1 files changed, 163 insertions, 46 deletions

diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
index 5e1b9f6e77f3..ed505c6de7ca 100644
--- a/kernel/trace/ring_buffer.c
+++ b/kernel/trace/ring_buffer.c
@@ -477,6 +477,7 @@ struct ring_buffer_per_cpu {
 	local_t				committing;
 	local_t				commits;
 	local_t				pages_touched;
+	local_t				pages_lost;
 	local_t				pages_read;
 	long				last_pages_touch;
 	size_t				shortest_full;
@@ -484,6 +485,8 @@ struct ring_buffer_per_cpu {
 	unsigned long			read_bytes;
 	u64				write_stamp;
 	u64				read_stamp;
+	/* pages removed since last reset */
+	unsigned long			pages_removed;
 	/* ring buffer pages to update, > 0 to add, < 0 to remove */
 	long				nr_pages_to_update;
 	struct list_head		new_pages; /* new pages to add */
@@ -519,6 +522,7 @@ struct ring_buffer_iter {
 	struct buffer_page		*head_page;
 	struct buffer_page		*cache_reader_page;
 	unsigned long			cache_read;
+	unsigned long			cache_pages_removed;
 	u64				read_stamp;
 };
 
@@ -544,10 +548,18 @@ size_t ring_buffer_nr_pages(struct ring_buffer *buffer, int cpu)
 size_t ring_buffer_nr_dirty_pages(struct ring_buffer *buffer, int cpu)
 {
 	size_t read;
+	size_t lost;
 	size_t cnt;
 
 	read = local_read(&buffer->buffers[cpu]->pages_read);
+	lost = local_read(&buffer->buffers[cpu]->pages_lost);
 	cnt = local_read(&buffer->buffers[cpu]->pages_touched);
+
+	if (WARN_ON_ONCE(cnt < lost))
+		return 0;
+
+	cnt -= lost;
+
 	/* The reader can read an empty page, but not more than that */
 	if (cnt < read) {
 		WARN_ON_ONCE(read > cnt + 1);
@@ -557,6 +569,26 @@ size_t ring_buffer_nr_dirty_pages(struct ring_buffer *buffer, int cpu)
 	return cnt - read;
 }
 
+static __always_inline bool full_hit(struct ring_buffer *buffer, int cpu, int full)
+{
+	struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu];
+	size_t nr_pages;
+	size_t dirty;
+
+	nr_pages = cpu_buffer->nr_pages;
+	if (!nr_pages || !full)
+		return true;
+
+	/*
+	 * Add one as dirty will never equal nr_pages, as the sub-buffer
+	 * that the writer is on is not counted as dirty.
+	 * This is needed if "buffer_percent" is set to 100.
+	 */
+	dirty = ring_buffer_nr_dirty_pages(buffer, cpu) + 1;
+
+	return (dirty * 100) >= (full * nr_pages);
+}
+
 /*
  * rb_wake_up_waiters - wake up tasks waiting for ring buffer input
  *
@@ -568,8 +600,9 @@ static void rb_wake_up_waiters(struct irq_work *work)
 	struct rb_irq_work *rbwork = container_of(work, struct rb_irq_work, work);
 
 	wake_up_all(&rbwork->waiters);
-	if (rbwork->wakeup_full) {
+	if (rbwork->full_waiters_pending || rbwork->wakeup_full) {
 		rbwork->wakeup_full = false;
+		rbwork->full_waiters_pending = false;
 		wake_up_all(&rbwork->full_waiters);
 	}
 }
@@ -586,7 +619,7 @@ static void rb_wake_up_waiters(struct irq_work *work)
  */
 int ring_buffer_wait(struct ring_buffer *buffer, int cpu, int full)
 {
-	struct ring_buffer_per_cpu *uninitialized_var(cpu_buffer);
+	struct ring_buffer_per_cpu *cpu_buffer;
 	DEFINE_WAIT(wait);
 	struct rb_irq_work *work;
 	int ret = 0;
@@ -651,22 +684,20 @@ int ring_buffer_wait(struct ring_buffer *buffer, int cpu, int full)
 		    !ring_buffer_empty_cpu(buffer, cpu)) {
 			unsigned long flags;
 			bool pagebusy;
-			size_t nr_pages;
-			size_t dirty;
+			bool done;
 
 			if (!full)
 				break;
 
