1 files changed, 253 insertions, 103 deletions

diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c
index b0b20a07f39..d7cbc579fc8 100644
--- a/kernel/perf_counter.c
+++ b/kernel/perf_counter.c
@@ -50,7 +50,7 @@ static atomic_t nr_task_counters __read_mostly;
  *  1 - disallow cpu counters to unpriv
  *  2 - disallow kernel profiling to unpriv
  */
-int sysctl_perf_counter_paranoid __read_mostly;
+int sysctl_perf_counter_paranoid __read_mostly = 1;
 
 static inline bool perf_paranoid_cpu(void)
 {
@@ -88,6 +88,7 @@ void __weak hw_perf_disable(void)		{ barrier(); }
 void __weak hw_perf_enable(void)		{ barrier(); }
 
 void __weak hw_perf_counter_setup(int cpu)	{ barrier(); }
+void __weak hw_perf_counter_setup_online(int cpu)	{ barrier(); }
 
 int __weak
 hw_perf_group_sched_in(struct perf_counter *group_leader,
@@ -306,6 +307,10 @@ counter_sched_out(struct perf_counter *counter,
 		return;
 
 	counter->state = PERF_COUNTER_STATE_INACTIVE;
+	if (counter->pending_disable) {
+		counter->pending_disable = 0;
+		counter->state = PERF_COUNTER_STATE_OFF;
+	}
 	counter->tstamp_stopped = ctx->time;
 	counter->pmu->disable(counter);
 	counter->oncpu = -1;
@@ -1498,10 +1503,21 @@ static void perf_counter_enable_on_exec(struct task_struct *task)
  */
 static void __perf_counter_read(void *info)
 {
+	struct perf_cpu_context *cpuctx = &__get_cpu_var(perf_cpu_context);
 	struct perf_counter *counter = info;
 	struct perf_counter_context *ctx = counter->ctx;
 	unsigned long flags;
 
+	/*
+	 * If this is a task context, we need to check whether it is
+	 * the current task context of this cpu.  If not it has been
+	 * scheduled out before the smp call arrived.  In that case
+	 * counter->count would have been updated to a recent sample
+	 * when the counter was scheduled out.
+	 */
+	if (ctx->task && cpuctx->task_ctx != ctx)
+		return;
+
 	local_irq_save(flags);
 	if (ctx->is_active)
 		update_context_time(ctx);
@@ -1691,7 +1707,32 @@ static int perf_release(struct inode *inode, struct file *file)
 	return 0;
 }
 
-static u64 perf_counter_read_tree(struct perf_counter *counter)
+static int perf_counter_read_size(struct perf_counter *counter)
+{
+	int entry = sizeof(u64); /* value */
+	int size = 0;
+	int nr = 1;
+
+	if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
+		size += sizeof(u64);
+
+	if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
+		size += sizeof(u64);
+
+	if (counter->attr.read_format & PERF_FORMAT_ID)
+		entry += sizeof(u64);
+
+	if (counter->attr.read_format & PERF_FORMAT_GROUP) {
+		nr += counter->group_leader->nr_siblings;
+		size += sizeof(u64);
+	}
+
+	size += entry * nr;
+
+	return size;
+}
+
+static u64 perf_counter_read_value(struct perf_counter *counter)
 {
 	struct perf_counter *child;
 	u64 total = 0;
@@ -1703,14 +1744,96 @@ static u64 perf_counter_read_tree(struct perf_counter *counter)
 	return total;
 }
 
+static int perf_counter_read_entry(struct perf_counter *counter,
+				   u64 read_format, char __user *buf)
+{
+	int n = 0, count = 0;
+	u64 values[2];
+
+	values[n++] = perf_counter_read_value(counter);
+	if (read_format & PERF_FORMAT_ID)
+		values[n++] = primary_counter_id(counter);
+
+	count = n * sizeof(u64);
+
+	if (copy_to_user(buf, values, count))
+		return -EFAULT;
+
+	return count;
+}
+
+static int perf_counter_read_group(struct perf_counter *counter,
+				   u64 read_format, char __user *buf)
+{
+	struct perf_counter *leader = counter->group_leader, *sub;
+	int n = 0, size = 0, err = -EFAULT;
+	u64 values[3];
+
+	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);
+	}
+
+	size = n * sizeof(u64);
+
+	if (copy_to_user(buf, values, size))
+		return -EFAULT;
+
+	err = perf_counter_read_entry(leader, read_format, buf + size);
+	if (err < 0)
+		return err;
+
+	size += err;
+
+	list_for_each_entry(sub, &leader->sibling_list, list_entry) {
+		err = perf_counter_read_entry(sub, read_format,
+				buf + size);
+		if (err < 0)
+			return err;
+
+		size += err;
+	}
+
+	return size;
+}
+
+static int perf_counter_read_one(struct perf_counter *counter,
+				 u64 read_format, char __user *buf)
+{
+	u64 values[4];
+	int n = 0;
+
+	values[n++] = perf_counter_read_value(counter);
+	if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) {
+		values[n++] = counter->total_time_enabled +
+			atomic64_read(&counter->child_total_time_enabled);
+	}
+	if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) {
+		values[n++] = counter->total_time_running +
+			atomic64_read(&counter->child_total_time_running);
+	}
+	if (read_format & PERF_FORMAT_ID)
+		values[n++] = primary_counter_id(counter);
+
+	if (copy_to_user(buf, values, n * sizeof(u64)))
+		return -EFAULT;
+
+	return n * sizeof(u64);
+}
+
 /*
  * Read the performance counter - simple non blocking version for now
  */
 static ssize_t
 perf_read_hw(struct perf_counter *counter, char __user *buf, size_t count)
 {
-	u64 values[4];
-	int n;
+	u64 read_format = counter->attr.read_format;
+	int ret;
 
