On Mon, Aug 29, 2016 at 08:08:31AM +0100, Chris Wilson wrote:
In order to be completely generic, we have to double check the read seqlock after acquiring a reference to the fence. If the driver is allocating fences from a SLAB_DESTROY_BY_RCU, or similar freelist, then within an RCU grace period a fence may be freed and reallocated. The RCU read side critical section does not prevent this reallocation, instead we have to inspect the reservation's seqlock to double check if the fences have been reassigned as we were acquiring our reference.
Signed-off-by: Chris Wilson chris@chris-wilson.co.uk Cc: Daniel Vetter daniel.vetter@ffwll.ch Cc: Maarten Lankhorst maarten.lankhorst@linux.intel.com Cc: Christian König christian.koenig@amd.com Cc: Alex Deucher alexander.deucher@amd.com Cc: Sumit Semwal sumit.semwal@linaro.org Cc: linux-media@vger.kernel.org Cc: dri-devel@lists.freedesktop.org Cc: linaro-mm-sig@lists.linaro.org
Reviewed-by: Daniel Vetter daniel.vetter@ffwll.ch
drivers/dma-buf/reservation.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-)
diff --git a/drivers/dma-buf/reservation.c b/drivers/dma-buf/reservation.c index 10fd441dd4ed..3369e4668e96 100644 --- a/drivers/dma-buf/reservation.c +++ b/drivers/dma-buf/reservation.c @@ -388,9 +388,6 @@ retry: if (fobj) shared_count = fobj->shared_count;
if (read_seqcount_retry(&obj->seq, seq))goto unlock_retry;- for (i = 0; i < shared_count; ++i) { struct fence *lfence = rcu_dereference(fobj->shared[i]);
@@ -413,9 +410,6 @@ retry: if (!shared_count) { struct fence *fence_excl = rcu_dereference(obj->fence_excl);
if (read_seqcount_retry(&obj->seq, seq))goto unlock_retry;- if (fence_excl && !test_bit(FENCE_FLAG_SIGNALED_BIT, &fence_excl->flags)) { if (!fence_get_rcu(fence_excl))
@@ -430,6 +424,11 @@ retry: rcu_read_unlock(); if (fence) {
if (read_seqcount_retry(&obj->seq, seq)) {fence_put(fence);goto retry;}- ret = fence_wait_timeout(fence, intr, ret); fence_put(fence); if (ret > 0 && wait_all && (i + 1 < shared_count))
-- 2.9.3