drm/amdgpu: add xcc index argument to soc15_grbm_select

To support grbm select for multiple XCD case.

v2: unify naming style
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c

@@ -50,12 +50,12 @@ static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
 			uint32_t queue, uint32_t vmid)
 {
 	mutex_lock(&adev->srbm_mutex);
-	soc15_grbm_select(adev, mec, pipe, queue, vmid);
+	soc15_grbm_select(adev, mec, pipe, queue, vmid, 0);
 }
 
 static void unlock_srbm(struct amdgpu_device *adev)
 {
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 }
 
@@ -700,7 +700,7 @@ static void get_wave_count(struct amdgpu_device *adev, int queue_idx,
 	*wave_cnt = 0;
 	pipe_idx = queue_idx / adev->gfx.mec.num_queue_per_pipe;
 	queue_slot = queue_idx % adev->gfx.mec.num_queue_per_pipe;
-	soc15_grbm_select(adev, 1, pipe_idx, queue_slot, 0);
+	soc15_grbm_select(adev, 1, pipe_idx, queue_slot, 0, 0);
 	reg_val = RREG32_SOC15_IP(GC, SOC15_REG_OFFSET(GC, 0, mmSPI_CSQ_WF_ACTIVE_COUNT_0) +
 			 queue_slot);
 	*wave_cnt = reg_val & SPI_CSQ_WF_ACTIVE_COUNT_0__COUNT_MASK;
@@ -772,7 +772,7 @@ void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, int pasid,
 	DECLARE_BITMAP(cp_queue_bitmap, KGD_MAX_QUEUES);
 
 	lock_spi_csq_mutexes(adev);
-	soc15_grbm_select(adev, 1, 0, 0, 0);
+	soc15_grbm_select(adev, 1, 0, 0, 0, 0);
 
 	/*
 	 * Iterate through the shader engines and arrays of the device
@@ -821,7 +821,7 @@ void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, int pasid,
 	}
 
 	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 	unlock_spi_csq_mutexes(adev);
 
 	/* Update the output parameters and return */

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

@@ -1831,7 +1831,7 @@ static void gfx_v9_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
 static void gfx_v9_0_select_me_pipe_q(struct amdgpu_device *adev,
 				  u32 me, u32 pipe, u32 q, u32 vm)
 {
-	soc15_grbm_select(adev, me, pipe, q, vm);
+	soc15_grbm_select(adev, me, pipe, q, vm, 0);
 }
 
 static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = {
@@ -2324,12 +2324,12 @@ static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev)
 
 	mutex_lock(&adev->srbm_mutex);
 	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
-		soc15_grbm_select(adev, 0, 0, 0, i);
+		soc15_grbm_select(adev, 0, 0, 0, i, 0);
 		/* CP and shaders */
 		WREG32_SOC15_RLC(GC, 0, mmSH_MEM_CONFIG, sh_mem_config);
 		WREG32_SOC15_RLC(GC, 0, mmSH_MEM_BASES, sh_mem_bases);
 	}
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 
 	/* Initialize all compute VMIDs to have no GDS, GWS, or OA
@@ -2394,7 +2394,7 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev)
 	/* where to put LDS, scratch, GPUVM in FSA64 space */
 	mutex_lock(&adev->srbm_mutex);
 	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
-		soc15_grbm_select(adev, 0, 0, 0, i);
+		soc15_grbm_select(adev, 0, 0, 0, i, 0);
 		/* CP and shaders */
 		if (i == 0) {
 			tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE,
@@ -2416,7 +2416,7 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev)
 			WREG32_SOC15_RLC(GC, 0, mmSH_MEM_BASES, tmp);
 		}
 	}
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 
 	mutex_unlock(&adev->srbm_mutex);
 
@@ -3540,9 +3540,9 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 		amdgpu_ring_clear_ring(ring);
 
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
 		gfx_v9_0_kiq_init_register(ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 	} else {
 		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
@@ -3551,10 +3551,10 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 		if (amdgpu_sriov_vf(adev) && adev->in_suspend)
 			amdgpu_ring_clear_ring(ring);
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
 		gfx_v9_0_mqd_init(ring);
 		gfx_v9_0_kiq_init_register(ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.kiq[0].mqd_backup)
@@ -3582,9 +3582,9 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring)
 		((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
 		((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
 		gfx_v9_0_mqd_init(ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
@@ -3791,9 +3791,9 @@ static int gfx_v9_0_hw_fini(void *handle)
 		mutex_lock(&adev->srbm_mutex);
 		soc15_grbm_select(adev, adev->gfx.kiq[0].ring.me,
 				adev->gfx.kiq[0].ring.pipe,
-				adev->gfx.kiq[0].ring.queue, 0);
+				adev->gfx.kiq[0].ring.queue, 0, 0);
 		gfx_v9_0_kiq_fini_register(&adev->gfx.kiq[0].ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 	}
 

drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c

@@ -761,7 +761,7 @@ void gfx_v9_4_2_debug_trap_config_init(struct amdgpu_device *adev,
 
 	for (i = first_vmid; i < last_vmid; i++) {
 		data = 0;
-		soc15_grbm_select(adev, 0, 0, 0, i);
+		soc15_grbm_select(adev, 0, 0, 0, i, 0);
 		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
 		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
 		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE,
@@ -769,7 +769,7 @@ void gfx_v9_4_2_debug_trap_config_init(struct amdgpu_device *adev,
 		WREG32(SOC15_REG_OFFSET(GC, 0, regSPI_GDBG_PER_VMID_CNTL), data);
 	}
 
