Re: [PATCH v2] perf/ring_buffer: Fix exposing a temporarily decreased data_head.

From: Peter Zijlstra
Date: Fri May 17 2019 - 04:49:23 EST


On Thu, May 16, 2019 at 11:40:10AM -0700, Yabin Cui wrote:
> In perf_output_put_handle(), an IRQ/NMI can happen in below location and
> write records to the same ring buffer:
> ...
> local_dec_and_test(&rb->nest)
> ... <-- an IRQ/NMI can happen here
> rb->user_page->data_head = head;
> ...
>
> In this case, a value A is written to data_head in the IRQ, then a value
> B is written to data_head after the IRQ. And A > B. As a result,
> data_head is temporarily decreased from A to B. And a reader may see
> data_head < data_tail if it read the buffer frequently enough, which
> creates unexpected behaviors.
>
> This can be fixed by moving dec(&rb->nest) to after updating data_head,
> which prevents the IRQ/NMI above from updating data_head.
>
> Signed-off-by: Yabin Cui <yabinc@xxxxxxxxxx>
> ---
>
> v1 -> v2: change rb->nest from local_t to unsigned int, and add barriers.
>
> ---
> kernel/events/internal.h | 2 +-
> kernel/events/ring_buffer.c | 24 ++++++++++++++++++------
> 2 files changed, 19 insertions(+), 7 deletions(-)
>
> diff --git a/kernel/events/internal.h b/kernel/events/internal.h
> index 79c47076700a..0a8c003b9bcf 100644
> --- a/kernel/events/internal.h
> +++ b/kernel/events/internal.h
> @@ -24,7 +24,7 @@ struct ring_buffer {
> atomic_t poll; /* POLL_ for wakeups */
>
> local_t head; /* write position */
> - local_t nest; /* nested writers */
> + unsigned int nest; /* nested writers */
> local_t events; /* event limit */
> local_t wakeup; /* wakeup stamp */
> local_t lost; /* nr records lost */
> diff --git a/kernel/events/ring_buffer.c b/kernel/events/ring_buffer.c
> index 674b35383491..c677beb01fb1 100644
> --- a/kernel/events/ring_buffer.c
> +++ b/kernel/events/ring_buffer.c
> @@ -38,7 +38,8 @@ static void perf_output_get_handle(struct perf_output_handle *handle)
> struct ring_buffer *rb = handle->rb;
>
> preempt_disable();
> - local_inc(&rb->nest);
> + rb->nest++;
> + barrier();

Urgh; almost but not quite. You just lost the 'volatile' qualifier and
now the compiler can mess things up for you.

> handle->wakeup = local_read(&rb->wakeup);
> }

What I'm going to do is split this into two patches, one fixes the
problem and marked for backport, and one changing away from local_t.