1 files changed, 163 insertions, 55 deletions

diff --git a/kernel/events/core.c b/kernel/events/core.c
index e6e11b598438..1f215aa8f30d 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -798,7 +798,7 @@ static DEFINE_PER_CPU(struct list_head, cgrp_cpuctx_list);
  */
 static void perf_cgroup_switch(struct task_struct *task, int mode)
 {
-	struct perf_cpu_context *cpuctx;
+	struct perf_cpu_context *cpuctx, *tmp;
 	struct list_head *list;
 	unsigned long flags;
 
@@ -809,7 +809,7 @@ static void perf_cgroup_switch(struct task_struct *task, int mode)
 	local_irq_save(flags);
 
 	list = this_cpu_ptr(&cgrp_cpuctx_list);
-	list_for_each_entry(cpuctx, list, cgrp_cpuctx_entry) {
+	list_for_each_entry_safe(cpuctx, tmp, list, cgrp_cpuctx_entry) {
 		WARN_ON_ONCE(cpuctx->ctx.nr_cgroups == 0);
 
 		perf_ctx_lock(cpuctx, cpuctx->task_ctx);
@@ -1133,6 +1133,11 @@ static int perf_mux_hrtimer_restart(struct perf_cpu_context *cpuctx)
 	return 0;
 }
 
+static int perf_mux_hrtimer_restart_ipi(void *arg)
+{
+	return perf_mux_hrtimer_restart(arg);
+}
+
 void perf_pmu_disable(struct pmu *pmu)
 {
 	int *count = this_cpu_ptr(pmu->pmu_disable_count);
@@ -1705,28 +1710,34 @@ static inline void perf_event__state_init(struct perf_event *event)
 					      PERF_EVENT_STATE_INACTIVE;
 }
 
-static void __perf_event_read_size(struct perf_event *event, int nr_siblings)
+static int __perf_event_read_size(u64 read_format, int nr_siblings)
 {
 	int entry = sizeof(u64); /* value */
 	int size = 0;
 	int nr = 1;
 
-	if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
+	if (read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
 		size += sizeof(u64);
 
-	if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
+	if (read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
 		size += sizeof(u64);
 
-	if (event->attr.read_format & PERF_FORMAT_ID)
+	if (read_format & PERF_FORMAT_ID)
+		entry += sizeof(u64);
+
+	if (read_format & PERF_FORMAT_LOST)
 		entry += sizeof(u64);
 
-	if (event->attr.read_format & PERF_FORMAT_GROUP) {
+	if (read_format & PERF_FORMAT_GROUP) {
 		nr += nr_siblings;
 		size += sizeof(u64);
 	}
 
-	size += entry * nr;
-	event->read_size = size;
+	/*
+	 * Since perf_event_validate_size() limits this to 16k and inhibits
+	 * adding more siblings, this will never overflow.
+	 */
+	return size + nr * entry;
 }
 
 static void __perf_event_header_size(struct perf_event *event, u64 sample_type)
@@ -1767,8 +1778,9 @@ static void __perf_event_header_size(struct perf_event *event, u64 sample_type)
  */
 static void perf_event__header_size(struct perf_event *event)
 {
-	__perf_event_read_size(event,
-			       event->group_leader->nr_siblings);
+	event->read_size =
+		__perf_event_read_size(event->attr.read_format,
+				       event->group_leader->nr_siblings);
 	__perf_event_header_size(event, event->attr.sample_type);
 }
 
@@ -1799,23 +1811,44 @@ static void perf_event__id_header_size(struct perf_event *event)
 	event->id_header_size = size;
 }
 
