From 5a4633c4b880cf8d1fe7df9c55766205cf9bc295 Mon Sep 17 00:00:00 2001 From: Christian König Date: Mon, 8 Jan 2018 14:48:11 +0100 Subject: drm/amdgpu: forward pasid to backend flush implementations MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit rd the pasid from the VM code to the emit_vm_flush function and update all implementations with the new parameter. Signed-off-by: Christian König Reviewed-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index 102dad3edf6a..12b9a06f4d21 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -126,7 +126,7 @@ struct amdgpu_ring_funcs { uint64_t seq, unsigned flags); void (*emit_pipeline_sync)(struct amdgpu_ring *ring); void (*emit_vm_flush)(struct amdgpu_ring *ring, unsigned vmid, - uint64_t pd_addr); + unsigned pasid, uint64_t pd_addr); void (*emit_hdp_flush)(struct amdgpu_ring *ring); void (*emit_hdp_invalidate)(struct amdgpu_ring *ring); void (*emit_gds_switch)(struct amdgpu_ring *ring, uint32_t vmid, -- cgit v1.2.3 From 2ee150cda7bdc766cf9baca3534f3a2c0b0e8357 Mon Sep 17 00:00:00 2001 From: Christian König Date: Fri, 19 Jan 2018 15:19:16 +0100 Subject: drm/amdgpu: remove now superflous *_hdp operation MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit All HDP invalidation and most flush can now be replaced by the generic ASIC function. Signed-off-by: Christian König Acked-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 - drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c | 7 +----- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 1 - drivers/gpu/drm/amd/amdgpu/cik_sdma.c | 10 +------- drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c | 39 ++------------------------------ drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 26 ++------------------- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 19 ++-------------- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 14 ++---------- drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 10 +------- drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 11 +-------- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 13 +---------- drivers/gpu/drm/amd/amdgpu/si_dma.c | 19 +--------------- drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c | 30 ------------------------ drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c | 30 ------------------------ drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c | 35 +--------------------------- drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c | 37 ++---------------------------- drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 18 +-------------- 17 files changed, 18 insertions(+), 302 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 8367f92de09a..b7c60bb4457d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1797,7 +1797,6 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_ring_emit_fence(r, addr, seq, flags) (r)->funcs->emit_fence((r), (addr), (seq), (flags)) #define amdgpu_ring_emit_gds_switch(r, v, db, ds, wb, ws, ab, as) (r)->funcs->emit_gds_switch((r), (v), (db), (ds), (wb), (ws), (ab), (as)) #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r)) -#define amdgpu_ring_emit_hdp_invalidate(r) (r)->funcs->emit_hdp_invalidate((r)) #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r)) #define amdgpu_ring_emit_cntxcntl(r, d) (r)->funcs->emit_cntxcntl((r), (d)) #define amdgpu_ring_emit_rreg(r, d) (r)->funcs->emit_rreg((r), (d)) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index e87c9952c901..8ea342dc6376 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -225,12 +225,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, #ifdef CONFIG_X86_64 if (!(adev->flags & AMD_IS_APU)) #endif - { - if (ring->funcs->emit_hdp_invalidate) - amdgpu_ring_emit_hdp_invalidate(ring); - else - amdgpu_asic_invalidate_hdp(adev, ring); - } + amdgpu_asic_invalidate_hdp(adev, ring); r = amdgpu_fence_emit(ring, f); if (r) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index 12b9a06f4d21..70d05ec7bc07 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -128,7 +128,6 @@ struct amdgpu_ring_funcs { void (*emit_vm_flush)(struct amdgpu_ring *ring, unsigned vmid, unsigned pasid, uint64_t pd_addr); void (*emit_hdp_flush)(struct amdgpu_ring *ring); - void (*emit_hdp_invalidate)(struct amdgpu_ring *ring); void (*emit_gds_switch)(struct amdgpu_ring *ring, uint32_t vmid, uint32_t gds_base, uint32_t gds_size, uint32_t gws_base, uint32_t gws_size, diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index 1d32dedb2534..5d18512cd090 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -261,13 +261,6 @@ static void cik_sdma_ring_emit_hdp_flush(struct amdgpu_ring *ring) amdgpu_ring_write(ring, (0xfff << 16) | 10); /* retry count, poll interval */ } -static void cik_sdma_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); - amdgpu_ring_write(ring, mmHDP_DEBUG0); - amdgpu_ring_write(ring, 1); -} - /** * cik_sdma_ring_emit_fence - emit a fence on the DMA ring * @@ -1277,7 +1270,7 @@ static const struct amdgpu_ring_funcs cik_sdma_ring_funcs = { .set_wptr = cik_sdma_ring_set_wptr, .emit_frame_size = 6 + /* cik_sdma_ring_emit_hdp_flush */ - 3 + /* cik_sdma_ring_emit_hdp_invalidate */ + 3 + /* hdp invalidate */ 6 + /* cik_sdma_ring_emit_pipeline_sync */ CIK_FLUSH_GPU_TLB_NUM_WREG * 3 + 6 + /* cik_sdma_ring_emit_vm_flush */ 9 + 9 + 9, /* cik_sdma_ring_emit_fence x3 for user fence, vm fence */ @@ -1287,7 +1280,6 @@ static const struct amdgpu_ring_funcs cik_sdma_ring_funcs = { .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, .test_ring = cik_sdma_ring_test_ring, .test_ib = cik_sdma_ring_test_ib, .insert_nop = cik_sdma_ring_insert_nop, diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c index e6c3a2465ba4..3517fd9e11c9 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c @@ -1809,17 +1809,6 @@ static int gfx_v6_0_ring_test_ring(struct amdgpu_ring *ring) return r; } -static void gfx_v6_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) -{ - /* flush hdp cache */ - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(1) | - WRITE_DATA_DST_SEL(0))); - amdgpu_ring_write(ring, mmHDP_MEM_COHERENCY_FLUSH_CNTL); - amdgpu_ring_write(ring, 0); - amdgpu_ring_write(ring, 0x1); -} - static void gfx_v6_0_ring_emit_vgt_flush(struct amdgpu_ring *ring) { amdgpu_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE, 0)); @@ -1827,24 +1816,6 @@ static void gfx_v6_0_ring_emit_vgt_flush(struct amdgpu_ring *ring) EVENT_INDEX(0)); } -/** - * gfx_v6_0_ring_emit_hdp_invalidate - emit an hdp invalidate on the cp - * - * @adev: amdgpu_device pointer - * @ridx: amdgpu ring index - * - * Emits an hdp invalidate on the cp. - */ -static void gfx_v6_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(1) | - WRITE_DATA_DST_SEL(0))); - amdgpu_ring_write(ring, mmHDP_DEBUG0); - amdgpu_ring_write(ring, 0); - amdgpu_ring_write(ring, 0x1); -} - static void gfx_v6_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq, unsigned flags) { @@ -3507,8 +3478,7 @@ static const struct amdgpu_ring_funcs gfx_v6_0_ring_funcs_gfx = { .get_wptr = gfx_v6_0_ring_get_wptr, .set_wptr = gfx_v6_0_ring_set_wptr_gfx, .emit_frame_size = - 5 + /* gfx_v6_0_ring_emit_hdp_flush */ - 5 + /* gfx_v6_0_ring_emit_hdp_invalidate */ + 5 + 5 + /* hdp flush / invalidate */ 14 + 14 + 14 + /* gfx_v6_0_ring_emit_fence x3 for user fence, vm fence */ 7 + 4 + /* gfx_v6_0_ring_emit_pipeline_sync */ SI_FLUSH_GPU_TLB_NUM_WREG * 5 + 7 + 6 + /* gfx_v6_0_ring_emit_vm_flush */ @@ -3518,8 +3488,6 @@ static const struct amdgpu_ring_funcs gfx_v6_0_ring_funcs_gfx = { .emit_fence = gfx_v6_0_ring_emit_fence, .emit_pipeline_sync = gfx_v6_0_ring_emit_pipeline_sync, .emit_vm_flush = gfx_v6_0_ring_emit_vm_flush, - .emit_hdp_flush = gfx_v6_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = gfx_v6_0_ring_emit_hdp_invalidate, .test_ring = gfx_v6_0_ring_test_ring, .test_ib = gfx_v6_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, @@ -3535,8 +3503,7 @@ static const struct amdgpu_ring_funcs gfx_v6_0_ring_funcs_compute = { .get_wptr = gfx_v6_0_ring_get_wptr, .set_wptr = gfx_v6_0_ring_set_wptr_compute, .emit_frame_size = - 5 + /* gfx_v6_0_ring_emit_hdp_flush */ - 5 + /* gfx_v6_0_ring_emit_hdp_invalidate */ + 5 + 5 + /* hdp flush / invalidate */ 7 + /* gfx_v6_0_ring_emit_pipeline_sync */ SI_FLUSH_GPU_TLB_NUM_WREG * 5 + 7 + /* gfx_v6_0_ring_emit_vm_flush */ 14 + 14 + 14, /* gfx_v6_0_ring_emit_fence x3 for user fence, vm fence */ @@ -3545,8 +3512,6 @@ static const struct amdgpu_ring_funcs gfx_v6_0_ring_funcs_compute = { .emit_fence = gfx_v6_0_ring_emit_fence, .emit_pipeline_sync = gfx_v6_0_ring_emit_pipeline_sync, .emit_vm_flush = gfx_v6_0_ring_emit_vm_flush, - .emit_hdp_flush = gfx_v6_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = gfx_v6_0_ring_emit_hdp_invalidate, .test_ring = gfx_v6_0_ring_test_ring, .test_ib = gfx_v6_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index 3c2b678436f2..764e068fc2dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -2147,26 +2147,6 @@ static void gfx_v7_0_ring_emit_vgt_flush(struct amdgpu_ring *ring) EVENT_INDEX(0)); } - -/** - * gfx_v7_0_ring_emit_hdp_invalidate - emit an hdp invalidate on the cp - * - * @adev: amdgpu_device pointer - * @ridx: amdgpu ring index - * - * Emits an hdp invalidate on the cp. - */ -static void gfx_v7_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | - WRITE_DATA_DST_SEL(0) | - WR_CONFIRM)); - amdgpu_ring_write(ring, mmHDP_DEBUG0); - amdgpu_ring_write(ring, 0); - amdgpu_ring_write(ring, 1); -} - /** * gfx_v7_0_ring_emit_fence_gfx - emit a fence on the gfx ring * @@ -5110,7 +5090,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = { .emit_frame_size = 20 + /* gfx_v7_0_ring_emit_gds_switch */ 7 + /* gfx_v7_0_ring_emit_hdp_flush */ - 5 + /* gfx_v7_0_ring_emit_hdp_invalidate */ + 5 + /* hdp invalidate */ 12 + 12 + 12 + /* gfx_v7_0_ring_emit_fence_gfx x3 for user fence, vm fence */ 7 + 4 + /* gfx_v7_0_ring_emit_pipeline_sync */ CIK_FLUSH_GPU_TLB_NUM_WREG * 5 + 7 + 6 + /* gfx_v7_0_ring_emit_vm_flush */ @@ -5122,7 +5102,6 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = { .emit_vm_flush = gfx_v7_0_ring_emit_vm_flush, .emit_gds_switch = gfx_v7_0_ring_emit_gds_switch, .emit_hdp_flush = gfx_v7_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = gfx_v7_0_ring_emit_hdp_invalidate, .test_ring = gfx_v7_0_ring_test_ring, .test_ib = gfx_v7_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, @@ -5142,7 +5121,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = { .emit_frame_size = 20 + /* gfx_v7_0_ring_emit_gds_switch */ 7 + /* gfx_v7_0_ring_emit_hdp_flush */ - 5 + /* gfx_v7_0_ring_emit_hdp_invalidate */ + 5 + /* hdp invalidate */ 7 + /* gfx_v7_0_ring_emit_pipeline_sync */ CIK_FLUSH_GPU_TLB_NUM_WREG * 5 + 7 + /* gfx_v7_0_ring_emit_vm_flush */ 7 + 7 + 7, /* gfx_v7_0_ring_emit_fence_compute x3 for user fence, vm fence */ @@ -5153,7 +5132,6 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = { .emit_vm_flush = gfx_v7_0_ring_emit_vm_flush, .emit_gds_switch = gfx_v7_0_ring_emit_gds_switch, .emit_hdp_flush = gfx_v7_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = gfx_v7_0_ring_emit_hdp_invalidate, .test_ring = gfx_v7_0_ring_test_ring, .test_ib = gfx_v7_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index e4d209b5c879..5a2e4d5a5bd1 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -6230,19 +6230,6 @@ static void gfx_v8_0_ring_emit_vgt_flush(struct amdgpu_ring *ring) EVENT_INDEX(0)); } - -static void gfx_v8_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | - WRITE_DATA_DST_SEL(0) | - WR_CONFIRM)); - amdgpu_ring_write(ring, mmHDP_DEBUG0); - amdgpu_ring_write(ring, 0); - amdgpu_ring_write(ring, 1); - -} - static void gfx_v8_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, struct amdgpu_ib *ib, unsigned vmid, bool ctx_switch) @@ -6887,7 +6874,6 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = { .emit_vm_flush = gfx_v8_0_ring_emit_vm_flush, .emit_gds_switch = gfx_v8_0_ring_emit_gds_switch, .emit_hdp_flush = gfx_v8_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = gfx_v8_0_ring_emit_hdp_invalidate, .test_ring = gfx_v8_0_ring_test_ring, .test_ib = gfx_v8_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, @@ -6910,7 +6896,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = { .emit_frame_size = 20 + /* gfx_v8_0_ring_emit_gds_switch */ 7 + /* gfx_v8_0_ring_emit_hdp_flush */ - 5 + /* gfx_v8_0_ring_emit_hdp_invalidate */ + 5 + /* hdp_invalidate */ 7 + /* gfx_v8_0_ring_emit_pipeline_sync */ VI_FLUSH_GPU_TLB_NUM_WREG * 5 + 7 + /* gfx_v8_0_ring_emit_vm_flush */ 7 + 7 + 7, /* gfx_v8_0_ring_emit_fence_compute x3 for user fence, vm fence */ @@ -6921,7 +6907,6 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = { .emit_vm_flush = gfx_v8_0_ring_emit_vm_flush, .emit_gds_switch = gfx_v8_0_ring_emit_gds_switch, .emit_hdp_flush = gfx_v8_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = gfx_v8_0_ring_emit_hdp_invalidate, .test_ring = gfx_v8_0_ring_test_ring, .test_ib = gfx_v8_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, @@ -6941,7 +6926,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_kiq = { .emit_frame_size = 20 + /* gfx_v8_0_ring_emit_gds_switch */ 7 + /* gfx_v8_0_ring_emit_hdp_flush */ - 5 + /* gfx_v8_0_ring_emit_hdp_invalidate */ + 5 + /* hdp_invalidate */ 7 + /* gfx_v8_0_ring_emit_pipeline_sync */ 17 + /* gfx_v8_0_ring_emit_vm_flush */ 7 + 7 + 7, /* gfx_v8_0_ring_emit_fence_kiq x3 for user fence, vm fence */ diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index afdf57206efa..cd2b24ce785b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -3585,14 +3585,6 @@ static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) ref_and_mask, ref_and_mask, 0x20); } -static void gfx_v9_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - struct amdgpu_device *adev = ring->adev; - - gfx_v9_0_write_data_to_reg(ring, 0, true, - SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 1); -} - static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, struct amdgpu_ib *ib, unsigned vmid, bool ctx_switch) @@ -4319,7 +4311,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = { .emit_vm_flush = gfx_v9_0_ring_emit_vm_flush, .emit_gds_switch = gfx_v9_0_ring_emit_gds_switch, .emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = gfx_v9_0_ring_emit_hdp_invalidate, .test_ring = gfx_v9_0_ring_test_ring, .test_ib = gfx_v9_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, @@ -4344,7 +4335,7 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = { .emit_frame_size = 20 + /* gfx_v9_0_ring_emit_gds_switch */ 7 + /* gfx_v9_0_ring_emit_hdp_flush */ - 5 + /* gfx_v9_0_ring_emit_hdp_invalidate */ + 5 + /* hdp invalidate */ 7 + /* gfx_v9_0_ring_emit_pipeline_sync */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 + 9 + /* gfx_v9_0_ring_emit_vm_flush */ 8 + 8 + 8, /* gfx_v9_0_ring_emit_fence x3 for user fence, vm fence */ @@ -4355,7 +4346,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = { .emit_vm_flush = gfx_v9_0_ring_emit_vm_flush, .emit_gds_switch = gfx_v9_0_ring_emit_gds_switch, .emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = gfx_v9_0_ring_emit_hdp_invalidate, .test_ring = gfx_v9_0_ring_test_ring, .test_ib = gfx_v9_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, @@ -4376,7 +4366,7 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = { .emit_frame_size = 20 + /* gfx_v9_0_ring_emit_gds_switch */ 7 + /* gfx_v9_0_ring_emit_hdp_flush */ - 5 + /* gfx_v9_0_ring_emit_hdp_invalidate */ + 5 + /* hdp invalidate */ 7 + /* gfx_v9_0_ring_emit_pipeline_sync */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 + 9 + /* gfx_v9_0_ring_emit_vm_flush */ 8 + 8 + 8, /* gfx_v9_0_ring_emit_fence_kiq x3 for user fence, vm fence */ diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index 0aa336371816..6a7a82a8c65d 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -289,13 +289,6 @@ 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 * @@ -1200,7 +1193,7 @@ static const struct amdgpu_ring_funcs sdma_v2_4_ring_funcs = { .set_wptr = sdma_v2_4_ring_set_wptr, .emit_frame_size = 6 + /* sdma_v2_4_ring_emit_hdp_flush */ - 3 + /* sdma_v2_4_ring_emit_hdp_invalidate */ + 3 + /* hdp invalidate */ 6 + /* sdma_v2_4_ring_emit_pipeline_sync */ VI_FLUSH_GPU_TLB_NUM_WREG * 3 + 6 + /* sdma_v2_4_ring_emit_vm_flush */ 10 + 10 + 10, /* sdma_v2_4_ring_emit_fence x3 for user fence, vm fence */ @@ -1210,7 +1203,6 @@ static const struct amdgpu_ring_funcs sdma_v2_4_ring_funcs = { .