, comm_event->task->comm, sizeof(comm));
size = ALIGN(strlen(comm)+1, sizeof(u64));
comm_event->comm = comm;
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);
- put_cpu_var(perf_cpu_context);
-
- rcu_read_lock();
- /*
- * doesn't really matter which of the child contexts the
- * events ends up in.
- */
ctx = rcu_dereference(current->perf_event_ctxp);
if (ctx)
perf_event_comm_ctx(ctx, comm_event);
+ put_cpu_var(perf_cpu_context);
rcu_read_unlock();
}
static int perf_event_mmap_match(struct perf_event *event,
struct perf_mmap_event *mmap_event)
{
+ if (event->state != PERF_EVENT_STATE_ACTIVE)
+ return 0;
+
+ if (event->cpu != -1 && event->cpu != smp_processor_id())
+ return 0;
+
if (event->attr.mmap)
return 1;
{
struct perf_event *event;
- if (system_state != SYSTEM_RUNNING || list_empty(&ctx->event_list))
- return;
-
- rcu_read_lock();
list_for_each_entry_rcu(event, &ctx->event_list, event_entry) {
if (perf_event_mmap_match(event, mmap_event))
perf_event_mmap_output(event, mmap_event);
}
- rcu_read_unlock();
}
static void perf_event_mmap_event(struct perf_mmap_event *mmap_event)
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);
- put_cpu_var(perf_cpu_context);
-
- rcu_read_lock();
- /*
- * doesn't really matter which of the child contexts the
- * events ends up in.
- */
ctx = rcu_dereference(current->perf_event_ctxp);
if (ctx)
perf_event_mmap_ctx(ctx, mmap_event);
+ put_cpu_var(perf_cpu_context);
rcu_read_unlock();
kfree(buf);
perf_event_disable(event);
}
- perf_event_output(event, nmi, data, regs);
+ if (event->overflow_handler)
+ event->overflow_handler(event, nmi, data, regs);
+ else
+ perf_event_output(event, nmi, data, regs);
+
return ret;
}
return nr;
}
-static void perf_swevent_overflow(struct perf_event *event,
+static void perf_swevent_overflow(struct perf_event *event, u64 overflow,
int nmi, struct perf_sample_data *data,
struct pt_regs *regs)
{
struct hw_perf_event *hwc = &event->hw;
int throttle = 0;
- u64 overflow;
data->period = event->hw.last_period;
- overflow = perf_swevent_set_period(event);
+ if (!overflow)
+ overflow = perf_swevent_set_period(event);
if (hwc->interrupts == MAX_INTERRUPTS)
return;
atomic64_add(nr, &event->count);
+ if (!regs)
+ return;
+
if (!hwc->sample_period)
return;
- if (!regs)
+ if (nr == 1 && hwc->sample_period == 1 && !event->attr.freq)
+ return perf_swevent_overflow(event, 1, nmi, data, regs);
+
+ if (atomic64_add_negative(nr, &hwc->period_left))
return;
- if (!atomic64_add_negative(nr, &hwc->period_left))
- perf_swevent_overflow(event, nmi, data, regs);
+ perf_swevent_overflow(event, 0, nmi, data, regs);
}
static int perf_swevent_is_counting(struct perf_event *event)
static int perf_tp_event_match(struct perf_event *event,
struct perf_sample_data *data);
