1 files changed, 478 insertions, 156 deletions

diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index 671f6c8c8a3..11847bf1e8c 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -13,6 +13,7 @@
 #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>
@@ -21,7 +22,9 @@
 #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>
@@ -133,6 +136,28 @@ static void unclone_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.
@@ -312,9 +337,84 @@ list_add_event(struct perf_event *event, struct perf_event_context *ctx)
 		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.
@@ -333,6 +433,11 @@ static void perf_group_attach(struct perf_event *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);
 }
 
 /*
@@ -391,7 +496,7 @@ static void perf_group_detach(struct perf_event *event)
 	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))
@@ -410,6 +515,12 @@ static void perf_group_detach(struct perf_event *event)
 		/* 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
@@ -1073,7 +1184,7 @@ static int perf_event_refresh(struct perf_event *event, int refresh)
 	/*
 	 * 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);
@@ -1287,8 +1398,6 @@ void __perf_event_task_sched_out(struct task_struct *task,
 {
 	int ctxn;
 
-	perf_sw_event(PERF_COUNT_SW_CONTEXT_SWITCHES, 1, 1, NULL, 0);
-
 	for_each_task_context_nr(ctxn)
 		perf_event_context_sched_out(task, ctxn, next);
 }
@@ -1622,8 +1731,12 @@ static void rotate_ctx(struct perf_event_context *ctx)
 {
 	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);
 }
@@ -2287,31 +2400,6 @@ static int perf_release(struct inode *inode, struct file *file)
 	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);
-	}
-
-	size += entry * nr;
-
-	return size;
-}
-
 u64 perf_event_read_value(struct perf_event *event, u64 *enabled, u64 *running)
 {
 	struct perf_event *child;
@@ -2426,7 +2514,7 @@ perf_read_hw(struct perf_event *event, char __user *buf, size_t count)
 	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);
@@ -2512,7 +2600,7 @@ static int perf_event_period(struct perf_event *event, u64 __user *arg)
 	int ret = 0;
 	u64 value;
 
-	if (!event->attr.sample_period)
+	if (!is_sampling_event(event))
 		return -EINVAL;
 
 	if (copy_from_user(&value, arg, sizeof(value)))
@@ -3303,6 +3391,73 @@ __always_inline void perf_output_copy(struct perf_output_handle *handle,
 	} 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)
@@ -3310,6 +3465,7 @@ int perf_output_begin(struct perf_output_handle *handle,
 	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;
@@ -3336,8 +3492,12 @@ int perf_output_begin(struct perf_output_handle *handle,
 		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);
 
@@ -3368,11 +3528,11 @@ int perf_output_begin(struct perf_output_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;
@@ -3405,28 +3565,6 @@ void perf_output_end(struct perf_output_handle *handle)
 	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,
 				 u64 enabled, u64 running)
@@ -3601,61 +3739,16 @@ void perf_prepare_sample(struct perf_event_header *header,
 {
 	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) {
-		data->ip = perf_instruction_pointer(regs);
-
-		header->size += sizeof(data->ip);
-	}
-
-	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);
+	__perf_event_header__init_id(header, data, event);
 
-		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_IP)
+		data->ip = perf_instruction_pointer(regs);
 
 	if (sample_type & PERF_SAMPLE_CALLCHAIN) {
 		int size = 1;
@@ -3720,23 +3813,26 @@ perf_event_read_event(struct perf_event *event,
 			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);
 }
@@ -3766,14 +3862,16 @@ static void perf_event_task_output(struct perf_event *event,
 				     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);
@@ -3783,7 +3881,11 @@ static void perf_event_task_output(struct perf_event *event,
 
 	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)
@@ -3822,6 +3924,8 @@ static void perf_event_task_event(struct perf_task_event *task_event)
 	rcu_read_lock();
 	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;
@@ -3896,11 +4000,16 @@ static void perf_event_comm_output(struct perf_event *event,
 				     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);
@@ -3908,7 +4017,12 @@ static void perf_event_comm_output(struct perf_event *event,
 	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)
@@ -3953,10 +4067,11 @@ static void perf_event_comm_event(struct perf_comm_event *comm_event)
 	comm_event->comm_size = size;
 
 	comm_event->event_id.header.size = sizeof(comm_event->event_id) + size;
-
 	rcu_read_lock();
 	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;
@@ -4032,11 +4147,15 @@ static void perf_event_mmap_output(struct perf_event *event,
 				     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);
@@ -4044,7 +4163,12 @@ static void perf_event_mmap_output(struct perf_event *event,
 	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,
@@ -4142,6 +4266,8 @@ got_name:
 	rcu_read_lock();
 	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);
 
@@ -4197,6 +4323,7 @@ void perf_event_mmap(struct vm_area_struct *vma)
 static void perf_log_throttle(struct perf_event *event, int enable)
 {
 	struct perf_output_handle handle;
+	struct perf_sample_data sample;
 	int ret;
 
 	struct {
@@ -4218,11 +4345,15 @@ static void perf_log_throttle(struct perf_event *event, int enable)
 	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);
 }
 