 	/*
 	 * Return end-of-file for a read on a counter that is in
@@ -1720,28 +1843,18 @@ perf_read_hw(struct perf_counter *counter, char __user *buf, size_t count)
 	if (counter->state == PERF_COUNTER_STATE_ERROR)
 		return 0;
 
+	if (count < perf_counter_read_size(counter))
+		return -ENOSPC;
+
 	WARN_ON_ONCE(counter->ctx->parent_ctx);
 	mutex_lock(&counter->child_mutex);
-	values[0] = perf_counter_read_tree(counter);
-	n = 1;
-	if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
-		values[n++] = counter->total_time_enabled +
-			atomic64_read(&counter->child_total_time_enabled);
-	if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
-		values[n++] = counter->total_time_running +
-			atomic64_read(&counter->child_total_time_running);
-	if (counter->attr.read_format & PERF_FORMAT_ID)
-		values[n++] = primary_counter_id(counter);
+	if (read_format & PERF_FORMAT_GROUP)
+		ret = perf_counter_read_group(counter, read_format, buf);
+	else
+		ret = perf_counter_read_one(counter, read_format, buf);
 	mutex_unlock(&counter->child_mutex);
 
-	if (count < n * sizeof(u64))
-		return -EINVAL;
-	count = n * sizeof(u64);
-
-	if (copy_to_user(buf, values, count))
-		return -EFAULT;
-
-	return count;
+	return ret;
 }
 
 static ssize_t
@@ -1906,6 +2019,10 @@ int perf_counter_task_disable(void)
 	return 0;
 }
 
+#ifndef PERF_COUNTER_INDEX_OFFSET
+# define PERF_COUNTER_INDEX_OFFSET 0
+#endif
+
 static int perf_counter_index(struct perf_counter *counter)
 {
 	if (counter->state != PERF_COUNTER_STATE_ACTIVE)
@@ -2245,7 +2362,7 @@ static void perf_pending_counter(struct perf_pending_entry *entry)
 
 	if (counter->pending_disable) {
 		counter->pending_disable = 0;
-		perf_counter_disable(counter);
+		__perf_counter_disable(counter);
 	}
 
 	if (counter->pending_wakeup) {
@@ -2630,7 +2747,80 @@ static u32 perf_counter_tid(struct perf_counter *counter, struct task_struct *p)
 	return task_pid_nr_ns(p, counter->ns);
 }
 