+/*
+ * Check that adding an event to the group does not result in anybody
+ * overflowing the 64k event limit imposed by the output buffer.
+ *
+ * Specifically, check that the read_size for the event does not exceed 16k,
+ * read_size being the one term that grows with groups size. Since read_size
+ * depends on per-event read_format, also (re)check the existing events.
+ *
+ * This leaves 48k for the constant size fields and things like callchains,
+ * branch stacks and register sets.
+ */
 static bool perf_event_validate_size(struct perf_event *event)
 {
-	/*
-	 * The values computed here will be over-written when we actually
-	 * attach the event.
-	 */
-	__perf_event_read_size(event, event->group_leader->nr_siblings + 1);
-	__perf_event_header_size(event, event->attr.sample_type & ~PERF_SAMPLE_READ);
-	perf_event__id_header_size(event);
+	struct perf_event *sibling, *group_leader = event->group_leader;
+
+	if (__perf_event_read_size(event->attr.read_format,
+				   group_leader->nr_siblings + 1) > 16*1024)
+		return false;
+
+	if (__perf_event_read_size(group_leader->attr.read_format,
+				   group_leader->nr_siblings + 1) > 16*1024)
+		return false;
 
 	/*
-	 * Sum the lot; should not exceed the 64k limit we have on records.
-	 * Conservative limit to allow for callchains and other variable fields.
+	 * When creating a new group leader, group_leader->ctx is initialized
+	 * after the size has been validated, but we cannot safely use
+	 * for_each_sibling_event() until group_leader->ctx is set. A new group
+	 * leader cannot have any siblings yet, so we can safely skip checking
+	 * the non-existent siblings.
 	 */
-	if (event->read_size + event->header_size +
-	    event->id_header_size + sizeof(struct perf_event_header) >= 16*1024)
-		return false;
+	if (event == group_leader)
+		return true;
+
+	for_each_sibling_event(sibling, group_leader) {
+		if (__perf_event_read_size(sibling->attr.read_format,
+					   group_leader->nr_siblings + 1) > 16*1024)
+			return false;
+	}
 
 	return true;
 }
@@ -1843,6 +1876,7 @@ static void perf_group_attach(struct perf_event *event)
 
 	list_add_tail(&event->sibling_list, &group_leader->sibling_list);
 	group_leader->nr_siblings++;
+	group_leader->group_generation++;
 
 	perf_event__header_size(group_leader);
 
@@ -1913,6 +1947,7 @@ static void perf_group_detach(struct perf_event *event)
 	if (event->group_leader != event) {
 		list_del_init(&event->sibling_list);
 		event->group_leader->nr_siblings--;
+		event->group_leader->group_generation++;
 		goto out;
 	}
 
@@ -4750,7 +4785,7 @@ static int __perf_read_group_add(struct perf_event *leader,
 					u64 read_format, u64 *values)
 {
 	struct perf_event_context *ctx = leader->ctx;
-	struct perf_event *sub;
+	struct perf_event *sub, *parent;
 	unsigned long flags;
 	int n = 1; /* skip @nr */
 	int ret;
@@ -4760,6 +4795,33 @@ static int __perf_read_group_add(struct perf_event *leader,
 		return ret;
 
 	raw_spin_lock_irqsave(&ctx->lock, flags);
+	/*
+	 * Verify the grouping between the parent and child (inherited)
+	 * events is still in tact.
+	 *
+	 * Specifically:
+	 *  - leader->ctx->lock pins leader->sibling_list
+	 *  - parent->child_mutex pins parent->child_list
+	 *  - parent->ctx->mutex pins parent->sibling_list
+	 *
+	 * Because parent->ctx != leader->ctx (and child_list nests inside
+	 * ctx->mutex), group destruction is not atomic between children, also
+	 * see perf_event_release_kernel(). Additionally, parent can grow the
+	 * group.
+	 *
+	 * Therefore it is possible to have parent and child groups in a
+	 * different configuration and summing over such a beast makes no sense
+	 * what so ever.
+	 *
+	 * Reject this.
+	 */
+	parent = leader->parent;
+	if (parent &&
+	    (parent->group_generation != leader->group_generation ||
+	     parent->nr_siblings != leader->nr_siblings)) {
+		ret = -ECHILD;
+		goto unlock;
+	}
 
