drm/amdgpu: forward pasid to backend flush implementations

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 <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/amdgpu.h

@@ -1056,8 +1056,9 @@ struct amdgpu_job {
 	void			*owner;
 	uint64_t		fence_ctx; /* the fence_context this job uses */
 	bool                    vm_needs_flush;
-	unsigned		vmid;
 	uint64_t		vm_pd_addr;
+	unsigned		vmid;
+	unsigned		pasid;
 	uint32_t		gds_base, gds_size;
 	uint32_t		gws_base, gws_size;
 	uint32_t		oa_base, oa_size;
@@ -1790,7 +1791,7 @@ 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, addr) (r)->funcs->emit_vm_flush((r), (vmid), (addr))
+#define amdgpu_ring_emit_vm_flush(r, vmid, pasid, addr) (r)->funcs->emit_vm_flush((r), (vmid), (pasid), (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))

drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c

@@ -237,6 +237,7 @@ static int amdgpu_vmid_grab_reserved_locked(struct amdgpu_vm *vm,
 		id->last_flush = NULL;
 	}
 	job->vmid = id - id_mgr->ids;
+	job->pasid = vm->pasid;
 	trace_amdgpu_vm_grab_id(vm, ring, job);
 out:
 	return r;
@@ -388,6 +389,7 @@ no_flush_needed:
 	list_move_tail(&id->list, &id_mgr->ids_lru);
 
 	job->vmid = id - id_mgr->ids;
+	job->pasid = vm->pasid;
 	trace_amdgpu_vm_grab_id(vm, ring, job);
 
 error:

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,

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c

@@ -513,7 +513,8 @@ 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->vm_pd_addr);
+		amdgpu_ring_emit_vm_flush(ring, job->vmid, job->pasid,
+					  job->vm_pd_addr);
 
 		r = amdgpu_fence_emit(ring, &fence);
 		if (r)

drivers/gpu/drm/amd/amdgpu/cik_sdma.c

@@ -880,7 +880,8 @@ 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, uint64_t pd_addr)
+					unsigned vmid, unsigned pasid,
+					uint64_t pd_addr)
 {
 	u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(0) |
 			  SDMA_POLL_REG_MEM_EXTRA_FUNC(0)); /* always */

drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c

@@ -2354,7 +2354,8 @@ 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, uint64_t pd_addr)
+					unsigned vmid, unsigned pasid,
+					uint64_t pd_addr)
 {
 	int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX);
 

drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c

@@ -3239,7 +3239,8 @@ 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, uint64_t pd_addr)
+					unsigned vmid, unsigned pasid,
+					uint64_t pd_addr)
 {
 	int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX);
 

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c

@@ -6328,7 +6328,8 @@ 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, uint64_t pd_addr)
+					unsigned vmid, unsigned pasid,
+					uint64_t pd_addr)
 {
 	int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX);
 

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c

@@ -3684,7 +3684,8 @@ 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, uint64_t pd_addr)
+					unsigned vmid, unsigned pasid,
+					uint64_t pd_addr)
 {
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX);

drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c

@@ -859,7 +859,8 @@ 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, uint64_t pd_addr)
+					 unsigned vmid, unsigned pasid,
+					 uint64_t pd_addr)
 {
 	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) |
 			  SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf));

drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c

@@ -1125,7 +1125,8 @@ 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, uint64_t pd_addr)
+					 unsigned vmid, unsigned pasid,
+					 uint64_t pd_addr)
 {
 	amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) |
 			  SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf));

drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c

@@ -1133,7 +1133,8 @@ 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, uint64_t pd_addr)
+					 unsigned vmid, unsigned pasid,
+					 uint64_t pd_addr)
 {
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gmc.gmc_funcs->get_invalidate_req(vmid);

drivers/gpu/drm/amd/amdgpu/si_dma.c

@@ -473,7 +473,8 @@ 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, uint64_t pd_addr)
+				      unsigned vmid, unsigned pasid,
+				      uint64_t pd_addr)
 {
 	amdgpu_ring_write(ring, DMA_PACKET(DMA_PACKET_SRBM_WRITE, 0, 0, 0, 0));
 	if (vmid < 8)

drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c

@@ -1073,7 +1073,8 @@ static void uvd_v6_0_enc_ring_emit_ib(struct amdgpu_ring *ring,
 }
 
 static void uvd_v6_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
-					 unsigned vmid, uint64_t pd_addr)
+					unsigned vmid, unsigned pasid,
+					uint64_t pd_addr)
 {
 	uint32_t reg;
 
@@ -1140,7 +1141,8 @@ 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, uint64_t pd_addr)
+					    unsigned int vmid, unsigned pasid,
+					    uint64_t pd_addr)
 {
 	amdgpu_ring_write(ring, HEVC_ENC_CMD_UPDATE_PTB);
 	amdgpu_ring_write(ring, vmid);

drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c

@@ -1291,7 +1291,8 @@ static void uvd_v7_0_vm_reg_wait(struct amdgpu_ring *ring,
 }
 
 static void uvd_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
-					unsigned vmid, uint64_t pd_addr)
+					unsigned vmid, unsigned pasid,
+					uint64_t pd_addr)
 {
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gmc.gmc_funcs->get_invalidate_req(vmid);
@@ -1343,7 +1344,8 @@ static void uvd_v7_0_enc_ring_insert_end(struct amdgpu_ring *ring)
 }
 
 static void uvd_v7_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
-			 unsigned int vmid, uint64_t pd_addr)
+					    unsigned int vmid, unsigned pasid,
+					    uint64_t pd_addr)
 {
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gmc.gmc_funcs->get_invalidate_req(vmid);

drivers/gpu/drm/amd/amdgpu/vce_v3_0.c

@@ -844,7 +844,8 @@ 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, uint64_t pd_addr)
+				   unsigned int vmid, unsigned pasid,
+				   uint64_t pd_addr)
 {
 	amdgpu_ring_write(ring, VCE_CMD_UPDATE_PTB);
 	amdgpu_ring_write(ring, vmid);

drivers/gpu/drm/amd/amdgpu/vce_v4_0.c

@@ -965,7 +965,8 @@ static void vce_v4_0_ring_insert_end(struct amdgpu_ring *ring)
 }
 
 static void vce_v4_0_emit_vm_flush(struct amdgpu_ring *ring,
-			 unsigned int vmid, uint64_t pd_addr)
+				   unsigned int vmid, unsigned pasid,
+				   uint64_t pd_addr)
 {
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gmc.gmc_funcs->get_invalidate_req(vmid);

drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c

@@ -888,7 +888,8 @@ static void vcn_v1_0_dec_vm_reg_wait(struct amdgpu_ring *ring,
 }
 
 static void vcn_v1_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring,
-					unsigned vmid, uint64_t pd_addr)
+					    unsigned vmid, unsigned pasid,
+					    uint64_t pd_addr)
 {
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gmc.gmc_funcs->get_invalidate_req(vmid);
@@ -1021,7 +1022,8 @@ static void vcn_v1_0_enc_ring_emit_ib(struct amdgpu_ring *ring,
 }
 
 static void vcn_v1_0_enc_ring_emit_vm_flush(struct amdgpu_ring *ring,
-			 unsigned int vmid, uint64_t pd_addr)
+					    unsigned int vmid, unsigned pasid,
+					    uint64_t pd_addr)
 {
 	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
 	uint32_t req = ring->adev->gmc.gmc_funcs->get_invalidate_req(vmid);