1 files changed, 186 insertions, 92 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
index c48207b377bc..586a58facca1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
@@ -28,13 +28,15 @@
  */
 #include <linux/seq_file.h>
 #include <linux/slab.h>
-#include <drm/drmP.h>
+
 #include <drm/amdgpu_drm.h>
+
 #include "amdgpu.h"
 #include "atom.h"
 #include "amdgpu_trace.h"
 
 #define AMDGPU_IB_TEST_TIMEOUT	msecs_to_jiffies(1000)
+#define AMDGPU_IB_TEST_GFX_XGMI_TIMEOUT	msecs_to_jiffies(2000)
 
 /*
  * IB
@@ -45,13 +47,14 @@
  * produce command buffers which are send to the kernel and
  * put in IBs for execution by the requested ring.
  */
-static int amdgpu_debugfs_sa_init(struct amdgpu_device *adev);
 
 /**
  * amdgpu_ib_get - request an IB (Indirect Buffer)
  *
- * @ring: ring index the IB is associated with
+ * @adev: amdgpu_device pointer
+ * @vm: amdgpu_vm pointer
  * @size: requested IB size
+ * @pool_type: IB pool type (delayed, immediate, direct)
  * @ib: IB object returned
  *
  * Request an IB (all asics).  IBs are allocated using the
@@ -59,19 +62,22 @@ static int amdgpu_debugfs_sa_init(struct amdgpu_device *adev);
  * Returns 0 on success, error on failure.
  */
 int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-		  unsigned size, struct amdgpu_ib *ib)
+		  unsigned int 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,
-				      &ib->sa_bo, size, 256);
+		r = amdgpu_sa_bo_new(&adev->ib_pools[pool_type],
+				     &ib->sa_bo, size);
 		if (r) {
 			dev_err(adev->dev, "failed to get a new IB (%d)\n", r);
 			return r;
 		}
 
 		ib->ptr = amdgpu_sa_bo_cpu_addr(ib->sa_bo);
+		/* flush the cache before commit the IB */
+		ib->flags = AMDGPU_IB_FLAG_EMIT_MEM_SYNC;
 
 		if (!vm)
 			ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo);
@@ -83,24 +89,23 @@ int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 /**
  * amdgpu_ib_free - free an IB (Indirect Buffer)
  *
- * @adev: amdgpu_device pointer
  * @ib: IB object to free
  * @f: the fence SA bo need wait on for the ib alloation
  *
  * Free an IB (all asics).
  */
-void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib,
-		    struct dma_fence *f)
+void amdgpu_ib_free(struct amdgpu_ib *ib, struct dma_fence *f)
 {
-	amdgpu_sa_bo_free(adev, &ib->sa_bo, f);
+	amdgpu_sa_bo_free(&ib->sa_bo, f);
 }
 
 /**
  * amdgpu_ib_schedule - schedule an IB (Indirect Buffer) on the ring
  *
- * @adev: amdgpu_device pointer
+ * @ring: ring index the IB is associated with
  * @num_ibs: number of IBs to schedule
  * @ibs: IB objects to schedule
+ * @job: job to schedule
  * @f: fence created during this submission
  *
  * Schedule an IB on the associated ring (all asics).
@@ -116,23 +121,26 @@ void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib,
  * a CONST_IB), it will be put on the ring prior to the DE IB.  Prior
  * to SI there was just a DE IB.
  */
-int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
+int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 		       struct amdgpu_ib *ibs, struct amdgpu_job *job,
 		       struct dma_fence **f)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_ib *ib = &ibs[0];
 	struct dma_fence *tmp = NULL;
-	bool skip_preamble, need_ctx_switch;
-	unsigned patch_offset = ~0;
+	struct amdgpu_fence *af;
+	bool need_ctx_switch;
 	struct amdgpu_vm *vm;
 	uint64_t fence_ctx;
 	uint32_t status = 0, alloc_size;
-	unsigned fence_flags = 0;
-
-	unsigned i;
-	int r = 0;
+	unsigned int fence_flags = 0;
+	bool secure, init_shadow;
+	u64 shadow_va, csa_va, gds_va;
+	int vmid = AMDGPU_JOB_GET_VMID(job);
 	bool need_pipe_sync = false;
+	unsigned int cond_exec;
+	unsigned int i;
+	int r = 0;
 