+static int perf_exclude_event(struct perf_event *event,
+ struct pt_regs *regs)
+{
+ if (regs) {
+ if (event->attr.exclude_user && user_mode(regs))
+ return 1;
+
+ if (event->attr.exclude_kernel && !user_mode(regs))
+ return 1;
+ }
+
+ return 0;
+}
+
static int perf_swevent_match(struct perf_event *event,
enum perf_type_id type,
u32 event_id,
struct perf_sample_data *data,
struct pt_regs *regs)
{
+ if (event->cpu != -1 && event->cpu != smp_processor_id())
+ return 0;
+
if (!perf_swevent_is_counting(event))
return 0;
if (event->attr.type != type)
return 0;
+
if (event->attr.config != event_id)
return 0;
- if (regs) {
- if (event->attr.exclude_user && user_mode(regs))
- return 0;
-
- if (event->attr.exclude_kernel && !user_mode(regs))
- return 0;
- }
+ if (perf_exclude_event(event, regs))
+ return 0;
if (event->attr.type == PERF_TYPE_TRACEPOINT &&
!perf_tp_event_match(event, data))
{
struct perf_event *event;
- if (system_state != SYSTEM_RUNNING || list_empty(&ctx->event_list))
- return;
-
- rcu_read_lock();
list_for_each_entry_rcu(event, &ctx->event_list, event_entry) {
if (perf_swevent_match(event, type, event_id, data, regs))
perf_swevent_add(event, nr, nmi, data, regs);
}
- rcu_read_unlock();
}
-static int *perf_swevent_recursion_context(struct perf_cpu_context *cpuctx)
+int perf_swevent_get_recursion_context(void)
{
+ struct perf_cpu_context *cpuctx = &get_cpu_var(perf_cpu_context);
+ int rctx;
+
if (in_nmi())
- return &cpuctx->recursion[3];
+ rctx = 3;
+ else if (in_irq())
+ rctx = 2;
+ else if (in_softirq())
+ rctx = 1;
+ else
+ rctx = 0;
- if (in_irq())
- return &cpuctx->recursion[2];
+ if (cpuctx->recursion[rctx]) {
+ put_cpu_var(perf_cpu_context);
+ return -1;
+ }
+
+ cpuctx->recursion[rctx]++;
+ barrier();
- if (in_softirq())
- return &cpuctx->recursion[1];
+ return rctx;
+}
+EXPORT_SYMBOL_GPL(perf_swevent_get_recursion_context);
- return &cpuctx->recursion[0];
+void perf_swevent_put_recursion_context(int rctx)
+{
+ struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
+ barrier();
+ cpuctx->recursion[rctx]--;
+ put_cpu_var(perf_cpu_context);
}
+EXPORT_SYMBOL_GPL(perf_swevent_put_recursion_context);
static void do_perf_sw_event(enum perf_type_id type, u32 event_id,
u64 nr, int nmi,
struct perf_sample_data *data,
struct pt_regs *regs)
{
- struct perf_cpu_context *cpuctx = &get_cpu_var(perf_cpu_context);
- int *recursion = perf_swevent_recursion_context(cpuctx);
+ struct perf_cpu_context *cpuctx;
struct perf_event_context *ctx;
- if (*recursion)
- goto out;
-
- (*recursion)++;
- barrier();
-
+ cpuctx = &__get_cpu_var(perf_cpu_context);
+ rcu_read_lock();
perf_swevent_ctx_event(&cpuctx->ctx, type, event_id,
nr, nmi, data, regs);
- rcu_read_lock();
/*
* doesn't really matter which of the child contexts the
* events ends up in.