 			raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags);
 			pagebusy = cpu_buffer->reader_page == cpu_buffer->commit_page;
-			nr_pages = cpu_buffer->nr_pages;
-			dirty = ring_buffer_nr_dirty_pages(buffer, cpu);
+			done = !pagebusy && full_hit(buffer, cpu, full);
+
 			if (!cpu_buffer->shortest_full ||
-			    cpu_buffer->shortest_full < full)
+			    cpu_buffer->shortest_full > full)
 				cpu_buffer->shortest_full = full;
 			raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);
-			if (!pagebusy &&
-			    (!nr_pages || (dirty * 100) > full * nr_pages))
+			if (done)
 				break;
 		}
 
@@ -687,6 +718,7 @@ int ring_buffer_wait(struct ring_buffer *buffer, int cpu, int full)
  * @cpu: the cpu buffer to wait on
  * @filp: the file descriptor
  * @poll_table: The poll descriptor
+ * @full: wait until the percentage of pages are available, if @cpu != RING_BUFFER_ALL_CPUS
  *
  * If @cpu == RING_BUFFER_ALL_CPUS then the task will wake up as soon
  * as data is added to any of the @buffer's cpu buffers. Otherwise
@@ -696,23 +728,33 @@ int ring_buffer_wait(struct ring_buffer *buffer, int cpu, int full)
  * zero otherwise.
  */
 __poll_t ring_buffer_poll_wait(struct ring_buffer *buffer, int cpu,
-			  struct file *filp, poll_table *poll_table)
+			  struct file *filp, poll_table *poll_table, int full)
 {
 	struct ring_buffer_per_cpu *cpu_buffer;
 	struct rb_irq_work *work;
 
-	if (cpu == RING_BUFFER_ALL_CPUS)
+	if (cpu == RING_BUFFER_ALL_CPUS) {
 		work = &buffer->irq_work;
-	else {
+		full = 0;
+	} else {
 		if (!cpumask_test_cpu(cpu, buffer->cpumask))
-			return -EINVAL;
+			return EPOLLERR;
 
 		cpu_buffer = buffer->buffers[cpu];
 		work = &cpu_buffer->irq_work;
 	}
 
-	poll_wait(filp, &work->waiters, poll_table);
-	work->waiters_pending = true;
+	if (full) {
+		poll_wait(filp, &work->full_waiters, poll_table);
+		work->full_waiters_pending = true;
+		if (!cpu_buffer->shortest_full ||
+		    cpu_buffer->shortest_full > full)
+			cpu_buffer->shortest_full = full;
+	} else {
+		poll_wait(filp, &work->waiters, poll_table);
+		work->waiters_pending = true;
+	}
+
 	/*
 	 * There's a tight race between setting the waiters_pending and
 	 * checking if the ring buffer is empty.  Once the waiters_pending bit
@@ -728,6 +770,9 @@ __poll_t ring_buffer_poll_wait(struct ring_buffer *buffer, int cpu,
 	 */
 	smp_mb();
 
+	if (full)
+		return full_hit(buffer, cpu, full) ? EPOLLIN | EPOLLRDNORM : 0;
+
 	if ((cpu == RING_BUFFER_ALL_CPUS && !ring_buffer_empty(buffer)) ||
 	    (cpu != RING_BUFFER_ALL_CPUS && !ring_buffer_empty_cpu(buffer, cpu)))
 		return EPOLLIN | EPOLLRDNORM;
@@ -1359,11 +1404,13 @@ static void rb_free_cpu_buffer(struct ring_buffer_per_cpu *cpu_buffer)
 	struct list_head *head = cpu_buffer->pages;
 	struct buffer_page *bpage, *tmp;
 
-	free_buffer_page(cpu_buffer->reader_page);
+	irq_work_sync(&cpu_buffer->irq_work.work);
 
-	rb_head_page_deactivate(cpu_buffer);
+	free_buffer_page(cpu_buffer->reader_page);
 
 	if (head) {
+		rb_head_page_deactivate(cpu_buffer);
+
 		list_for_each_entry_safe(bpage, tmp, head, list) {
 			list_del_init(&bpage->list);
 			free_buffer_page(bpage);
@@ -1372,6 +1419,8 @@ static void rb_free_cpu_buffer(struct ring_buffer_per_cpu *cpu_buffer)
 		free_buffer_page(bpage);
 	}
 
