Skip to contentSkip to conversation

*** Dedicated vmid per process v5 ***

PATCH In review Chunming Zhou 2 participants 8 messages Apr 27, 2017 1 message not in the archive

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v5 cover *** Dedicated vmid per process v5 ***

The current kernel implementation, which grabs the idle VMID from pool when emitting the job may:

The back-to-back submission from one process could use different VMID. The submission to different queues from single process could use different VMID

It works well in most case but cannot work for the SQ thread trace capture.

The VMID for the submission that set the {SQTT}_BASE, which refers to the address of the trace buffer, is stored in shader engine.

If the profiling application have to use different VMIDs to submit IBs in its life cycle:

Some trace is not captured since it actually uses different VMID to submit jobs.
Some part of captured trace may come from different application since they are accidentally uses the owner’s VMID to submit jobs.

V2:
1. address Christian's comments:
a. drop context flags for tag process, instead, add vm ioctl. b. change order of patches.
c. sync waiting only when vm flush needs.

2. address Alex's comments;
bump module version

V3:
address Jerry and Christian's comments.
and only reserve gfxhub vmid

v4:
address Jerry and Christian's comments.
fix some race condistions.

v5:
patch#5: peek_fence instead of get_fence.
fix potential context starved.

Chunming Zhou (6):
drm/amdgpu: add vm ioctl
drm/amdgpu: add reserved vmid field in vm struct v2 drm/amdgpu: reserve/unreserve vmid by vm ioctl v4 drm/amdgpu: add limitation for dedicated vm number v4 drm/amdgpu: implement grab reserved vmid V4 drm/amdgpu: bump module verion for reserved vmid

 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c |   3 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c |   1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c  | 159 +++++++++++++++++++++++++++++++-
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h  |   6 ++
 include/uapi/drm/amdgpu_drm.h           |  22 +++++
 5 files changed, 185 insertions(+), 6 deletions(-)
Signature · 6 lines
1.9.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 1493298597-28517-1-git-send-email-David1.Zhou-5C7GfCeVMHo@public.gmane.org

#2 lore raw

PATCH v5 1/6 drm/amdgpu: add vm ioctl

It will be used for reserving vmid.

Change-Id Ib7169ea999690c8e82d0dcbccdd2d97760c0270a
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c |  1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c  | 18 ++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h  |  1 +
 include/uapi/drm/amdgpu_drm.h           | 22 ++++++++++++++++++++++
 4 files changed, 42 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +1 -0
index 052e6a5..8ea33c8 100644
@@ -1062,6 +1062,7 @@ int amdgpu_get_vblank_timestamp_kms(struct drm_device *dev, unsigned int pipe,
 const struct drm_ioctl_desc amdgpu_ioctls_kms[] = {
 	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_CREATE, amdgpu_gem_create_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_CTX, amdgpu_ctx_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_VM, amdgpu_vm_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_BO_LIST, amdgpu_bo_list_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	/* KMS */
 	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_MMAP, amdgpu_gem_mmap_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +18 -0
index a8cd23b..8698177 100644
@@ -2379,3 +2379,21 @@ void amdgpu_vm_manager_fini(struct amdgpu_device *adev)
 		}
 	}
 }
+
+int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
+{
+	union drm_amdgpu_vm *args = data;
+	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+
+	switch (args->in.op) {
+	case AMDGPU_VM_OP_RESERVE_VMID:
+	case AMDGPU_VM_OP_UNRESERVE_VMID:
+		return -EINVAL;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +1 -0
index 8eba2d3..16e0aaa 100644
@@ -248,5 +248,6 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
 void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
 		      struct amdgpu_bo_va *bo_va);
 void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size);
+int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp);
 
 #endif
include/uapi/drm/amdgpu_drm.h +22 -0
index b98b371..1c95775 100644
@@ -51,6 +51,7 @@
 #define DRM_AMDGPU_GEM_OP		0x10
 #define DRM_AMDGPU_GEM_USERPTR		0x11
 #define DRM_AMDGPU_WAIT_FENCES		0x12
+#define DRM_AMDGPU_VM			0x13
 