 	if (num_ibs == 0)
 		return -EINVAL;
@@ -140,20 +148,49 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 	/* ring tests don't use a job */
 	if (job) {
 		vm = job->vm;
-		fence_ctx = job->base.s_fence->scheduled.context;
+		fence_ctx = job->base.s_fence ?
+			job->base.s_fence->finished.context : 0;
+		shadow_va = job->shadow_va;
+		csa_va = job->csa_va;
+		gds_va = job->gds_va;
+		init_shadow = job->init_shadow;
+		af = job->hw_fence;
+		/* Save the context of the job for reset handling.
+		 * The driver needs this so it can skip the ring
+		 * contents for guilty contexts.
+		 */
+		af->context = fence_ctx;
+		/* the vm fence is also part of the job's context */
+		job->hw_vm_fence->context = fence_ctx;
 	} else {
 		vm = NULL;
 		fence_ctx = 0;
+		shadow_va = 0;
+		csa_va = 0;
+		gds_va = 0;
+		init_shadow = false;
+		af = kzalloc(sizeof(*af), GFP_ATOMIC);
+		if (!af)
+			return -ENOMEM;
 	}
 
 	if (!ring->sched.ready) {
 		dev_err(adev->dev, "couldn't schedule ib on ring <%s>\n", ring->name);
-		return -EINVAL;
+		r = -EINVAL;
+		goto free_fence;
 	}
 
 	if (vm && !job->vmid) {
 		dev_err(adev->dev, "VM IB without ID\n");
-		return -EINVAL;
+		r = -EINVAL;
+		goto free_fence;
+	}
+
+	if ((ib->flags & AMDGPU_IB_FLAGS_SECURE) &&
+	    (!ring->funcs->secure_submission_supported)) {
+		dev_err(adev->dev, "secure submissions not supported on ring <%s>\n", ring->name);
+		r = -EINVAL;
+		goto free_fence;
 	}
 
 	alloc_size = ring->funcs->emit_frame_size + num_ibs *
@@ -162,14 +199,14 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 	r = amdgpu_ring_alloc(ring, alloc_size);
 	if (r) {
 		dev_err(adev->dev, "scheduling IB failed (%d).\n", r);
-		return r;
+		goto free_fence;
 	}
 
 	need_ctx_switch = ring->current_ctx != fence_ctx;
 	if (ring->funcs->emit_pipeline_sync && job &&
-	    ((tmp = amdgpu_sync_get_fence(&job->sched_sync, NULL)) ||
-	     (amdgpu_sriov_vf(adev) && need_ctx_switch) ||
-	     amdgpu_vm_need_pipeline_sync(ring, job))) {
+	    ((tmp = amdgpu_sync_get_fence(&job->explicit_sync)) ||
+	     need_ctx_switch || amdgpu_vm_need_pipeline_sync(ring, job))) {
+
 		need_pipe_sync = true;
 
 		if (tmp)
@@ -178,6 +215,13 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 		dma_fence_put(tmp);
 	}
 
+	if ((ib->flags & AMDGPU_IB_FLAG_EMIT_MEM_SYNC) && ring->funcs->emit_mem_sync)
+		ring->funcs->emit_mem_sync(ring);
+
+	if (ring->funcs->emit_wave_limit &&
+	    ring->hw_prio == AMDGPU_GFX_PIPE_PRIO_HIGH)
+		ring->funcs->emit_wave_limit(ring, true);
+
 	if (ring->funcs->insert_start)
 		ring->funcs->insert_start(ring);
 
@@ -189,49 +233,54 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 		}
 	}
 
-	if (job && ring->funcs->init_cond_exec)
-		patch_offset = amdgpu_ring_init_cond_exec(ring);
+	amdgpu_ring_ib_begin(ring);
 
-#ifdef CONFIG_X86_64
-	if (!(adev->flags & AMD_IS_APU))
-#endif
-	{
-		if (ring->funcs->emit_hdp_flush)
-			amdgpu_ring_emit_hdp_flush(ring);
-		else
-			amdgpu_asic_flush_hdp(adev, ring);
-	}
+	if (ring->funcs->emit_gfx_shadow)
+		amdgpu_ring_emit_gfx_shadow(ring, shadow_va, csa_va, gds_va,
+					    init_shadow, vmid);
+
+	if (ring->funcs->init_cond_exec)
+		cond_exec = amdgpu_ring_init_cond_exec(ring,
+						       ring->cond_exe_gpu_addr);
+
+	amdgpu_device_flush_hdp(adev, ring);
+
+	if (need_ctx_switch)
+		status |= AMDGPU_HAVE_CTX_SWITCH;
 
-	skip_preamble = ring->current_ctx == fence_ctx;
 	if (job && ring->funcs->emit_cntxcntl) {
-		if (need_ctx_switch)
-			status |= AMDGPU_HAVE_CTX_SWITCH;
 		status |= job->preamble_status;
-
+		status |= job->preemption_status;
 		amdgpu_ring_emit_cntxcntl(ring, status);
 	}
 
