diff options
author | Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com> | 2019-11-28 12:08:58 +0100 |
---|---|---|
committer | Alex Deucher <alexander.deucher@amd.com> | 2019-12-05 16:24:58 -0500 |
commit | b456c93253bde374057bfca8f74ada527250d5d9 (patch) | |
tree | 302f4ace2549533b8535891d450e8c659014290a /drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | |
parent | 80c5a807a51f2d19df4a5a5d12cf9d4997993999 (diff) | |
download | linux-stable-b456c93253bde374057bfca8f74ada527250d5d9.tar.gz linux-stable-b456c93253bde374057bfca8f74ada527250d5d9.tar.bz2 linux-stable-b456c93253bde374057bfca8f74ada527250d5d9.zip |
drm/amdgpu: add cache flush workaround to gfx8 emit_fence
The same workaround is used for gfx7.
Both PAL and Mesa use it for gfx8 too, so port this commit to
gfx_v8_0_ring_emit_fence_gfx.
Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c')
-rw-r--r-- | drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 22 |
1 files changed, 19 insertions, 3 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 9b0b22522e83..cfc1403fc855 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -6145,7 +6145,23 @@ static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr, bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT; bool int_sel = flags & AMDGPU_FENCE_FLAG_INT; - /* EVENT_WRITE_EOP - flush caches, send int */ + /* Workaround for cache flush problems. First send a dummy EOP + * event down the pipe with seq one below. + */ + amdgpu_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4)); + amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN | + EOP_TC_ACTION_EN | + EOP_TC_WB_ACTION_EN | + EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) | + EVENT_INDEX(5))); + amdgpu_ring_write(ring, addr & 0xfffffffc); + amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xffff) | + DATA_SEL(1) | INT_SEL(0)); + amdgpu_ring_write(ring, lower_32_bits(seq - 1)); + amdgpu_ring_write(ring, upper_32_bits(seq - 1)); + + /* Then send the real EOP event down the pipe: + * EVENT_WRITE_EOP - flush caches, send int */ amdgpu_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4)); amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN | EOP_TC_ACTION_EN | @@ -6887,7 +6903,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = { 5 + /* COND_EXEC */ 7 + /* PIPELINE_SYNC */ VI_FLUSH_GPU_TLB_NUM_WREG * 5 + 9 + /* VM_FLUSH */ - 8 + /* FENCE for VM_FLUSH */ + 12 + /* FENCE for VM_FLUSH */ 20 + /* GDS switch */ 4 + /* double SWITCH_BUFFER, the first COND_EXEC jump to the place just @@ -6899,7 +6915,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = { 31 + /* DE_META */ 3 + /* CNTX_CTRL */ 5 + /* HDP_INVL */ - 8 + 8 + /* FENCE x2 */ + 12 + 12 + /* FENCE x2 */ 2, /* SWITCH_BUFFER */ .emit_ib_size = 4, /* gfx_v8_0_ring_emit_ib_gfx */ .emit_ib = gfx_v8_0_ring_emit_ib_gfx, |