 /* hybrid specific ioctls */
 #define DRM_AMDGPU_SEM			0x5b
@@ -71,6 +72,7 @@
 #define DRM_IOCTL_AMDGPU_GEM_OP		DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_OP, struct drm_amdgpu_gem_op)
 #define DRM_IOCTL_AMDGPU_GEM_USERPTR	DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_USERPTR, struct drm_amdgpu_gem_userptr)
 #define DRM_IOCTL_AMDGPU_WAIT_FENCES	DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_WAIT_FENCES, union drm_amdgpu_wait_fences)
+#define DRM_IOCTL_AMDGPU_VM		DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_VM, union drm_amdgpu_vm)
 
 /* hybrid specific ioctls */
 #define DRM_IOCTL_AMDGPU_GEM_DGMA	DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_DGMA, struct drm_amdgpu_gem_dgma)
@@ -212,6 +214,26 @@ struct drm_amdgpu_ctx_in {
 	union drm_amdgpu_ctx_out out;
 };
 
+/* vm ioctl */
+#define AMDGPU_VM_OP_RESERVE_VMID	1
+#define AMDGPU_VM_OP_UNRESERVE_VMID	2
+
+struct drm_amdgpu_vm_in {
+	/** AMDGPU_VM_OP_* */
+	__u32	op;
+	__u32	flags;
+};
+
+struct drm_amdgpu_vm_out {
+	/** For future use, no flags defined so far */
+	__u64	flags;
+};
+
+union drm_amdgpu_vm {
+	struct drm_amdgpu_vm_in in;
+	struct drm_amdgpu_vm_out out;
+};
+
 /* sem related */
 #define AMDGPU_SEM_OP_CREATE_SEM        1
 #define AMDGPU_SEM_OP_WAIT_SEM	        2
Signature · 6 lines
1.9.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#3 lore raw

PATCH v5 2/6 drm/amdgpu: add reserved vmid field in vm struct v2

In reply to a message not in the archive

v2: rename dedicated_vmid to reserved_vmid

Change-Id Id728e20366c8a1ae90d4e901dc80e136e2a613bb
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 17 ++++++++++++++++-
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h |  2 ++
 2 files changed, 18 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +16 -1
index 8698177..261aaea 100644
@@ -2163,10 +2163,12 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	unsigned ring_instance;
 	struct amdgpu_ring *ring;
 	struct amd_sched_rq *rq;
-	int r;
+	int r, i;
 
 	vm->va = RB_ROOT;
 	vm->client_id = atomic64_inc_return(&adev->vm_manager.client_counter);
+	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
+		vm->reserved_vmid[i] = NULL;
 	spin_lock_init(&vm->status_lock);
 	INIT_LIST_HEAD(&vm->invalidated);
 	INIT_LIST_HEAD(&vm->cleared);
@@ -2270,6 +2272,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 {
 	struct amdgpu_bo_va_mapping *mapping, *tmp;
 	bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt;
+	int i;
 
 	if (vm->is_kfd_vm) {
 		struct amdgpu_vm_id_manager *id_mgr =
@@ -2313,6 +2316,18 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 
 	amdgpu_vm_free_levels(&vm->root);
 	fence_put(vm->last_dir_update);
+	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++) {
+		struct amdgpu_vm_id_manager *id_mgr =
+			&adev->vm_manager.id_mgr[i];
+
+		mutex_lock(&id_mgr->lock);
+		if (vm->reserved_vmid[i]) {
+			list_add(&vm->reserved_vmid[i]->list,
+				 &id_mgr->ids_lru);
+			vm->reserved_vmid[i] = NULL;
+		}
+		mutex_unlock(&id_mgr->lock);
+	}
 }
 
 /**
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +2 -0
index 16e0aaa..8eedca1 100644
@@ -123,6 +123,8 @@ struct amdgpu_vm {
 
 	/* client id */
 	u64                     client_id;
+	/* dedicated to vm */
+	struct amdgpu_vm_id	*reserved_vmid[AMDGPU_MAX_VMHUBS];
 	/* each VM will map on CSA */
 	struct amdgpu_bo_va *csa_bo_va;
 
Signature · 6 lines
1.9.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#4 lore raw

PATCH v5 3/6 drm/amdgpu: reserve/unreserve vmid by vm ioctl v4

In reply to a message not in the archive

add reserve/unreserve vmid funtions.
v3:
only reserve vmid from gfxhub
v4:
fix racy condition

Change-Id I5f80dc39dc9d44660a96a2b710b0dbb4d3b9039d
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 62 +++++++++++++++++++++++++++-------
 1 file changed, 49 insertions(+), 13 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +49 -13
index 261aaea..e37421e 100644
@@ -546,6 +546,45 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 	return r;
 }
 