if (ctx)
perf_swevent_ctx_event(ctx, type, event_id, nr, nmi, data, regs);
rcu_read_unlock();
-
- barrier();
- (*recursion)--;
-
-out:
- put_cpu_var(perf_cpu_context);
}
void __perf_sw_event(u32 event_id, u64 nr, int nmi,
struct pt_regs *regs, u64 addr)
{
- struct perf_sample_data data = {
- .addr = addr,
- };
+ struct perf_sample_data data;
+ int rctx;
- do_perf_sw_event(PERF_TYPE_SOFTWARE, event_id, nr, nmi,
- &data, regs);
+ rctx = perf_swevent_get_recursion_context();
+ if (rctx < 0)
+ return;
+
+ data.addr = addr;
+ data.raw = NULL;
+
+ do_perf_sw_event(PERF_TYPE_SOFTWARE, event_id, nr, nmi, &data, regs);
+
+ perf_swevent_put_recursion_context(rctx);
}
static void perf_swevent_read(struct perf_event *event)
event->pmu->read(event);
data.addr = 0;
+ data.raw = NULL;
+ data.period = event->hw.last_period;
regs = get_irq_regs();
/*
* In case we exclude kernel IPs or are somehow not in interrupt
regs = task_pt_regs(current);
if (regs) {
- if (perf_event_overflow(event, 0, &data, regs))
- ret = HRTIMER_NORESTART;
+ if (!(event->attr.exclude_idle && current->pid == 0))
+ if (perf_event_overflow(event, 0, &data, regs))
+ ret = HRTIMER_NORESTART;
}
period = max_t(u64, 10000, event->hw.sample_period);
return ret;
}
+static void perf_swevent_start_hrtimer(struct perf_event *event)
+{
+ struct hw_perf_event *hwc = &event->hw;
+
+ hrtimer_init(&hwc->hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
+ hwc->hrtimer.function = perf_swevent_hrtimer;
+ if (hwc->sample_period) {
+ u64 period;
+
+ if (hwc->remaining) {
+ if (hwc->remaining < 0)
+ period = 10000;
+ else
+ period = hwc->remaining;
+ hwc->remaining = 0;
+ } else {
+ period = max_t(u64, 10000, hwc->sample_period);
+ }
+ __hrtimer_start_range_ns(&hwc->hrtimer,
+ ns_to_ktime(period), 0,
+ HRTIMER_MODE_REL, 0);
+ }
+}
+
+static void perf_swevent_cancel_hrtimer(struct perf_event *event)
+{
+ struct hw_perf_event *hwc = &event->hw;
+
+ if (hwc->sample_period) {
+ ktime_t remaining = hrtimer_get_remaining(&hwc->hrtimer);
+ hwc->remaining = ktime_to_ns(remaining);
+
+ hrtimer_cancel(&hwc->hrtimer);
+ }
+}
+
/*
* Software event: cpu wall time clock
*/
u64 now;
now = cpu_clock(cpu);
- prev = atomic64_read(&event->hw.prev_count);
- atomic64_set(&event->hw.prev_count, now);
+ prev = atomic64_xchg(&event->hw.prev_count, now);
atomic64_add(now - prev, &event->count);
}
int cpu = raw_smp_processor_id();
atomic64_set(&hwc->prev_count, cpu_clock(cpu));
- hrtimer_init(&hwc->hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
- hwc->hrtimer.function = perf_swevent_hrtimer;
- if (hwc->sample_period) {
- u64 period = max_t(u64, 10000, hwc->sample_period);
- __hrtimer_start_range_ns(&hwc->hrtimer,
- ns_to_ktime(period), 0,
- HRTIMER_MODE_REL, 0);
- }
+ perf_swevent_start_hrtimer(event);
return 0;
}
static void cpu_clock_perf_event_disable(struct perf_event *event)
{
- if (event->hw.sample_period)
- hrtimer_cancel(&event->hw.hrtimer);
+ perf_swevent_cancel_hrtimer(event);
cpu_clock_perf_event_update(event);
}
now = event->ctx->time;
atomic64_set(&hwc->prev_count, now);
- hrtimer_init(&hwc->hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
- hwc->hrtimer.function = perf_swevent_hrtimer;
- if (hwc->sample_period) {
- u64 period = max_t(u64, 10000, hwc->sample_period);
- __hrtimer_start_range_ns(&hwc->hrtimer,
- ns_to_ktime(period), 0,
- HRTIMER_MODE_REL, 0);
- }
+
+ perf_swevent_start_hrtimer(event);
return 0;
}
static void task_clock_perf_event_disable(struct perf_event *event)
{
- if (event->hw.sample_period)
- hrtimer_cancel(&event->hw.hrtimer);
+ perf_swevent_cancel_hrtimer(event);
task_clock_perf_event_update(event, event->ctx->time);
}
if (!regs)
regs = task_pt_regs(current);
+ /* Trace events already protected against recursion */