+	/* Setup initial TMZiness and send it off.
+	 */
+	secure = false;
+	if (job && ring->funcs->emit_frame_cntl) {
+		secure = ib->flags & AMDGPU_IB_FLAGS_SECURE;
+		amdgpu_ring_emit_frame_cntl(ring, true, secure);
+	}
+
 	for (i = 0; i < num_ibs; ++i) {
 		ib = &ibs[i];
 
-		/* drop preamble IBs if we don't have a context switch */
-		if ((ib->flags & AMDGPU_IB_FLAG_PREAMBLE) &&
-			skip_preamble &&
-			!(status & AMDGPU_PREAMBLE_IB_PRESENT_FIRST) &&
-			!amdgpu_sriov_vf(adev)) /* for SRIOV preemption, Preamble CE ib must be inserted anyway */
-			continue;
+		if (job && ring->funcs->emit_frame_cntl) {
+			if (secure != !!(ib->flags & AMDGPU_IB_FLAGS_SECURE)) {
+				amdgpu_ring_emit_frame_cntl(ring, false, secure);
+				secure = !secure;
+				amdgpu_ring_emit_frame_cntl(ring, true, secure);
+			}
+		}
 
-		amdgpu_ring_emit_ib(ring, job, ib, need_ctx_switch);
-		need_ctx_switch = false;
+		amdgpu_ring_emit_ib(ring, job, ib, status);
+		status &= ~AMDGPU_HAVE_CTX_SWITCH;
 	}
 
-	if (ring->funcs->emit_tmz)
-		amdgpu_ring_emit_tmz(ring, false);
+	if (job && ring->funcs->emit_frame_cntl)
+		amdgpu_ring_emit_frame_cntl(ring, false, secure);
 
-#ifdef CONFIG_X86_64
-	if (!(adev->flags & AMD_IS_APU))
-#endif
-		amdgpu_asic_invalidate_hdp(adev, ring);
+	amdgpu_device_invalidate_hdp(adev, ring);
 
 	if (ib->flags & AMDGPU_IB_FLAG_TC_WB_NOT_INVALIDATE)
 		fence_flags |= AMDGPU_FENCE_FLAG_TC_WB_ONLY;
@@ -242,26 +291,54 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 				       fence_flags | AMDGPU_FENCE_FLAG_64BIT);
 	}
 
-	r = amdgpu_fence_emit(ring, f, fence_flags);
+	if (ring->funcs->emit_gfx_shadow && ring->funcs->init_cond_exec) {
+		amdgpu_ring_emit_gfx_shadow(ring, 0, 0, 0, false, 0);
+		amdgpu_ring_init_cond_exec(ring, ring->cond_exe_gpu_addr);
+	}
+
+	r = amdgpu_fence_emit(ring, af, fence_flags);
 	if (r) {
 		dev_err(adev->dev, "failed to emit fence (%d)\n", r);
 		if (job && job->vmid)
-			amdgpu_vmid_reset(adev, ring->funcs->vmhub, job->vmid);
+			amdgpu_vmid_reset(adev, ring->vm_hub, job->vmid);
 		amdgpu_ring_undo(ring);
 		return r;
 	}
+	*f = &af->base;
+	/* get a ref for the job */
+	if (job)
+		dma_fence_get(*f);
 
 	if (ring->funcs->insert_end)
 		ring->funcs->insert_end(ring);
 
-	if (patch_offset != ~0 && ring->funcs->patch_cond_exec)
-		amdgpu_ring_patch_cond_exec(ring, patch_offset);
+	amdgpu_ring_patch_cond_exec(ring, cond_exec);
 
 	ring->current_ctx = fence_ctx;
-	if (vm && ring->funcs->emit_switch_buffer)
+	if (job && ring->funcs->emit_switch_buffer)
 		amdgpu_ring_emit_switch_buffer(ring);
+
+	if (ring->funcs->emit_wave_limit &&
+	    ring->hw_prio == AMDGPU_GFX_PIPE_PRIO_HIGH)
+		ring->funcs->emit_wave_limit(ring, false);
+
+	/* Save the wptr associated with this fence.
+	 * This must be last for resets to work properly
+	 * as we need to save the wptr associated with this
+	 * fence so we know what rings contents to backup
+	 * after we reset the queue.
+	 */
+	amdgpu_fence_save_wptr(af);
+
+	amdgpu_ring_ib_end(ring);
 	amdgpu_ring_commit(ring);
+
 	return 0;
+
+free_fence:
+	if (!job)
+		kfree(af);
+	return r;
 }
 
