Merge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull perf updates from Ingo Molnar:
 "Bigger kernel side changes:

   - Add backwards writing capability to the perf ring-buffer code,
     which is preparation for future advanced features like robust
     'overwrite support' and snapshot mode.  (Wang Nan)

   - Add pause and resume ioctls for the perf ringbuffer (Wang Nan)

   - x86 Intel cstate code cleanups and reorgnization (Thomas Gleixner)

   - x86 Intel uncore and CPU PMU driver updates (Kan Liang, Peter
     Zijlstra)

   - x86 AUX (Intel PT) related enhancements and updates (Alexander
     Shishkin)

   - x86 MSR PMU driver enhancements and updates (Huang Rui)

   - ... and lots of other changes spread out over 40+ commits.

  Biggest tooling side changes:

   - 'perf trace' features and enhancements.  (Arnaldo Carvalho de Melo)

   - BPF tooling updates (Wang Nan)

   - 'perf sched' updates (Jiri Olsa)

   - 'perf probe' updates (Masami Hiramatsu)

   - ... plus 200+ other enhancements, fixes and cleanups to tools/

  The merge commits, the shortlog and the changelogs contain a lot more
  details"

* 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (249 commits)
  perf/core: Disable the event on a truncated AUX record
  perf/x86/intel/pt: Generate PMI in the STOP region as well
  perf buildid-cache: Use lsdir() for looking up buildid caches
  perf symbols: Use lsdir() for the search in kcore cache directory
  perf tools: Use SBUILD_ID_SIZE where applicable
  perf tools: Fix lsdir to set errno correctly
  perf trace: Move seccomp args beautifiers to tools/perf/trace/beauty/
  perf trace: Move flock op beautifier to tools/perf/trace/beauty/
  perf build: Add build-test for debug-frame on arm/arm64
  perf build: Add build-test for libunwind cross-platforms support
  perf script: Fix export of callchains with recursion in db-export
  perf script: Fix callchain addresses in db-export
  perf script: Fix symbol insertion behavior in db-export
  perf symbols: Add dso__insert_symbol function
  perf scripting python: Use Py_FatalError instead of die()
  perf tools: Remove xrealloc and ALLOC_GROW
  perf help: Do not use ALLOC_GROW in add_cmd_list
  perf pmu: Make pmu_formats_string to check return value of strbuf
  perf header: Make topology checkers to check return value of strbuf
  perf tools: Make alias handler to check return value of strbuf
  ...

7 files changed, 1005 insertions, 92 deletions

diff --git a/kernel/bpf/stackmap.c b/kernel/bpf/stackmap.c
index 499d9e933f8e..f5a19548be12 100644
--- a/kernel/bpf/stackmap.c
+++ b/kernel/bpf/stackmap.c
@@ -66,7 +66,7 @@ static struct bpf_map *stack_map_alloc(union bpf_attr *attr)
 	/* check sanity of attributes */
 	if (attr->max_entries == 0 || attr->key_size != 4 ||
 	    value_size < 8 || value_size % 8 ||
-	    value_size / 8 > PERF_MAX_STACK_DEPTH)
+	    value_size / 8 > sysctl_perf_event_max_stack)
 		return ERR_PTR(-EINVAL);
 
 	/* hash table size must be power of 2 */
@@ -124,8 +124,8 @@ static u64 bpf_get_stackid(u64 r1, u64 r2, u64 flags, u64 r4, u64 r5)
 	struct perf_callchain_entry *trace;
 	struct stack_map_bucket *bucket, *new_bucket, *old_bucket;
 	u32 max_depth = map->value_size / 8;
-	/* stack_map_alloc() checks that max_depth <= PERF_MAX_STACK_DEPTH */
-	u32 init_nr = PERF_MAX_STACK_DEPTH - max_depth;
+	/* stack_map_alloc() checks that max_depth <= sysctl_perf_event_max_stack */
+	u32 init_nr = sysctl_perf_event_max_stack - max_depth;
 	u32 skip = flags & BPF_F_SKIP_FIELD_MASK;
 	u32 hash, id, trace_nr, trace_len;
 	bool user = flags & BPF_F_USER_STACK;
@@ -143,7 +143,7 @@ static u64 bpf_get_stackid(u64 r1, u64 r2, u64 flags, u64 r4, u64 r5)
 		return -EFAULT;
 
 	/* get_perf_callchain() guarantees that trace->nr >= init_nr
-	 * and trace-nr <= PERF_MAX_STACK_DEPTH, so trace_nr <= max_depth
+	 * and trace-nr <= sysctl_perf_event_max_stack, so trace_nr <= max_depth
 	 */
 	trace_nr = trace->nr - init_nr;
 
diff --git a/kernel/events/callchain.c b/kernel/events/callchain.c
index 343c22f5e867..b9325e7dcba1 100644
--- a/kernel/events/callchain.c
+++ b/kernel/events/callchain.c
@@ -18,6 +18,14 @@ struct callchain_cpus_entries {
 	struct perf_callchain_entry	*cpu_entries[0];
 };
 
+int sysctl_perf_event_max_stack __read_mostly = PERF_MAX_STACK_DEPTH;
+
+static inline size_t perf_callchain_entry__sizeof(void)
+{
+	return (sizeof(struct perf_callchain_entry) +
+		sizeof(__u64) * sysctl_perf_event_max_stack);
+}
+
 static DEFINE_PER_CPU(int, callchain_recursion[PERF_NR_CONTEXTS]);
 static atomic_t nr_callchain_events;
 static DEFINE_MUTEX(callchain_mutex);
@@ -73,7 +81,7 @@ static int alloc_callchain_buffers(void)
 	if (!entries)
 		return -ENOMEM;
 
-	size = sizeof(struct perf_callchain_entry) * PERF_NR_CONTEXTS;
+	size = perf_callchain_entry__sizeof() * PERF_NR_CONTEXTS;
 
 	for_each_possible_cpu(cpu) {
 		entries->cpu_entries[cpu] = kmalloc_node(size, GFP_KERNEL,
@@ -147,7 +155,8 @@ static struct perf_callchain_entry *get_callchain_entry(int *rctx)
 
 	cpu = smp_processor_id();
 
-	return &entries->cpu_entries[cpu][*rctx];
+	return (((void *)entries->cpu_entries[cpu]) +
+		(*rctx * perf_callchain_entry__sizeof()));
 }
 
 static void
@@ -215,3 +224,25 @@ exit_put:
 
 	return entry;
 }
+
+int perf_event_max_stack_handler(struct ctl_table *table, int write,
+				 void __user *buffer, size_t *lenp, loff_t *ppos)
+{
+	int new_value = sysctl_perf_event_max_stack, ret;
+	struct ctl_table new_table = *table;
+
+	new_table.data = &new_value;
+	ret = proc_dointvec_minmax(&new_table, write, buffer, lenp, ppos);
+	if (ret || !write)
+		return ret;
+
+	mutex_lock(&callchain_mutex);
+	if (atomic_read(&nr_callchain_events))
+		ret = -EBUSY;
+	else
+		sysctl_perf_event_max_stack = new_value;
+
+	mutex_unlock(&callchain_mutex);
+
+	return ret;
+}
diff --git a/kernel/events/core.c b/kernel/events/core.c
index c0ded2416615..050a290c72c7 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -44,6 +44,8 @@
 #include <linux/compat.h>
 #include <linux/bpf.h>
 #include <linux/filter.h>
+#include <linux/namei.h>
+#include <linux/parser.h>
 