+static void amdgpu_vm_free_reserved_vmid(struct amdgpu_device *adev,
+					  struct amdgpu_vm *vm,
+					  unsigned vmhub)
+{
+	struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[vmhub];
+
+	mutex_lock(&id_mgr->lock);
+	if (vm->reserved_vmid[vmhub]) {
+		list_add(&vm->reserved_vmid[vmhub]->list,
+			&id_mgr->ids_lru);
+		vm->reserved_vmid[vmhub] = NULL;
+	}
+	mutex_unlock(&id_mgr->lock);
+}
+
+static int amdgpu_vm_alloc_reserved_vmid(struct amdgpu_device *adev,
+					 struct amdgpu_vm *vm,
+					 unsigned vmhub)
+{
+	struct amdgpu_vm_id_manager *id_mgr;
+	struct amdgpu_vm_id *idle;
+	int r = 0;
+
+	id_mgr = &adev->vm_manager.id_mgr[vmhub];
+	mutex_lock(&id_mgr->lock);
+	if (vm->reserved_vmid[vmhub])
+		goto unlock;
+	/* Select the first entry VMID */
+	idle = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vm_id, list);
+	list_del_init(&idle->list);
+	vm->reserved_vmid[vmhub] = idle;
+	mutex_unlock(&id_mgr->lock);
+
+	return 0;
+unlock:
+	mutex_unlock(&id_mgr->lock);
+	return r;
+}
+
 static bool amdgpu_vm_ring_has_compute_vm_bug(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
@@ -2316,18 +2355,8 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 
 	amdgpu_vm_free_levels(&vm->root);
 	fence_put(vm->last_dir_update);
-	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++) {
-		struct amdgpu_vm_id_manager *id_mgr =
-			&adev->vm_manager.id_mgr[i];
-
-		mutex_lock(&id_mgr->lock);
-		if (vm->reserved_vmid[i]) {
-			list_add(&vm->reserved_vmid[i]->list,
-				 &id_mgr->ids_lru);
-			vm->reserved_vmid[i] = NULL;
-		}
-		mutex_unlock(&id_mgr->lock);
-	}
+	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
+		amdgpu_vm_free_reserved_vmid(adev, vm, i);
 }
 
 /**
@@ -2400,11 +2429,18 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	union drm_amdgpu_vm *args = data;
 	struct amdgpu_device *adev = dev->dev_private;
 	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	int r;
 
 	switch (args->in.op) {
 	case AMDGPU_VM_OP_RESERVE_VMID:
+		/* current, we only have requirement to reserve vmid from gfxhub */
+		r = amdgpu_vm_alloc_reserved_vmid(adev, &fpriv->vm,
+						  AMDGPU_GFXHUB);
+		if (r)
+			return r;
+		break;
 	case AMDGPU_VM_OP_UNRESERVE_VMID:
-		return -EINVAL;
+		amdgpu_vm_free_reserved_vmid(adev, &fpriv->vm, AMDGPU_GFXHUB);
 		break;
 	default:
 		return -EINVAL;
Signature · 6 lines
1.9.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#5 lore raw

PATCH v5 4/6 drm/amdgpu: add limitation for dedicated vm number v4

In reply to a message not in the archive

v2: move #define to amdgpu_vm.h
v3: move reserved vmid counter to id_manager, and increase counter before allocating vmid
v4: rename to reserved_vmid_num

Change-Id Ie5958cf6dbdc1c8278e61d9158483472d6f5c6e3
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 9 +++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 3 +++
 2 files changed, 12 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +9 -0
index e37421e..e6fdfa4 100644
@@ -557,6 +557,7 @@ static void amdgpu_vm_free_reserved_vmid(struct amdgpu_device *adev,
 		list_add(&vm->reserved_vmid[vmhub]->list,
 			&id_mgr->ids_lru);
 		vm->reserved_vmid[vmhub] = NULL;
+		atomic_dec(&id_mgr->reserved_vmid_num);
 	}
 	mutex_unlock(&id_mgr->lock);
 }
@@ -573,6 +574,13 @@ static int amdgpu_vm_alloc_reserved_vmid(struct amdgpu_device *adev,
 	mutex_lock(&id_mgr->lock);
 	if (vm->reserved_vmid[vmhub])
 		goto unlock;
+	if (atomic_inc_return(&id_mgr->reserved_vmid_num) >
+	    AMDGPU_VM_MAX_RESERVED_VMID) {
+		DRM_ERROR("Over limitation of reserved vmid\n");
+		atomic_dec(&id_mgr->reserved_vmid_num);
+		r = -EINVAL;
+		goto unlock;
+	}
 	/* Select the first entry VMID */
 	idle = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vm_id, list);
 	list_del_init(&idle->list);
@@ -2376,6 +2384,7 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev)
 
 		mutex_init(&id_mgr->lock);
 		INIT_LIST_HEAD(&id_mgr->ids_lru);
+		atomic_set(&id_mgr->reserved_vmid_num, 0);
 
