Skip to contentSkip to conversation

drm/i915/gvt: Introduce per object locking in GVT scheduler.

PATCH In review Zhi Wang 4 participants 8 messages Sep 7, 2020 – Sep 9, 2020

Subsystems (3)
lore
Share & tools
lore

Zhi Wang

author
#0 lore raw

PATCH 1/1 drm/i915/gvt: Introduce per object locking in GVT scheduler.

To support ww locking and per-object implemented in i915, GVT scheduler needs to be refined. Most of the changes are located in shadow batch buffer, shadow wa context in GVT-g, where use quite a lot of i915 gem object APIs.

Cc Maarten Lankhorst <[email protected]>
Cc Joonas Lahtinen <[email protected]>
Cc Zhenyu Wang <[email protected]>
---
 drivers/gpu/drm/i915/gvt/scheduler.c | 68 ++++++++++++++++++++++++++++++------
 1 file changed, 57 insertions(+), 11 deletions(-)
drivers/gpu/drm/i915/gvt/scheduler.c +57 -11
index 1570eb8..fe7ee10 100644
@@ -396,7 +396,9 @@ static void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
 	if (!wa_ctx->indirect_ctx.obj)
 		return;
 
+	i915_gem_object_lock(wa_ctx->indirect_ctx.obj, NULL);
 	i915_gem_object_unpin_map(wa_ctx->indirect_ctx.obj);
+	i915_gem_object_unlock(wa_ctx->indirect_ctx.obj);
 	i915_gem_object_put(wa_ctx->indirect_ctx.obj);
 
 	wa_ctx->indirect_ctx.obj = NULL;
@@ -504,6 +506,7 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
 	struct intel_gvt *gvt = workload->vgpu->gvt;
 	const int gmadr_bytes = gvt->device_info.gmadr_bytes_in_cmd;
 	struct intel_vgpu_shadow_bb *bb;
+	struct i915_gem_ww_ctx ww;
 	int ret;
 
 	list_for_each_entry(bb, &workload->shadow_bb, list) {
@@ -528,10 +531,19 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
 		 * directly
 		 */
 		if (!bb->ppgtt) {
-			bb->vma = i915_gem_object_ggtt_pin(bb->obj,
-							   NULL, 0, 0, 0);
+			i915_gem_ww_ctx_init(&ww, false);
+retry:
+			i915_gem_object_lock(bb->obj, &ww);
+
+			bb->vma = i915_gem_object_ggtt_pin_ww(bb->obj, &ww,
+							      NULL, 0, 0, 0);
 			if (IS_ERR(bb->vma)) {
 				ret = PTR_ERR(bb->vma);
+				if (ret == -EDEADLK) {
+					ret = i915_gem_ww_ctx_backoff(&ww);
+					if (!ret)
+						goto retry;
+				}
 				goto err;
 			}
 
@@ -545,13 +557,18 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
 						      0);
 			if (ret)
 				goto err;
+
+			/* No one is going to touch shadow bb from now on. */
+			i915_gem_object_flush_map(bb->obj);
+
+			i915_gem_object_unlock(bb->obj);
+			i915_gem_ww_ctx_fini(&ww);
 		}
 
-		/* No one is going to touch shadow bb from now on. */
-		i915_gem_object_flush_map(bb->obj);
 	}
 	return 0;
 err:
+	i915_gem_ww_ctx_fini(&ww);
 	release_shadow_batch_buffer(workload);
 	return ret;
 }
@@ -578,14 +595,30 @@ static int prepare_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
 	unsigned char *per_ctx_va =
 		(unsigned char *)wa_ctx->indirect_ctx.shadow_va +
 		wa_ctx->indirect_ctx.size;
+	struct i915_gem_ww_ctx ww;
+	int ret;
 
 	if (wa_ctx->indirect_ctx.size == 0)
 		return 0;
 
-	vma = i915_gem_object_ggtt_pin(wa_ctx->indirect_ctx.obj, NULL,
-				       0, CACHELINE_BYTES, 0);
-	if (IS_ERR(vma))
-		return PTR_ERR(vma);
+	i915_gem_ww_ctx_init(&ww, false);
+retry:
+	i915_gem_object_lock(wa_ctx->indirect_ctx.obj, &ww);
+
+	vma = i915_gem_object_ggtt_pin_ww(wa_ctx->indirect_ctx.obj, &ww, NULL,
+					  0, CACHELINE_BYTES, 0);
+	if (IS_ERR(vma)) {
+		ret = PTR_ERR(vma);
+		if (ret == -EDEADLK) {
+			ret = i915_gem_ww_ctx_backoff(&ww);
+			if (!ret)
+				goto retry;
+		}
+		return ret;
+	}
+
+	i915_gem_object_unlock(wa_ctx->indirect_ctx.obj);
+	i915_gem_ww_ctx_fini(&ww);
 