 #include "internal.h"
 
@@ -1927,8 +1929,13 @@ event_sched_in(struct perf_event *event,
 	if (event->state <= PERF_EVENT_STATE_OFF)
 		return 0;
 
-	event->state = PERF_EVENT_STATE_ACTIVE;
-	event->oncpu = smp_processor_id();
+	WRITE_ONCE(event->oncpu, smp_processor_id());
+	/*
+	 * Order event::oncpu write to happen before the ACTIVE state
+	 * is visible.
+	 */
+	smp_wmb();
+	WRITE_ONCE(event->state, PERF_EVENT_STATE_ACTIVE);
 
 	/*
 	 * Unthrottle events, since we scheduled we might have missed several
@@ -2360,6 +2367,112 @@ void perf_event_enable(struct perf_event *event)
 }
 EXPORT_SYMBOL_GPL(perf_event_enable);
 
+struct stop_event_data {
+	struct perf_event	*event;
+	unsigned int		restart;
+};
+
+static int __perf_event_stop(void *info)
+{
+	struct stop_event_data *sd = info;
+	struct perf_event *event = sd->event;
+
+	/* if it's already INACTIVE, do nothing */
+	if (READ_ONCE(event->state) != PERF_EVENT_STATE_ACTIVE)
+		return 0;
+
+	/* matches smp_wmb() in event_sched_in() */
+	smp_rmb();
+
+	/*
+	 * There is a window with interrupts enabled before we get here,
+	 * so we need to check again lest we try to stop another CPU's event.
+	 */
+	if (READ_ONCE(event->oncpu) != smp_processor_id())
+		return -EAGAIN;
+
+	event->pmu->stop(event, PERF_EF_UPDATE);
+
+	/*
+	 * May race with the actual stop (through perf_pmu_output_stop()),
+	 * but it is only used for events with AUX ring buffer, and such
+	 * events will refuse to restart because of rb::aux_mmap_count==0,
+	 * see comments in perf_aux_output_begin().
+	 *
+	 * Since this is happening on a event-local CPU, no trace is lost
+	 * while restarting.
+	 */
+	if (sd->restart)
+		event->pmu->start(event, PERF_EF_START);
+
+	return 0;
+}
+
+static int perf_event_restart(struct perf_event *event)
+{
+	struct stop_event_data sd = {
+		.event		= event,
+		.restart	= 1,
+	};
+	int ret = 0;
+
+	do {
+		if (READ_ONCE(event->state) != PERF_EVENT_STATE_ACTIVE)
+			return 0;
+
+		/* matches smp_wmb() in event_sched_in() */
+		smp_rmb();
+
+		/*
+		 * We only want to restart ACTIVE events, so if the event goes
+		 * inactive here (event->oncpu==-1), there's nothing more to do;
+		 * fall through with ret==-ENXIO.
+		 */
+		ret = cpu_function_call(READ_ONCE(event->oncpu),
+					__perf_event_stop, &sd);
+	} while (ret == -EAGAIN);
+
+	return ret;
+}
+
+/*
+ * In order to contain the amount of racy and tricky in the address filter
+ * configuration management, it is a two part process:
+ *
+ * (p1) when userspace mappings change as a result of (1) or (2) or (3) below,
+ *      we update the addresses of corresponding vmas in
+ *	event::addr_filters_offs array and bump the event::addr_filters_gen;
+ * (p2) when an event is scheduled in (pmu::add), it calls
+ *      perf_event_addr_filters_sync() which calls pmu::addr_filters_sync()
+ *      if the generation has changed since the previous call.
+ *
+ * If (p1) happens while the event is active, we restart it to force (p2).
+ *
+ * (1) perf_addr_filters_apply(): adjusting filters' offsets based on
+ *     pre-existing mappings, called once when new filters arrive via SET_FILTER
+ *     ioctl;
+ * (2) perf_addr_filters_adjust(): adjusting filters' offsets based on newly
+ *     registered mapping, called for every new mmap(), with mm::mmap_sem down
+ *     for reading;
+ * (3) perf_event_addr_filters_exec(): clearing filters' offsets in the process
+ *     of exec.
+ */
+void perf_event_addr_filters_sync(struct perf_event *event)
+{
+	struct perf_addr_filters_head *ifh = perf_event_addr_filters(event);
+
+	if (!has_addr_filter(event))
+		return;
+
+	raw_spin_lock(&ifh->lock);
+	if (event->addr_filters_gen != event->hw.addr_filters_gen) {
+		event->pmu->addr_filters_sync(event);
+		event->hw.addr_filters_gen = event->addr_filters_gen;
+	}
+	raw_spin_unlock(&ifh->lock);
+}
+EXPORT_SYMBOL_GPL(perf_event_addr_filters_sync);
+
 static int _perf_event_refresh(struct perf_event *event, int refresh)
 {
 	/*
@@ -3209,16 +3322,6 @@ out:
 		put_ctx(clone_ctx);
 }
 
-void perf_event_exec(void)
-{
-	int ctxn;
-
-	rcu_read_lock();
-	for_each_task_context_nr(ctxn)
-		perf_event_enable_on_exec(ctxn);
-	rcu_read_unlock();
-}
-
 struct perf_read_data {
 	struct perf_event *event;
 	bool group;
@@ -3720,6 +3823,9 @@ static bool exclusive_event_installable(struct perf_event *event,
 	return true;
 }
 
+static void perf_addr_filters_splice(struct perf_event *event,
+				       struct list_head *head);
+
 static void _free_event(struct perf_event *event)
 {
 	irq_work_sync(&event->pending);
@@ -3747,6 +3853,8 @@ static void _free_event(struct perf_event *event)
 	}
 
 	perf_event_free_bpf_prog(event);
+	perf_addr_filters_splice(event, NULL);
+	kfree(event->addr_filters_offs);
 
 	if (event->destroy)
 		event->destroy(event);
@@ -4343,6 +4451,19 @@ static long _perf_ioctl(struct perf_event *event, unsigned int cmd, unsigned lon
 	case PERF_EVENT_IOC_SET_BPF:
 		return perf_event_set_bpf_prog(event, arg);
 