 	/*
 	 * Since we co-schedule groups, {enabled,running} times of siblings
@@ -4782,15 +4844,20 @@ static int __perf_read_group_add(struct perf_event *leader,
 	values[n++] += perf_event_count(leader);
 	if (read_format & PERF_FORMAT_ID)
 		values[n++] = primary_event_id(leader);
+	if (read_format & PERF_FORMAT_LOST)
+		values[n++] = atomic64_read(&leader->lost_samples);
 
 	for_each_sibling_event(sub, leader) {
 		values[n++] += perf_event_count(sub);
 		if (read_format & PERF_FORMAT_ID)
 			values[n++] = primary_event_id(sub);
+		if (read_format & PERF_FORMAT_LOST)
+			values[n++] = atomic64_read(&sub->lost_samples);
 	}
 
+unlock:
 	raw_spin_unlock_irqrestore(&ctx->lock, flags);
-	return 0;
+	return ret;
 }
 
 static int perf_read_group(struct perf_event *event,
@@ -4809,10 +4876,6 @@ static int perf_read_group(struct perf_event *event,
 
 	values[0] = 1 + leader->nr_siblings;
 
-	/*
-	 * By locking the child_mutex of the leader we effectively
-	 * lock the child list of all siblings.. XXX explain how.
-	 */
 	mutex_lock(&leader->child_mutex);
 
 	ret = __perf_read_group_add(leader, read_format, values);
@@ -4843,7 +4906,7 @@ static int perf_read_one(struct perf_event *event,
 				 u64 read_format, char __user *buf)
 {
 	u64 enabled, running;
-	u64 values[4];
+	u64 values[5];
 	int n = 0;
 
 	values[n++] = __perf_event_read_value(event, &enabled, &running);
@@ -4853,6 +4916,8 @@ static int perf_read_one(struct perf_event *event,
 		values[n++] = running;
 	if (read_format & PERF_FORMAT_ID)
 		values[n++] = primary_event_id(event);
+	if (read_format & PERF_FORMAT_LOST)
+		values[n++] = atomic64_read(&event->lost_samples);
 
 	if (copy_to_user(buf, values, n * sizeof(u64)))
 		return -EFAULT;
@@ -5719,10 +5784,10 @@ again:
 
 		if (!atomic_inc_not_zero(&event->rb->mmap_count)) {
 			/*
-			 * Raced against perf_mmap_close() through
-			 * perf_event_set_output(). Try again, hope for better
-			 * luck.
+			 * Raced against perf_mmap_close(); remove the
+			 * event and try again.
 			 */
+			ring_buffer_attach(event, NULL);
 			mutex_unlock(&event->mmap_mutex);
 			goto again;
 		}
@@ -6162,7 +6227,7 @@ static void perf_output_read_one(struct perf_output_handle *handle,
 				 u64 enabled, u64 running)
 {
 	u64 read_format = event->attr.read_format;
-	u64 values[4];
+	u64 values[5];
 	int n = 0;
 
 	values[n++] = perf_event_count(event);
@@ -6176,6 +6241,8 @@ static void perf_output_read_one(struct perf_output_handle *handle,
 	}
 	if (read_format & PERF_FORMAT_ID)
 		values[n++] = primary_event_id(event);
+	if (read_format & PERF_FORMAT_LOST)
+		values[n++] = atomic64_read(&event->lost_samples);
 
 	__output_copy(handle, values, n * sizeof(u64));
 }
@@ -6186,7 +6253,7 @@ static void perf_output_read_group(struct perf_output_handle *handle,
 {
 	struct perf_event *leader = event->group_leader, *sub;
 	u64 read_format = event->attr.read_format;
-	u64 values[5];
+	u64 values[6];
 	int n = 0;
 
 	values[n++] = 1 + leader->nr_siblings;
@@ -6204,6 +6271,8 @@ static void perf_output_read_group(struct perf_output_handle *handle,
 	values[n++] = perf_event_count(leader);
 	if (read_format & PERF_FORMAT_ID)
 		values[n++] = primary_event_id(leader);
+	if (read_format & PERF_FORMAT_LOST)
+		values[n++] = atomic64_read(&leader->lost_samples);
 