 	/* FIXME: we are not tracking our pinned VMA leaving it
 	 * up to the core to fix up the stray pin_count upon
@@ -619,12 +652,14 @@ static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload)
 
 	list_for_each_entry_safe(bb, pos, &workload->shadow_bb, list) {
 		if (bb->obj) {
+			i915_gem_object_lock(bb->obj, NULL);
 			if (bb->va && !IS_ERR(bb->va))
 				i915_gem_object_unpin_map(bb->obj);
 
 			if (bb->vma && !IS_ERR(bb->vma))
 				i915_vma_unpin(bb->vma);
 
+			i915_gem_object_unlock(bb->obj);
 			i915_gem_object_put(bb->obj);
 		}
 		list_del(&bb->list);
@@ -1337,6 +1372,7 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
 	struct intel_vgpu_submission *s = &vgpu->submission;
 	struct intel_engine_cs *engine;
 	struct i915_ppgtt *ppgtt;
+	struct i915_gem_ww_ctx ww;
 	enum intel_engine_id i;
 	int ret;
 
@@ -1368,11 +1404,20 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
 
 			ce->ring = __intel_context_ring_size(ring_size);
 		}
+		i915_gem_ww_ctx_init(&ww, false);
+retry:
+		ret = intel_context_pin_ww(ce, &ww);
+		if (ret) {
+			if (ret == -EDEADLK) {
+				ret = i915_gem_ww_ctx_backoff(&ww);
+				if (!ret)
+					goto retry;
+			}
+			goto out_shadow_ctx;
+		}
 
-		ret = intel_context_pin(ce);
 		intel_context_put(ce);
-		if (ret)
-			goto out_shadow_ctx;
+		i915_gem_ww_ctx_fini(&ww);
 
 		s->shadow[i] = ce;
 	}
@@ -1400,6 +1445,7 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
 	return 0;
 
 out_shadow_ctx:
+	i915_gem_ww_ctx_fini(&ww);
 	i915_context_ppgtt_root_restore(s, ppgtt);
 	for_each_engine(engine, vgpu->gvt->gt, i) {
 		if (IS_ERR(s->shadow[i]))
Signature · 6 lines
2.7.4

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

[Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for drm/i915/gvt: Introduce per object locking in GVT scheduler.

== Series Details ==

== Series Details ==

Series: drm/i915/gvt: Introduce per object locking in GVT scheduler.
URL : https://patchwork.freedesktop.org/series/81434/
State : warning

== Summary ==

$ dim checkpatch origin/drm-tip
b9e9ccc05038 drm/i915/gvt: Introduce per object locking in GVT scheduler.
-:6: WARNING:COMMIT_LOG_LONG_LINE: Possible unwrapped commit description (prefer a maximum 75 chars per line)
#6: 
To support ww locking and per-object implemented in i915, GVT scheduler needs

total: 0 errors, 1 warnings, 0 checks, 142 lines checked

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

#2 lore raw

[Intel-gfx] ✓ Fi.CI.BAT: success for drm/i915/gvt: Introduce per object locking in GVT scheduler.

In reply to Zhi Wang
== Series Details ==

== Series Details ==

Series: drm/i915/gvt: Introduce per object locking in GVT scheduler.
URL : https://patchwork.freedesktop.org/series/81434/
State : success

== Summary ==

CI Bug Log - changes from CI_DRM_8973 -> Patchwork_18450
====================================================

Summary
-------

**SUCCESS**

No regressions found.