+	free_page((unsigned long)cpu_buffer->free_page);
+
 	kfree(cpu_buffer);
 }
 
@@ -1464,6 +1513,8 @@ ring_buffer_free(struct ring_buffer *buffer)
 
 	cpuhp_state_remove_instance(CPUHP_TRACE_RB_PREPARE, &buffer->node);
 
+	irq_work_sync(&buffer->irq_work.work);
+
 	for_each_buffer_cpu(buffer, cpu)
 		rb_free_cpu_buffer(buffer->buffers[cpu]);
 
@@ -1543,6 +1594,8 @@ rb_remove_pages(struct ring_buffer_per_cpu *cpu_buffer, unsigned long nr_pages)
 		to_remove = rb_list_head(to_remove)->next;
 		head_bit |= (unsigned long)to_remove & RB_PAGE_HEAD;
 	}
+	/* Read iterators need to reset themselves when some pages removed */
+	cpu_buffer->pages_removed += nr_removed;
 
 	next_page = rb_list_head(to_remove)->next;
 
@@ -1564,12 +1617,6 @@ rb_remove_pages(struct ring_buffer_per_cpu *cpu_buffer, unsigned long nr_pages)
 		cpu_buffer->head_page = list_entry(next_page,
 						struct buffer_page, list);
 
-	/*
-	 * change read pointer to make sure any read iterators reset
-	 * themselves
-	 */
-	cpu_buffer->read = 0;
-
 	/* pages are removed, resume tracing and then free the pages */
 	atomic_dec(&cpu_buffer->record_disabled);
 	raw_spin_unlock_irq(&cpu_buffer->reader_lock);
@@ -1598,6 +1645,7 @@ rb_remove_pages(struct ring_buffer_per_cpu *cpu_buffer, unsigned long nr_pages)
 			 */
 			local_add(page_entries, &cpu_buffer->overrun);
 			local_sub(BUF_PAGE_SIZE, &cpu_buffer->entries_bytes);
+			local_inc(&cpu_buffer->pages_lost);
 		}
 
 		/*
@@ -1783,6 +1831,8 @@ int ring_buffer_resize(struct ring_buffer *buffer, unsigned long size,
 				err = -ENOMEM;
 				goto out_err;
 			}
+
+			cond_resched();
 		}
 
 		get_online_cpus();
@@ -2022,6 +2072,7 @@ rb_handle_head_page(struct ring_buffer_per_cpu *cpu_buffer,
 		 */
 		local_add(entries, &cpu_buffer->overrun);
 		local_sub(BUF_PAGE_SIZE, &cpu_buffer->entries_bytes);
+		local_inc(&cpu_buffer->pages_lost);
 
 		/*
 		 * The entries will be zeroed out when we move the
@@ -2190,6 +2241,9 @@ rb_reset_tail(struct ring_buffer_per_cpu *cpu_buffer,
 		/* Mark the rest of the page with padding */
 		rb_event_set_padding(event);
 
+		/* Make sure the padding is visible before the write update */
+		smp_wmb();
+
 		/* Set the write back to the previous setting */
 		local_sub(length, &tail_page->write);
 		return;
@@ -2201,6 +2255,9 @@ rb_reset_tail(struct ring_buffer_per_cpu *cpu_buffer,
 	/* time delta must be non zero */
 	event->time_delta = 1;
 
+	/* Make sure the padding is visible before the tail_page->write update */
+	smp_wmb();
+
 	/* Set write to end of buffer */
 	length = (tail + length) - BUF_PAGE_SIZE;
 	local_sub(length, &tail_page->write);
@@ -2490,6 +2547,10 @@ rb_set_commit_to_write(struct ring_buffer_per_cpu *cpu_buffer)
 		if (RB_WARN_ON(cpu_buffer,
 			       rb_is_reader_page(cpu_buffer->tail_page)))
 			return;
+		/*
+		 * No need for a memory barrier here, as the update
+		 * of the tail_page did it for this page.
+		 */
 		local_set(&cpu_buffer->commit_page->page->commit,
 			  rb_page_write(cpu_buffer->commit_page));
 		rb_inc_page(cpu_buffer, &cpu_buffer->commit_page);
@@ -2503,6 +2564,8 @@ rb_set_commit_to_write(struct ring_buffer_per_cpu *cpu_buffer)
 	while (rb_commit_index(cpu_buffer) !=
 	       rb_page_write(cpu_buffer->commit_page)) {
 