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, .test_ring = sdma_v2_4_ring_test_ring, .test_ib = sdma_v2_4_ring_test_ib, .insert_nop = sdma_v2_4_ring_insert_nop, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index e417546e2048..88178d81bd5a 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -460,14 +460,6 @@ static void sdma_v3_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); /* retry count, poll interval */ } -static void sdma_v3_0_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_v3_0_ring_emit_fence - emit a fence on the DMA ring * @@ -1634,7 +1626,7 @@ static const struct amdgpu_ring_funcs sdma_v3_0_ring_funcs = { .set_wptr = sdma_v3_0_ring_set_wptr, .emit_frame_size = 6 + /* sdma_v3_0_ring_emit_hdp_flush */ - 3 + /* sdma_v3_0_ring_emit_hdp_invalidate */ + 3 + /* hdp invalidate */ 6 + /* sdma_v3_0_ring_emit_pipeline_sync */ 12 + /* sdma_v3_0_ring_emit_vm_flush */ 10 + 10 + 10, /* sdma_v3_0_ring_emit_fence x3 for user fence, vm fence */ @@ -1644,7 +1636,6 @@ static const struct amdgpu_ring_funcs sdma_v3_0_ring_funcs = { .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, .test_ring = sdma_v3_0_ring_test_ring, .test_ib = sdma_v3_0_ring_test_ib, .insert_nop = sdma_v3_0_ring_insert_nop, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index ee919477d7ed..e9b1b834fee1 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -375,16 +375,6 @@ static void sdma_v4_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); /* retry count, poll interval */ } -static void sdma_v4_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - struct amdgpu_device *adev = ring->adev; - - amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) | - SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf)); - amdgpu_ring_write(ring, SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE)); - amdgpu_ring_write(ring, 1); -} - /** * sdma_v4_0_ring_emit_fence - emit a fence on the DMA ring * @@ -1583,7 +1573,7 @@ static const struct amdgpu_ring_funcs sdma_v4_0_ring_funcs = { .set_wptr = sdma_v4_0_ring_set_wptr, .emit_frame_size = 6 + /* sdma_v4_0_ring_emit_hdp_flush */ - 3 + /* sdma_v4_0_ring_emit_hdp_invalidate */ + 3 + /* hdp invalidate */ 6 + /* sdma_v4_0_ring_emit_pipeline_sync */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 + 6 + /* sdma_v4_0_ring_emit_vm_flush */ 10 + 10 + 10, /* sdma_v4_0_ring_emit_fence x3 for user fence, vm fence */ @@ -1593,7 +1583,6 @@ static const struct amdgpu_ring_funcs sdma_v4_0_ring_funcs = { .emit_pipeline_sync = sdma_v4_0_ring_emit_pipeline_sync, .emit_vm_flush = sdma_v4_0_ring_emit_vm_flush, .emit_hdp_flush = sdma_v4_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = sdma_v4_0_ring_emit_hdp_invalidate, .test_ring = sdma_v4_0_ring_test_ring, .test_ib = sdma_v4_0_ring_test_ib, .insert_nop = sdma_v4_0_ring_insert_nop, diff --git a/drivers/gpu/drm/amd/amdgpu/si_dma.c b/drivers/gpu/drm/amd/amdgpu/si_dma.c index 8f9509f6f15b..e59521bacf0b 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dma.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dma.c @@ -75,20 +75,6 @@ static void si_dma_ring_emit_ib(struct amdgpu_ring *ring, } -static void si_dma_ring_emit_hdp_flush(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0)); - amdgpu_ring_write(ring, (0xf << 16) | (HDP_MEM_COHERENCY_FLUSH_CNTL)); - amdgpu_ring_write(ring, 1); -} - -static void si_dma_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0)); - amdgpu_ring_write(ring, (0xf << 16) | (HDP_DEBUG0)); - amdgpu_ring_write(ring, 1); -} - /** * si_dma_ring_emit_fence - emit a fence on the DMA ring * @@ -772,8 +758,7 @@ static const struct amdgpu_ring_funcs si_dma_ring_funcs = { .get_wptr = si_dma_ring_get_wptr, .set_wptr = si_dma_ring_set_wptr, .emit_frame_size = - 3 + /* si_dma_ring_emit_hdp_flush */ - 3 + /* si_dma_ring_emit_hdp_invalidate */ + 3 + 3 + /* hdp flush / invalidate */ 6 + /* si_dma_ring_emit_pipeline_sync */ SI_FLUSH_GPU_TLB_NUM_WREG * 3 + 6 + /* si_dma_ring_emit_vm_flush */ 9 + 9 + 9, /* si_dma_ring_emit_fence x3 for user fence, vm fence */ @@ -782,8 +767,6 @@ static const struct amdgpu_ring_funcs si_dma_ring_funcs = { .emit_fence = si_dma_ring_emit_fence, .emit_pipeline_sync = si_dma_ring_emit_pipeline_sync, .emit_vm_flush = si_dma_ring_emit_vm_flush, - .emit_hdp_flush = si_dma_ring_emit_hdp_flush, - .emit_hdp_invalidate = si_dma_ring_emit_hdp_invalidate, .test_ring = si_dma_ring_test_ring, .test_ib = si_dma_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c index 8ab10c220910..948bb9437757 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c @@ -463,32 +463,6 @@ static void uvd_v4_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq amdgpu_ring_write(ring, 2); } -/** - * uvd_v4_2_ring_emit_hdp_flush - emit an hdp flush - * - * @ring: amdgpu_ring pointer - * - * Emits an hdp flush. - */ -static void uvd_v4_2_ring_emit_hdp_flush(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, PACKET0(mmHDP_MEM_COHERENCY_FLUSH_CNTL, 0)); - amdgpu_ring_write(ring, 0); -} - -/** - * uvd_v4_2_ring_hdp_invalidate - emit an hdp invalidate - * - * @ring: amdgpu_ring pointer - * - * Emits an hdp invalidate. - */ -static void uvd_v4_2_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, PACKET0(mmHDP_DEBUG0, 0)); - amdgpu_ring_write(ring, 1); -} - /** * uvd_v4_2_ring_test_ring - register write test * @@ -765,14 +739,10 @@ static const struct amdgpu_ring_funcs uvd_v4_2_ring_funcs = { .set_wptr = uvd_v4_2_ring_set_wptr, .parse_cs = amdgpu_uvd_ring_parse_cs, .emit_frame_size = - 2 + /* uvd_v4_2_ring_emit_hdp_flush */ - 2 + /* uvd_v4_2_ring_emit_hdp_invalidate */ 14, /* uvd_v4_2_ring_emit_fence x1 no user fence */ .emit_ib_size = 4, /* uvd_v4_2_ring_emit_ib */ .emit_ib = uvd_v4_2_ring_emit_ib, .emit_fence = uvd_v4_2_ring_emit_fence, - .emit_hdp_flush = uvd_v4_2_ring_emit_hdp_flush, - .emit_hdp_invalidate = uvd_v4_2_ring_emit_hdp_invalidate, .test_ring = uvd_v4_2_ring_test_ring, .test_ib = amdgpu_uvd_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c index c1fe30cdba32..6445d55e7d5a 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c @@ -478,32 +478,6 @@ static void uvd_v5_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq amdgpu_ring_write(ring, 2); } -/** - * uvd_v5_0_ring_emit_hdp_flush - emit an hdp flush - * - * @ring: amdgpu_ring pointer - * - * Emits an hdp flush. - */ -static void uvd_v5_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, PACKET0(mmHDP_MEM_COHERENCY_FLUSH_CNTL, 0)); - amdgpu_ring_write(ring, 0); -} - -/** - * uvd_v5_0_ring_hdp_invalidate - emit an hdp invalidate - * - * @ring: amdgpu_ring pointer - * - * Emits an hdp invalidate. - */ -static void uvd_v5_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, PACKET0(mmHDP_DEBUG0, 0)); - amdgpu_ring_write(ring, 1); -} - /** * uvd_v5_0_ring_test_ring - register write test * @@ -873,14 +847,10 @@ static const struct amdgpu_ring_funcs uvd_v5_0_ring_funcs = { .set_wptr = uvd_v5_0_ring_set_wptr, .parse_cs = amdgpu_uvd_ring_parse_cs, .emit_frame_size = - 2 + /* uvd_v5_0_ring_emit_hdp_flush */ - 2 + /* uvd_v5_0_ring_emit_hdp_invalidate */ 14, /* uvd_v5_0_ring_emit_fence x1 no user fence */ .emit_ib_size = 6, /* uvd_v5_0_ring_emit_ib */ .emit_ib = uvd_v5_0_ring_emit_ib, .emit_fence = uvd_v5_0_ring_emit_fence, - .emit_hdp_flush = uvd_v5_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = uvd_v5_0_ring_emit_hdp_invalidate, .test_ring = uvd_v5_0_ring_test_ring, .test_ib = amdgpu_uvd_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c index 5f499e663e2a..e7546d5b301c 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c @@ -963,32 +963,6 @@ static void uvd_v6_0_enc_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, amdgpu_ring_write(ring, HEVC_ENC_CMD_TRAP); } -/** - * uvd_v6_0_ring_emit_hdp_flush - emit an hdp flush - * - * @ring: amdgpu_ring pointer - * - * Emits an hdp flush. - */ -static void uvd_v6_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, PACKET0(mmHDP_MEM_COHERENCY_FLUSH_CNTL, 0)); - amdgpu_ring_write(ring, 0); -} - -/** - * uvd_v6_0_ring_hdp_invalidate - emit an hdp invalidate - * - * @ring: amdgpu_ring pointer - * - * Emits an hdp invalidate. - */ -static void uvd_v6_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, PACKET0(mmHDP_DEBUG0, 0)); - amdgpu_ring_write(ring, 1); -} - /** * uvd_v6_0_ring_test_ring - register write test * @@ -1556,15 +1530,11 @@ static