External URL: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/index.html

Known issues
------------

Here are the changes found in Patchwork_18450 that come from known issues:

### IGT changes ###

#### Issues hit ####

* igt@i915_pm_rpm@basic-pci-d3-state:
- fi-byt-j1900: [PASS][1] -> [DMESG-WARN][2] ([i915#1982])
[1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/fi-byt-j1900/igt@[email protected]
[2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/fi-byt-j1900/igt@[email protected]

* igt@i915_selftest@live@execlists:
- fi-icl-y: [PASS][3] -> [INCOMPLETE][4] ([i915#2276])
[3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/fi-icl-y/igt@i915_selftest@[email protected]
[4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/fi-icl-y/igt@i915_selftest@[email protected]

* igt@kms_cursor_legacy@basic-busy-flip-before-cursor-atomic:
- fi-bsw-n3050: [PASS][5] -> [DMESG-WARN][6] ([i915#1982])
[5]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/fi-bsw-n3050/igt@[email protected]
[6]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/fi-bsw-n3050/igt@[email protected]

* igt@kms_flip@basic-flip-vs-wf_vblank@b-edp1:
- fi-icl-u2: [PASS][7] -> [DMESG-WARN][8] ([i915#1982]) +1 similar issue
[7]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/fi-icl-u2/igt@kms_flip@[email protected]
[8]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/fi-icl-u2/igt@kms_flip@[email protected]

* igt@prime_vgem@basic-fence-flip:
- fi-tgl-y: [PASS][9] -> [DMESG-WARN][10] ([i915#402]) +1 similar issue
[9]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/fi-tgl-y/igt@[email protected]
[10]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/fi-tgl-y/igt@[email protected]

#### Possible fixes ####

* igt@gem_sync@basic-all:
- fi-tgl-y: [DMESG-WARN][11] ([i915#402]) -> [PASS][12] +1 similar issue
[11]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/fi-tgl-y/igt@[email protected]
[12]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/fi-tgl-y/igt@[email protected]

* igt@i915_selftest@live@gem_contexts:
- fi-tgl-u2: [INCOMPLETE][13] ([i915#2045]) -> [PASS][14]
[13]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/fi-tgl-u2/igt@i915_selftest@live@gem_contexts.html
[14]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/fi-tgl-u2/igt@i915_selftest@live@gem_contexts.html

* igt@kms_flip@basic-flip-vs-dpms@d-edp1:
- fi-tgl-y: [DMESG-WARN][15] ([i915#1982]) -> [PASS][16] +1 similar issue
[15]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/fi-tgl-y/igt@kms_flip@[email protected]
[16]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/fi-tgl-y/igt@kms_flip@[email protected]

#### Warnings ####

* igt@gem_exec_suspend@basic-s0:
- fi-kbl-x1275: [DMESG-WARN][17] ([i915#62] / [i915#92]) -> [DMESG-WARN][18] ([i915#62] / [i915#92] / [i915#95]) +3 similar issues
[17]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/fi-kbl-x1275/igt@[email protected]
[18]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/fi-kbl-x1275/igt@[email protected]

* igt@kms_force_connector_basic@prune-stale-modes:
- fi-kbl-x1275: [DMESG-WARN][19] ([i915#62] / [i915#92] / [i915#95]) -> [DMESG-WARN][20] ([i915#62] / [i915#92]) +4 similar issues
[19]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/fi-kbl-x1275/igt@[email protected]
[20]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/fi-kbl-x1275/igt@[email protected]

{name}: This element is suppressed. This means it is ignored when computing
the status of the difference (SUCCESS, WARNING, or FAILURE).

[i915#1982]: https://gitlab.freedesktop.org/drm/intel/issues/1982
[i915#2045]: https://gitlab.freedesktop.org/drm/intel/issues/2045
[i915#2276]: https://gitlab.freedesktop.org/drm/intel/issues/2276
[i915#402]: https://gitlab.freedesktop.org/drm/intel/issues/402
[i915#62]: https://gitlab.freedesktop.org/drm/intel/issues/62
[i915#92]: https://gitlab.freedesktop.org/drm/intel/issues/92
[i915#95]: https://gitlab.freedesktop.org/drm/intel/issues/95

Participating hosts (45 -> 39)
------------------------------

Missing (6): fi-ilk-m540 fi-hsw-4200u fi-byt-squawks fi-bsw-cyan fi-byt-clapper fi-bdw-samus

Build changes
-------------

* Linux: CI_DRM_8973 -> Patchwork_18450

CI-20190529: 20190529
CI_DRM_8973: df57d413475ccb334a5d964f5b830731ecdfc994 @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_5779: f52bf19b5f02d52fc3e201c6467ec3f511227fba @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
Patchwork_18450: b9e9ccc05038546dd23317803456f94d0b9c3640 @ git://anongit.freedesktop.org/gfx-ci/linux

== Linux commits ==

b9e9ccc05038 drm/i915/gvt: Introduce per object locking in GVT scheduler.