+		/* Make sure the readers see the content of what is committed. */
+		smp_wmb();
 		local_set(&cpu_buffer->commit_page->page->commit,
 			  rb_page_write(cpu_buffer->commit_page));
 		RB_WARN_ON(cpu_buffer,
@@ -2622,10 +2685,6 @@ static void rb_commit(struct ring_buffer_per_cpu *cpu_buffer,
 static __always_inline void
 rb_wakeups(struct ring_buffer *buffer, struct ring_buffer_per_cpu *cpu_buffer)
 {
-	size_t nr_pages;
-	size_t dirty;
-	size_t full;
-
 	if (buffer->irq_work.waiters_pending) {
 		buffer->irq_work.waiters_pending = false;
 		/* irq_work_queue() supplies it's own memory barriers */
@@ -2649,10 +2708,7 @@ rb_wakeups(struct ring_buffer *buffer, struct ring_buffer_per_cpu *cpu_buffer)
 
 	cpu_buffer->last_pages_touch = local_read(&cpu_buffer->pages_touched);
 
-	full = cpu_buffer->shortest_full;
-	nr_pages = cpu_buffer->nr_pages;
-	dirty = ring_buffer_nr_dirty_pages(buffer, cpu_buffer->cpu);
-	if (full && nr_pages && (dirty * 100) <= full * nr_pages)
+	if (!full_hit(buffer, cpu_buffer->cpu, cpu_buffer->shortest_full))
 		return;
 
 	cpu_buffer->irq_work.wakeup_full = true;
@@ -2920,6 +2976,12 @@ rb_reserve_next_event(struct ring_buffer *buffer,
 	int nr_loops = 0;
 	u64 diff;
 
+	/* ring buffer does cmpxchg, make sure it is safe in NMI context */
+	if (!IS_ENABLED(CONFIG_ARCH_HAVE_NMI_SAFE_CMPXCHG) &&
+	    (unlikely(in_nmi()))) {
+		return NULL;
+	}
+
 	rb_start_commit(cpu_buffer);
 
 #ifdef CONFIG_RING_BUFFER_ALLOW_SWAP
@@ -3614,6 +3676,7 @@ static void rb_iter_reset(struct ring_buffer_iter *iter)
 
 	iter->cache_reader_page = iter->head_page;
 	iter->cache_read = cpu_buffer->read;
+	iter->cache_pages_removed = cpu_buffer->pages_removed;
 
 	if (iter->head)
 		iter->read_stamp = cpu_buffer->read_stamp;
@@ -3863,6 +3926,38 @@ rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
 	arch_spin_unlock(&cpu_buffer->lock);
 	local_irq_restore(flags);
 
+	/*
+	 * The writer has preempt disable, wait for it. But not forever
+	 * Although, 1 second is pretty much "forever"
+	 */
+#define USECS_WAIT	1000000
+        for (nr_loops = 0; nr_loops < USECS_WAIT; nr_loops++) {
+		/* If the write is past the end of page, a writer is still updating it */
+		if (likely(!reader || rb_page_write(reader) <= BUF_PAGE_SIZE))
+			break;
+
+		udelay(1);
+
+		/* Get the latest version of the reader write value */
+		smp_rmb();
+	}
+
+	/* The writer is not moving forward? Something is wrong */
+	if (RB_WARN_ON(cpu_buffer, nr_loops == USECS_WAIT))
+		reader = NULL;
+
+	/*
+	 * Make sure we see any padding after the write update
+	 * (see rb_reset_tail()).
+	 *
+	 * In addition, a writer may be writing on the reader page
+	 * if the page has not been fully filled, so the read barrier
+	 * is also needed to make sure we see the content of what is
+	 * committed by the writer (see rb_set_commit_to_write()).
+	 */
+	smp_rmb();
+
+
 	return reader;
 }
 
@@ -4024,12 +4119,13 @@ rb_iter_peek(struct ring_buffer_iter *iter, u64 *ts)
 	buffer = cpu_buffer->buffer;
 
