From 5c55db83b759b079ead93aa0767fc97048b1f899 Mon Sep 17 00:00:00 2001 From: Chunming Zhou Date: Wed, 2 Mar 2016 11:30:31 +0800 Subject: drm/amdgpu: wait engine idle before vm flush for sdma MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Chunming Zhou Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c') diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index dddb8d6a81f3..0e246ff6734c 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -885,6 +885,21 @@ static void sdma_v2_4_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib static void sdma_v2_4_ring_emit_vm_flush(struct amdgpu_ring *ring, unsigned vm_id, uint64_t pd_addr) { + uint32_t seq = ring->fence_drv.sync_seq; + uint64_t addr = ring->fence_drv.gpu_addr; + + /* wait for idle */ + amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) | + SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(0) | + SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3) | /* equal */ + SDMA_PKT_POLL_REGMEM_HEADER_MEM_POLL(1)); + amdgpu_ring_write(ring, addr & 0xfffffffc); + amdgpu_ring_write(ring, upper_32_bits(addr) & 0xffffffff); + amdgpu_ring_write(ring, seq); /* reference */ + amdgpu_ring_write(ring, 0xfffffff); /* mask */ + amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) | + SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(4)); /* retry count, poll interval */ + amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) | SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf)); if (vm_id < 8) { -- cgit v1.2.3 From 6ad550c35d8c5327a548bc59bba406ef2db51f1b Mon Sep 17 00:00:00 2001 From: Chunming Zhou Date: Thu, 3 Mar 2016 12:06:34 +0800 Subject: drm/amdgpu: add hdp invalidation for sdma v2_4 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Chunming Zhou Reviewed-by: Alex Deucher Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c') diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index 0e246ff6734c..134d5d0bb77f 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -300,6 +300,13 @@ static void sdma_v2_4_ring_emit_hdp_flush(struct amdgpu_ring *ring) SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); /* retry count, poll interval */ } +static void sdma_v2_4_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) +{ + amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) | + SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf)); + amdgpu_ring_write(ring, mmHDP_DEBUG0); + amdgpu_ring_write(ring, 1); +} /** * sdma_v2_4_ring_emit_fence - emit a fence on the DMA ring * @@ -1291,6 +1298,7 @@ static const struct amdgpu_ring_funcs sdma_v2_4_ring_funcs = { .emit_fence = sdma_v2_4_ring_emit_fence, .emit_vm_flush = sdma_v2_4_ring_emit_vm_flush, .emit_hdp_flush = sdma_v2_4_ring_emit_hdp_flush, + .emit_hdp_invalidate = sdma_v2_4_ring_emit_hdp_invalidate, .test_ring = sdma_v2_4_ring_test_ring, .test_ib = sdma_v2_4_ring_test_ib, .insert_nop = sdma_v2_4_ring_insert_nop, -- cgit v1.2.3 From 336d1f5efe93db3d997a6d105760dd613d7ecdce Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 16 Feb 2016 10:57:10 +0100 Subject: drm/amdgpu: remove HW fence owner MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Not used any more since we now always use the sheduler. Signed-off-by: Christian König Reviewed-by: Alex Deucher Reviewed-by: Chunming Zhou --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 9 ++------- drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c | 6 +----- drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 6 ++---- drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c | 8 +------- drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 3 +-- drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 7 ++----- drivers/gpu/drm/amd/amdgpu/cik_sdma.c | 3 +-- drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 3 +-- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 6 ++---- drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 3 +-- drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 3 +-- 12 files changed, 16 insertions(+), 43 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 11a9f5899995..fab6ddb26b5b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -409,9 +409,6 @@ struct amdgpu_fence { struct amdgpu_ring *ring; uint64_t seq; - /* filp or special value for fence creator */ - void *owner; - wait_queue_t fence_wake; }; @@ -432,8 +429,7 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring, unsigned irq_type); void amdgpu_fence_driver_suspend(struct amdgpu_device *adev); void amdgpu_fence_driver_resume(struct amdgpu_device *adev); -int amdgpu_fence_emit(struct amdgpu_ring *ring, void *owner, - struct amdgpu_fence **fence); +int amdgpu_fence_emit(struct amdgpu_ring *ring, struct amdgpu_fence **fence); void amdgpu_fence_process(struct amdgpu_ring *ring); int amdgpu_fence_wait_next(struct