const struct amdgpu_ring_funcs uvd_v6_0_ring_phys_funcs = { .set_wptr = uvd_v6_0_ring_set_wptr, .parse_cs = amdgpu_uvd_ring_parse_cs, .emit_frame_size = - 2 + /* uvd_v6_0_ring_emit_hdp_flush */ - 2 + /* uvd_v6_0_ring_emit_hdp_invalidate */ 10 + /* uvd_v6_0_ring_emit_pipeline_sync */ 14, /* uvd_v6_0_ring_emit_fence x1 no user fence */ .emit_ib_size = 8, /* uvd_v6_0_ring_emit_ib */ .emit_ib = uvd_v6_0_ring_emit_ib, .emit_fence = uvd_v6_0_ring_emit_fence, - .emit_hdp_flush = uvd_v6_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = uvd_v6_0_ring_emit_hdp_invalidate, .test_ring = uvd_v6_0_ring_test_ring, .test_ib = amdgpu_uvd_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, @@ -1582,8 +1552,7 @@ static const struct amdgpu_ring_funcs uvd_v6_0_ring_vm_funcs = { .get_wptr = uvd_v6_0_ring_get_wptr, .set_wptr = uvd_v6_0_ring_set_wptr, .emit_frame_size = - 2 + /* uvd_v6_0_ring_emit_hdp_flush */ - 2 + /* uvd_v6_0_ring_emit_hdp_invalidate */ + 6 + 6 + /* hdp flush / invalidate */ 10 + /* uvd_v6_0_ring_emit_pipeline_sync */ 20 + /* uvd_v6_0_ring_emit_vm_flush */ 14 + 14, /* uvd_v6_0_ring_emit_fence x2 vm fence */ @@ -1592,8 +1561,6 @@ static const struct amdgpu_ring_funcs uvd_v6_0_ring_vm_funcs = { .emit_fence = uvd_v6_0_ring_emit_fence, .emit_vm_flush = uvd_v6_0_ring_emit_vm_flush, .emit_pipeline_sync = uvd_v6_0_ring_emit_pipeline_sync, - .emit_hdp_flush = uvd_v6_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = uvd_v6_0_ring_emit_hdp_invalidate, .test_ring = uvd_v6_0_ring_test_ring, .test_ib = amdgpu_uvd_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c index 241e73022cd7..d317c764cc91 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c @@ -1135,37 +1135,6 @@ static void uvd_v7_0_enc_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, amdgpu_ring_write(ring, HEVC_ENC_CMD_TRAP); } -/** - * uvd_v7_0_ring_emit_hdp_flush - emit an hdp flush - * - * @ring: amdgpu_ring pointer - * - * Emits an hdp flush. - */ -static void uvd_v7_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) -{ - struct amdgpu_device *adev = ring->adev; - - amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(NBIF, 0, - mmHDP_MEM_COHERENCY_FLUSH_CNTL), 0)); - amdgpu_ring_write(ring, 0); -} - -/** - * uvd_v7_0_ring_hdp_invalidate - emit an hdp invalidate - * - * @ring: amdgpu_ring pointer - * - * Emits an hdp invalidate. - */ -static void uvd_v7_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - struct amdgpu_device *adev = ring->adev; - - amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 0)); - amdgpu_ring_write(ring, 1); -} - /** * uvd_v7_0_ring_test_ring - register write test * @@ -1693,16 +1662,13 @@ static const struct amdgpu_ring_funcs uvd_v7_0_ring_vm_funcs = { .get_wptr = uvd_v7_0_ring_get_wptr, .set_wptr = uvd_v7_0_ring_set_wptr, .emit_frame_size = - 2 + /* uvd_v7_0_ring_emit_hdp_flush */ - 2 + /* uvd_v7_0_ring_emit_hdp_invalidate */ + 6 + 6 + /* hdp flush / invalidate */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + 16 + /* uvd_v7_0_ring_emit_vm_flush */ 14 + 14, /* uvd_v7_0_ring_emit_fence x2 vm fence */ .emit_ib_size = 8, /* uvd_v7_0_ring_emit_ib */ .emit_ib = uvd_v7_0_ring_emit_ib, .emit_fence = uvd_v7_0_ring_emit_fence, .emit_vm_flush = uvd_v7_0_ring_emit_vm_flush, - .emit_hdp_flush = uvd_v7_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = uvd_v7_0_ring_emit_hdp_invalidate, .test_ring = uvd_v7_0_ring_test_ring, .test_ib = amdgpu_uvd_ring_test_ib, .insert_nop = uvd_v7_0_ring_insert_nop, @@ -1722,6 +1688,7 @@ static const struct amdgpu_ring_funcs uvd_v7_0_enc_ring_vm_funcs = { .get_wptr = uvd_v7_0_enc_ring_get_wptr, .set_wptr = uvd_v7_0_enc_ring_set_wptr, .emit_frame_size = + 3 + 3 + /* hdp flush / invalidate */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 + 8 + /* uvd_v7_0_enc_ring_emit_vm_flush */ 5 + 5 + /* uvd_v7_0_enc_ring_emit_fence x2 vm fence */ 1, /* uvd_v7_0_enc_ring_insert_end */ diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c index 76cdef29b9d1..44c041a1fe68 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c @@ -809,21 +809,6 @@ static void vcn_v1_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 amdgpu_ring_write(ring, VCN_DEC_CMD_TRAP << 1); } -/** - * vcn_v1_0_dec_ring_hdp_invalidate - emit an hdp invalidate - * - * @ring: amdgpu_ring pointer - * - * Emits an hdp invalidate. - */ -static void vcn_v1_0_dec_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - struct amdgpu_device *adev = ring->adev; - - amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 0)); - amdgpu_ring_write(ring, 1); -} - /** * vcn_v1_0_dec_ring_emit_ib - execute indirect buffer * @@ -1096,7 +1081,7 @@ static const struct amdgpu_ring_funcs vcn_v1_0_dec_ring_vm_funcs = { .get_wptr = vcn_v1_0_dec_ring_get_wptr, .set_wptr = vcn_v1_0_dec_ring_set_wptr, .emit_frame_size = - 2 + /* vcn_v1_0_dec_ring_emit_hdp_invalidate */ + 6 + 6 + /* hdp invalidate / flush */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + 16 + /* vcn_v1_0_dec_ring_emit_vm_flush */ 14 + 14 + /* vcn_v1_0_dec_ring_emit_fence x2 vm fence */ 6, @@ -1104,7 +1089,6 @@ static const struct amdgpu_ring_funcs vcn_v1_0_dec_ring_vm_funcs = { .emit_ib = vcn_v1_0_dec_ring_emit_ib, .emit_fence = vcn_v1_0_dec_ring_emit_fence, .emit_vm_flush = vcn_v1_0_dec_ring_emit_vm_flush, - .emit_hdp_invalidate = vcn_v1_0_dec_ring_emit_hdp_invalidate, .test_ring = amdgpu_vcn_dec_ring_test_ring, .test_ib = amdgpu_vcn_dec_ring_test_ib, .insert_nop = vcn_v1_0_ring_insert_nop, -- cgit v1.2.3 From c1e877da570cf3bb4cf88c3306ac8d68ca51b6ad Mon Sep 17 00:00:00 2001 From: Christian König Date: Fri, 26 Jan 2018 12:45:32 +0100 Subject: drm/amdgpu: add new emit_reg_wait callback MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Allows us to wait for a register value/mask on a ring. Signed-off-by: Christian König Reviewed-by: Felix Kuehling Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 2 ++ 2 files changed, 3 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index a1f738170d37..36ac07aee0eb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1794,6 +1794,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_ring_emit_cntxcntl(r, d) (r)->funcs->emit_cntxcntl((r), (d)) #define amdgpu_ring_emit_rreg(r, d) (r)->funcs->emit_rreg((r), (d)) #define amdgpu_ring_emit_wreg(r, d, v) (r)->funcs->emit_wreg((r), (d), (v)) +#define amdgpu_ring_emit_reg_wait(r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m)) #define amdgpu_ring_emit_tmz(r, b) (r)->funcs->emit_tmz((r), (b)) #define amdgpu_ring_pad_ib(r, ib) ((r)->funcs->pad_ib((r), (ib))) #define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r)) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index 70d05ec7bc07..867f53332305 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -150,6 +150,8 @@ struct amdgpu_ring_funcs { void (*emit_cntxcntl) (struct amdgpu_ring *ring, uint32_t flags); void (*emit_rreg)(struct amdgpu_ring *ring, uint32_t reg); void (*emit_wreg)(struct amdgpu_ring *ring, uint32_t reg, uint32_t val); + void (*emit_reg_wait)(struct amdgpu_ring *ring, uint32_t reg, + uint32_t val, uint32_t mask); void (*emit_tmz)(struct amdgpu_ring *ring, bool start); /* priority functions */ void (*set_priority) (struct amdgpu_ring *ring, -- cgit v1.2.3 From 3af81440a9b02e2c6e244539ee567063c3d62292 Mon Sep 17 00:00:00 2001 From: Christian König Date: Wed, 31 Jan 2018 16:03:19 +0100 Subject: drm/amdgpu: cache the fence to wait for a VMID MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Beneficial when a lot of processes are waiting for VMIDs. Signed-off-by: Christian König Reviewed-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c | 7 +++++-- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c | 3 +++ drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 1 + 3 files changed, 9 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c index 156e026046b5..7d2805729c20 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c @@ -205,6 +205,9 @@ static int amdgpu_vmid_grab_idle(struct amdgpu_vm *vm, unsigned i; int r; + if (ring->vmid_wait && !dma_fence_is_signaled(ring->vmid_wait)) + return amdgpu_sync_fence(adev, sync, ring->vmid_wait, false); + fences = kmalloc_array(sizeof(void *), id_mgr->num_ids, GFP_KERNEL); if (!fences) return -ENOMEM; @@ -239,9 +242,9 @@ static