diff --git a/kernel/events/core.c b/kernel/events/core.c
index 8ff42177710f7ef745fe5c538a8f1892b18b536e..f53d7c6ad513122944153b1fe0080bcf6c0f7587 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -1259,8 +1259,9 @@ static void put_ctx(struct perf_event_context *ctx)
  *	  perf_event_context::mutex
  *	    perf_event::child_mutex;
  *	      perf_event_context::lock
- *	    perf_event::mmap_mutex
  *	    mmap_lock
+ *	      perf_event::mmap_mutex
+ *	        perf_buffer::aux_mutex
  *	      perf_addr_filters_head::lock
  *
  *    cpu_hotplug_lock
@@ -6355,17 +6356,17 @@ static void perf_mmap_close(struct vm_area_struct *vma)
 	int mmap_locked = rb->mmap_locked;
 	unsigned long size = perf_data_size(rb);
 	bool detach_rest = false;
+	struct perf_buffer_ext *rb_ext = container_of(rb, struct perf_buffer_ext, perf_buffer);
 
 	if (event->pmu->event_unmapped)
 		event->pmu->event_unmapped(event, vma->vm_mm);
 
 	/*
-	 * rb->aux_mmap_count will always drop before rb->mmap_count and
-	 * event->mmap_count, so it is ok to use event->mmap_mutex to
-	 * serialize with perf_mmap here.
+	 * The AUX buffer is strictly a sub-buffer, serialize using aux_mutex
+	 * to avoid complications.
 	 */
 	if (rb_has_aux(rb) && vma->vm_pgoff == rb->aux_pgoff &&
-	    atomic_dec_and_mutex_lock(&rb->aux_mmap_count, &event->mmap_mutex)) {
+	    atomic_dec_and_mutex_lock(&rb->aux_mmap_count, &rb_ext->aux_mutex)) {
 		/*
 		 * Stop all AUX events that are writing to this buffer,
 		 * so that we can free its AUX pages and corresponding PMU
@@ -6382,7 +6383,7 @@ static void perf_mmap_close(struct vm_area_struct *vma)
 		rb_free_aux(rb);
 		WARN_ON_ONCE(refcount_read(&rb->aux_refcount));
 
-		mutex_unlock(&event->mmap_mutex);
+		mutex_unlock(&rb_ext->aux_mutex);
 	}
 
 	if (atomic_dec_and_test(&rb->mmap_count))
@@ -6470,7 +6471,9 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma)
 	struct perf_event *event = file->private_data;
 	unsigned long user_locked, user_lock_limit;
 	struct user_struct *user = current_user();
+	struct mutex *aux_mutex = NULL;
 	struct perf_buffer *rb = NULL;
+	struct perf_buffer_ext *rb_ext = NULL;
 	unsigned long locked, lock_limit;
 	unsigned long vma_size;
 	unsigned long nr_pages;
@@ -6518,6 +6521,10 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma)
 		if (!rb)
 			goto aux_unlock;
 
+		rb_ext = container_of(rb, struct perf_buffer_ext, perf_buffer);
+		aux_mutex = &rb_ext->aux_mutex;
+		mutex_lock(aux_mutex);
+
 		aux_offset = READ_ONCE(rb->user_page->aux_offset);
 		aux_size = READ_ONCE(rb->user_page->aux_size);
 
@@ -6668,6 +6675,8 @@ static int perf_mmap(struct file *file, struct vm_area_struct *vma)
 		atomic_dec(&rb->mmap_count);
 	}
 aux_unlock:
+	if (aux_mutex)
+		mutex_unlock(aux_mutex);
 	mutex_unlock(&event->mmap_mutex);
 
 	/*
diff --git a/kernel/events/internal.h b/kernel/events/internal.h
index 386d21c7edfa038d51880be8789cb4c4206595e3..d2e6e6144c54273ef0ecde17aca5975b42c03af4 100644
--- a/kernel/events/internal.h
+++ b/kernel/events/internal.h
@@ -58,6 +58,11 @@ struct perf_buffer {
 	void				*data_pages[];
 };
 