amdgpu_ring *ring); int amdgpu_fence_wait_empty(struct amdgpu_ring *ring); @@ -1177,8 +1173,7 @@ int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned size, struct amdgpu_ib *ib); void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib); int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, - struct amdgpu_ib *ib, void *owner, - struct fence *last_vm_update, + struct amdgpu_ib *ib, struct fence *last_vm_update, struct fence **f); int amdgpu_ib_pool_init(struct amdgpu_device *adev); void amdgpu_ib_pool_fini(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index 97db196dc6f8..d94b13ac290f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -91,25 +91,21 @@ static u32 amdgpu_fence_read(struct amdgpu_ring *ring) * amdgpu_fence_emit - emit a fence on the requested ring * * @ring: ring the fence is associated with - * @owner: creator of the fence * @fence: amdgpu fence object * * Emits a fence command on the requested ring (all asics). * Returns 0 on success, -ENOMEM on failure. */ -int amdgpu_fence_emit(struct amdgpu_ring *ring, void *owner, - struct amdgpu_fence **fence) +int amdgpu_fence_emit(struct amdgpu_ring *ring, struct amdgpu_fence **fence) { struct amdgpu_device *adev = ring->adev; - /* we are protected by the ring emission mutex */ *fence = kmem_cache_alloc(amdgpu_fence_slab, GFP_KERNEL); if ((*fence) == NULL) { return -ENOMEM; } (*fence)->seq = ++ring->fence_drv.sync_seq; (*fence)->ring = ring; - (*fence)->owner = owner; fence_init(&(*fence)->base, &amdgpu_fence_ops, &ring->fence_drv.fence_queue.lock, adev->fence_context + ring->idx, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index 4e978e7aa1b8..9550247b030d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -101,7 +101,6 @@ void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib) * @adev: amdgpu_device pointer * @num_ibs: number of IBs to schedule * @ibs: IB objects to schedule - * @owner: owner for creating the fences * @f: fence created during this submission * * Schedule an IB on the associated ring (all asics). @@ -118,8 +117,7 @@ void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib) * to SI there was just a DE IB. */ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, - struct amdgpu_ib *ibs, void *owner, - struct fence *last_vm_update, + struct amdgpu_ib *ibs, struct fence *last_vm_update, struct fence **f) { struct amdgpu_device *adev = ring->adev; @@ -183,7 +181,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, amdgpu_ring_emit_hdp_invalidate(ring); } - r = amdgpu_fence_emit(ring, owner, &ib->fence); + r = amdgpu_fence_emit(ring, &ib->fence); if (r) { dev_err(adev->dev, "failed to emit fence (%d)\n", r); ring->current_ctx = old_ctx; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index 90e52f7e17a0..692b45560d0a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -148,7 +148,7 @@ static struct fence *amdgpu_job_run(struct amd_sched_job *sched_job) } trace_amdgpu_sched_run_job(job); - r = amdgpu_ib_schedule(job->ring, job->num_ibs, job->ibs, job->owner, + r = amdgpu_ib_schedule(job->ring, job->num_ibs, job->ibs, job->sync.last_vm_update, &fence); if (r) { DRM_ERROR("Error scheduling IBs (%d)\n", r); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c index 87690cc57206..e3673422aac8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c @@ -60,12 +60,8 @@ void amdgpu_sync_create(struct amdgpu_sync *sync) */ static bool amdgpu_sync_same_dev(struct amdgpu_device *adev, struct fence *f) { - struct amdgpu_fence *a_fence = to_amdgpu_fence(f); struct amd_sched_fence *s_fence = to_amd_sched_fence(f); - if (a_fence) - return a_fence->ring->adev == adev; - if (s_fence) { struct amdgpu_ring *ring; @@ -85,13 +81,11 @@ static bool amdgpu_sync_same_dev(struct amdgpu_device *adev, struct fence *f) */ static void *amdgpu_sync_get_owner(struct fence *f) { - struct amdgpu_fence *a_fence = to_amdgpu_fence(f); struct amd_sched_fence *s_fence = to_amd_sched_fence(f); if (s_fence) return s_fence->owner; - else if (a_fence) - return a_fence->owner; + return AMDGPU_FENCE_OWNER_UNDEFINED; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index 1de82bf4fc79..e5f0a5e29551 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -886,8 +886,7 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, ib->length_dw = 16; if (direct) { - r = amdgpu_ib_schedule(ring, 1, ib, - AMDGPU_FENCE_OWNER_UNDEFINED, NULL, &f); + r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); if (r) goto err_free; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index 39c3aa60381a..6d191fb40b38 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -425,8 +425,7 @@ int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; - r = amdgpu_ib_schedule(ring, 1, ib, AMDGPU_FENCE_OWNER_UNDEFINED, - NULL, &f); + r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); if (r) goto err; @@ -487,9 +486,7 @@ int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, ib->ptr[i] = 0x0; if (direct) { - r = amdgpu_ib_schedule(ring, 1, ib, - AMDGPU_FENCE_OWNER_UNDEFINED, - NULL, &f); + r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); if (r) goto err; diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index c4101ebd2b3f..b5b4220a6141 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -643,8 +643,7 @@ static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring) ib.ptr[3] = 1; ib.ptr[4] = 0xDEADBEEF; ib.length_dw = 5; - r = amdgpu_ib_schedule(ring, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED, - NULL, &f); + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); if (r) goto err1; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index 1b713009c5d2..46c2436d74bd 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -2136,8 +2136,7 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring) ib.ptr[2] = 0xDEADBEEF; ib.length_dw = 3; - r = amdgpu_ib_schedule(ring, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED, - NULL, &f); + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); if (r) goto err2; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index aeb35f504999..e0b64de9b5af 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -706,8 +706,7 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring) ib.ptr[2] = 0xDEADBEEF; ib.length_dw = 3; - r = amdgpu_ib_schedule(ring, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED, - NULL, &f); + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); if (r) goto err2; @@ -1262,8 +1261,7 @@ static int gfx_v8_0_do_edc_gpr_workarounds(struct amdgpu_device *adev) ib.ptr[ib.length_dw++] = EVENT_TYPE(7) | EVENT_INDEX(4); /* shedule the ib on the ring */ - r = amdgpu_ib_schedule(ring, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED, - NULL, &f); + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); if (r) { DRM_ERROR("amdgpu: ib submit failed (%d).\n", r); goto fail; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index 134d5d0bb77f..747ef558033a 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -701,8 +701,7 @@ static int sdma_v2_4_ring_test_ib(struct amdgpu_ring *ring) ib.ptr[7] = SDMA_PKT_HEADER_OP(SDMA_OP_NOP); ib.length_dw = 8; - r = amdgpu_ib_schedule(ring, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED, - NULL, &f); + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); if (r) goto err1; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index ca12d6faa8c2..11a544fbf197 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -853,8 +853,7 @@ static int sdma_v3_0_ring_test_ib(struct amdgpu_ring *ring) ib.ptr[7] = SDMA_PKT_NOP_HEADER_OP(SDMA_OP_NOP); ib.length_dw = 8; - r = amdgpu_ib_schedule(ring, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED, - NULL, &f); + r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); if (r) goto err1; -- cgit v1.2.3 From 00b7c4ff7d482d287a591f047e0963d494569b46 Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 8 Mar 2016 14:11:00 +0100 Subject: drm/amdgpu: split pipeline sync out of SDMA vm_flush() as well MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Code it similar to how we did it for the gfx and compute engines. Signed-off-by: Christian König Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/cik_sdma.c | 27 +++++++++++++++++++-------- drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 23 +++++++++++++++++------ drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 23 +++++++++++++++++------ 3 files changed, 53 insertions(+), 20 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c') diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index b5b4220a6141..8663a27018a4 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -822,19 +822,14 @@ static void cik_sdma_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib) } /** - * cik_sdma_ring_emit_vm_flush - cik vm flush using sDMA + * cik_sdma_ring_emit_pipeline_sync - sync the pipeline * * @ring: amdgpu_ring pointer - * @vm: amdgpu_vm pointer * - * Update the page table base and flush the VM TLB - * using sDMA (CIK). + * Make sure all previous operations are completed (CIK). */ -static void cik_sdma_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vm_id, uint64_t pd_addr) +static void cik_sdma_ring_emit_pipeline_sync(struct amdgpu_ring *ring) { - u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(0) | - SDMA_POLL_REG_MEM_EXTRA_FUNC(0)); /* always */ uint32_t seq = ring->fence_drv.sync_seq; uint64_t addr = ring->fence_drv.gpu_addr; @@ -848,7 +843,22 @@ static void cik_sdma_ring_emit_vm_flush(struct amdgpu_ring *ring, amdgpu_ring_write(ring, seq); /* reference */ amdgpu_ring_write(ring, 0xfffffff); /* mask */ amdgpu_ring_write(ring, (0xfff << 16) | 4); /* retry count, poll