int amdgpu_vmid_grab_idle(struct amdgpu_vm *vm, } r = amdgpu_sync_fence(adev, sync, &array->base, false); - dma_fence_put(&array->base); + dma_fence_put(ring->vmid_wait); + ring->vmid_wait = &array->base; return r; - } kfree(fences); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index 13044e66dcaf..e223b0f6417b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -360,6 +360,9 @@ void amdgpu_ring_fini(struct amdgpu_ring *ring) amdgpu_debugfs_ring_fini(ring); + dma_fence_put(ring->vmid_wait); + ring->vmid_wait = NULL; + ring->adev->rings[ring->idx] = NULL; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index 867f53332305..075976855651 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -196,6 +196,7 @@ struct amdgpu_ring { u64 cond_exe_gpu_addr; volatile u32 *cond_exe_cpu_addr; unsigned vm_inv_eng; + struct dma_fence *vmid_wait; bool has_compute_vm_bug; atomic_t num_jobs[DRM_SCHED_PRIORITY_MAX]; -- cgit v1.2.3 From c633c00bf06779ec6d5e2c01748d4753ede98f8a Mon Sep 17 00:00:00 2001 From: Christian König Date: Sun, 4 Feb 2018 10:32:35 +0100 Subject: drm/amdgpu: separate PASID mapping from VM flush v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Stuffing the PASID mapping into the VM flush isn't flexible enough since the PASID mapping changes not as often as we need a VM flush. v2: add missing use of gmc_v7_0_emit_pasid_mapping Signed-off-by: Christian König Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 5 +++-- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h | 5 ++++- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 7 +++++-- drivers/gpu/drm/amd/amdgpu/cik_sdma.c | 5 ++--- drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c | 5 ++--- drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 5 ++--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 5 ++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 5 ++--- drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c | 3 +-- drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c | 12 ++++++++---- drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c | 12 ++++++++---- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 28 +++++++++++++++++----------- drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 5 ++--- drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 5 ++--- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 5 ++--- drivers/gpu/drm/amd/amdgpu/si_dma.c | 5 ++--- drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c | 8 +++----- drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c | 10 ++++------ drivers/gpu/drm/amd/amdgpu/vce_v3_0.c | 3 +-- drivers/gpu/drm/amd/amdgpu/vce_v4_0.c | 5 ++--- drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 10 ++++------ 22 files changed, 79 insertions(+), 76 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 44cf4b9a5703..c6123e586589 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1774,7 +1774,8 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_asic_flush_hdp(adev, r) (adev)->asic_funcs->flush_hdp((adev), (r)) #define amdgpu_asic_invalidate_hdp(adev, r) (adev)->asic_funcs->invalidate_hdp((adev), (r)) #define amdgpu_gmc_flush_gpu_tlb(adev, vmid) (adev)->gmc.gmc_funcs->flush_gpu_tlb((adev), (vmid)) -#define amdgpu_gmc_emit_flush_gpu_tlb(r, vmid, pasid, addr) (r)->adev->gmc.gmc_funcs->emit_flush_gpu_tlb((r), (vmid), (pasid), (addr)) +#define amdgpu_gmc_emit_flush_gpu_tlb(r, vmid, addr) (r)->adev->gmc.gmc_funcs->emit_flush_gpu_tlb((r), (vmid), (addr)) +#define amdgpu_gmc_emit_pasid_mapping(r, vmid, pasid) (r)->adev->gmc.gmc_funcs->emit_pasid_mapping((r), (vmid), (pasid)) #define amdgpu_gmc_set_pte_pde(adev, pt, idx, addr, flags) (adev)->gmc.gmc_funcs->set_pte_pde((adev), (pt), (idx), (addr), (flags)) #define amdgpu_gmc_get_vm_pde(adev, level, dst, flags) (adev)->gmc.gmc_funcs->get_vm_pde((adev), (level), (dst), (flags)) #define amdgpu_gmc_get_pte_flags(adev, flags) (adev)->gmc.gmc_funcs->get_vm_pte_flags((adev),(flags)) @@ -1789,7 +1790,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_ring_set_wptr(r) (r)->funcs->set_wptr((r)) #define amdgpu_ring_emit_ib(r, ib, vmid, c) (r)->funcs->emit_ib((r), (ib), (vmid), (c)) #define amdgpu_ring_emit_pipeline_sync(r) (r)->funcs->emit_pipeline_sync((r)) -#define amdgpu_ring_emit_vm_flush(r, vmid, pasid, addr) (r)->funcs->emit_vm_flush((r), (vmid), (pasid), (addr)) +#define amdgpu_ring_emit_vm_flush(r, vmid, addr) (r)->funcs->emit_vm_flush((r), (vmid), (addr)) #define amdgpu_ring_emit_fence(r, addr, seq, flags) (r)->funcs->emit_fence((r), (addr), (seq), (flags)) #define amdgpu_ring_emit_gds_switch(r, v, db, ds, wb, ws, ab, as) (r)->funcs->emit_gds_switch((r), (v), (db), (ds), (wb), (ws), (ab), (as)) #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r)) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h index b3d1bd2f51cf..893c2490b783 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h @@ -54,7 +54,10 @@ struct amdgpu_gmc_funcs { uint32_t vmid); /* flush the vm tlb via ring */ uint64_t (*emit_flush_gpu_tlb)(struct amdgpu_ring *ring, unsigned vmid, - unsigned pasid, uint64_t pd_addr); + uint64_t pd_addr); + /* Change the VMID -> PASID mapping */ + void (*emit_pasid_mapping)(struct amdgpu_ring *ring, unsigned vmid, + unsigned pasid); /* write pte/pde updates using the cpu */ int (*set_pte_pde)(struct amdgpu_device *adev, void *cpu_pt_addr, /* cpu addr of page table */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index 075976855651..1d0d250cbfdf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -126,7 +126,7 @@ struct amdgpu_ring_funcs { uint64_t seq, unsigned flags); void (*emit_pipeline_sync)(struct amdgpu_ring *ring); void (*emit_vm_flush)(struct amdgpu_ring *ring, unsigned vmid, - unsigned pasid, uint64_t pd_addr); + uint64_t pd_addr); void (*emit_hdp_flush)(struct amdgpu_ring *ring); void (*emit_gds_switch)(struct amdgpu_ring *ring, uint32_t vmid, uint32_t gds_base, uint32_t gds_size, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 0572d6072baa..afa16a862eaa 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -612,8 +612,11 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_ struct dma_fence *fence; trace_amdgpu_vm_flush(ring, job->vmid, job->vm_pd_addr); - amdgpu_ring_emit_vm_flush(ring, job->vmid, job->pasid, - job->vm_pd_addr); + amdgpu_ring_emit_vm_flush(ring, job->vmid, job->vm_pd_addr); + if (adev->gmc.gmc_funcs->emit_pasid_mapping && + ring->funcs->emit_wreg) + amdgpu_gmc_emit_pasid_mapping(ring, job->vmid, + job->pasid); r = amdgpu_fence_emit(ring, &fence); if (r) diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index d78bf183488b..69568cd1bb99 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -873,13 +873,12 @@ static void cik_sdma_ring_emit_pipeline_sync(struct amdgpu_ring *ring) * using sDMA (CIK). */ static void cik_sdma_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(0) | SDMA_POLL_REG_MEM_EXTRA_FUNC(0)); /* always */ - amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_POLL_REG_MEM, 0, extra_bits)); amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST << 2); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c index 3517fd9e11c9..0fff5b8cd318 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c @@ -2326,12 +2326,11 @@ static void gfx_v6_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) } static void gfx_v6_0_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX); - amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* wait for the invalidate to complete */ amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5)); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index 764e068fc2dd..972d421caada 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -3219,12 +3219,11 @@ static void gfx_v7_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) * using the CP (CIK). */ static void gfx_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX); - amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* wait for the invalidate to complete */ amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5)); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 8a65b5327a02..27943e57681c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -6311,12 +6311,11 @@ static void