-static void perf_counter_output(struct perf_counter *counter, int nmi,
+static void perf_output_read_one(struct perf_output_handle *handle,
+				 struct perf_counter *counter)
+{
+	u64 read_format = counter->attr.read_format;
+	u64 values[4];
+	int n = 0;
+
+	values[n++] = atomic64_read(&counter->count);
+	if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) {
+		values[n++] = counter->total_time_enabled +
+			atomic64_read(&counter->child_total_time_enabled);
+	}
+	if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) {
+		values[n++] = counter->total_time_running +
+			atomic64_read(&counter->child_total_time_running);
+	}
+	if (read_format & PERF_FORMAT_ID)
+		values[n++] = primary_counter_id(counter);
+
+	perf_output_copy(handle, values, n * sizeof(u64));
+}
+
+/*
+ * XXX PERF_FORMAT_GROUP vs inherited counters seems difficult.
+ */
+static void perf_output_read_group(struct perf_output_handle *handle,
+			    struct perf_counter *counter)
+{
+	struct perf_counter *leader = counter->group_leader, *sub;
+	u64 read_format = counter->attr.read_format;
+	u64 values[5];
+	int n = 0;
+
+	values[n++] = 1 + leader->nr_siblings;
+
+	if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
+		values[n++] = leader->total_time_enabled;
+
+	if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
+		values[n++] = leader->total_time_running;
+
+	if (leader != counter)
+		leader->pmu->read(leader);
+
+	values[n++] = atomic64_read(&leader->count);
+	if (read_format & PERF_FORMAT_ID)
+		values[n++] = primary_counter_id(leader);
+
+	perf_output_copy(handle, values, n * sizeof(u64));
+
+	list_for_each_entry(sub, &leader->sibling_list, list_entry) {
+		n = 0;
+
+		if (sub != counter)
+			sub->pmu->read(sub);
+
+		values[n++] = atomic64_read(&sub->count);
+		if (read_format & PERF_FORMAT_ID)
+			values[n++] = primary_counter_id(sub);
+
+		perf_output_copy(handle, values, n * sizeof(u64));
+	}
+}
+
+static void perf_output_read(struct perf_output_handle *handle,
+			     struct perf_counter *counter)
+{
+	if (counter->attr.read_format & PERF_FORMAT_GROUP)
+		perf_output_read_group(handle, counter);
+	else
+		perf_output_read_one(handle, counter);
+}
+
+void perf_counter_output(struct perf_counter *counter, int nmi,
 				struct perf_sample_data *data)
 {
 	int ret;
@@ -2641,10 +2831,6 @@ static void perf_counter_output(struct perf_counter *counter, int nmi,
 	struct {
 		u32 pid, tid;
 	} tid_entry;
-	struct {
-		u64 id;
-		u64 counter;
-	} group_entry;
 	struct perf_callchain_entry *callchain = NULL;
 	int callchain_size = 0;
 	u64 time;
@@ -2699,10 +2885,8 @@ static void perf_counter_output(struct perf_counter *counter, int nmi,
 	if (sample_type & PERF_SAMPLE_PERIOD)
 		header.size += sizeof(u64);
 
-	if (sample_type & PERF_SAMPLE_GROUP) {
-		header.size += sizeof(u64) +
-			counter->nr_siblings * sizeof(group_entry);
-	}
+	if (sample_type & PERF_SAMPLE_READ)
+		header.size += perf_counter_read_size(counter);
 
 	if (sample_type & PERF_SAMPLE_CALLCHAIN) {
 		callchain = perf_callchain(data->regs);
@@ -2759,26 +2943,8 @@ static void perf_counter_output(struct perf_counter *counter, int nmi,
 	if (sample_type & PERF_SAMPLE_PERIOD)
 		perf_output_put(&handle, data->period);
 
-	/*
-	 * XXX PERF_SAMPLE_GROUP vs inherited counters seems difficult.
-	 */
-	if (sample_type & PERF_SAMPLE_GROUP) {
-		struct perf_counter *leader, *sub;
-		u64 nr = counter->nr_siblings;
-
-		perf_output_put(&handle, nr);
-
-		leader = counter->group_leader;
-		list_for_each_entry(sub, &leader->sibling_list, list_entry) {
-			if (sub != counter)
-				sub->pmu->read(sub);
-
-			group_entry.id = primary_counter_id(sub);
-			group_entry.counter = atomic64_read(&sub->count);
-
-			perf_output_put(&handle, group_entry);
-		}
-	}
+	if (sample_type & PERF_SAMPLE_READ)
+		perf_output_read(&handle, counter);
 
 	if (sample_type & PERF_SAMPLE_CALLCHAIN) {
 		if (callchain)
@@ -2817,8 +2983,6 @@ struct perf_read_event {
 
 	u32				pid;
 	u32				tid;
-	u64				value;
-	u64				format[3];
 };
 
 static void
@@ -2830,34 +2994,20 @@ perf_counter_read_event(struct perf_counter *counter,
 		.header = {
 			.type = PERF_EVENT_READ,
 			.misc = 0,
-			.size = sizeof(event) - sizeof(event.format),
+			.size = sizeof(event) + perf_counter_read_size(counter),
 		},
 		.pid = perf_counter_pid(counter, task),
 		.tid = perf_counter_tid(counter, task),
-		.value = atomic64_read(&counter->count),
 	};
-	int ret, i = 0;
-
-	if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) {
-		event.header.size += sizeof(u64);
-		event.format[i++] = counter->total_time_enabled;
-	}
-
-	if (counter->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) {
-		event.header.size += sizeof(u64);
-		event.format[i++] = counter->total_time_running;
-	}
-
-	if (counter->attr.read_format & PERF_FORMAT_ID) {
-		event.header.size += sizeof(u64);
-		event.format[i++] = primary_counter_id(counter);
-	}
+	int ret;
 
 	ret = perf_output_begin(&handle, counter, event.header.size, 0, 0);
 	if (ret)
 		return;
 