 /**
@@ -275,24 +352,26 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
  */
 int amdgpu_ib_pool_init(struct amdgpu_device *adev)
 {
-	int r;
+	int r, i;
 
-	if (adev->ib_pool_ready) {
+	if (adev->ib_pool_ready)
 		return 0;
-	}
-	r = amdgpu_sa_bo_manager_init(adev, &adev->ring_tmp_bo,
-				      AMDGPU_IB_POOL_SIZE*64*1024,
-				      AMDGPU_GPU_PAGE_SIZE,
-				      AMDGPU_GEM_DOMAIN_GTT);
-	if (r) {
-		return r;
-	}
 
-	adev->ib_pool_ready = true;
-	if (amdgpu_debugfs_sa_init(adev)) {
-		dev_err(adev->dev, "failed to register debugfs file for SA\n");
+	for (i = 0; i < AMDGPU_IB_POOL_MAX; i++) {
+		r = amdgpu_sa_bo_manager_init(adev, &adev->ib_pools[i],
+					      AMDGPU_IB_POOL_SIZE, 256,
+					      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;
 }
 
 /**
@@ -305,10 +384,14 @@ int amdgpu_ib_pool_init(struct amdgpu_device *adev)
  */
 void amdgpu_ib_pool_fini(struct amdgpu_device *adev)
 {
-	if (adev->ib_pool_ready) {
-		amdgpu_sa_bo_manager_fini(adev, &adev->ring_tmp_bo);
-		adev->ib_pool_ready = false;
-	}
+	int i;
+
+	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;
 }
 
 /**
@@ -323,9 +406,9 @@ 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 int i;
 
 	tmo_mm = tmo_gfx = AMDGPU_IB_TEST_TIMEOUT;
 	if (amdgpu_sriov_vf(adev)) {
@@ -342,8 +425,10 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
 		/* for CP & SDMA engines since they are scheduled together so
 		 * need to make the timeout width enough to cover the time
 		 * cost waiting for it coming back under RUNTIME only
-		*/
+		 */
 		tmo_gfx = 8 * AMDGPU_IB_TEST_TIMEOUT;
+	} else if (adev->gmc.xgmi.hive_id) {
+		tmo_gfx = AMDGPU_IB_TEST_GFX_XGMI_TIMEOUT;
 	}
 
 	for (i = 0; i < adev->num_rings; ++i) {
@@ -356,6 +441,10 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
 		if (!ring->sched.ready || !ring->funcs->test_ib)
 			continue;
 
+		if (adev->enable_mes &&
+		    ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
+			continue;
+
 		/* MM engine need more time */
 		if (ring->funcs->type == AMDGPU_RING_TYPE_UVD ||
 			ring->funcs->type == AMDGPU_RING_TYPE_VCE ||
@@ -395,29 +484,34 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
  */
 #if defined(CONFIG_DEBUG_FS)
 
-static int amdgpu_debugfs_sa_info(struct seq_file *m, void *data)
+static int amdgpu_debugfs_sa_info_show(struct seq_file *m, void *unused)
 {
-	struct drm_info_node *node = (struct drm_info_node *) m->private;
-	struct drm_device *dev = node->minor->dev;
-	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_device *adev = m->private;
 
-	amdgpu_sa_bo_dump_debug_info(&adev->ring_tmp_bo, m);
+	seq_puts(m, "--------------------- DELAYED ---------------------\n");
+	amdgpu_sa_bo_dump_debug_info(&adev->ib_pools[AMDGPU_IB_POOL_DELAYED],
+				     m);
+	seq_puts(m, "-------------------- IMMEDIATE --------------------\n");
+	amdgpu_sa_bo_dump_debug_info(&adev->ib_pools[AMDGPU_IB_POOL_IMMEDIATE],
+				     m);
+	seq_puts(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[] = {
-	{"amdgpu_sa_info", &amdgpu_debugfs_sa_info, 0, NULL},
-};
+DEFINE_SHOW_ATTRIBUTE(amdgpu_debugfs_sa_info);
 
 #endif
 
-static int amdgpu_debugfs_sa_init(struct amdgpu_device *adev)
+void amdgpu_debugfs_sa_init(struct amdgpu_device *adev)
 {
 #if defined(CONFIG_DEBUG_FS)
-	return amdgpu_debugfs_add_files(adev, amdgpu_debugfs_sa_list, 1);
-#else
-	return 0;
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+
+	debugfs_create_file("amdgpu_sa_info", 0444, root, adev,
+			    &amdgpu_debugfs_sa_info_fops);
+
 #endif
 }