@@ -4238,6 +4369,13 @@ static int __perf_event_overflow(struct perf_event *event, int nmi,
 	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 {
@@ -4383,7 +4521,7 @@ static void perf_swevent_event(struct perf_event *event, u64 nr,
 	if (!regs)
 		return;
 
-	if (!hwc->sample_period)
+	if (!is_sampling_event(event))
 		return;
 
 	if (nr == 1 && hwc->sample_period == 1 && !event->attr.freq)
@@ -4546,7 +4684,7 @@ static int perf_swevent_add(struct perf_event *event, int flags)
 	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);
 	}
@@ -4711,7 +4849,7 @@ static int perf_swevent_init(struct perf_event *event)
 		break;
 	}
 
-	if (event_id > PERF_COUNT_SW_MAX)
+	if (event_id >= PERF_COUNT_SW_MAX)
 		return -ENOENT;
 
 	if (!event->parent) {
@@ -4803,15 +4941,6 @@ static int perf_tp_event_init(struct perf_event *event)
 	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;
@@ -4834,7 +4963,7 @@ static struct pmu perf_tracepoint = {
 
 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)
@@ -4924,31 +5053,33 @@ static enum hrtimer_restart perf_swevent_hrtimer(struct hrtimer *hrtimer)
 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));
 
@@ -5143,25 +5274,94 @@ static void *find_pmu_context(int ctxn)
 	return NULL;
 }
 
-static void free_pmu_context(void * __percpu cpu_context)
+static void update_pmu_context(struct pmu *pmu, struct pmu *old_pmu)
 {
-	struct pmu *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(pmu, &pmus, entry) {
-		if (pmu->pmu_cpu_context == cpu_context)
+	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(cpu_context);
+	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,
+};
 
-int perf_pmu_register(struct pmu *pmu)
+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;
+}
+
+int perf_pmu_register(struct pmu *pmu, char *name, int type)
 {
 	int cpu, ret;
 
@@ -5171,13 +5371,38 @@ int perf_pmu_register(struct pmu *pmu)
 	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_pdc;
+		goto free_dev;
 
 	for_each_possible_cpu(cpu) {
 		struct perf_cpu_context *cpuctx;
@@ -5188,6 +5413,7 @@ int perf_pmu_register(struct pmu *pmu)
 		cpuctx->ctx.pmu = pmu;
 		cpuctx->jiffies_interval = 1;
 		INIT_LIST_HEAD(&cpuctx->rotation_list);
+		cpuctx->active_pmu = pmu;
 	}
 
 got_cpu_context:
@@ -5220,6 +5446,14 @@ unlock:
 
 	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;
@@ -5239,7 +5473,11 @@ void perf_pmu_unregister(struct pmu *pmu)
 	synchronize_rcu();
 
 	free_percpu(pmu->pmu_disable_count);
-	free_pmu_context(pmu->pmu_cpu_context);
+	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)
@@ -5248,6 +5486,13 @@ 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)
@@ -5713,6 +5958,12 @@ SYSCALL_DEFINE5(perf_event_open,
 	mutex_unlock(&current->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
@@ -6065,6 +6316,12 @@ inherit_event(struct perf_event *parent_event,
 	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);
@@ -6162,6 +6419,7 @@ int perf_event_init_context(struct task_struct *child, int ctxn)
 	struct perf_event *event;
 	struct task_struct *parent = current;
 	int inherited_all = 1;
+	unsigned long flags;
 	int ret = 0;
 
 	child->perf_event_ctxp[ctxn] = NULL;
@@ -6202,6 +6460,15 @@ int perf_event_init_context(struct task_struct *child, int ctxn)
 			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, ctxn, &inherited_all);
@@ -6209,6 +6476,10 @@ int perf_event_init_context(struct task_struct *child, int ctxn)
 			break;
 	}
 
+	raw_spin_lock_irqsave(&parent_ctx->lock, flags);
+	parent_ctx->rotate_disable = 0;
+	raw_spin_unlock_irqrestore(&parent_ctx->lock, flags);
+
 	child_ctx = child->perf_event_ctxp[ctxn];
 
 	if (child_ctx && inherited_all) {
@@ -6281,7 +6552,7 @@ static void __cpuinit perf_event_init_cpu(int cpu)
 	mutex_unlock(&swhash->hlist_mutex);
 }
 
-#ifdef CONFIG_HOTPLUG_CPU
+#if defined CONFIG_HOTPLUG_CPU || defined CONFIG_KEXEC
 static void perf_pmu_rotate_stop(struct pmu *pmu)
 {
 	struct perf_cpu_context *cpuctx = this_cpu_ptr(pmu->pmu_cpu_context);
@@ -6335,6 +6606,26 @@ static void perf_event_exit_cpu(int cpu)
 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)
 {
@@ -6363,14 +6654,45 @@ 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);