+struct perf_buffer_ext {
+	struct mutex		aux_mutex;
+	struct perf_buffer	perf_buffer;
+};
+
 extern void rb_free(struct perf_buffer *rb);
 
 static inline void rb_free_rcu(struct rcu_head *rcu_head)
diff --git a/kernel/events/ring_buffer.c b/kernel/events/ring_buffer.c
index f1f4a627f93db1fda1113c473dca1f6f753773b4..cfd9448ce28fb53495ebea58c6595697cf03c236 100644
--- a/kernel/events/ring_buffer.c
+++ b/kernel/events/ring_buffer.c
@@ -310,6 +310,7 @@ static void
 ring_buffer_init(struct perf_buffer *rb, long watermark, int flags)
 {
 	long max_size = perf_data_size(rb);
+	struct perf_buffer_ext *rb_ext = container_of(rb, struct perf_buffer_ext, perf_buffer);
 
 	if (watermark)
 		rb->watermark = min(max_size, watermark);
@@ -333,6 +334,8 @@ ring_buffer_init(struct perf_buffer *rb, long watermark, int flags)
 	 */
 	if (!rb->nr_pages)
 		rb->paused = 1;
+
+	mutex_init(&rb_ext->aux_mutex);
 }
 
 void perf_aux_output_flag(struct perf_output_handle *handle, u64 flags)
@@ -816,21 +819,23 @@ static void perf_mmap_free_page(void *addr)
 
 struct perf_buffer *rb_alloc(int nr_pages, long watermark, int cpu, int flags)
 {
+	struct perf_buffer_ext *rb_ext;
 	struct perf_buffer *rb;
 	unsigned long size;
 	int i, node;
 
-	size = sizeof(struct perf_buffer);
+	size = sizeof(struct perf_buffer_ext);
 	size += nr_pages * sizeof(void *);
 
 	if (order_base_2(size) > PAGE_SHIFT+MAX_ORDER)
 		goto fail;
 
 	node = (cpu == -1) ? cpu : cpu_to_node(cpu);
-	rb = kzalloc_node(size, GFP_KERNEL, node);
-	if (!rb)
+	rb_ext = kzalloc_node(size, GFP_KERNEL, node);
+	if (!rb_ext)
 		goto fail;
 
+	rb = &rb_ext->perf_buffer;
 	rb->user_page = perf_mmap_alloc_page(cpu);
 	if (!rb->user_page)
 		goto fail_user_page;
@@ -854,7 +859,7 @@ struct perf_buffer *rb_alloc(int nr_pages, long watermark, int cpu, int flags)
 	perf_mmap_free_page(rb->user_page);
 
 fail_user_page:
-	kfree(rb);
+	kfree(rb_ext);
 
 fail:
 	return NULL;
@@ -863,11 +868,12 @@ struct perf_buffer *rb_alloc(int nr_pages, long watermark, int cpu, int flags)
 void rb_free(struct perf_buffer *rb)
 {
 	int i;
+	struct perf_buffer_ext *rb_ext = container_of(rb, struct perf_buffer_ext, perf_buffer);
 
 	perf_mmap_free_page(rb->user_page);
 	for (i = 0; i < rb->nr_pages; i++)
 		perf_mmap_free_page(rb->data_pages[i]);
-	kfree(rb);
+	kfree(rb_ext);
 }
 
 #else
@@ -890,6 +896,7 @@ static void perf_mmap_unmark_page(void *addr)
 
 static void rb_free_work(struct work_struct *work)
 {
+	struct perf_buffer_ext *rb_ext;
 	struct perf_buffer *rb;
 	void *base;
 	int i, nr;
@@ -903,7 +910,8 @@ static void rb_free_work(struct work_struct *work)
 		perf_mmap_unmark_page(base + (i * PAGE_SIZE));
 
 	vfree(base);
-	kfree(rb);
+	rb_ext = container_of(rb, struct perf_buffer_ext, perf_buffer);
+	kfree(rb_ext);
 }
 
 void rb_free(struct perf_buffer *rb)
@@ -913,19 +921,21 @@ void rb_free(struct perf_buffer *rb)
 
 struct perf_buffer *rb_alloc(int nr_pages, long watermark, int cpu, int flags)
 {
+	struct perf_buffer_ext *rb_ext;
 	struct perf_buffer *rb;
 	unsigned long size;
 	void *all_buf;
 	int node;
 
-	size = sizeof(struct perf_buffer);
+	size = sizeof(struct perf_buffer_ext);
 	size += sizeof(void *);
 
 	node = (cpu == -1) ? cpu : cpu_to_node(cpu);
-	rb = kzalloc_node(size, GFP_KERNEL, node);
-	if (!rb)
+	rb_ext = kzalloc_node(size, GFP_KERNEL, node);
+	if (!rb_ext)
 		goto fail;
 
+	rb = &rb_ext->perf_buffer;
 	INIT_WORK(&rb->work, rb_free_work);
 
 	all_buf = vmalloc_user((nr_pages + 1) * PAGE_SIZE);
@@ -944,7 +954,7 @@ struct perf_buffer *rb_alloc(int nr_pages, long watermark, int cpu, int flags)
 	return rb;
 
 fail_all_buf:
-	kfree(rb);
+	kfree(rb_ext);
 
 fail:
 	return NULL;