gfx_v8_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) } static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX); - amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* wait for the invalidate to complete */ amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5)); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index f7363f821cff..848008ef46b8 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -3676,10 +3676,9 @@ static void gfx_v9_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) } static void gfx_v9_0_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { - amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* compute doesn't have PFP */ if (ring->funcs->type == AMDGPU_RING_TYPE_GFX) { diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c index 1945fe842188..2c0ed9dd0c91 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c @@ -363,8 +363,7 @@ static void gmc_v6_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid) } static uint64_t gmc_v6_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { uint32_t reg; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c index 761def04f93f..4edd17059868 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -436,8 +436,7 @@ static void gmc_v7_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid) } static uint64_t gmc_v7_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { uint32_t reg; @@ -447,14 +446,18 @@ static uint64_t gmc_v7_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring, reg = mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8; amdgpu_ring_emit_wreg(ring, reg, pd_addr >> 12); - amdgpu_ring_emit_wreg(ring, mmIH_VMID_0_LUT + vmid, pasid); - /* bits 0-15 are the VM contexts0-15 */ amdgpu_ring_emit_wreg(ring, mmVM_INVALIDATE_REQUEST, 1 << vmid); return pd_addr; } +static void gmc_v7_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid, + unsigned pasid) +{ + amdgpu_ring_emit_wreg(ring, mmIH_VMID_0_LUT + vmid, pasid); +} + /** * gmc_v7_0_set_pte_pde - update the page tables using MMIO * @@ -1327,6 +1330,7 @@ static const struct amd_ip_funcs gmc_v7_0_ip_funcs = { static const struct amdgpu_gmc_funcs gmc_v7_0_gmc_funcs = { .flush_gpu_tlb = gmc_v7_0_flush_gpu_tlb, .emit_flush_gpu_tlb = gmc_v7_0_emit_flush_gpu_tlb, + .emit_pasid_mapping = gmc_v7_0_emit_pasid_mapping, .set_pte_pde = gmc_v7_0_set_pte_pde, .set_prt = gmc_v7_0_set_prt, .get_vm_pte_flags = gmc_v7_0_get_vm_pte_flags, diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index 2489be7ad62b..1e0ad0657e96 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -612,8 +612,7 @@ static void gmc_v8_0_flush_gpu_tlb(struct amdgpu_device *adev, } static uint64_t gmc_v8_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { uint32_t reg; @@ -623,14 +622,18 @@ static uint64_t gmc_v8_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring, reg = mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8; amdgpu_ring_emit_wreg(ring, reg, pd_addr >> 12); - amdgpu_ring_emit_wreg(ring, mmIH_VMID_0_LUT + vmid, pasid); - /* bits 0-15 are the VM contexts0-15 */ amdgpu_ring_emit_wreg(ring, mmVM_INVALIDATE_REQUEST, 1 << vmid); return pd_addr; } +static void gmc_v8_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid, + unsigned pasid) +{ + amdgpu_ring_emit_wreg(ring, mmIH_VMID_0_LUT + vmid, pasid); +} + /** * gmc_v8_0_set_pte_pde - update the page tables using MMIO * @@ -1662,6 +1665,7 @@ static const struct amd_ip_funcs gmc_v8_0_ip_funcs = { static const struct amdgpu_gmc_funcs gmc_v8_0_gmc_funcs = { .flush_gpu_tlb = gmc_v8_0_flush_gpu_tlb, .emit_flush_gpu_tlb = gmc_v8_0_emit_flush_gpu_tlb, + .emit_pasid_mapping = gmc_v8_0_emit_pasid_mapping, .set_pte_pde = gmc_v8_0_set_pte_pde, .set_prt = gmc_v8_0_set_prt, .get_vm_pte_flags = gmc_v8_0_get_vm_pte_flags, diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index d5b6d00d83a6..bc4bd5e7ac94 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -368,17 +368,15 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, } static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { struct amdgpu_device *adev = ring->adev; struct amdgpu_vmhub *hub = &adev->vmhub[ring->funcs->vmhub]; uint32_t req = gmc_v9_0_get_invalidate_req(vmid); uint64_t flags = AMDGPU_PTE_VALID; unsigned eng = ring->vm_inv_eng; - uint32_t reg; - amdgpu_gmc_get_vm_pde(ring->adev, -1, &pd_addr, &flags); + amdgpu_gmc_get_vm_pde(adev, -1, &pd_addr, &flags); pd_addr |= flags; amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_lo32 + (2 * vmid), @@ -387,13 +385,6 @@ static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring, amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_hi32 + (2 * vmid), upper_32_bits(pd_addr)); - if (ring->funcs->vmhub == AMDGPU_GFXHUB) - reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid; - else - reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT_MM) + vmid; - - amdgpu_ring_emit_wreg(ring, reg, pasid); - amdgpu_ring_emit_wreg(ring, hub->vm_inv_eng0_req + eng, req); /* wait for the invalidate to complete */ @@ -403,6 +394,20 @@ static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring, return pd_addr; } +static void gmc_v9_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid, + unsigned pasid) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t reg; + + if (ring->funcs->vmhub == AMDGPU_GFXHUB) + reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid; + else + reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT_MM) + vmid; + + amdgpu_ring_emit_wreg(ring, reg, pasid); +} + /** * gmc_v9_0_set_pte_pde - update the page tables using MMIO * @@ -529,6 +534,7 @@ static void gmc_v9_0_get_vm_pde(struct amdgpu_device *adev, int level, static const struct amdgpu_gmc_funcs gmc_v9_0_gmc_funcs = { .flush_gpu_tlb = gmc_v9_0_flush_gpu_tlb, .emit_flush_gpu_tlb = gmc_v9_0_emit_flush_gpu_tlb, + .emit_pasid_mapping = gmc_v9_0_emit_pasid_mapping, .set_pte_pde = gmc_v9_0_set_pte_pde, .get_vm_pte_flags = gmc_v9_0_get_vm_pte_flags, .get_vm_pde = gmc_v9_0_get_vm_pde diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index 792774eee909..6ccc9d43a7b8 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -852,10 +852,9 @@ static void sdma_v2_4_ring_emit_pipeline_sync(struct amdgpu_ring *ring) * using sDMA (VI). */ static void sdma_v2_4_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { - amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* wait for flush */ amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) | diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 5680ced69359..0c2b12ec0e9f 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -1117,10 +1117,9 @@ static void sdma_v3_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) * using sDMA (VI). */ static void sdma_v3_0_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { - amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* wait for flush */ amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) | diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index ce599fd24412..3d5385dda34c 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -1123,10 +1123,9 @@ static void sdma_v4_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) * using sDMA (VEGA10). */ static void sdma_v4_0_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { - amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); } static void sdma_v4_0_ring_emit_wreg(struct amdgpu_ring *ring, diff --git a/drivers/gpu/drm/amd/amdgpu/si_dma.c b/drivers/gpu/drm/amd/amdgpu/si_dma.c index 2db5bfba771e..acbf5afa4f38 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dma.