+	case PERF_EVENT_IOC_PAUSE_OUTPUT: {
+		struct ring_buffer *rb;
+
+		rcu_read_lock();
+		rb = rcu_dereference(event->rb);
+		if (!rb || !rb->nr_pages) {
+			rcu_read_unlock();
+			return -EINVAL;
+		}
+		rb_toggle_paused(rb, !!arg);
+		rcu_read_unlock();
+		return 0;
+	}
 	default:
 		return -ENOTTY;
 	}
@@ -4659,6 +4780,8 @@ static void perf_mmap_open(struct vm_area_struct *vma)
 		event->pmu->event_mapped(event);
 }
 
+static void perf_pmu_output_stop(struct perf_event *event);
+
 /*
  * A buffer can be mmap()ed multiple times; either directly through the same
  * event, or through other events by use of perf_event_set_output().
@@ -4686,10 +4809,22 @@ static void perf_mmap_close(struct vm_area_struct *vma)
 	 */
 	if (rb_has_aux(rb) && vma->vm_pgoff == rb->aux_pgoff &&
 	    atomic_dec_and_mutex_lock(&rb->aux_mmap_count, &event->mmap_mutex)) {
+		/*
+		 * Stop all AUX events that are writing to this buffer,
+		 * so that we can free its AUX pages and corresponding PMU
+		 * data. Note that after rb::aux_mmap_count dropped to zero,
+		 * they won't start any more (see perf_aux_output_begin()).
+		 */
+		perf_pmu_output_stop(event);
+
+		/* now it's safe to free the pages */
 		atomic_long_sub(rb->aux_nr_pages, &mmap_user->locked_vm);
 		vma->vm_mm->pinned_vm -= rb->aux_mmap_locked;
 
+		/* this has to be the last one */
 		rb_free_aux(rb);
+		WARN_ON_ONCE(atomic_read(&rb->aux_refcount));
+
 		mutex_unlock(&event->mmap_mutex);
 	}
 
@@ -5630,9 +5765,13 @@ void perf_prepare_sample(struct perf_event_header *header,
 	}
 }
 
-void perf_event_output(struct perf_event *event,
-			struct perf_sample_data *data,
-			struct pt_regs *regs)
+static void __always_inline
+__perf_event_output(struct perf_event *event,
+		    struct perf_sample_data *data,
+		    struct pt_regs *regs,
+		    int (*output_begin)(struct perf_output_handle *,
+					struct perf_event *,
+					unsigned int))
 {
 	struct perf_output_handle handle;
 	struct perf_event_header header;
@@ -5642,7 +5781,7 @@ void perf_event_output(struct perf_event *event,
 
 	perf_prepare_sample(&header, data, event, regs);
 
-	if (perf_output_begin(&handle, event, header.size))
+	if (output_begin(&handle, event, header.size))
 		goto exit;
 
 	perf_output_sample(&handle, &header, data, event);
@@ -5653,6 +5792,30 @@ exit:
 	rcu_read_unlock();
 }
 
+void
+perf_event_output_forward(struct perf_event *event,
+			 struct perf_sample_data *data,
+			 struct pt_regs *regs)
+{
+	__perf_event_output(event, data, regs, perf_output_begin_forward);
+}
+
+void
+perf_event_output_backward(struct perf_event *event,
+			   struct perf_sample_data *data,
+			   struct pt_regs *regs)
+{
+	__perf_event_output(event, data, regs, perf_output_begin_backward);
+}
+
+void
+perf_event_output(struct perf_event *event,
+		  struct perf_sample_data *data,
+		  struct pt_regs *regs)
+{
+	__perf_event_output(event, data, regs, perf_output_begin);
+}
+
 /*
  * read event_id
  */
@@ -5698,15 +5861,18 @@ typedef void (perf_event_aux_output_cb)(struct perf_event *event, void *data);
 static void
 perf_event_aux_ctx(struct perf_event_context *ctx,
 		   perf_event_aux_output_cb output,
-		   void *data)
+		   void *data, bool all)
 {
 	struct perf_event *event;
 
 	list_for_each_entry_rcu(event, &ctx->event_list, event_entry) {
-		if (event->state < PERF_EVENT_STATE_INACTIVE)
-			continue;
-		if (!event_filter_match(event))
-			continue;
+		if (!all) {
+			if (event->state < PERF_EVENT_STATE_INACTIVE)
+				continue;
+			if (!event_filter_match(event))
+				continue;
+		}
+
 		output(event, data);
 	}
 }
@@ -5717,7 +5883,7 @@ perf_event_aux_task_ctx(perf_event_aux_output_cb output, void *data,
 {
 	rcu_read_lock();
 	preempt_disable();
-	perf_event_aux_ctx(task_ctx, output, data);
+	perf_event_aux_ctx(task_ctx, output, data, false);
 	preempt_enable();
 	rcu_read_unlock();
 }
@@ -5747,13 +5913,13 @@ perf_event_aux(perf_event_aux_output_cb output, void *data,
 		cpuctx = get_cpu_ptr(pmu->pmu_cpu_context);
 		if (cpuctx->unique_pmu != pmu)
 			goto next;
-		perf_event_aux_ctx(&cpuctx->ctx, output, data);
+		perf_event_aux_ctx(&cpuctx->ctx, output, data, false);
 		ctxn = pmu->task_ctx_nr;
 		if (ctxn < 0)
 			goto next;
 		ctx = rcu_dereference(current->perf_event_ctxp[ctxn]);
 		if (ctx)
-			perf_event_aux_ctx(ctx, output, data);
+			perf_event_aux_ctx(ctx, output, data, false);
 next:
 		put_cpu_ptr(pmu->pmu_cpu_context);
 	}
@@ -5761,6 +5927,134 @@ next:
 }
 
