drm/amdgpu: cleanup IB pool handling a bit

Fix the coding style, move and rename the definitions to
better match what they are supposed to be doing.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

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

@@ -204,8 +204,6 @@ extern int amdgpu_cik_support;
 #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS	        3000
 #define AMDGPU_MAX_USEC_TIMEOUT			100000	/* 100 ms */
 #define AMDGPU_FENCE_JIFFIES_TIMEOUT		(HZ / 2)
-/* AMDGPU_IB_POOL_SIZE must be a power of 2 */
-#define AMDGPU_IB_POOL_SIZE			16
 #define AMDGPU_DEBUGFS_MAX_COMPONENTS		32
 #define AMDGPUFB_CONN_LIMIT			4
 #define AMDGPU_BIOS_NUM_SCRATCH			16
@@ -402,13 +400,6 @@ struct amdgpu_sa_bo {
 int amdgpu_fence_slab_init(void);
 void amdgpu_fence_slab_fini(void);
 
-enum amdgpu_ib_pool_type {
-	AMDGPU_IB_POOL_NORMAL = 0,
-	AMDGPU_IB_POOL_VM,
-	AMDGPU_IB_POOL_DIRECT,
-
-	AMDGPU_IB_POOL_MAX
-};
 /*
  * IRQS.
  */
@@ -866,7 +857,7 @@ struct amdgpu_device {
 	unsigned			num_rings;
 	struct amdgpu_ring		*rings[AMDGPU_MAX_RINGS];
 	bool				ib_pool_ready;
-	struct amdgpu_sa_manager	ring_tmp_bo[AMDGPU_IB_POOL_MAX];
+	struct amdgpu_sa_manager	ib_pools[AMDGPU_IB_POOL_MAX];
 	struct amdgpu_sched		gpu_sched[AMDGPU_HW_IP_NUM][AMDGPU_RING_PRIO_MAX];
 
 	/* interrupts */

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

@@ -924,7 +924,8 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev,
 
 		ring = to_amdgpu_ring(entity->rq->sched);
 		r =  amdgpu_ib_get(adev, vm, ring->funcs->parse_cs ?
-				   chunk_ib->ib_bytes : 0, AMDGPU_IB_POOL_NORMAL, ib);
+				   chunk_ib->ib_bytes : 0,
+				   AMDGPU_IB_POOL_DELAYED, ib);
 		if (r) {
 			DRM_ERROR("Failed to get ib !\n");
 			return r;

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

@@ -61,14 +61,13 @@
  * Returns 0 on success, error on failure.
  */
 int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-		unsigned size,
-		enum amdgpu_ib_pool_type pool_type,
-		struct amdgpu_ib *ib)
+		  unsigned size, enum amdgpu_ib_pool_type pool_type,
+		  struct amdgpu_ib *ib)
 {
 	int r;
 
 	if (size) {
-		r = amdgpu_sa_bo_new(&adev->ring_tmp_bo[pool_type],
+		r = amdgpu_sa_bo_new(&adev->ib_pools[pool_type],
 				      &ib->sa_bo, size, 256);
 		if (r) {
 			dev_err(adev->dev, "failed to get a new IB (%d)\n", r);
@@ -305,30 +304,32 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
  */
 int amdgpu_ib_pool_init(struct amdgpu_device *adev)
 {
-	int r, i;
 	unsigned size;
+	int r, i;
 
-	if (adev->ib_pool_ready) {
+	if (adev->ib_pool_ready)
 		return 0;
-	}
+
 	for (i = 0; i < AMDGPU_IB_POOL_MAX; i++) {
 		if (i == AMDGPU_IB_POOL_DIRECT)
 			size = PAGE_SIZE * 2;
 		else
-			size = AMDGPU_IB_POOL_SIZE*64*1024;
-		r = amdgpu_sa_bo_manager_init(adev, &adev->ring_tmp_bo[i],
-				size,
-				AMDGPU_GPU_PAGE_SIZE,
-				AMDGPU_GEM_DOMAIN_GTT);
-		if (r) {
-			for (i--; i >= 0; i--)
-				amdgpu_sa_bo_manager_fini(adev, &adev->ring_tmp_bo[i]);
-			return r;
-		}
+			size = AMDGPU_IB_POOL_SIZE;
+
+		r = amdgpu_sa_bo_manager_init(adev, &adev->ib_pools[i],
+					      size, AMDGPU_GPU_PAGE_SIZE,
+					      AMDGPU_GEM_DOMAIN_GTT);
+		if (r)
+			goto error;
 	}
 	adev->ib_pool_ready = true;
 
 	return 0;
+
+error:
+	while (i--)
+		amdgpu_sa_bo_manager_fini(adev, &adev->ib_pools[i]);
+	return r;
 }
 
 /**
@@ -343,11 +344,12 @@ void amdgpu_ib_pool_fini(struct amdgpu_device *adev)
 {
 	int i;
 