interval */ +} +/** + * cik_sdma_ring_emit_vm_flush - cik vm flush using sDMA + * + * @ring: amdgpu_ring pointer + * @vm: amdgpu_vm pointer + * + * Update the page table base and flush the VM TLB + * using sDMA (CIK). + */ +static void cik_sdma_ring_emit_vm_flush(struct amdgpu_ring *ring, + unsigned vm_id, uint64_t pd_addr) +{ + u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(0) | + SDMA_POLL_REG_MEM_EXTRA_FUNC(0)); /* always */ amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); if (vm_id < 8) { @@ -1290,6 +1300,7 @@ static const struct amdgpu_ring_funcs cik_sdma_ring_funcs = { .parse_cs = NULL, .emit_ib = cik_sdma_ring_emit_ib, .emit_fence = cik_sdma_ring_emit_fence, + .emit_pipeline_sync = cik_sdma_ring_emit_pipeline_sync, .emit_vm_flush = cik_sdma_ring_emit_vm_flush, .emit_hdp_flush = cik_sdma_ring_emit_hdp_flush, .emit_hdp_invalidate = cik_sdma_ring_emit_hdp_invalidate, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index 747ef558033a..ab9ff89a3096 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -880,16 +880,13 @@ static void sdma_v2_4_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib } /** - * sdma_v2_4_ring_emit_vm_flush - cik vm flush using sDMA + * sdma_v2_4_ring_emit_pipeline_sync - sync the pipeline * * @ring: amdgpu_ring pointer - * @vm: amdgpu_vm pointer * - * Update the page table base and flush the VM TLB - * using sDMA (VI). + * Make sure all previous operations are completed (CIK). */ -static void sdma_v2_4_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vm_id, uint64_t pd_addr) +static void sdma_v2_4_ring_emit_pipeline_sync(struct amdgpu_ring *ring) { uint32_t seq = ring->fence_drv.sync_seq; uint64_t addr = ring->fence_drv.gpu_addr; @@ -905,7 +902,20 @@ static void sdma_v2_4_ring_emit_vm_flush(struct amdgpu_ring *ring, amdgpu_ring_write(ring, 0xfffffff); /* mask */ amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) | SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(4)); /* retry count, poll interval */ +} +/** + * sdma_v2_4_ring_emit_vm_flush - cik vm flush using sDMA + * + * @ring: amdgpu_ring pointer + * @vm: amdgpu_vm pointer + * + * Update the page table base and flush the VM TLB + * using sDMA (VI). + */ +static void sdma_v2_4_ring_emit_vm_flush(struct amdgpu_ring *ring, + unsigned vm_id, uint64_t pd_addr) +{ amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) | SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf)); if (vm_id < 8) { @@ -1295,6 +1305,7 @@ static const struct amdgpu_ring_funcs sdma_v2_4_ring_funcs = { .parse_cs = NULL, .emit_ib = sdma_v2_4_ring_emit_ib, .emit_fence = sdma_v2_4_ring_emit_fence, + .emit_pipeline_sync = sdma_v2_4_ring_emit_pipeline_sync, .emit_vm_flush = sdma_v2_4_ring_emit_vm_flush, .emit_hdp_flush = sdma_v2_4_ring_emit_hdp_flush, .emit_hdp_invalidate = sdma_v2_4_ring_emit_hdp_invalidate, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 11a544fbf197..4c24c371fec7 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -1031,16 +1031,13 @@ static void sdma_v3_0_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib } /** - * sdma_v3_0_ring_emit_vm_flush - cik vm flush using sDMA + * sdma_v3_0_ring_emit_pipeline_sync - sync the pipeline * * @ring: amdgpu_ring pointer - * @vm: amdgpu_vm pointer * - * Update the page table base and flush the VM TLB - * using sDMA (VI). + * Make sure all previous operations are completed (CIK). */ -static void sdma_v3_0_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vm_id, uint64_t pd_addr) +static void sdma_v3_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) { uint32_t seq = ring->fence_drv.sync_seq; uint64_t addr = ring->fence_drv.gpu_addr; @@ -1056,7 +1053,20 @@ static void sdma_v3_0_ring_emit_vm_flush(struct amdgpu_ring *ring, amdgpu_ring_write(ring, 0xfffffff); /* mask */ amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) | SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(4)); /* retry count, poll interval */ +} +/** + * sdma_v3_0_ring_emit_vm_flush - cik vm flush using sDMA + * + * @ring: amdgpu_ring pointer + * @vm: amdgpu_vm pointer + * + * Update the page table base and flush the VM TLB + * using sDMA (VI). + */ +static void sdma_v3_0_ring_emit_vm_flush(struct amdgpu_ring *ring, + unsigned vm_id, uint64_t pd_addr) +{ amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) | SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf)); if (vm_id < 8) { @@ -1563,6 +1573,7 @@ static const struct amdgpu_ring_funcs sdma_v3_0_ring_funcs = { .parse_cs = NULL, .emit_ib = sdma_v3_0_ring_emit_ib, .emit_fence = sdma_v3_0_ring_emit_fence, + .emit_pipeline_sync = sdma_v3_0_ring_emit_pipeline_sync, .emit_vm_flush = sdma_v3_0_ring_emit_vm_flush, .emit_hdp_flush = sdma_v3_0_ring_emit_hdp_flush, .emit_hdp_invalidate = sdma_v3_0_ring_emit_hdp_invalidate, -- cgit v1.2.3