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/index.html

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

Reviewed-by on v1 1/1
#3 lore raw
In reply to Zhi Wang
Quoted 11 lines · Zhi Wang

To support ww locking and per-object implemented in i915, GVT scheduler needs to be refined. Most of the changes are located in shadow batch buffer, shadow wa context in GVT-g, where use quite a lot of i915 gem object APIs.

Cc Maarten Lankhorst <[email protected]>
Cc Joonas Lahtinen <[email protected]>
Cc Zhenyu Wang <[email protected]>
---
 drivers/gpu/drm/i915/gvt/scheduler.c | 68 ++++++++++++++++++++++++++++++------
 1 file changed, 57 insertions(+), 11 deletions(-)

Looks ok to me.

Pls verify against latest gvt-staging then create pull against gt-next for 5.10.

Thanks

Quoted 159 lines
drivers/gpu/drm/i915/gvt/scheduler.c +57 -11
index 1570eb8..fe7ee10 100644
@@ -396,7 +396,9 @@ static void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
 	if (!wa_ctx->indirect_ctx.obj)
 		return;
 
+	i915_gem_object_lock(wa_ctx->indirect_ctx.obj, NULL);
 	i915_gem_object_unpin_map(wa_ctx->indirect_ctx.obj);
+	i915_gem_object_unlock(wa_ctx->indirect_ctx.obj);
 	i915_gem_object_put(wa_ctx->indirect_ctx.obj);
 
 	wa_ctx->indirect_ctx.obj = NULL;
@@ -504,6 +506,7 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
 	struct intel_gvt *gvt = workload->vgpu->gvt;
 	const int gmadr_bytes = gvt->device_info.gmadr_bytes_in_cmd;
 	struct intel_vgpu_shadow_bb *bb;
+	struct i915_gem_ww_ctx ww;
 	int ret;
 
 	list_for_each_entry(bb, &workload->shadow_bb, list) {
@@ -528,10 +531,19 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
 		 * directly
 		 */
 		if (!bb->ppgtt) {
-			bb->vma = i915_gem_object_ggtt_pin(bb->obj,
-							   NULL, 0, 0, 0);
+			i915_gem_ww_ctx_init(&ww, false);
+retry:
+			i915_gem_object_lock(bb->obj, &ww);
+
+			bb->vma = i915_gem_object_ggtt_pin_ww(bb->obj, &ww,
+							      NULL, 0, 0, 0);
 			if (IS_ERR(bb->vma)) {
 				ret = PTR_ERR(bb->vma);
+				if (ret == -EDEADLK) {
+					ret = i915_gem_ww_ctx_backoff(&ww);
+					if (!ret)
+						goto retry;
+				}
 				goto err;
 			}
 
@@ -545,13 +557,18 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
 						      0);
 			if (ret)
 				goto err;
+
+			/* No one is going to touch shadow bb from now on. */
+			i915_gem_object_flush_map(bb->obj);
+
+			i915_gem_object_unlock(bb->obj);
+			i915_gem_ww_ctx_fini(&ww);
 		}
 
-		/* No one is going to touch shadow bb from now on. */
-		i915_gem_object_flush_map(bb->obj);
 	}
 	return 0;
 err:
+	i915_gem_ww_ctx_fini(&ww);
 	release_shadow_batch_buffer(workload);
 	return ret;
 }
@@ -578,14 +595,30 @@ static int prepare_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
 	unsigned char *per_ctx_va =
 		(unsigned char *)wa_ctx->indirect_ctx.shadow_va +
 		wa_ctx->indirect_ctx.size;
+	struct i915_gem_ww_ctx ww;
+	int ret;
 
 	if (wa_ctx->indirect_ctx.size == 0)
 		return 0;
 
