/*
* 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);
}
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;
}
*/
if (!list_empty(&event->group_entry))
list_del_event(event, ctx);
- spin_unlock_irq(&ctx->lock);
+ 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;
}
perf_enable();
out:
- spin_unlock(&ctx->lock);
+ raw_spin_unlock(&ctx->lock);
}
/*
* 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);
+ 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:
if (ctx->task && cpuctx->task_ctx != ctx)
return;
- spin_lock(&ctx->lock);
+ raw_spin_lock(&ctx->lock);
update_context_time(ctx);
update_event_times(event);
- spin_unlock(&ctx->lock);
+ raw_spin_unlock(&ctx->lock);
event->pmu->read(event);
}
struct perf_event_context *ctx = event->ctx;
unsigned long flags;
- spin_lock_irqsave(&ctx->lock, flags);
+ raw_spin_lock_irqsave(&ctx->lock, flags);
update_context_time(ctx);
update_event_times(event);
- spin_unlock_irqrestore(&ctx->lock, flags);
+ 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;
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;
}
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;
rcu_read_lock();
cpuctx = &get_cpu_var(perf_cpu_context);
perf_event_task_ctx(&cpuctx->ctx, task_event);
- put_cpu_var(perf_cpu_context);
-
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;
rcu_read_lock();
cpuctx = &get_cpu_var(perf_cpu_context);
perf_event_comm_ctx(&cpuctx->ctx, comm_event);
- put_cpu_var(perf_cpu_context);
-
- /*
- * 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;
rcu_read_lock();
cpuctx = &get_cpu_var(perf_cpu_context);
perf_event_mmap_ctx(&cpuctx->ctx, mmap_event);
- put_cpu_var(perf_cpu_context);
-
- /*
- * 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);
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;
event->pmu->read(event);
data.addr = 0;
+ data.raw = NULL;
data.period = event->hw.last_period;
regs = get_irq_regs();
/*
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);
}
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))
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;
* 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
*/
unclone_ctx(child_ctx);
update_context_time(child_ctx);
- spin_unlock_irqrestore(&child_ctx->lock, flags);
+ 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