-	if (adev->ib_pool_ready) {
-		for (i = 0; i < AMDGPU_IB_POOL_MAX; i++)
-			amdgpu_sa_bo_manager_fini(adev, &adev->ring_tmp_bo[i]);
-		adev->ib_pool_ready = false;
-	}
+	if (!adev->ib_pool_ready)
+		return;
+
+	for (i = 0; i < AMDGPU_IB_POOL_MAX; i++)
+		amdgpu_sa_bo_manager_fini(adev, &adev->ib_pools[i]);
+	adev->ib_pool_ready = false;
 }
 
 /**
@@ -362,9 +364,9 @@ void amdgpu_ib_pool_fini(struct amdgpu_device *adev)
  */
 int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
 {
-	unsigned i;
-	int r, ret = 0;
 	long tmo_gfx, tmo_mm;
+	int r, ret = 0;
+	unsigned i;
 
 	tmo_mm = tmo_gfx = AMDGPU_IB_TEST_TIMEOUT;
 	if (amdgpu_sriov_vf(adev)) {
@@ -442,15 +444,16 @@ static int amdgpu_debugfs_sa_info(struct seq_file *m, void *data)
 	struct drm_device *dev = node->minor->dev;
 	struct amdgpu_device *adev = dev->dev_private;
 
-	seq_printf(m, "-------------------- NORMAL -------------------- \n");
-	amdgpu_sa_bo_dump_debug_info(&adev->ring_tmp_bo[AMDGPU_IB_POOL_NORMAL], m);
-	seq_printf(m, "---------------------- VM ---------------------- \n");
-	amdgpu_sa_bo_dump_debug_info(&adev->ring_tmp_bo[AMDGPU_IB_POOL_VM], m);
-	seq_printf(m, "-------------------- DIRECT--------------------- \n");
-	amdgpu_sa_bo_dump_debug_info(&adev->ring_tmp_bo[AMDGPU_IB_POOL_DIRECT], m);
+	seq_printf(m, "--------------------- DELAYED --------------------- \n");
+	amdgpu_sa_bo_dump_debug_info(&adev->ib_pools[AMDGPU_IB_POOL_DELAYED],
+				     m);
+	seq_printf(m, "-------------------- IMMEDIATE -------------------- \n");
+	amdgpu_sa_bo_dump_debug_info(&adev->ib_pools[AMDGPU_IB_POOL_IMMEDIATE],
+				     m);
+	seq_printf(m, "--------------------- DIRECT ---------------------- \n");
+	amdgpu_sa_bo_dump_debug_info(&adev->ib_pools[AMDGPU_IB_POOL_DIRECT], m);
 
 	return 0;
-
 }
 
 static const struct drm_info_list amdgpu_debugfs_sa_list[] = {

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

@@ -50,6 +50,8 @@
 
 #define to_amdgpu_ring(s) container_of((s), struct amdgpu_ring, sched)
 
+#define AMDGPU_IB_POOL_SIZE	(1024 * 1024)
+
 enum amdgpu_ring_type {
 	AMDGPU_RING_TYPE_GFX		= AMDGPU_HW_IP_GFX,
 	AMDGPU_RING_TYPE_COMPUTE	= AMDGPU_HW_IP_COMPUTE,
@@ -63,6 +65,17 @@ enum amdgpu_ring_type {
 	AMDGPU_RING_TYPE_KIQ
 };
 
+enum amdgpu_ib_pool_type {
+	/* Normal submissions to the top of the pipeline. */
+	AMDGPU_IB_POOL_DELAYED,
+	/* Immediate submissions to the bottom of the pipeline. */
+	AMDGPU_IB_POOL_IMMEDIATE,
+	/* Direct submission to the ring buffer during init and reset. */
+	AMDGPU_IB_POOL_DIRECT,
+
+	AMDGPU_IB_POOL_MAX
+};
+
 struct amdgpu_device;
 struct amdgpu_ring;
 struct amdgpu_ib;

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

@@ -44,7 +44,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev)
 	/* Number of tests =
 	 * (Total GTT - IB pool - writeback page - ring buffers) / test size
 	 */
-	n = adev->gmc.gart_size - AMDGPU_IB_POOL_SIZE*64*1024;
+	n = adev->gmc.gart_size - AMDGPU_IB_POOL_SIZE;
 	for (i = 0; i < AMDGPU_MAX_RINGS; ++i)
 		if (adev->rings[i])
 			n -= adev->rings[i]->ring_size;

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

@@ -333,7 +333,7 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo,
 	num_bytes = num_pages * 8;
 
 	r = amdgpu_job_alloc_with_ib(adev, num_dw * 4 + num_bytes,
-				     AMDGPU_IB_POOL_NORMAL, &job);
+				     AMDGPU_IB_POOL_DELAYED, &job);
 	if (r)
 		return r;
 