 /*
+ * Clear all file-based filters at exec, they'll have to be
+ * re-instated when/if these objects are mmapped again.
+ */
+static void perf_event_addr_filters_exec(struct perf_event *event, void *data)
+{
+	struct perf_addr_filters_head *ifh = perf_event_addr_filters(event);
+	struct perf_addr_filter *filter;
+	unsigned int restart = 0, count = 0;
+	unsigned long flags;
+
+	if (!has_addr_filter(event))
+		return;
+
+	raw_spin_lock_irqsave(&ifh->lock, flags);
+	list_for_each_entry(filter, &ifh->list, entry) {
+		if (filter->inode) {
+			event->addr_filters_offs[count] = 0;
+			restart++;
+		}
+
+		count++;
+	}
+
+	if (restart)
+		event->addr_filters_gen++;
+	raw_spin_unlock_irqrestore(&ifh->lock, flags);
+
+	if (restart)
+		perf_event_restart(event);
+}
+
+void perf_event_exec(void)
+{
+	struct perf_event_context *ctx;
+	int ctxn;
+
+	rcu_read_lock();
+	for_each_task_context_nr(ctxn) {
+		ctx = current->perf_event_ctxp[ctxn];
+		if (!ctx)
+			continue;
+
+		perf_event_enable_on_exec(ctxn);
+
+		perf_event_aux_ctx(ctx, perf_event_addr_filters_exec, NULL,
+				   true);
+	}
+	rcu_read_unlock();
+}
+
+struct remote_output {
+	struct ring_buffer	*rb;
+	int			err;
+};
+
+static void __perf_event_output_stop(struct perf_event *event, void *data)
+{
+	struct perf_event *parent = event->parent;
+	struct remote_output *ro = data;
+	struct ring_buffer *rb = ro->rb;
+	struct stop_event_data sd = {
+		.event	= event,
+	};
+
+	if (!has_aux(event))
+		return;
+
+	if (!parent)
+		parent = event;
+
+	/*
+	 * In case of inheritance, it will be the parent that links to the
+	 * ring-buffer, but it will be the child that's actually using it:
+	 */
+	if (rcu_dereference(parent->rb) == rb)
+		ro->err = __perf_event_stop(&sd);
+}
+
+static int __perf_pmu_output_stop(void *info)
+{
+	struct perf_event *event = info;
+	struct pmu *pmu = event->pmu;
+	struct perf_cpu_context *cpuctx = get_cpu_ptr(pmu->pmu_cpu_context);
+	struct remote_output ro = {
+		.rb	= event->rb,
+	};
+
+	rcu_read_lock();
+	perf_event_aux_ctx(&cpuctx->ctx, __perf_event_output_stop, &ro, false);
+	if (cpuctx->task_ctx)
+		perf_event_aux_ctx(cpuctx->task_ctx, __perf_event_output_stop,
+				   &ro, false);
+	rcu_read_unlock();
+
+	return ro.err;
+}
+
+static void perf_pmu_output_stop(struct perf_event *event)
+{
+	struct perf_event *iter;
+	int err, cpu;
+
+restart:
+	rcu_read_lock();
+	list_for_each_entry_rcu(iter, &event->rb->event_list, rb_entry) {
+		/*
+		 * For per-CPU events, we need to make sure that neither they
+		 * nor their children are running; for cpu==-1 events it's
+		 * sufficient to stop the event itself if it's active, since
+		 * it can't have children.
+		 */
+		cpu = iter->cpu;
+		if (cpu == -1)
+			cpu = READ_ONCE(iter->oncpu);
+
+		if (cpu == -1)
+			continue;
+
+		err = cpu_function_call(cpu, __perf_pmu_output_stop, event);
+		if (err == -EAGAIN) {
+			rcu_read_unlock();
+			goto restart;
+		}
+	}
+	rcu_read_unlock();
+}
+
+/*
  * task tracking -- fork/exit
  *
  * enabled by: attr.comm | attr.mmap | attr.mmap2 | attr.mmap_data | attr.task
@@ -6169,6 +6463,87 @@ got_name:
 	kfree(buf);
 }
 
+/*
+ * Whether this @filter depends on a dynamic object which is not loaded
+ * yet or its load addresses are not known.
+ */
+static bool perf_addr_filter_needs_mmap(struct perf_addr_filter *filter)
+{
+	return filter->filter && filter->inode;
+}
+
+/*
+ * Check whether inode and address range match filter criteria.
+ */
+static bool perf_addr_filter_match(struct perf_addr_filter *filter,
+				     struct file *file, unsigned long offset,
+				     unsigned long size)
+{
+	if (filter->inode != file->f_inode)
+		return false;
+
+	if (filter->offset > offset + size)
+		return false;
+
+	if (filter->offset + filter->size < offset)
+		return false;
+
+	return true;
+}
+
+static void __perf_addr_filters_adjust(struct perf_event *event, void *data)
+{
+	struct perf_addr_filters_head *ifh = perf_event_addr_filters(event);
+	struct vm_area_struct *vma = data;
+	unsigned long off = vma->vm_pgoff << PAGE_SHIFT, flags;
+	struct file *file = vma->vm_file;
+	struct perf_addr_filter *filter;
+	unsigned int restart = 0, count = 0;
+
+	if (!has_addr_filter(event))
+		return;
+
+	if (!file)
+		return;
+
+	raw_spin_lock_irqsave(&ifh->lock, flags);
+	list_for_each_entry(filter, &ifh->list, entry) {
+		if (perf_addr_filter_match(filter, file, off,
+					     vma->vm_end - vma->vm_start)) {
+			event->addr_filters_offs[count] = vma->vm_start;
+			restart++;
+		}
+
+		count++;
+	}
+
+	if (restart)
+		event->addr_filters_gen++;
+	raw_spin_unlock_irqrestore(&ifh->lock, flags);
+
+	if (restart)
+		perf_event_restart(event);
+}
+
+/*
+ * Adjust all task's events' filters to the new vma
+ */
+static void perf_addr_filters_adjust(struct vm_area_struct *vma)
+{
+	struct perf_event_context *ctx;
+	int ctxn;
+
+	rcu_read_lock();
+	for_each_task_context_nr(ctxn) {
+		ctx = rcu_dereference(current->perf_event_ctxp[ctxn]);
+		if (!ctx)
+			continue;
+
+		perf_event_aux_ctx(ctx, __perf_addr_filters_adjust, vma, true);
+	}
+	rcu_read_unlock();
+}
+
 void perf_event_mmap(struct vm_area_struct *vma)
 {
 	struct perf_mmap_event mmap_event;
@@ -6200,6 +6575,7 @@ void perf_event_mmap(struct vm_area_struct *vma)
 		/* .flags (attr_mmap2 only) */
 	};
 
+	perf_addr_filters_adjust(vma);
 	perf_event_mmap_event(&mmap_event);
 }
 
@@ -6491,10 +6867,7 @@ static int __perf_event_overflow(struct perf_event *event,
 		irq_work_queue(&event->pending);
 	}
 
-	if (event->overflow_handler)
-		event->overflow_handler(event, data, regs);
-	else
-		perf_event_output(event, data, regs);
+	event->overflow_handler(event, data, regs);
 
 	if (*perf_event_fasync(event) && event->pending_kill) {
 		event->pending_wakeup = 1;
@@ -7081,24 +7454,6 @@ static inline void perf_tp_register(void)
 	perf_pmu_register(&perf_tracepoint, "tracepoint", PERF_TYPE_TRACEPOINT);
 }
 
-static int perf_event_set_filter(struct perf_event *event, void __user *arg)
-{
-	char *filter_str;
-	int ret;
-
-	if (event->attr.type != PERF_TYPE_TRACEPOINT)
-		return -EINVAL;
-
-	filter_str = strndup_user(arg, PAGE_SIZE);
-	if (IS_ERR(filter_str))
-		return PTR_ERR(filter_str);
-
-	ret = ftrace_profile_set_filter(event, event->attr.config, filter_str);
-
-	kfree(filter_str);
-	return ret;
-}
-
 static void perf_event_free_filter(struct perf_event *event)
 {
 	ftrace_profile_free_filter(event);
@@ -7153,11 +7508,6 @@ static inline void perf_tp_register(void)
 {
 }
 
-static int perf_event_set_filter(struct perf_event *event, void __user *arg)
-{
-	return -ENOENT;
-}
-
 static void perf_event_free_filter(struct perf_event *event)
 {
 }
@@ -7186,6 +7536,387 @@ void perf_bp_event(struct perf_event *bp, void *data)
 #endif
 