 		/* skip over VMID 0, since it is the system VM */
 		for (j = 1; j < id_mgr->num_ids; ++j) {
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +3 -0
index 8eedca1..9828fcd 100644
@@ -84,6 +84,8 @@
 
 /* hardcode that limit for now */
 #define AMDGPU_VA_RESERVED_SIZE			(8 << 20)
+/* max vmids dedicated for process */
+#define AMDGPU_VM_MAX_RESERVED_VMID	1
 
 struct amdgpu_vm_pt {
 	struct amdgpu_bo	*bo;
@@ -157,6 +159,7 @@ struct amdgpu_vm_id_manager {
 	unsigned		num_ids;
 	struct list_head	ids_lru;
 	struct amdgpu_vm_id	ids[AMDGPU_NUM_VM];
+	atomic_t		reserved_vmid_num;
 };
 
 struct amdgpu_vm_manager {
Signature · 6 lines
1.9.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#6 lore raw

PATCH v5 5/6 drm/amdgpu: implement grab reserved vmid V4

In reply to a message not in the archive

v2: move sync waiting only when flush needs
v3: fix racy
v4: peek fence instead of get fence, and fix potential context starved.

Change-Id I64da2701c9fdcf986afb90ba1492a78d5bef1b6c
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 79 ++++++++++++++++++++++++++++++++--
 1 file changed, 75 insertions(+), 4 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +75 -4
index e6fdfa4..b35b853 100644
@@ -397,6 +397,72 @@ static bool amdgpu_vm_had_gpu_reset(struct amdgpu_device *adev,
 		atomic_read(&adev->gpu_reset_counter);
 }
 
+static bool amdgpu_vm_reserved_vmid_ready(struct amdgpu_vm *vm, unsigned vmhub)
+{
+	return !!vm->reserved_vmid[vmhub];
+}
+
+/* idr_mgr->lock must be held */
+static int amdgpu_vm_grab_reserved_vmid_locked(struct amdgpu_vm *vm,
+					       struct amdgpu_ring *ring,
+					       struct amdgpu_sync *sync,
+					       struct fence *fence,
+					       struct amdgpu_job *job)
+{
+	struct amdgpu_device *adev = ring->adev;
+	unsigned vmhub = ring->funcs->vmhub;
+	uint64_t fence_context = adev->fence_context + ring->idx;
+	struct amdgpu_vm_id *id = vm->reserved_vmid[vmhub];
+	struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[vmhub];
+	struct fence *updates = sync->last_vm_update;
+	int r = 0;
+	struct fence *flushed, *tmp;
+	bool needs_flush = false;
+
+	flushed  = id->flushed_updates;
+	if ((amdgpu_vm_had_gpu_reset(adev, id)) ||
+	    (atomic64_read(&id->owner) != vm->client_id) ||
+	    (job->vm_pd_addr != id->pd_gpu_addr) ||
+	    (updates && (!flushed || updates->context != flushed->context ||
+			fence_is_later(updates, flushed))) ||
+	    (!id->last_flush || (id->last_flush->context != fence_context &&
+				 !fence_is_signaled(id->last_flush)))) {
+		needs_flush = true;
+		/* to prevent one context starved by another context */
+		id->pd_gpu_addr = 0;
+		tmp = amdgpu_sync_peek_fence(&id->active, ring);
+		if (tmp) {
+			r = amdgpu_sync_fence(adev, sync, tmp);
+			return r;
+		}
+	}
+
+	/* Good we can use this VMID. Remember this submission as
+	* user of the VMID.
+	*/
+	r = amdgpu_sync_fence(ring->adev, &id->active, fence);
+	if (r)
+		goto out;
+
+	if (updates && (!flushed || updates->context != flushed->context ||
+			fence_is_later(updates, flushed))) {
+		fence_put(id->flushed_updates);
+		id->flushed_updates = fence_get(updates);
+	}
+	id->pd_gpu_addr = job->vm_pd_addr;
+	id->current_gpu_reset_count = atomic_read(&adev->gpu_reset_counter);
+	atomic64_set(&id->owner, vm->client_id);
+	job->vm_needs_flush = needs_flush;
+	if (needs_flush) {
+		fence_put(id->last_flush);
+		id->last_flush = NULL;
+	}
+	job->vm_id = id - id_mgr->ids;
+	trace_amdgpu_vm_grab_id(vm, ring, job);
+out:
+	return r;
+}
+
 /**
  * amdgpu_vm_grab_id - allocate the next free VMID
  *
@@ -421,12 +487,17 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 	unsigned i;
 	int r = 0;
 
