[PATCH] dma-buf/fence-array: get signaled state when signaling is disabled
Christian König
christian.koenig at amd.com
Wed Sep 21 18:47:36 UTC 2016
Am 21.09.2016 um 13:36 schrieb Gustavo Padovan:
> From: Gustavo Padovan <gustavo.padovan at collabora.co.uk>
>
> If the fences in the fence_array signal on the fence_array does not have
> signalling enabled num_pending will not be updated accordingly.
>
> So when signaling is disabled check the signal of every fence with
> fence_is_signaled() and then compare with num_pending to learn if the
> fence_array was signalled or not.
>
> If we want to keep the poll_does_not_wait optimization I think we need
> something like this. It keeps the same behaviour if signalling is enabled
> but tries to calculated the state otherwise.
>
> Signed-off-by: Gustavo Padovan <gustavo.padovan at collabora.co.uk>
> Reviewed-by: Chris Wilson <chris at chris-wilson.co.uk>
First of all the patch is horrible wrong because fence_array_signaled()
is called without any locks held. So you can run into a race condition
between checking the fences here and enable signaling.
Additional to that I'm not sure if that is such a good idea or not,
cause fence_array_signaled() should be light weight and without calling
enable_signaling there is not guarantee that fences will ever signal.
Regards,
Christian.
> ---
> drivers/dma-buf/fence-array.c | 19 ++++++++++++++++++-
> 1 file changed, 18 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/dma-buf/fence-array.c b/drivers/dma-buf/fence-array.c
> index f1989fc..1eec271 100644
> --- a/drivers/dma-buf/fence-array.c
> +++ b/drivers/dma-buf/fence-array.c
> @@ -75,8 +75,25 @@ static bool fence_array_enable_signaling(struct fence *fence)
> static bool fence_array_signaled(struct fence *fence)
> {
> struct fence_array *array = to_fence_array(fence);
> + int i, num_pending;
> +
> + num_pending = atomic_read(&array->num_pending);
> +
> + /*
> + * Before signaling is enabled, num_pending is static (set during array
> + * construction as a count of all fences or set to 1 if signal_on_any
> + * flag is passed. To ensure forward progress, i.e. a while
> + * (!fence_is_signaled()) ; busy-loop eventually proceeds, we need to
> + * check the current status of our fences.
> + */
> + if (!test_bit(FENCE_FLAG_ENABLE_SIGNAL_BIT, &fence->flags)) {
> + for (i = 0 ; i < array->num_fences; ++i) {
> + if (fence_is_signaled(array->fences[i]))
> + num_pending--;
> + }
> + }
>
> - return atomic_read(&array->num_pending) <= 0;
> + return num_pending <= 0;
> }
>
> static void fence_array_release(struct fence *fence)
More information about the dri-devel
mailing list