do_perf_sw_event(PERF_TYPE_TRACEPOINT, event_id, count, 1,
&data, regs);
}
#endif /* CONFIG_EVENT_PROFILE */
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+static void bp_perf_event_destroy(struct perf_event *event)
+{
+ release_bp_slot(event);
+}
+
+static const struct pmu *bp_perf_event_init(struct perf_event *bp)
+{
+ int err;
+
+ err = register_perf_hw_breakpoint(bp);
+ if (err)
+ return ERR_PTR(err);
+
+ bp->destroy = bp_perf_event_destroy;
+
+ return &perf_ops_bp;
+}
+
+void perf_bp_event(struct perf_event *bp, void *data)
+{
+ struct perf_sample_data sample;
+ struct pt_regs *regs = data;
+
+ sample.raw = NULL;
+ sample.addr = bp->attr.bp_addr;
+
+ if (!perf_exclude_event(bp, regs))
+ perf_swevent_add(bp, 1, 1, &sample, regs);
+}
+#else
+static const struct pmu *bp_perf_event_init(struct perf_event *bp)
+{
+ return NULL;
+}
+
+void perf_bp_event(struct perf_event *bp, void *regs)
+{
+}
+#endif
+
atomic_t perf_swevent_enabled[PERF_COUNT_SW_MAX];
static void sw_perf_event_destroy(struct perf_event *event)
case PERF_COUNT_SW_PAGE_FAULTS_MAJ:
case PERF_COUNT_SW_CONTEXT_SWITCHES:
case PERF_COUNT_SW_CPU_MIGRATIONS:
+ case PERF_COUNT_SW_ALIGNMENT_FAULTS:
+ case PERF_COUNT_SW_EMULATION_FAULTS:
if (!event->parent) {
atomic_inc(&perf_swevent_enabled[event_id]);
event->destroy = sw_perf_event_destroy;
struct perf_event_context *ctx,
struct perf_event *group_leader,
struct perf_event *parent_event,
- perf_callback_t callback,
+ perf_overflow_handler_t overflow_handler,
gfp_t gfpflags)
{
const struct pmu *pmu;
event->state = PERF_EVENT_STATE_INACTIVE;
- if (!callback && parent_event)
- callback = parent_event->callback;
+ if (!overflow_handler && parent_event)
+ overflow_handler = parent_event->overflow_handler;
- event->callback = callback;
+ event->overflow_handler = overflow_handler;
if (attr->disabled)
event->state = PERF_EVENT_STATE_OFF;
pmu = tp_perf_event_init(event);
break;
+ case PERF_TYPE_BREAKPOINT:
+ pmu = bp_perf_event_init(event);
+ break;
+
+
default:
break;
}
if (attr->type >= PERF_TYPE_MAX)
return -EINVAL;
- if (attr->__reserved_1 || attr->__reserved_2 || attr->__reserved_3)
+ if (attr->__reserved_1 || attr->__reserved_2)
return -EINVAL;
if (attr->sample_type & ~(PERF_SAMPLE_MAX-1))
if (IS_ERR(event))
goto err_put_context;
- err = anon_inode_getfd("[perf_event]", &perf_fops, event, 0);
+ err = anon_inode_getfd("[perf_event]", &perf_fops, event, O_RDWR);
if (err < 0)
goto err_free_put_context;
*/
struct perf_event *
perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu,
- pid_t pid, perf_callback_t callback)
+ pid_t pid,
+ perf_overflow_handler_t overflow_handler)
{
struct perf_event *event;
struct perf_event_context *ctx;
*/
ctx = find_get_context(pid, cpu);
- if (IS_ERR(ctx))
- return NULL ;
+ if (IS_ERR(ctx)) {
+ err = PTR_ERR(ctx);
+ goto err_exit;
+ }
event = perf_event_alloc(attr, cpu, ctx, NULL,
- NULL, callback, GFP_KERNEL);
- err = PTR_ERR(event);
- if (IS_ERR(event))
+ NULL, overflow_handler, GFP_KERNEL);
+ if (IS_ERR(event)) {
+ err = PTR_ERR(event);
goto err_put_context;
+ }
event->filp = NULL;
WARN_ON_ONCE(ctx->parent_ctx);
return event;
-err_put_context:
- if (err < 0)
- put_ctx(ctx);
-
- return NULL;
+ err_put_context:
+ put_ctx(ctx);
+ err_exit:
+ return ERR_PTR(err);
}
EXPORT_SYMBOL_GPL(perf_event_create_kernel_counter);
if (parent_event->attr.freq)
child_event->hw.sample_period = parent_event->hw.sample_period;
+ child_event->overflow_handler = parent_event->overflow_handler;
+
/*
* Link it up in the child's context:
*/
{
struct perf_event *parent_event;
- update_event_times(child_event);
perf_event_remove_from_context(child_event);
parent_event = child_event->parent;
* reading child->perf_event_ctxp, we wait until it has
* incremented the context's refcount before we do put_ctx below.