+	mutex_lock(&id_mgr->lock);
+	if (amdgpu_vm_reserved_vmid_ready(vm, vmhub)) {
+		r = amdgpu_vm_grab_reserved_vmid_locked(vm, ring, sync, fence, job);
+		mutex_unlock(&id_mgr->lock);
+		return r;
+	}
 	fences = kmalloc_array(sizeof(void *), id_mgr->num_ids, GFP_KERNEL);
-	if (!fences)
+	if (!fences) {
+		mutex_unlock(&id_mgr->lock);
 		return -ENOMEM;
-
-	mutex_lock(&id_mgr->lock);
-
+	}
 	/* Check if we have an idle VMID */
 	i = 0;
 	list_for_each_entry(idle, &id_mgr->ids_lru, list) {
Signature · 6 lines
1.9.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#7 lore raw

PATCH v5 6/6 drm/amdgpu: bump module verion for reserved vmid

In reply to a message not in the archive
Change-Id I1065e0430ed44f7ee6c29214b72e35a7343ea02b
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +2 -1
index 55322b4..6799829 100644
@@ -64,9 +64,10 @@
  * - 3.12.0 - Add query for double offchip LDS buffers
  * - 3.13.0 - Add PRT support
  * - 3.14.0 - Fix race in amdgpu_ctx_get_fence() and note new functionality
+ * - 3.15.0 - Add reserved vmid support
  */
 #define KMS_DRIVER_MAJOR	3
-#define KMS_DRIVER_MINOR	14
+#define KMS_DRIVER_MINOR	15
 #define KMS_DRIVER_PATCHLEVEL	0
 
 int amdgpu_vram_limit = 0;
Signature · 6 lines
1.9.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Reviewed-by
#8 lore raw
In reply to a message not in the archive

Am 27.04.2017 um 15:09 schrieb Chunming Zhou:

Quoted 43 lines

The current kernel implementation, which grabs the idle VMID from pool when emitting the job may:

The back-to-back submission from one process could use different VMID. The submission to different queues from single process could use different VMID

It works well in most case but cannot work for the SQ thread trace capture.

The VMID for the submission that set the {SQTT}_BASE, which refers to the address of the trace buffer, is stored in shader engine.

If the profiling application have to use different VMIDs to submit IBs in its life cycle:

Some trace is not captured since it actually uses different VMID to submit jobs.
Some part of captured trace may come from different application since they are accidentally uses the owner’s VMID to submit jobs.

V2:
1. address Christian's comments:
a. drop context flags for tag process, instead, add vm ioctl. b. change order of patches.
c. sync waiting only when vm flush needs.

2. address Alex's comments;
bump module version

V3:
address Jerry and Christian's comments.
and only reserve gfxhub vmid

v4:
address Jerry and Christian's comments.
fix some race condistions.

v5:
patch#5: peek_fence instead of get_fence.
fix potential context starved.

Chunming Zhou (6):
drm/amdgpu: add vm ioctl
drm/amdgpu: add reserved vmid field in vm struct v2 drm/amdgpu: reserve/unreserve vmid by vm ioctl v4 drm/amdgpu: add limitation for dedicated vm number v4 drm/amdgpu: implement grab reserved vmid V4 drm/amdgpu: bump module verion for reserved vmid

Quoted 7 lines
  drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c |   3 +-
  drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c |   1 +
  drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c  | 159 +++++++++++++++++++++++++++++++-
  drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h  |   6 ++
  include/uapi/drm/amdgpu_drm.h           |  22 +++++
  5 files changed, 185 insertions(+), 6 deletions(-)

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx