ring-buffer: Fixes for the changes in the last merge window

Ingo,

This patch series has a couple of fixes for bugs that snuck into the last merge window. The first is a crash that can happen because of an uninitialized list, and failing to allocate the requested amount. This can happen by changing the buffer_size_kb to a huge amount as the first change that is done to it after booting up (note, the bug does not happen if a successful change happens first).

The second patch can cause the ring buffer accounting (which is used by the infrastructure of the ring buffer) to get out of sync. This happens if you fill the buffer and then remove a few pages from it while it is full.

The fixes are pretty simple, and should go into the 3.5-rc.

Please pull the latest tip/perf/urgent tree, which can be found at:

git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace.git
tip/perf/urgent

Head SHA1: 48fdc72f23ad9a9956e524a47843135d0bbc3317

Vaibhav Nagarnaik (2):
ring-buffer: Fix crash due to uninitialized new_pages list head
ring-buffer: Fix accounting of entries when removing pages

----

 kernel/trace/ring_buffer.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

From: Vaibhav Nagarnaik <[email protected]>

The new_pages list head in the cpu_buffer is not initialized. When adding pages to the ring buffer, if the memory allocation fails in ring_buffer_resize, the clean up handler tries to free up the allocated pages from all the cpu buffers. The panic is caused by referencing the uninitialized new_pages list head.

Initializing the new_pages list head in rb_allocate_cpu_buffer fixes this.

---
 kernel/trace/ring_buffer.c |    1 +
 1 file changed, 1 insertion(+)

index 1d0f6a8..ba39cba 100644
@@ -1075,6 +1075,7 @@ rb_allocate_cpu_buffer(struct ring_buffer *buffer, int nr_pages, int cpu)
 	rb_init_page(bpage->page);
 
 	INIT_LIST_HEAD(&cpu_buffer->reader_page->list);
+	INIT_LIST_HEAD(&cpu_buffer->new_pages);
 
 	ret = rb_allocate_pages(cpu_buffer, nr_pages);
 	if (ret < 0)

1.7.10

From: Vaibhav Nagarnaik <[email protected]>

When removing pages from the ring buffer, its state is not reset. This means that the counters need to be correctly updated to account for the pages removed.

Update the overrun counter to reflect the removed events from the pages.

---
 kernel/trace/ring_buffer.c |    5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

index ba39cba..f765465 100644
@@ -1347,10 +1347,9 @@ rb_remove_pages(struct ring_buffer_per_cpu *cpu_buffer, unsigned int nr_pages)
 			 * If something was added to this page, it was full
 			 * since it is not the tail page. So we deduct the
 			 * bytes consumed in ring buffer from here.
-			 * No need to update overruns, since this page is
-			 * deleted from ring buffer and its entries are
-			 * already accounted for.
+			 * Increment overrun to account for the lost events.
 			 */
+			local_add(page_entries, &cpu_buffer->overrun);
 			local_sub(BUF_PAGE_SIZE, &cpu_buffer->entries_bytes);
 		}
 

1.7.10

Pulled, thanks!

Ingo