-	perf_output_copy(&handle, &event, event.header.size);
+	perf_output_put(&handle, event);
+	perf_output_read(&handle, counter);
+
 	perf_output_end(&handle);
 }
 
@@ -2893,10 +3043,10 @@ static void perf_counter_task_output(struct perf_counter *counter,
 		return;
 
 	task_event->event.pid = perf_counter_pid(counter, task);
-	task_event->event.ppid = perf_counter_pid(counter, task->real_parent);
+	task_event->event.ppid = perf_counter_pid(counter, current);
 
 	task_event->event.tid = perf_counter_tid(counter, task);
-	task_event->event.ptid = perf_counter_tid(counter, task->real_parent);
+	task_event->event.ptid = perf_counter_tid(counter, current);
 
 	perf_output_put(&handle, task_event->event);
 	perf_output_end(&handle);
@@ -3443,40 +3593,32 @@ static void perf_swcounter_add(struct perf_counter *counter, u64 nr,
 
 static int perf_swcounter_is_counting(struct perf_counter *counter)
 {
-	struct perf_counter_context *ctx;
-	unsigned long flags;
-	int count;
-
+	/*
+	 * The counter is active, we're good!
+	 */
 	if (counter->state == PERF_COUNTER_STATE_ACTIVE)
 		return 1;
 
+	/*
+	 * The counter is off/error, not counting.
+	 */
 	if (counter->state != PERF_COUNTER_STATE_INACTIVE)
 		return 0;
 
 	/*
-	 * If the counter is inactive, it could be just because
-	 * its task is scheduled out, or because it's in a group
-	 * which could not go on the PMU.  We want to count in
-	 * the first case but not the second.  If the context is
-	 * currently active then an inactive software counter must
-	 * be the second case.  If it's not currently active then
-	 * we need to know whether the counter was active when the
-	 * context was last active, which we can determine by
-	 * comparing counter->tstamp_stopped with ctx->time.
-	 *
-	 * We are within an RCU read-side critical section,
-	 * which protects the existence of *ctx.
+	 * The counter is inactive, if the context is active
+	 * we're part of a group that didn't make it on the 'pmu',
+	 * not counting.
 	 */
-	ctx = counter->ctx;
-	spin_lock_irqsave(&ctx->lock, flags);
-	count = 1;
-	/* Re-check state now we have the lock */
-	if (counter->state < PERF_COUNTER_STATE_INACTIVE ||
-	    counter->ctx->is_active ||
-	    counter->tstamp_stopped < ctx->time)
-		count = 0;
-	spin_unlock_irqrestore(&ctx->lock, flags);
-	return count;
+	if (counter->ctx->is_active)
+		return 0;
+
+	/*
+	 * We're inactive and the context is too, this means the
+	 * task is scheduled out, we're counting events that happen
+	 * to us, like migration events.
+	 */
+	return 1;
 }
 
 static int perf_swcounter_match(struct perf_counter *counter,
@@ -3924,13 +4066,14 @@ perf_counter_alloc(struct perf_counter_attr *attr,
 	hwc->sample_period = attr->sample_period;
 	if (attr->freq && attr->sample_freq)
 		hwc->sample_period = 1;
+	hwc->last_period = hwc->sample_period;
 
 	atomic64_set(&hwc->period_left, hwc->sample_period);
 
 	/*
-	 * we currently do not support PERF_SAMPLE_GROUP on inherited counters
+	 * we currently do not support PERF_FORMAT_GROUP on inherited counters
 	 */
-	if (attr->inherit && (attr->sample_type & PERF_SAMPLE_GROUP))
+	if (attr->inherit && (attr->read_format & PERF_FORMAT_GROUP))
 		goto done;
 
 	switch (attr->type) {
@@ -4592,6 +4735,11 @@ perf_cpu_notify(struct notifier_block *self, unsigned long action, void *hcpu)
 		perf_counter_init_cpu(cpu);
 		break;
 
+	case CPU_ONLINE:
+	case CPU_ONLINE_FROZEN:
+		hw_perf_counter_setup_online(cpu);
+		break;
+
 	case CPU_DOWN_PREPARE:
 	case CPU_DOWN_PREPARE_FROZEN:
 		perf_counter_exit_cpu(cpu);
@@ -4616,6 +4764,8 @@ void __init perf_counter_init(void)
 {
 	perf_cpu_notify(&perf_cpu_nb, (unsigned long)CPU_UP_PREPARE,
 			(void *)(long)smp_processor_id());
+	perf_cpu_notify(&perf_cpu_nb, (unsigned long)CPU_ONLINE,
+			(void *)(long)smp_processor_id());
 	register_cpu_notifier(&perf_cpu_nb);
 }