 	__output_copy(handle, values, n * sizeof(u64));
 
@@ -6217,6 +6286,8 @@ static void perf_output_read_group(struct perf_output_handle *handle,
 		values[n++] = perf_event_count(sub);
 		if (read_format & PERF_FORMAT_ID)
 			values[n++] = primary_event_id(sub);
+		if (read_format & PERF_FORMAT_LOST)
+			values[n++] = atomic64_read(&sub->lost_samples);
 
 		__output_copy(handle, values, n * sizeof(u64));
 	}
@@ -7843,8 +7914,8 @@ __perf_event_account_interrupt(struct perf_event *event, int throttle)
 		hwc->interrupts = 1;
 	} else {
 		hwc->interrupts++;
-		if (unlikely(throttle
-			     && hwc->interrupts >= max_samples_per_tick)) {
+		if (unlikely(throttle &&
+			     hwc->interrupts > max_samples_per_tick)) {
 			__this_cpu_inc(perf_throttled_count);
 			tick_dep_set_cpu(smp_processor_id(), TICK_DEP_BIT_PERF_EVENTS);
 			hwc->interrupts = MAX_INTERRUPTS;
@@ -9137,8 +9208,11 @@ perf_event_parse_addr_filter(struct perf_event *event, char *fstr,
 			}
 
 			/* ready to consume more filters */
+			kfree(filename);
+			filename = NULL;
 			state = IF_STATE_ACTION;
 			filter = NULL;
+			kernel = 0;
 		}
 	}
 
@@ -9641,8 +9715,7 @@ perf_event_mux_interval_ms_store(struct device *dev,
 		cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu);
 		cpuctx->hrtimer_interval = ns_to_ktime(NSEC_PER_MSEC * timer);
 
-		cpu_function_call(cpu,
-			(remote_function_f)perf_mux_hrtimer_restart, cpuctx);
+		cpu_function_call(cpu, perf_mux_hrtimer_restart_ipi, cpuctx);
 	}
 	cpus_read_unlock();
 	mutex_unlock(&mux_interval_mutex);
@@ -9679,13 +9752,15 @@ static int pmu_dev_alloc(struct pmu *pmu)
 
 	pmu->dev->groups = pmu->attr_groups;
 	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 = dev_set_name(pmu->dev, "%s", pmu->name);
+	if (ret)
+		goto free_dev;
+
 	ret = device_add(pmu->dev);
 	if (ret)
 		goto free_dev;
@@ -10393,14 +10468,25 @@ err_size:
 	goto out;
 }
 
+static void mutex_lock_double(struct mutex *a, struct mutex *b)
+{
+	if (b < a)
+		swap(a, b);
+
+	mutex_lock(a);
+	mutex_lock_nested(b, SINGLE_DEPTH_NESTING);
+}
+
 static int
 perf_event_set_output(struct perf_event *event, struct perf_event *output_event)
 {
 	struct ring_buffer *rb = NULL;
 	int ret = -EINVAL;
 
-	if (!output_event)
+	if (!output_event) {
+		mutex_lock(&event->mmap_mutex);
 		goto set;
+	}
 
 	/* don't allow circular references */
 	if (event == output_event)
@@ -10415,7 +10501,7 @@ perf_event_set_output(struct perf_event *event, struct perf_event *output_event)
 	/*
 	 * If its not a per-cpu rb, it must be the same task.
 	 */
-	if (output_event->cpu == -1 && output_event->ctx != event->ctx)
+	if (output_event->cpu == -1 && output_event->hw.target != event->hw.target)
 		goto out;
 