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dma.c @@ -460,10 +460,9 @@ static void si_dma_ring_emit_pipeline_sync(struct amdgpu_ring *ring) * using sDMA (VI). */ static void si_dma_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { - amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* wait for invalidate to complete */ amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_POLL_REG_MEM, 0, 0, 0, 0)); diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c index 0f192ab71205..a3e64e22c93c 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c @@ -1058,10 +1058,9 @@ static void uvd_v6_0_ring_emit_wreg(struct amdgpu_ring *ring, } static void uvd_v6_0_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { - amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA0, 0)); amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST << 2); @@ -1107,8 +1106,7 @@ static void uvd_v6_0_enc_ring_insert_end(struct amdgpu_ring *ring) } static void uvd_v6_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned int vmid, unsigned pasid, - uint64_t pd_addr) + unsigned int vmid, uint64_t pd_addr) { amdgpu_ring_write(ring, HEVC_ENC_CMD_UPDATE_PTB); amdgpu_ring_write(ring, vmid); diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c index bf16440e7258..e54cc3ca2303 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c @@ -1261,13 +1261,12 @@ static void uvd_v7_0_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg, } static void uvd_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub]; uint32_t data0, data1, mask; - pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* wait for reg writes */ data0 = hub->ctx0_ptb_addr_lo32 + vmid * 2; @@ -1302,12 +1301,11 @@ static void uvd_v7_0_enc_ring_emit_reg_wait(struct amdgpu_ring *ring, } static void uvd_v7_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned int vmid, unsigned pasid, - uint64_t pd_addr) + unsigned int vmid, uint64_t pd_addr) { struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub]; - pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* wait for reg writes */ uvd_v7_0_enc_ring_emit_reg_wait(ring, hub->ctx0_ptb_addr_lo32 + vmid * 2, diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c index 6d616015085b..428d1928e44e 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c @@ -844,8 +844,7 @@ static void vce_v3_0_ring_emit_ib(struct amdgpu_ring *ring, } static void vce_v3_0_emit_vm_flush(struct amdgpu_ring *ring, - unsigned int vmid, unsigned pasid, - uint64_t pd_addr) + unsigned int vmid, uint64_t pd_addr) { amdgpu_ring_write(ring, VCE_CMD_UPDATE_PTB); amdgpu_ring_write(ring, vmid); diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c index 22c2067bd849..2329b310ccf2 100755 --- a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c @@ -975,12 +975,11 @@ static void vce_v4_0_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg, } static void vce_v4_0_emit_vm_flush(struct amdgpu_ring *ring, - unsigned int vmid, unsigned pasid, - uint64_t pd_addr) + unsigned int vmid, uint64_t pd_addr) { struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub]; - pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* wait for reg writes */ vce_v4_0_emit_reg_wait(ring, hub->ctx0_ptb_addr_lo32 + vmid * 2, diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c index d9f597c36b63..fdf4ac9313cf 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c @@ -859,13 +859,12 @@ static void vcn_v1_0_dec_ring_emit_reg_wait(struct amdgpu_ring *ring, } static void vcn_v1_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned vmid, unsigned pasid, - uint64_t pd_addr) + unsigned vmid, uint64_t pd_addr) { struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub]; uint32_t data0, data1, mask; - pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* wait for register write */ data0 = hub->ctx0_ptb_addr_lo32 + vmid * 2; @@ -997,12 +996,11 @@ static void vcn_v1_0_enc_ring_emit_reg_wait(struct amdgpu_ring *ring, } static void vcn_v1_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring, - unsigned int vmid, unsigned pasid, - uint64_t pd_addr) + unsigned int vmid, uint64_t pd_addr) { struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub]; - pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pasid, pd_addr); + pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* wait for reg writes */ vcn_v1_0_enc_ring_emit_reg_wait(ring, hub->ctx0_ptb_addr_lo32 + vmid * 2, -- cgit v1.2.3 From 61b100e98f16e02df44862bba7798c7654b565f2 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Tue, 6 Feb 2018 20:32:32 -0500 Subject: drm/amdgpu: Fix header file dependencies MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Felix Kuehling Reviewed-by: Christian König Signed-off-by: Oded Gabbay --- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 1 + 2 files changed, 2 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index 1d0d250cbfdf..06436c3ebd2b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -26,6 +26,7 @@ #include #include +#include /* max number of rings */ #define AMDGPU_MAX_RINGS 18 diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index fabf44b262be..e9841518343e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -28,6 +28,7 @@ #include #include #include +#include #include "amdgpu_sync.h" #include "amdgpu_ring.h" -- cgit v1.2.3 From d8d019ccffb838bb0dd98e583b5c25ccc0bc6ece Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Tue, 6 Feb 2018 20:32:35 -0500 Subject: drm/amdgpu: Add KFD eviction fence This fence is used by KFD to keep memory resident while user mode queues are enabled. Trying to evict memory will trigger the enable_signaling callback, which starts a KFD eviction, which involves preempting user mode queues before signaling the fence. There is one such fence per process. v2: * Grab a reference to mm_struct * Dereference fence after NULL check * Simplify fence release, no need to signal without anyone waiting * Added signed-off-by Harish, who is the original author of this code v3: * update MAINTAINERS file * change amd_kfd_ prefix to amdkfd_ * remove useless initialization of variable to NULL v4: * set amdkfd_fence_ops to be static * Suggested by: Fengguang Wu Signed-off-by: Harish Kasiviswanathan Signed-off-by: Felix Kuehling Reviewed-by: Oded Gabbay Signed-off-by: Oded Gabbay --- MAINTAINERS | 1 + drivers/gpu/drm/amd/amdgpu/Makefile | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 15 ++ drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c | 179 +++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h | 5 +- drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c | 21 ++- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 18 +++ drivers/gpu/drm/amd/include/kgd_kfd_interface.h | 6 + 8 files changed, 242 insertions(+), 4 deletions(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h') diff --git a/MAINTAINERS b/MAINTAINERS index 13c8ec11135a..9e376c7f15b1 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -766,6 +766,7 @@ F: drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c F: drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h F: drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c F: drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +F: drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c F: drivers/gpu/drm/amd/amdkfd/ F: drivers/gpu/drm/amd/include/cik_structs.h F: drivers/gpu/drm/amd/include/kgd_kfd_interface.h diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index 353c937d947d..5dd317579e77 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -129,6 +129,7 @@ amdgpu-y += \ # add amdkfd interfaces amdgpu-y += \ amdgpu_amdkfd.o \ + amdgpu_amdkfd_fence.o \ amdgpu_amdkfd_gfx_v8.o # add cgs diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index 2a519f9062ee..833dc26d402f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -29,6 +29,8 @@ #include #include +extern const struct kgd2kfd_calls *kgd2kfd; + struct amdgpu_device; struct kgd_mem { @@ -37,6 +39,19 @@ struct kgd_mem { void *cpu_ptr; }; +/* KFD Memory Eviction */ +struct amdgpu_amdkfd_fence { + struct dma_fence base; + struct mm_struct *mm; + spinlock_t lock; + char timeline_name[TASK_COMM_LEN]; +}; + +struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context, + struct mm_struct *mm); +bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm); +struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f); + int amdgpu_amdkfd_init(void); void amdgpu_amdkfd_fini(void); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c new file mode 100644 index 000000000000..2c14025e5e76 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c @@ -0,0 +1,179 @@ +/* + * Copyright 2016-2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include +#include +#include +#include +#include +#include +#include "amdgpu_amdkfd.h" + +static const struct dma_fence_ops amdkfd_fence_ops; +static atomic_t fence_seq = ATOMIC_INIT(0); + +/* Eviction Fence + * Fence helper functions to deal with KFD memory eviction. + * Big Idea - Since KFD submissions are done by user