@@ -2122,6 +2122,8 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 		       struct dma_fence **fence, bool direct_submit,
 		       bool vm_needs_flush, bool tmz)
 {
+	enum amdgpu_ib_pool_type pool = direct_submit ? AMDGPU_IB_POOL_DIRECT :
+		AMDGPU_IB_POOL_DELAYED;
 	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_job *job;
 
@@ -2139,8 +2141,7 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 	num_loops = DIV_ROUND_UP(byte_count, max_bytes);
 	num_dw = ALIGN(num_loops * adev->mman.buffer_funcs->copy_num_dw, 8);
 
-	r = amdgpu_job_alloc_with_ib(adev, num_dw * 4,
-			direct_submit ? AMDGPU_IB_POOL_DIRECT : AMDGPU_IB_POOL_NORMAL, &job);
+	r = amdgpu_job_alloc_with_ib(adev, num_dw * 4, pool, &job);
 	if (r)
 		return r;
 
@@ -2229,7 +2230,8 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 	/* for IB padding */
 	num_dw += 64;
 
-	r = amdgpu_job_alloc_with_ib(adev, num_dw * 4, AMDGPU_IB_POOL_NORMAL, &job);
+	r = amdgpu_job_alloc_with_ib(adev, num_dw * 4, AMDGPU_IB_POOL_DELAYED,
+				     &job);
 	if (r)
 		return r;
 

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

@@ -1056,8 +1056,8 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo,
 			goto err;
 	}
 
-	r = amdgpu_job_alloc_with_ib(adev, 64,
-			direct ? AMDGPU_IB_POOL_DIRECT : AMDGPU_IB_POOL_NORMAL, &job);
+	r = amdgpu_job_alloc_with_ib(adev, 64, direct ? AMDGPU_IB_POOL_DIRECT :
+				     AMDGPU_IB_POOL_DELAYED, &job);
 	if (r)
 		goto err;
 

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

@@ -447,7 +447,7 @@ static int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
 	int i, r;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4,
-					AMDGPU_IB_POOL_DIRECT, &job);
+				     AMDGPU_IB_POOL_DIRECT, &job);
 	if (r)
 		return r;
 
@@ -526,7 +526,8 @@ static int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
 	int i, r;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4,
-			direct ? AMDGPU_IB_POOL_DIRECT : AMDGPU_IB_POOL_NORMAL, &job);
+				     direct ? AMDGPU_IB_POOL_DIRECT :
+				     AMDGPU_IB_POOL_DELAYED, &job);
 	if (r)
 		return r;
 

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

@@ -61,11 +61,12 @@ static int amdgpu_vm_sdma_prepare(struct amdgpu_vm_update_params *p,
 				  struct dma_resv *resv,
 				  enum amdgpu_sync_mode sync_mode)
 {
+	enum amdgpu_ib_pool_type pool = p->direct ? AMDGPU_IB_POOL_IMMEDIATE :
+		AMDGPU_IB_POOL_DELAYED;
 	unsigned int ndw = AMDGPU_VM_SDMA_MIN_NUM_DW;
 	int r;
 
-	r = amdgpu_job_alloc_with_ib(p->adev, ndw * 4,
-			p->direct ? AMDGPU_IB_POOL_VM : AMDGPU_IB_POOL_NORMAL, &p->job);
+	r = amdgpu_job_alloc_with_ib(p->adev, ndw * 4, pool, &p->job);
 	if (r)
 		return r;
 
@@ -199,6 +200,8 @@ static int amdgpu_vm_sdma_update(struct amdgpu_vm_update_params *p,
 				 uint64_t addr, unsigned count, uint32_t incr,
 				 uint64_t flags)
 {
+	enum amdgpu_ib_pool_type pool = p->direct ? AMDGPU_IB_POOL_IMMEDIATE :
+		AMDGPU_IB_POOL_DELAYED;
 	unsigned int i, ndw, nptes;
 	uint64_t *pte;
 	int r;
@@ -224,8 +227,8 @@ static int amdgpu_vm_sdma_update(struct amdgpu_vm_update_params *p,
 			ndw = max(ndw, AMDGPU_VM_SDMA_MIN_NUM_DW);
 			ndw = min(ndw, AMDGPU_VM_SDMA_MAX_NUM_DW);
 
-			r = amdgpu_job_alloc_with_ib(p->adev, ndw * 4,
-					p->direct ? AMDGPU_IB_POOL_VM : AMDGPU_IB_POOL_NORMAL, &p->job);
+			r = amdgpu_job_alloc_with_ib(p->adev, ndw * 4, pool,
+						     &p->job);
 			if (r)
 				return r;
 

drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c

@@ -372,7 +372,8 @@ static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 	 * translation. Avoid this by doing the invalidation from the SDMA
 	 * itself.
 	 */
-	r = amdgpu_job_alloc_with_ib(adev, 16 * 4, AMDGPU_IB_POOL_VM, &job);
+	r = amdgpu_job_alloc_with_ib(adev, 16 * 4, AMDGPU_IB_POOL_IMMEDIATE,
+				     &job);
 	if (r)
 		goto error_alloc;