 /*
+ * Allocate a new address filter
+ */
+static struct perf_addr_filter *
+perf_addr_filter_new(struct perf_event *event, struct list_head *filters)
+{
+	int node = cpu_to_node(event->cpu == -1 ? 0 : event->cpu);
+	struct perf_addr_filter *filter;
+
+	filter = kzalloc_node(sizeof(*filter), GFP_KERNEL, node);
+	if (!filter)
+		return NULL;
+
+	INIT_LIST_HEAD(&filter->entry);
+	list_add_tail(&filter->entry, filters);
+
+	return filter;
+}
+
+static void free_filters_list(struct list_head *filters)
+{
+	struct perf_addr_filter *filter, *iter;
+
+	list_for_each_entry_safe(filter, iter, filters, entry) {
+		if (filter->inode)
+			iput(filter->inode);
+		list_del(&filter->entry);
+		kfree(filter);
+	}
+}
+
+/*
+ * Free existing address filters and optionally install new ones
+ */
+static void perf_addr_filters_splice(struct perf_event *event,
+				     struct list_head *head)
+{
+	unsigned long flags;
+	LIST_HEAD(list);
+
+	if (!has_addr_filter(event))
+		return;
+
+	/* don't bother with children, they don't have their own filters */
+	if (event->parent)
+		return;
+
+	raw_spin_lock_irqsave(&event->addr_filters.lock, flags);
+
+	list_splice_init(&event->addr_filters.list, &list);
+	if (head)
+		list_splice(head, &event->addr_filters.list);
+
+	raw_spin_unlock_irqrestore(&event->addr_filters.lock, flags);
+
+	free_filters_list(&list);
+}
+
+/*
+ * Scan through mm's vmas and see if one of them matches the
+ * @filter; if so, adjust filter's address range.
+ * Called with mm::mmap_sem down for reading.
+ */
+static unsigned long perf_addr_filter_apply(struct perf_addr_filter *filter,
+					    struct mm_struct *mm)
+{
+	struct vm_area_struct *vma;
+
+	for (vma = mm->mmap; vma; vma = vma->vm_next) {
+		struct file *file = vma->vm_file;
+		unsigned long off = vma->vm_pgoff << PAGE_SHIFT;
+		unsigned long vma_size = vma->vm_end - vma->vm_start;
+
+		if (!file)
+			continue;
+
+		if (!perf_addr_filter_match(filter, file, off, vma_size))
+			continue;
+
+		return vma->vm_start;
+	}
+
+	return 0;
+}
+
+/*
+ * Update event's address range filters based on the
+ * task's existing mappings, if any.
+ */
+static void perf_event_addr_filters_apply(struct perf_event *event)
+{
+	struct perf_addr_filters_head *ifh = perf_event_addr_filters(event);
+	struct task_struct *task = READ_ONCE(event->ctx->task);
+	struct perf_addr_filter *filter;
+	struct mm_struct *mm = NULL;
+	unsigned int count = 0;
+	unsigned long flags;
+
+	/*
+	 * We may observe TASK_TOMBSTONE, which means that the event tear-down
+	 * will stop on the parent's child_mutex that our caller is also holding
+	 */
+	if (task == TASK_TOMBSTONE)
+		return;
+
+	mm = get_task_mm(event->ctx->task);
+	if (!mm)
+		goto restart;
+
+	down_read(&mm->mmap_sem);
+
+	raw_spin_lock_irqsave(&ifh->lock, flags);
+	list_for_each_entry(filter, &ifh->list, entry) {
+		event->addr_filters_offs[count] = 0;
+
+		if (perf_addr_filter_needs_mmap(filter))
+			event->addr_filters_offs[count] =
+				perf_addr_filter_apply(filter, mm);
+
+		count++;
+	}
+
+	event->addr_filters_gen++;
+	raw_spin_unlock_irqrestore(&ifh->lock, flags);
+
+	up_read(&mm->mmap_sem);
+
+	mmput(mm);
+
+restart:
+	perf_event_restart(event);
+}
+
+/*
+ * Address range filtering: limiting the data to certain
+ * instruction address ranges. Filters are ioctl()ed to us from
+ * userspace as ascii strings.
+ *
+ * Filter string format:
+ *
+ * ACTION RANGE_SPEC
+ * where ACTION is one of the
+ *  * "filter": limit the trace to this region
+ *  * "start": start tracing from this address
+ *  * "stop": stop tracing at this address/region;
+ * RANGE_SPEC is
+ *  * for kernel addresses: <start address>[/<size>]
+ *  * for object files:     <start address>[/<size>]@</path/to/object/file>
+ *
+ * if <size> is not specified, the range is treated as a single address.
+ */
+enum {
+	IF_ACT_FILTER,
+	IF_ACT_START,
+	IF_ACT_STOP,
+	IF_SRC_FILE,
+	IF_SRC_KERNEL,
+	IF_SRC_FILEADDR,
+	IF_SRC_KERNELADDR,
+};
+
+enum {
+	IF_STATE_ACTION = 0,
+	IF_STATE_SOURCE,
+	IF_STATE_END,
+};
+
+static const match_table_t if_tokens = {
+	{ IF_ACT_FILTER,	"filter" },
+	{ IF_ACT_START,		"start" },
+	{ IF_ACT_STOP,		"stop" },
+	{ IF_SRC_FILE,		"%u/%u@%s" },
+	{ IF_SRC_KERNEL,	"%u/%u" },
+	{ IF_SRC_FILEADDR,	"%u@%s" },
+	{ IF_SRC_KERNELADDR,	"%u" },
+};
+
+/*
+ * Address filter string parser
+ */
+static int
+perf_event_parse_addr_filter(struct perf_event *event, char *fstr,
+			     struct list_head *filters)
+{
+	struct perf_addr_filter *filter = NULL;
+	char *start, *orig, *filename = NULL;
+	struct path path;
+	substring_t args[MAX_OPT_ARGS];
+	int state = IF_STATE_ACTION, token;
+	unsigned int kernel = 0;
+	int ret = -EINVAL;
+
+	orig = fstr = kstrdup(fstr, GFP_KERNEL);
+	if (!fstr)
+		return -ENOMEM;
+
+	while ((start = strsep(&fstr, " ,\n")) != NULL) {
+		ret = -EINVAL;
+
+		if (!*start)
+			continue;
+
+		/* filter definition begins */
+		if (state == IF_STATE_ACTION) {
+			filter = perf_addr_filter_new(event, filters);
+			if (!filter)
+				goto fail;
+		}
+
+		token = match_token(start, if_tokens, args);
+		switch (token) {
+		case IF_ACT_FILTER:
+		case IF_ACT_START:
+			filter->filter = 1;
+
+		case IF_ACT_STOP:
+			if (state != IF_STATE_ACTION)
+				goto fail;
+
+			state = IF_STATE_SOURCE;
+			break;
+
+		case IF_SRC_KERNELADDR:
+		case IF_SRC_KERNEL:
+			kernel = 1;
+
+		case IF_SRC_FILEADDR:
+		case IF_SRC_FILE:
+			if (state != IF_STATE_SOURCE)
+				goto fail;
+
+			if (token == IF_SRC_FILE || token == IF_SRC_KERNEL)
+				filter->range = 1;
+
+			*args[0].to = 0;
+			ret = kstrtoul(args[0].from, 0, &filter->offset);
+			if (ret)
+				goto fail;
+
+			if (filter->range) {
+				*args[1].to = 0;
+				ret = kstrtoul(args[1].from, 0, &filter->size);
+				if (ret)
+					goto fail;
+			}
+
+			if (token == IF_SRC_FILE) {
+				filename = match_strdup(&args[2]);
+				if (!filename) {
+					ret = -ENOMEM;
+					goto fail;
+				}
+			}
+
+			state = IF_STATE_END;
+			break;
+
+		default:
+			goto fail;
+		}
+
+		/*
+		 * Filter definition is fully parsed, validate and install it.
+		 * Make sure that it doesn't contradict itself or the event's
+		 * attribute.
+		 */
+		if (state == IF_STATE_END) {
+			if (kernel && event->attr.exclude_kernel)
+				goto fail;
+
+			if (!kernel) {
+				if (!filename)
+					goto fail;
+
+				/* look up the path and grab its inode */
+				ret = kern_path(filename, LOOKUP_FOLLOW, &path);
+				if (ret)
+					goto fail_free_name;
+
+				filter->inode = igrab(d_inode(path.dentry));
+				path_put(&path);
+				kfree(filename);
+				filename = NULL;
+
+				ret = -EINVAL;
+				if (!filter->inode ||
+				    !S_ISREG(filter->inode->i_mode))
+					/* free_filters_list() will iput() */
+					goto fail;
+			}
+
+			/* ready to consume more filters */
+			state = IF_STATE_ACTION;
+			filter = NULL;
+		}
+	}
+
+	if (state != IF_STATE_ACTION)
+		goto fail;
+
+	kfree(orig);
+
+	return 0;
+
+fail_free_name:
+	kfree(filename);
+fail:
+	free_filters_list(filters);
+	kfree(orig);
+
+	return ret;
+}
+
+static int
+perf_event_set_addr_filter(struct perf_event *event, char *filter_str)
+{
+	LIST_HEAD(filters);
+	int ret;
+
+	/*
+	 * Since this is called in perf_ioctl() path, we're already holding
+	 * ctx::mutex.
+	 */
+	lockdep_assert_held(&event->ctx->mutex);
+
+	if (WARN_ON_ONCE(event->parent))
+		return -EINVAL;
+
+	/*
+	 * For now, we only support filtering in per-task events; doing so
+	 * for CPU-wide events requires additional context switching trickery,
+	 * since same object code will be mapped at different virtual
+	 * addresses in different processes.
+	 */
+	if (!event->ctx->task)
+		return -EOPNOTSUPP;
+
+	ret = perf_event_parse_addr_filter(event, filter_str, &filters);
+	if (ret)
+		return ret;
+
+	ret = event->pmu->addr_filters_validate(&filters);
+	if (ret) {
+		free_filters_list(&filters);
+		return ret;
+	}
+
+	/* remove existing filters, if any */
+	perf_addr_filters_splice(event, &filters);
+
+	/* install new filters */
+	perf_event_for_each_child(event, perf_event_addr_filters_apply);
+
+	return ret;
+}
+
+static int perf_event_set_filter(struct perf_event *event, void __user *arg)
+{
+	char *filter_str;
+	int ret = -EINVAL;
+
+	if ((event->attr.type != PERF_TYPE_TRACEPOINT ||
+	    !IS_ENABLED(CONFIG_EVENT_TRACING)) &&
+	    !has_addr_filter(event))
+		return -EINVAL;
+
+	filter_str = strndup_user(arg, PAGE_SIZE);
+	if (IS_ERR(filter_str))
+		return PTR_ERR(filter_str);
+
+	if (IS_ENABLED(CONFIG_EVENT_TRACING) &&
+	    event->attr.type == PERF_TYPE_TRACEPOINT)
+		ret = ftrace_profile_set_filter(event, event->attr.config,
+						filter_str);
+	else if (has_addr_filter(event))
+		ret = perf_event_set_addr_filter(event, filter_str);
+
+	kfree(filter_str);
+	return ret;
+}
+
+/*
  * hrtimer based swevent callback
  */
 