queues, a BO cannot be + * evicted unless all the user queues for that process are evicted. + * + * All the BOs in a process share an eviction fence. When process X wants + * to map VRAM memory but TTM can't find enough space, TTM will attempt to + * evict BOs from its LRU list. TTM checks if the BO is valuable to evict + * by calling ttm_bo_driver->eviction_valuable(). + * + * ttm_bo_driver->eviction_valuable() - will return false if the BO belongs + * to process X. Otherwise, it will return true to indicate BO can be + * evicted by TTM. + * + * If ttm_bo_driver->eviction_valuable returns true, then TTM will continue + * the evcition process for that BO by calling ttm_bo_evict --> amdgpu_bo_move + * --> amdgpu_copy_buffer(). This sets up job in GPU scheduler. + * + * GPU Scheduler (amd_sched_main) - sets up a cb (fence_add_callback) to + * nofity when the BO is free to move. fence_add_callback --> enable_signaling + * --> amdgpu_amdkfd_fence.enable_signaling + * + * amdgpu_amdkfd_fence.enable_signaling - Start a work item that will quiesce + * user queues and signal fence. The work item will also start another delayed + * work item to restore BOs + */ + +struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context, + struct mm_struct *mm) +{ + struct amdgpu_amdkfd_fence *fence; + + fence = kzalloc(sizeof(*fence), GFP_KERNEL); + if (fence == NULL) + return NULL; + + /* This reference gets released in amdkfd_fence_release */ + mmgrab(mm); + fence->mm = mm; + get_task_comm(fence->timeline_name, current); + spin_lock_init(&fence->lock); + + dma_fence_init(&fence->base, &amdkfd_fence_ops, &fence->lock, + context, atomic_inc_return(&fence_seq)); + + return fence; +} + +struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f) +{ + struct amdgpu_amdkfd_fence *fence; + + if (!f) + return NULL; + + fence = container_of(f, struct amdgpu_amdkfd_fence, base); + if (fence && f->ops == &amdkfd_fence_ops) + return fence; + + return NULL; +} + +static const char *amdkfd_fence_get_driver_name(struct dma_fence *f) +{ + return "amdgpu_amdkfd_fence"; +} + +static const char *amdkfd_fence_get_timeline_name(struct dma_fence *f) +{ + struct amdgpu_amdkfd_fence *fence = to_amdgpu_amdkfd_fence(f); + + return fence->timeline_name; +} + +/** + * amdkfd_fence_enable_signaling - This gets called when TTM wants to evict + * a KFD BO and schedules a job to move the BO. + * If fence is already signaled return true. + * If fence is not signaled schedule a evict KFD process work item. + */ +static bool amdkfd_fence_enable_signaling(struct dma_fence *f) +{ + struct amdgpu_amdkfd_fence *fence = to_amdgpu_amdkfd_fence(f); + + if (!fence) + return false; + + if (dma_fence_is_signaled(f)) + return true; + + if (!kgd2kfd->schedule_evict_and_restore_process(fence->mm, f)) + return true; + + return false; +} + +/** + * amdkfd_fence_release - callback that fence can be freed + * + * @fence: fence + * + * This function is called when the reference count becomes zero. + * Drops the mm_struct reference and RCU schedules freeing up the fence. + */ +static void amdkfd_fence_release(struct dma_fence *f) +{ + struct amdgpu_amdkfd_fence *fence = to_amdgpu_amdkfd_fence(f); + + /* Unconditionally signal the fence. The process is getting + * terminated. + */ + if (WARN_ON(!fence)) + return; /* Not an amdgpu_amdkfd_fence */ + + mmdrop(fence->mm); + kfree_rcu(f, rcu); +} + +/** + * amdkfd_fence_check_mm - Check if @mm is same as that of the fence @f + * if same return TRUE else return FALSE. + * + * @f: [IN] fence + * @mm: [IN] mm that needs to be verified + */ +bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm) +{ + struct amdgpu_amdkfd_fence *fence = to_amdgpu_amdkfd_fence(f); + + if (!fence) + return false; + else if (fence->mm == mm) + return true; + + return false; +} + +static const struct dma_fence_ops amdkfd_fence_ops = { + .get_driver_name = amdkfd_fence_get_driver_name, + .get_timeline_name = amdkfd_fence_get_timeline_name, + .enable_signaling = amdkfd_fence_enable_signaling, + .signaled = NULL, + .wait = dma_fence_default_wait, + .release = amdkfd_fence_release, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index 06436c3ebd2b..1a5911882657 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -36,8 +36,9 @@ #define AMDGPU_MAX_UVD_ENC_RINGS 2 /* some special values for the owner field */ -#define AMDGPU_FENCE_OWNER_UNDEFINED ((void*)0ul) -#define AMDGPU_FENCE_OWNER_VM ((void*)1ul) +#define AMDGPU_FENCE_OWNER_UNDEFINED ((void *)0ul) +#define AMDGPU_FENCE_OWNER_VM ((void *)1ul) +#define AMDGPU_FENCE_OWNER_KFD ((void *)2ul) #define AMDGPU_FENCE_FLAG_64BIT (1 << 0) #define AMDGPU_FENCE_FLAG_INT (1 << 1) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c index df65c66dc956..b8d3b87fd43e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c @@ -31,6 +31,7 @@ #include #include "amdgpu.h" #include "amdgpu_trace.h" +#include "amdgpu_amdkfd.h" struct amdgpu_sync_entry { struct hlist_node node; @@ -85,11 +86,20 @@ static bool amdgpu_sync_same_dev(struct amdgpu_device *adev, */ static void *amdgpu_sync_get_owner(struct dma_fence *f) { - struct drm_sched_fence *s_fence = to_drm_sched_fence(f); + struct drm_sched_fence *s_fence; + struct amdgpu_amdkfd_fence *kfd_fence; + + if (!f) + return AMDGPU_FENCE_OWNER_UNDEFINED; + s_fence = to_drm_sched_fence(f); if (s_fence) return s_fence->owner; + kfd_fence = to_amdgpu_amdkfd_fence(f); + if (kfd_fence) + return AMDGPU_FENCE_OWNER_KFD; + return AMDGPU_FENCE_OWNER_UNDEFINED; } @@ -204,11 +214,18 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, for (i = 0; i < flist->shared_count; ++i) { f = rcu_dereference_protected(flist->shared[i], reservation_object_held(resv)); + /* We only want to trigger KFD eviction fences on + * evict or move jobs. Skip KFD fences otherwise. + */ + fence_owner = amdgpu_sync_get_owner(f); + if (fence_owner == AMDGPU_FENCE_OWNER_KFD && + owner != AMDGPU_FENCE_OWNER_UNDEFINED) + continue; + if (amdgpu_sync_same_dev(adev, f)) { /* VM updates are only interesting * for other VM updates and moves. */ - fence_owner = amdgpu_sync_get_owner(f); if ((owner != AMDGPU_FENCE_OWNER_UNDEFINED) && (fence_owner != AMDGPU_FENCE_OWNER_UNDEFINED) && ((owner == AMDGPU_FENCE_OWNER_VM) != diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 28c33d711bab..5fcb3488a595 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -46,6 +46,7 @@ #include "amdgpu.h" #include "amdgpu_object.h" #include "amdgpu_trace.h" +#include "amdgpu_amdkfd.h" #include "bif/bif_4_1_d.h" #define DRM_FILE_PAGE_OFFSET (0x100000000ULL >> PAGE_SHIFT) @@ -1171,6 +1172,23 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo, { unsigned long num_pages = bo->mem.num_pages; struct drm_mm_node *node = bo->mem.mm_node; + struct reservation_object_list *flist; + struct dma_fence *f; + int i; + + /* If bo is a KFD BO, check if the bo belongs to the current process. + * If true, then return false as any KFD process needs all its BOs to + * be resident to run successfully + */ + flist = reservation_object_get_list(bo->resv); + if (flist) { + for (i = 0; i < flist->shared_count; ++i) { + f = rcu_dereference_protected(flist->shared[i], + reservation_object_held(bo->resv)); + if (amdkfd_fence_check_mm(f, current->mm)) + return false; + } + } switch (bo->mem.mem_type) { case TTM_PL_TT: diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h index 94eab54864bf..9e352499c409 100644 --- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h @@ -30,6 +30,7 @@ #include #include +#include struct pci_dev; @@ -286,6 +287,9 @@ struct kfd2kgd_calls { * * @resume: Notifies amdkfd about a resume action done to a kgd device * + * @schedule_evict_and_restore_process: Schedules work queue that will prepare + * for safe eviction of KFD BOs that belong to the specified process. + * * This structure contains function callback pointers so the kgd driver * will notify to the amdkfd about certain status changes. * @@ -300,6 +304,8 @@ struct kgd2kfd_calls { void (*interrupt)(struct kfd_dev *kfd, const void *ih_ring_entry); void (*suspend)(struct kfd_dev *kfd); int (*resume)(struct kfd_dev *kfd); + int (*schedule_evict_and_restore_process)(struct mm_struct *mm, + struct dma_fence *fence); }; int kgd2kfd_init(unsigned interface_version, -- cgit v1.2.3