 	/*
@@ -10438,8 +10524,15 @@ perf_event_set_output(struct perf_event *event, struct perf_event *output_event)
 	    event->pmu != output_event->pmu)
 		goto out;
 
+	/*
+	 * Hold both mmap_mutex to serialize against perf_mmap_close().  Since
+	 * output_event is already on rb->event_list, and the list iteration
+	 * restarts after every removal, it is guaranteed this new event is
+	 * observed *OR* if output_event is already removed, it's guaranteed we
+	 * observe !rb->mmap_count.
+	 */
+	mutex_lock_double(&event->mmap_mutex, &output_event->mmap_mutex);
 set:
-	mutex_lock(&event->mmap_mutex);
 	/* Can't redirect output if we've got an active mmap() */
 	if (atomic_read(&event->mmap_count))
 		goto unlock;
@@ -10449,6 +10542,12 @@ set:
 		rb = ring_buffer_get(output_event);
 		if (!rb)
 			goto unlock;
+
+		/* did we race against perf_mmap_close() */
+		if (!atomic_read(&rb->mmap_count)) {
+			ring_buffer_put(rb);
+			goto unlock;
+		}
 	}
 
 	ring_buffer_attach(event, rb);
@@ -10456,20 +10555,13 @@ set:
 	ret = 0;
 unlock:
 	mutex_unlock(&event->mmap_mutex);
+	if (output_event)
+		mutex_unlock(&output_event->mmap_mutex);
 
 out:
 	return ret;
 }
 
-static void mutex_lock_double(struct mutex *a, struct mutex *b)
-{
-	if (b < a)
-		swap(a, b);
-
-	mutex_lock(a);
-	mutex_lock_nested(b, SINGLE_DEPTH_NESTING);
-}
-
 static int perf_event_set_clock(struct perf_event *event, clockid_t clk_id)
 {
 	bool nmi_safe = false;
@@ -10553,7 +10645,7 @@ SYSCALL_DEFINE5(perf_event_open,
 	struct perf_event *group_leader = NULL, *output_event = NULL;
 	struct perf_event *event, *sibling;
 	struct perf_event_attr attr;
-	struct perf_event_context *ctx, *uninitialized_var(gctx);
+	struct perf_event_context *ctx, *gctx;
 	struct file *event_file = NULL;
 	struct fd group = {NULL, 0};
 	struct task_struct *task = NULL;
@@ -10755,6 +10847,9 @@ SYSCALL_DEFINE5(perf_event_open,
 		 * Do not allow to attach to a group in a different task
 		 * or CPU context. If we're moving SW events, we'll fix
 		 * this up later, so allow that.
+		 *
+		 * Racy, not holding group_leader->ctx->mutex, see comment with
+		 * perf_event_ctx_lock().
 		 */
 		if (!move_group && group_leader->ctx != ctx)
 			goto err_context;
@@ -10804,6 +10899,7 @@ SYSCALL_DEFINE5(perf_event_open,
 			} else {
 				perf_event_ctx_unlock(group_leader, gctx);
 				move_group = 0;
+				goto not_move_group;
 			}
 		}
 
@@ -10820,7 +10916,17 @@ SYSCALL_DEFINE5(perf_event_open,
 		}
 	} else {
 		mutex_lock(&ctx->mutex);
+
+		/*
+		 * Now that we hold ctx->lock, (re)validate group_leader->ctx == ctx,
+		 * see the group_leader && !move_group test earlier.
+		 */
+		if (group_leader && group_leader->ctx != ctx) {
+			err = -EINVAL;
+			goto err_locked;
+		}
 	}
+not_move_group:
 
 	if (ctx->task == TASK_TOMBSTONE) {
 		err = -ESRCH;
@@ -11563,6 +11669,8 @@ static int inherit_group(struct perf_event *parent_event,
 		if (IS_ERR(child_ctr))
 			return PTR_ERR(child_ctr);
 	}
+	if (leader)
+		leader->group_generation = parent_event->group_generation;
 	return 0;
 }