 	/*
-	 * Check if someone performed a consuming read to
-	 * the buffer. A consuming read invalidates the iterator
-	 * and we need to reset the iterator in this case.
+	 * Check if someone performed a consuming read to the buffer
+	 * or removed some pages from the buffer. In these cases,
+	 * iterator was invalidated and we need to reset it.
 	 */
 	if (unlikely(iter->cache_read != cpu_buffer->read ||
-		     iter->cache_reader_page != cpu_buffer->reader_page))
+		     iter->cache_reader_page != cpu_buffer->reader_page ||
+		     iter->cache_pages_removed != cpu_buffer->pages_removed))
 		rb_iter_reset(iter);
 
  again:
@@ -4410,28 +4506,34 @@ unsigned long ring_buffer_size(struct ring_buffer *buffer, int cpu)
 }
 EXPORT_SYMBOL_GPL(ring_buffer_size);
 
+static void rb_clear_buffer_page(struct buffer_page *page)
+{
+	local_set(&page->write, 0);
+	local_set(&page->entries, 0);
+	rb_init_page(page->page);
+	page->read = 0;
+}
+
 static void
 rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer)
 {
+	struct buffer_page *page;
+
 	rb_head_page_deactivate(cpu_buffer);
 
 	cpu_buffer->head_page
 		= list_entry(cpu_buffer->pages, struct buffer_page, list);
-	local_set(&cpu_buffer->head_page->write, 0);
-	local_set(&cpu_buffer->head_page->entries, 0);
-	local_set(&cpu_buffer->head_page->page->commit, 0);
-
-	cpu_buffer->head_page->read = 0;
+	rb_clear_buffer_page(cpu_buffer->head_page);
+	list_for_each_entry(page, cpu_buffer->pages, list) {
+		rb_clear_buffer_page(page);
+	}
 
 	cpu_buffer->tail_page = cpu_buffer->head_page;
 	cpu_buffer->commit_page = cpu_buffer->head_page;
 
 	INIT_LIST_HEAD(&cpu_buffer->reader_page->list);
 	INIT_LIST_HEAD(&cpu_buffer->new_pages);
-	local_set(&cpu_buffer->reader_page->write, 0);
-	local_set(&cpu_buffer->reader_page->entries, 0);
-	local_set(&cpu_buffer->reader_page->page->commit, 0);
-	cpu_buffer->reader_page->read = 0;
+	rb_clear_buffer_page(cpu_buffer->reader_page);
 
 	local_set(&cpu_buffer->entries_bytes, 0);
 	local_set(&cpu_buffer->overrun, 0);
@@ -4441,6 +4543,7 @@ rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer)
 	local_set(&cpu_buffer->committing, 0);
 	local_set(&cpu_buffer->commits, 0);
 	local_set(&cpu_buffer->pages_touched, 0);
+	local_set(&cpu_buffer->pages_lost, 0);
 	local_set(&cpu_buffer->pages_read, 0);
 	cpu_buffer->last_pages_touch = 0;
 	cpu_buffer->shortest_full = 0;
@@ -4454,6 +4557,7 @@ rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer)
 	cpu_buffer->last_overrun = 0;
 
 	rb_head_page_activate(cpu_buffer);
+	cpu_buffer->pages_removed = 0;
 }
 
 /**
@@ -4706,11 +4810,16 @@ EXPORT_SYMBOL_GPL(ring_buffer_alloc_read_page);
  */
 void ring_buffer_free_read_page(struct ring_buffer *buffer, int cpu, void *data)
 {
-	struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu];
+	struct ring_buffer_per_cpu *cpu_buffer;
 	struct buffer_data_page *bpage = data;
 	struct page *page = virt_to_page(bpage);
 	unsigned long flags;
 
+	if (!buffer || !buffer->buffers || !buffer->buffers[cpu])
+		return;
+
+	cpu_buffer = buffer->buffers[cpu];
+
 	/* If the page is still in use someplace else, we can't reuse it */
 	if (page_ref_count(page) > 1)
 		goto out;
@@ -4825,7 +4934,15 @@ int ring_buffer_read_page(struct ring_buffer *buffer,
 		unsigned int pos = 0;
 		unsigned int size;
 
-		if (full)
+		/*
+		 * If a full page is expected, this can still be returned
+		 * if there's been a previous partial read and the
+		 * rest of the page can be read and the commit page is off
+		 * the reader page.
+		 */
+		if (full &&
+		    (!read || (len < (commit - read)) ||
+		     cpu_buffer->reader_page == cpu_buffer->commit_page))
 			goto out_unlock;
 
 		if (len > (commit - read))