-	vma = i915_gem_object_ggtt_pin(wa_ctx->indirect_ctx.obj, NULL,
-				       0, CACHELINE_BYTES, 0);
-	if (IS_ERR(vma))
-		return PTR_ERR(vma);
+	i915_gem_ww_ctx_init(&ww, false);
+retry:
+	i915_gem_object_lock(wa_ctx->indirect_ctx.obj, &ww);
+
+	vma = i915_gem_object_ggtt_pin_ww(wa_ctx->indirect_ctx.obj, &ww, NULL,
+					  0, CACHELINE_BYTES, 0);
+	if (IS_ERR(vma)) {
+		ret = PTR_ERR(vma);
+		if (ret == -EDEADLK) {
+			ret = i915_gem_ww_ctx_backoff(&ww);
+			if (!ret)
+				goto retry;
+		}
+		return ret;
+	}
+
+	i915_gem_object_unlock(wa_ctx->indirect_ctx.obj);
+	i915_gem_ww_ctx_fini(&ww);
 
 	/* FIXME: we are not tracking our pinned VMA leaving it
 	 * up to the core to fix up the stray pin_count upon
@@ -619,12 +652,14 @@ static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload)
 
 	list_for_each_entry_safe(bb, pos, &workload->shadow_bb, list) {
 		if (bb->obj) {
+			i915_gem_object_lock(bb->obj, NULL);
 			if (bb->va && !IS_ERR(bb->va))
 				i915_gem_object_unpin_map(bb->obj);
 
 			if (bb->vma && !IS_ERR(bb->vma))
 				i915_vma_unpin(bb->vma);
 
+			i915_gem_object_unlock(bb->obj);
 			i915_gem_object_put(bb->obj);
 		}
 		list_del(&bb->list);
@@ -1337,6 +1372,7 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
 	struct intel_vgpu_submission *s = &vgpu->submission;
 	struct intel_engine_cs *engine;
 	struct i915_ppgtt *ppgtt;
+	struct i915_gem_ww_ctx ww;
 	enum intel_engine_id i;
 	int ret;
 
@@ -1368,11 +1404,20 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
 
 			ce->ring = __intel_context_ring_size(ring_size);
 		}
+		i915_gem_ww_ctx_init(&ww, false);
+retry:
+		ret = intel_context_pin_ww(ce, &ww);
+		if (ret) {
+			if (ret == -EDEADLK) {
+				ret = i915_gem_ww_ctx_backoff(&ww);
+				if (!ret)
+					goto retry;
+			}
+			goto out_shadow_ctx;
+		}
 
-		ret = intel_context_pin(ce);
 		intel_context_put(ce);
-		if (ret)
-			goto out_shadow_ctx;
+		i915_gem_ww_ctx_fini(&ww);
 
 		s->shadow[i] = ce;
 	}
@@ -1400,6 +1445,7 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
 	return 0;
 
 out_shadow_ctx:
+	i915_gem_ww_ctx_fini(&ww);
 	i915_context_ppgtt_root_restore(s, ppgtt);
 	for_each_engine(engine, vgpu->gvt->gt, i) {
 		if (IS_ERR(s->shadow[i]))
Signature · 1 line
2.7.4
Signature · 1 line
$gpg --keyserver wwwkeys.pgp.net --recv-keys 4D781827

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

#4 lore raw

[Intel-gfx] ✓ Fi.CI.IGT: success for drm/i915/gvt: Introduce per object locking in GVT scheduler.

In reply to Zhi Wang
== Series Details ==

== Series Details ==

Series: drm/i915/gvt: Introduce per object locking in GVT scheduler.
URL : https://patchwork.freedesktop.org/series/81434/
State : success

== Summary ==

CI Bug Log - changes from CI_DRM_8973_full -> Patchwork_18450_full
====================================================

Summary
-------

**SUCCESS**

No regressions found.

Known issues
------------

Here are the changes found in Patchwork_18450_full that come from known issues:

### IGT changes ###

#### Issues hit ####

* igt@gem_ctx_isolation@preservation-s3@vcs0:
- shard-kbl: [PASS][1] -> [DMESG-WARN][2] ([i915#180]) +1 similar issue
[1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-kbl6/igt@gem_ctx_isolation@[email protected]
[2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-kbl3/igt@gem_ctx_isolation@[email protected]

* igt@gem_exec_reloc@basic-many-active@rcs0:
- shard-glk: [PASS][3] -> [FAIL][4] ([i915#2389])
[3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-glk5/igt@gem_exec_reloc@[email protected]
[4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-glk2/igt@gem_exec_reloc@[email protected]

* igt@gem_exec_whisper@basic-queues-forked-all:
- shard-glk: [PASS][5] -> [DMESG-WARN][6] ([i915#118] / [i915#95]) +1 similar issue
[5]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-glk5/igt@[email protected]
[6]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-glk2/igt@[email protected]

