[PATCH AUTOSEL 5.4 21/34] drm/amdgpu/gfx10: fix race condition for kiq

From: Sasha Levin
Date: Mon Jul 20 2020 - 17:39:04 EST


From: Jack Xiao <Jack.Xiao@xxxxxxx>

[ Upstream commit 7d65a577bb58d4f27a3398a4c0cb0b00ab7d0511 ]

During preemption test for gfx10, it uses kiq to trigger
gfx preemption, which would result in race condition
with flushing TLB for kiq.

Signed-off-by: Jack Xiao <Jack.Xiao@xxxxxxx>
Reviewed-by: Hawking Zhang <Hawking.Zhang@xxxxxxx>
Acked-by: Christian KÃnig <christian.koenig@xxxxxxx>
Signed-off-by: Alex Deucher <alexander.deucher@xxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
---
drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c | 9 ++++++++-
1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index 6f118292e40fb..64d96eb0a2337 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
@@ -4683,12 +4683,17 @@ static int gfx_v10_0_ring_preempt_ib(struct amdgpu_ring *ring)
struct amdgpu_device *adev = ring->adev;
struct amdgpu_kiq *kiq = &adev->gfx.kiq;
struct amdgpu_ring *kiq_ring = &kiq->ring;
+ unsigned long flags;

if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
return -EINVAL;

- if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size))
+ spin_lock_irqsave(&kiq->ring_lock, flags);
+
+ if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) {
+ spin_unlock_irqrestore(&kiq->ring_lock, flags);
return -ENOMEM;
+ }

/* assert preemption condition */
amdgpu_ring_set_preempt_cond_exec(ring, false);
@@ -4699,6 +4704,8 @@ static int gfx_v10_0_ring_preempt_ib(struct amdgpu_ring *ring)
++ring->trail_seq);
amdgpu_ring_commit(kiq_ring);

+ spin_unlock_irqrestore(&kiq->ring_lock, flags);
+
/* poll the trailing fence */
for (i = 0; i < adev->usec_timeout; i++) {
if (ring->trail_seq ==
--
2.25.1