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 }
 

drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c

@@ -658,7 +658,7 @@ static void gfx_v9_4_3_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd
 static void gfx_v9_4_3_select_me_pipe_q(struct amdgpu_device *adev,
 					u32 me, u32 pipe, u32 q, u32 vm)
 {
-	soc15_grbm_select(adev, me, pipe, q, vm);
+	soc15_grbm_select(adev, me, pipe, q, vm, 0);
 }
 
 static const struct amdgpu_gfx_funcs gfx_v9_4_3_gfx_funcs = {
@@ -926,12 +926,12 @@ static void gfx_v9_4_3_init_compute_vmid(struct amdgpu_device *adev)
 
 	mutex_lock(&adev->srbm_mutex);
 	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
-		soc15_grbm_select(adev, 0, 0, 0, i);
+		soc15_grbm_select(adev, 0, 0, 0, i, 0);
 		/* CP and shaders */
 		WREG32_SOC15_RLC(GC, 0, regSH_MEM_CONFIG, sh_mem_config);
 		WREG32_SOC15_RLC(GC, 0, regSH_MEM_BASES, sh_mem_bases);
 	}
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 
 	/* Initialize all compute VMIDs to have no GDS, GWS, or OA
@@ -977,7 +977,7 @@ static void gfx_v9_4_3_constants_init(struct amdgpu_device *adev)
 	/* where to put LDS, scratch, GPUVM in FSA64 space */
 	mutex_lock(&adev->srbm_mutex);
 	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
-		soc15_grbm_select(adev, 0, 0, 0, i);
+		soc15_grbm_select(adev, 0, 0, 0, i, 0);
 		/* CP and shaders */
 		if (i == 0) {
 			tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE,
@@ -999,7 +999,7 @@ static void gfx_v9_4_3_constants_init(struct amdgpu_device *adev)
 			WREG32_SOC15_RLC(GC, 0, regSH_MEM_BASES, tmp);
 		}
 	}
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 
 	mutex_unlock(&adev->srbm_mutex);
 
@@ -1706,19 +1706,19 @@ static int gfx_v9_4_3_kiq_init_queue(struct amdgpu_ring *ring)
 		amdgpu_ring_clear_ring(ring);
 
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
 		gfx_v9_4_3_kiq_init_register(ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 	} else {
 		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
 		((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
 		((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
 		gfx_v9_4_3_mqd_init(ring);
 		gfx_v9_4_3_kiq_init_register(ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.kiq[0].mqd_backup)
@@ -1746,9 +1746,9 @@ static int gfx_v9_4_3_kcq_init_queue(struct amdgpu_ring *ring)
 		((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
 		((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
 		gfx_v9_4_3_mqd_init(ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
@@ -1896,9 +1896,9 @@ static int gfx_v9_4_3_hw_fini(void *handle)
 		mutex_lock(&adev->srbm_mutex);
 		soc15_grbm_select(adev, adev->gfx.kiq[0].ring.me,
 				adev->gfx.kiq[0].ring.pipe,
-				adev->gfx.kiq[0].ring.queue, 0);
+				adev->gfx.kiq[0].ring.queue, 0, 0);
 		gfx_v9_4_3_kiq_fini_register(&adev->gfx.kiq[0].ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 	}
 

drivers/gpu/drm/amd/amdgpu/soc15.c

@@ -311,7 +311,7 @@ static u32 soc15_get_xclk(struct amdgpu_device *adev)
 
 
 void soc15_grbm_select(struct amdgpu_device *adev,
-		     u32 me, u32 pipe, u32 queue, u32 vmid)
+		     u32 me, u32 pipe, u32 queue, u32 vmid, int xcc_id)
 {
 	u32 grbm_gfx_cntl = 0;
 	grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, PIPEID, pipe);
@@ -319,7 +319,7 @@ void soc15_grbm_select(struct amdgpu_device *adev,
 	grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, VMID, vmid);
 	grbm_gfx_cntl = REG_SET_FIELD(grbm_gfx_cntl, GRBM_GFX_CNTL, QUEUEID, queue);
 
-	WREG32_SOC15_RLC_SHADOW(GC, 0, mmGRBM_GFX_CNTL, grbm_gfx_cntl);
+	WREG32_SOC15_RLC_SHADOW(GC, xcc_id, mmGRBM_GFX_CNTL, grbm_gfx_cntl);
 }
 
 static void soc15_vga_set_state(struct amdgpu_device *adev, bool state)

drivers/gpu/drm/amd/amdgpu/soc15.h

@@ -100,7 +100,7 @@ struct soc15_ras_field_entry {
 #define SOC15_RAS_REG_FIELD_VAL(val, entry, field) SOC15_REG_FIELD_VAL((val), (entry).field##_count_mask, (entry).field##_count_shift)
 
 void soc15_grbm_select(struct amdgpu_device *adev,
-		    u32 me, u32 pipe, u32 queue, u32 vmid);
+		    u32 me, u32 pipe, u32 queue, u32 vmid, int xcc_id);
 void soc15_set_virt_ops(struct amdgpu_device *adev);
 
 void soc15_program_register_sequence(struct amdgpu_device *adev,