* igt@gem_huc_copy@huc-copy:
- shard-tglb: [PASS][7] -> [SKIP][8] ([i915#2190])
[7]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-tglb1/igt@[email protected]
[8]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-tglb6/igt@[email protected]

* igt@gem_userptr_blits@sync-unmap-cycles:
- shard-skl: [PASS][9] -> [TIMEOUT][10] ([i915#1958] / [i915#2424])
[9]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-skl1/igt@[email protected]
[10]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-skl4/igt@[email protected]

* igt@kms_atomic_transition@plane-all-transition-nonblocking-fencing@edp-1-pipe-a:
- shard-skl: [PASS][11] -> [DMESG-WARN][12] ([i915#1982]) +6 similar issues
[11]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-skl4/igt@kms_atomic_transition@[email protected]
[12]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-skl2/igt@kms_atomic_transition@[email protected]

* igt@kms_flip@flip-vs-suspend@a-edp1:
- shard-skl: [PASS][13] -> [INCOMPLETE][14] ([i915#198])
[13]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-skl8/igt@kms_flip@[email protected]
[14]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-skl6/igt@kms_flip@[email protected]

* igt@kms_flip@flip-vs-suspend@c-hdmi-a1:
- shard-hsw: [PASS][15] -> [INCOMPLETE][16] ([i915#2055])
[15]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-hsw5/igt@kms_flip@[email protected]
[16]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-hsw1/igt@kms_flip@[email protected]

* igt@kms_flip@plain-flip-fb-recreate-interruptible@c-edp1:
- shard-skl: [PASS][17] -> [FAIL][18] ([i915#2122]) +1 similar issue
[17]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-skl7/igt@kms_flip@[email protected]
[18]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-skl5/igt@kms_flip@[email protected]

* igt@kms_flip_tiling@flip-changes-tiling-y:
- shard-skl: [PASS][19] -> [FAIL][20] ([i915#699])
[19]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-skl9/igt@[email protected]
[20]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-skl7/igt@[email protected]

* igt@kms_frontbuffer_tracking@fbc-1p-primscrn-shrfb-msflip-blt:
- shard-iclb: [PASS][21] -> [DMESG-WARN][22] ([i915#1982]) +1 similar issue
[21]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-iclb3/igt@[email protected]
[22]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-iclb3/igt@[email protected]

* igt@kms_frontbuffer_tracking@fbc-farfromfence:
- shard-kbl: [PASS][23] -> [DMESG-WARN][24] ([i915#1982]) +1 similar issue
[23]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-kbl1/igt@[email protected]
[24]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-kbl4/igt@[email protected]

* igt@kms_hdr@bpc-switch-suspend:
- shard-skl: [PASS][25] -> [FAIL][26] ([i915#1188])
[25]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-skl4/igt@[email protected]
[26]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-skl7/igt@[email protected]

* igt@kms_plane_alpha_blend@pipe-c-coverage-7efc:
- shard-skl: [PASS][27] -> [FAIL][28] ([fdo#108145] / [i915#265]) +1 similar issue
[27]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-skl9/igt@[email protected]
[28]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-skl7/igt@[email protected]

* igt@kms_psr@psr2_cursor_mmap_gtt:
- shard-iclb: [PASS][29] -> [SKIP][30] ([fdo#109441])
[29]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-iclb2/igt@kms_psr@psr2_cursor_mmap_gtt.html
[30]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-iclb1/igt@kms_psr@psr2_cursor_mmap_gtt.html

* igt@perf_pmu@module-unload:
- shard-tglb: [PASS][31] -> [DMESG-WARN][32] ([i915#1982])
[31]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-tglb8/igt@[email protected]
[32]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-tglb8/igt@[email protected]

* igt@sysfs_timeslice_duration@timeout@rcs0:
- shard-skl: [PASS][33] -> [FAIL][34] ([i915#1732])
[33]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-skl6/igt@sysfs_timeslice_duration@[email protected]
[34]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-skl3/igt@sysfs_timeslice_duration@[email protected]

#### Possible fixes ####

* igt@gem_exec_whisper@basic-contexts-priority-all:
- shard-skl: [DMESG-WARN][35] ([i915#1982]) -> [PASS][36] +1 similar issue
[35]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-skl4/igt@[email protected]
[36]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-skl2/igt@[email protected]

