[PATCH v2] dma-buf: Discard old fence_excl on retrying get_fences_rcu for realloc

[Chris Wilson <chris@chris-wilson.co.uk>]

If we have to drop the seqcount & rcu lock to perform a krealloc, we
have to restart the loop. In doing so, be careful not to lose track of
the already acquired exclusive fence.

Fixes: fedf54132d24 ("dma-buf: Restart reservation_object_get_fences_rcu() after writes") #v4.10
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: stable@vger.kernel.org
---
 drivers/dma-buf/reservation.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/dma-buf/reservation.c b/drivers/dma-buf/reservation.c
index 4d32e2c67862..704503df4892 100644
--- a/drivers/dma-buf/reservation.c
+++ b/drivers/dma-buf/reservation.c
@@ -365,6 +365,12 @@ int reservation_object_get_fences_rcu(struct reservation_object *obj,
 					   GFP_NOWAIT | __GFP_NOWARN);
 			if (!nshared) {
 				rcu_read_unlock();
+
+				if (fence_excl) {
+					dma_fence_put(fence_excl);
+					fence_excl = NULL;
+				}
+
 				nshared = krealloc(shared, sz, GFP_KERNEL);
 				if (nshared) {
 					shared = nshared;
-- 
2.20.1