*/
- spin_lock(&child_ctx->lock);
+ raw_spin_lock(&child_ctx->lock);
child->perf_event_ctxp = NULL;
/*
* If this context is a clone; unclone it so it can't get
* the events from it.
*/
unclone_ctx(child_ctx);
- spin_unlock_irqrestore(&child_ctx->lock, flags);
+ update_context_time(child_ctx);
+ raw_spin_unlock_irqrestore(&child_ctx->lock, flags);
/*
* Report the task dead after unscheduling the events so that we
*/
int perf_event_init_task(struct task_struct *child)
{
- struct perf_event_context *child_ctx, *parent_ctx;
+ struct perf_event_context *child_ctx = NULL, *parent_ctx;
struct perf_event_context *cloned_ctx;
struct perf_event *event;
struct task_struct *parent = current;
return 0;
/*
- * This is executed from the parent task context, so inherit
- * events that have been marked for cloning.
- * First allocate and initialize a context for the child.
- */
-
- child_ctx = kmalloc(sizeof(struct perf_event_context), GFP_KERNEL);
- if (!child_ctx)
- return -ENOMEM;
-
- __perf_event_init_context(child_ctx, child);
- child->perf_event_ctxp = child_ctx;
- get_task_struct(child);
-
- /*
* If the parent's context is a clone, pin it so it won't get
* swapped under us.
*/
continue;
}
+ if (!child->perf_event_ctxp) {
+ /*
+ * This is executed from the parent task context, so
+ * inherit events that have been marked for cloning.
+ * First allocate and initialize a context for the
+ * child.
+ */
+
+ child_ctx = kzalloc(sizeof(struct perf_event_context),
+ GFP_KERNEL);
+ if (!child_ctx) {
+ ret = -ENOMEM;
+ break;
+ }
+
+ __perf_event_init_context(child_ctx, child);
+ child->perf_event_ctxp = child_ctx;
+ get_task_struct(child);
+ }
+
ret = inherit_group(event, parent, parent_ctx,
child, child_ctx);
if (ret) {
}
}
- if (inherited_all) {
+ 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.
perf_reserved_percpu = val;
for_each_online_cpu(cpu) {
cpuctx = &per_cpu(perf_cpu_context, cpu);
- spin_lock_irq(&cpuctx->ctx.lock);
+ raw_spin_lock_irq(&cpuctx->ctx.lock);
mpt = min(perf_max_events - cpuctx->ctx.nr_events,
perf_max_events - perf_reserved_percpu);
cpuctx->max_pertask = mpt;
- spin_unlock_irq(&cpuctx->ctx.lock);
+ raw_spin_unlock_irq(&cpuctx->ctx.lock);
}
spin_unlock(&perf_resource_lock);
git://ftp.safe.ca / safe / jmp / linux-2.6 / blobdiff