@@ -7542,6 +8273,20 @@ static void free_pmu_context(struct pmu *pmu)
 out:
 	mutex_unlock(&pmus_lock);
 }
+
+/*
+ * Let userspace know that this PMU supports address range filtering:
+ */
+static ssize_t nr_addr_filters_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->nr_addr_filters);
+}
+DEVICE_ATTR_RO(nr_addr_filters);
+
 static struct idr pmu_idr;
 
 static ssize_t
@@ -7643,9 +8388,19 @@ static int pmu_dev_alloc(struct pmu *pmu)
 	if (ret)
 		goto free_dev;
 
+	/* For PMUs with address filters, throw in an extra attribute: */
+	if (pmu->nr_addr_filters)
+		ret = device_create_file(pmu->dev, &dev_attr_nr_addr_filters);
+
+	if (ret)
+		goto del_dev;
+
 out:
 	return ret;
 
+del_dev:
+	device_del(pmu->dev);
+
 free_dev:
 	put_device(pmu->dev);
 	goto out;
@@ -7685,6 +8440,21 @@ int perf_pmu_register(struct pmu *pmu, const char *name, int type)
 	}
 
 skip_type:
+	if (pmu->task_ctx_nr == perf_hw_context) {
+		static int hw_context_taken = 0;
+
+		/*
+		 * Other than systems with heterogeneous CPUs, it never makes
+		 * sense for two PMUs to share perf_hw_context. PMUs which are
+		 * uncore must use perf_invalid_context.
+		 */
+		if (WARN_ON_ONCE(hw_context_taken &&
+		    !(pmu->capabilities & PERF_PMU_CAP_HETEROGENEOUS_CPUS)))
+			pmu->task_ctx_nr = perf_invalid_context;
+
+		hw_context_taken = 1;
+	}
+
 	pmu->pmu_cpu_context = find_pmu_context(pmu->task_ctx_nr);
 	if (pmu->pmu_cpu_context)
 		goto got_cpu_context;
@@ -7772,6 +8542,8 @@ void perf_pmu_unregister(struct pmu *pmu)
 	free_percpu(pmu->pmu_disable_count);
 	if (pmu->type >= PERF_TYPE_MAX)
 		idr_remove(&pmu_idr, pmu->type);
+	if (pmu->nr_addr_filters)
+		device_remove_file(pmu->dev, &dev_attr_nr_addr_filters);
 	device_del(pmu->dev);
 	put_device(pmu->dev);
 	free_pmu_context(pmu);
@@ -7965,6 +8737,7 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
 	INIT_LIST_HEAD(&event->sibling_list);
 	INIT_LIST_HEAD(&event->rb_entry);
 	INIT_LIST_HEAD(&event->active_entry);
+	INIT_LIST_HEAD(&event->addr_filters.list);
 	INIT_HLIST_NODE(&event->hlist_entry);
 
 
@@ -7972,6 +8745,7 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
 	init_irq_work(&event->pending, perf_pending_event);
 
 	mutex_init(&event->mmap_mutex);
+	raw_spin_lock_init(&event->addr_filters.lock);
 
 	atomic_long_set(&event->refcount, 1);
 	event->cpu		= cpu;
@@ -8006,8 +8780,16 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
 		context = parent_event->overflow_handler_context;
 	}
 
-	event->overflow_handler	= overflow_handler;
-	event->overflow_handler_context = context;
+	if (overflow_handler) {
+		event->overflow_handler	= overflow_handler;
+		event->overflow_handler_context = context;
+	} else if (is_write_backward(event)){
+		event->overflow_handler = perf_event_output_backward;
+		event->overflow_handler_context = NULL;
+	} else {
+		event->overflow_handler = perf_event_output_forward;
+		event->overflow_handler_context = NULL;
+	}
 
 	perf_event__state_init(event);
 
@@ -8048,11 +8830,22 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
 	if (err)
 		goto err_pmu;
 
+	if (has_addr_filter(event)) {
+		event->addr_filters_offs = kcalloc(pmu->nr_addr_filters,
+						   sizeof(unsigned long),
+						   GFP_KERNEL);
+		if (!event->addr_filters_offs)
+			goto err_per_task;
+
+		/* force hw sync on the address filters */
+		event->addr_filters_gen = 1;
+	}
+
 	if (!event->parent) {
 		if (event->attr.sample_type & PERF_SAMPLE_CALLCHAIN) {
 			err = get_callchain_buffers();
 			if (err)
-				goto err_per_task;
+				goto err_addr_filters;
 		}
 	}
 
@@ -8061,6 +8854,9 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,
 
 	return event;
 
+err_addr_filters:
+	kfree(event->addr_filters_offs);
+
 err_per_task:
 	exclusive_event_destroy(event);
 
@@ -8240,6 +9036,13 @@ perf_event_set_output(struct perf_event *event, struct perf_event *output_event)
 		goto out;
 
 	/*
+	 * Either writing ring buffer from beginning or from end.
+	 * Mixing is not allowed.
+	 */
+	if (is_write_backward(output_event) != is_write_backward(event))
+		goto out;
+
+	/*
 	 * If both events generate aux data, they must be on the same PMU
 	 */
 	if (has_aux(event) && has_aux(output_event) &&
diff --git a/kernel/events/internal.h b/kernel/events/internal.h
index 4199b6d193f5..05f9f6d626df 100644
--- a/kernel/events/internal.h
+++ b/kernel/events/internal.h
@@ -11,13 +11,13 @@
 struct ring_buffer {
 	atomic_t			refcount;
 	struct rcu_head			rcu_head;
-	struct irq_work			irq_work;
 #ifdef CONFIG_PERF_USE_VMALLOC
 	struct work_struct		work;
 	int				page_order;	/* allocation order  */
 #endif
 	int				nr_pages;	/* nr of data pages  */
 	int				overwrite;	/* can overwrite itself */
+	int				paused;		/* can write into ring buffer */
 
 	atomic_t			poll;		/* POLL_ for wakeups */
 
@@ -65,6 +65,14 @@ static inline void rb_free_rcu(struct rcu_head *rcu_head)
 	rb_free(rb);
 }
 
+static inline void rb_toggle_paused(struct ring_buffer *rb, bool pause)
+{
+	if (!pause && rb->nr_pages)
+		rb->paused = 0;
+	else
+		rb->paused = 1;
+}
+
 extern struct ring_buffer *
 rb_alloc(int nr_pages, long watermark, int cpu, int flags);
 extern void perf_event_wakeup(struct perf_event *event);
diff --git a/kernel/events/ring_buffer.c b/kernel/events/ring_buffer.c
index 7611d0f66cf8..ae9b90dc9a5a 100644
--- a/kernel/events/ring_buffer.c
+++ b/kernel/events/ring_buffer.c
@@ -102,8 +102,21 @@ out:
 	preempt_enable();
 }
 
-int perf_output_begin(struct perf_output_handle *handle,
-		      struct perf_event *event, unsigned int size)
+static bool __always_inline
+ring_buffer_has_space(unsigned long head, unsigned long tail,
+		      unsigned long data_size, unsigned int size,
+		      bool backward)
+{
+	if (!backward)
+		return CIRC_SPACE(head, tail, data_size) >= size;
+	else
+		return CIRC_SPACE(tail, head, data_size) >= size;
+}
+
+static int __always_inline
+__perf_output_begin(struct perf_output_handle *handle,
+		    struct perf_event *event, unsigned int size,
+		    bool backward)
 {
 	struct ring_buffer *rb;
 	unsigned long tail, offset, head;
@@ -125,8 +138,11 @@ int perf_output_begin(struct perf_output_handle *handle,
 	if (unlikely(!rb))
 		goto out;
 
-	if (unlikely(!rb->nr_pages))
+	if (unlikely(rb->paused)) {
+		if (rb->nr_pages)
+			local_inc(&rb->lost);
 		goto out;
+	}
 
 	handle->rb    = rb;
 	handle->event = event;
@@ -143,9 +159,12 @@ int perf_output_begin(struct perf_output_handle *handle,
 	do {
 		tail = READ_ONCE(rb->user_page->data_tail);
 		offset = head = local_read(&rb->head);
-		if (!rb->overwrite &&
-		    unlikely(CIRC_SPACE(head, tail, perf_data_size(rb)) < size))
-			goto fail;
+		if (!rb->overwrite) {
+			if (unlikely(!ring_buffer_has_space(head, tail,
+							    perf_data_size(rb),
+							    size, backward)))
+				goto fail;
+		}
 
 		/*
 		 * The above forms a control dependency barrier separating the
@@ -159,9 +178,17 @@ int perf_output_begin(struct perf_output_handle *handle,
 		 * See perf_output_put_handle().
 		 */
 
-		head += size;
+		if (!backward)
+			head += size;
+		else
+			head -= size;
 	} while (local_cmpxchg(&rb->head, offset, head) != offset);
 
+	if (backward) {
+		offset = head;
+		head = (u64)(-head);
+	}
+
 	/*
 	 * We rely on the implied barrier() by local_cmpxchg() to ensure
 	 * none of the data stores below can be lifted up by the compiler.
@@ -203,6 +230,26 @@ out:
 	return -ENOSPC;
 }
 
+int perf_output_begin_forward(struct perf_output_handle *handle,
+			     struct perf_event *event, unsigned int size)
+{
+	return __perf_output_begin(handle, event, size, false);
+}
+
+int perf_output_begin_backward(struct perf_output_handle *handle,
+			       struct perf_event *event, unsigned int size)
+{
+	return __perf_output_begin(handle, event, size, true);
+}
+
+int perf_output_begin(struct perf_output_handle *handle,
+		      struct perf_event *event, unsigned int size)
+{
+
+	return __perf_output_begin(handle, event, size,
+				   unlikely(is_write_backward(event)));
+}
+
 unsigned int perf_output_copy(struct perf_output_handle *handle,
 		      const void *buf, unsigned int len)
 {
@@ -221,8 +268,6 @@ void perf_output_end(struct perf_output_handle *handle)
 	rcu_read_unlock();
 }
 
-static void rb_irq_work(struct irq_work *work);
-
 static void
 ring_buffer_init(struct ring_buffer *rb, long watermark, int flags)
 {
@@ -243,16 +288,13 @@ ring_buffer_init(struct ring_buffer *rb, long watermark, int flags)
 
 	INIT_LIST_HEAD(&rb->event_list);
 	spin_lock_init(&rb->event_lock);
-	init_irq_work(&rb->irq_work, rb_irq_work);
-}
 
-static void ring_buffer_put_async(struct ring_buffer *rb)
-{
-	if (!atomic_dec_and_test(&rb->refcount))
-		return;
-
-	rb->rcu_head.next = (void *)rb;
-	irq_work_queue(&rb->irq_work);
+	/*
+	 * perf_output_begin() only checks rb->paused, therefore
+	 * rb->paused must be true if we have no pages for output.
+	 */
+	if (!rb->nr_pages)
+		rb->paused = 1;
 }
 
 /*
@@ -264,6 +306,10 @@ static void ring_buffer_put_async(struct ring_buffer *rb)
  * The ordering is similar to that of perf_output_{begin,end}, with
  * the exception of (B), which should be taken care of by the pmu
  * driver, since ordering rules will differ depending on hardware.
+ *
+ * Call this from pmu::start(); see the comment in perf_aux_output_end()
+ * about its use in pmu callbacks. Both can also be called from the PMI
+ * handler if needed.
  */
 void *perf_aux_output_begin(struct perf_output_handle *handle,
 			    struct perf_event *event)
@@ -288,6 +334,13 @@ void *perf_aux_output_begin(struct perf_output_handle *handle,
 		goto err;
 
 	/*
+	 * If rb::aux_mmap_count is zero (and rb_has_aux() above went through),
+	 * the aux buffer is in perf_mmap_close(), about to get freed.
+	 */
+	if (!atomic_read(&rb->aux_mmap_count))
+		goto err_put;
+
+	/*
 	 * Nesting is not supported for AUX area, make sure nested
 	 * writers are caught early
 	 */
@@ -328,10 +381,11 @@ void *perf_aux_output_begin(struct perf_output_handle *handle,
 	return handle->rb->aux_priv;
 
 err_put:
+	/* can't be last */
 	rb_free_aux(rb);
 
 err:
-	ring_buffer_put_async(rb);
+	ring_buffer_put(rb);
 	handle->event = NULL;
 
 	return NULL;
@@ -342,6 +396,10 @@ err:
  * aux_head and posting a PERF_RECORD_AUX into the perf buffer. It is the
  * pmu driver's responsibility to observe ordering rules of the hardware,
  * so that all the data is externally visible before this is called.
+ *
+ * Note: this has to be called from pmu::stop() callback, as the assumption
+ * of the AUX buffer management code is that after pmu::stop(), the AUX
+ * transaction must be stopped and therefore drop the AUX reference count.
  */
 void perf_aux_output_end(struct perf_output_handle *handle, unsigned long size,
 			 bool truncated)
@@ -389,8 +447,9 @@ void perf_aux_output_end(struct perf_output_handle *handle, unsigned long size,
 	handle->event = NULL;
 
 	local_set(&rb->aux_nest, 0);
+	/* can't be last */
 	rb_free_aux(rb);
-	ring_buffer_put_async(rb);
+	ring_buffer_put(rb);
 }
 
 /*
@@ -471,6 +530,14 @@ static void __rb_free_aux(struct ring_buffer *rb)
 {
 	int pg;
 
+	/*
+	 * Should never happen, the last reference should be dropped from
+	 * perf_mmap_close() path, which first stops aux transactions (which
+	 * in turn are the atomic holders of aux_refcount) and then does the
+	 * last rb_free_aux().
+	 */
+	WARN_ON_ONCE(in_atomic());
+
 	if (rb->aux_priv) {
 		rb->free_aux(rb->aux_priv);
 		rb->free_aux = NULL;
@@ -582,18 +649,7 @@ out:
 void rb_free_aux(struct ring_buffer *rb)
 {
 	if (atomic_dec_and_test(&rb->aux_refcount))
-		irq_work_queue(&rb->irq_work);
-}
-
-static void rb_irq_work(struct irq_work *work)
-{
-	struct ring_buffer *rb = container_of(work, struct ring_buffer, irq_work);
-
-	if (!atomic_read(&rb->aux_refcount))
 		__rb_free_aux(rb);
-
-	if (rb->rcu_head.next == (void *)rb)
-		call_rcu(&rb->rcu_head, rb_free_rcu);
 }
 
 #ifndef CONFIG_PERF_USE_VMALLOC
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 725587f10667..c8b318663525 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -130,6 +130,9 @@ static int one_thousand = 1000;
 #ifdef CONFIG_PRINTK
 static int ten_thousand = 10000;
 #endif
+#ifdef CONFIG_PERF_EVENTS
+static int six_hundred_forty_kb = 640 * 1024;
+#endif
 
 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
@@ -1144,6 +1147,15 @@ static struct ctl_table kern_table[] = {
 		.extra1		= &zero,
 		.extra2		= &one_hundred,
 	},
+	{
+		.procname	= "perf_event_max_stack",
+		.data		= NULL, /* filled in by handler */
+		.maxlen		= sizeof(sysctl_perf_event_max_stack),
+		.mode		= 0644,
+		.proc_handler	= perf_event_max_stack_handler,
+		.extra1		= &zero,
+		.extra2		= &six_hundred_forty_kb,
+	},
 #endif
 #ifdef CONFIG_KMEMCHECK
 	{
diff --git a/kernel/trace/trace_event_perf.c b/kernel/trace/trace_event_perf.c
index 00df25fd86ef..e11108f1d197 100644
--- a/kernel/trace/trace_event_perf.c
+++ b/kernel/trace/trace_event_perf.c
@@ -47,6 +47,9 @@ static int perf_trace_event_perm(struct trace_event_call *tp_event,
 		if (perf_paranoid_tracepoint_raw() && !capable(CAP_SYS_ADMIN))
 			return -EPERM;
 
+		if (!is_sampling_event(p_event))
+			return 0;
+
 		/*
 		 * We don't allow user space callchains for  function trace
 		 * event, due to issues with page faults while tracing page