* igt@gem_softpin@noreloc-s3:
- shard-kbl: [INCOMPLETE][37] ([i915#1959]) -> [PASS][38]
[37]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-kbl3/igt@[email protected]
[38]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-kbl7/igt@[email protected]

* igt@gem_userptr_blits@unsync-unmap-cycles:
- shard-skl: [TIMEOUT][39] ([i915#1958]) -> [PASS][40]
[39]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-skl6/igt@[email protected]
[40]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-skl3/igt@[email protected]

* igt@i915_pm_rc6_residency@rc6-idle:
- shard-hsw: [WARN][41] ([i915#1519]) -> [PASS][42]
[41]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-hsw2/igt@[email protected]
[42]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-hsw8/igt@[email protected]

* igt@kms_cursor_edge_walk@pipe-b-256x256-left-edge:
- shard-glk: [DMESG-WARN][43] ([i915#1982]) -> [PASS][44]
[43]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-glk3/igt@[email protected]
[44]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-glk8/igt@[email protected]
- shard-apl: [DMESG-WARN][45] ([i915#1635] / [i915#1982]) -> [PASS][46] +1 similar issue
[45]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-apl7/igt@[email protected]
[46]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-apl1/igt@[email protected]

* igt@kms_cursor_legacy@2x-long-cursor-vs-flip-legacy:
- shard-hsw: [FAIL][47] ([i915#96]) -> [PASS][48]
[47]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-hsw1/igt@[email protected]
[48]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-hsw4/igt@[email protected]

* igt@kms_cursor_legacy@pipe-d-torture-move:
- shard-tglb: [DMESG-WARN][49] ([i915#128]) -> [PASS][50]
[49]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-tglb8/igt@[email protected]
[50]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-tglb8/igt@[email protected]

* igt@kms_cursor_legacy@short-flip-after-cursor-atomic-transitions-varying-size:
- shard-kbl: [DMESG-WARN][51] ([i915#1982]) -> [PASS][52]
[51]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-kbl4/igt@kms_cursor_legacy@short-flip-after-cursor-atomic-transitions-varying-size.html
[52]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-kbl4/igt@kms_cursor_legacy@short-flip-after-cursor-atomic-transitions-varying-size.html

* igt@kms_frontbuffer_tracking@fbcpsr-slowdraw:
- shard-tglb: [DMESG-WARN][53] ([i915#1982]) -> [PASS][54] +1 similar issue
[53]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-tglb5/igt@[email protected]
[54]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-tglb7/igt@[email protected]

* igt@kms_frontbuffer_tracking@psr-rgb565-draw-mmap-cpu:
- shard-skl: [FAIL][55] ([i915#49]) -> [PASS][56]
[55]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-skl4/igt@[email protected]
[56]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-skl7/igt@[email protected]

* igt@kms_plane@plane-panning-bottom-right-suspend-pipe-a-planes:
- shard-kbl: [DMESG-WARN][57] ([i915#180]) -> [PASS][58] +3 similar issues
[57]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-kbl4/igt@[email protected]
[58]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-kbl4/igt@[email protected]
- shard-iclb: [INCOMPLETE][59] ([i915#1185] / [i915#250]) -> [PASS][60]
[59]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-iclb3/igt@[email protected]
[60]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-iclb5/igt@[email protected]

* igt@kms_plane_alpha_blend@pipe-a-coverage-7efc:
- shard-skl: [FAIL][61] ([fdo#108145] / [i915#265]) -> [PASS][62]
[61]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-skl4/igt@[email protected]
[62]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-skl7/igt@[email protected]

* igt@kms_psr@psr2_primary_mmap_cpu:
- shard-iclb: [SKIP][63] ([fdo#109441]) -> [PASS][64] +1 similar issue
[63]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-iclb7/igt@kms_psr@psr2_primary_mmap_cpu.html
[64]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-iclb2/igt@kms_psr@psr2_primary_mmap_cpu.html

* igt@perf@polling-parameterized:
- shard-apl: [FAIL][65] ([i915#1542] / [i915#1635]) -> [PASS][66]
[65]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8973/shard-apl2/igt@[email protected]
[66]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/shard-apl2/igt@[email protected]

[fdo#108145]: https://bugs.freedesktop.org/show_bug.cgi?id=108145
[fdo#109441]: https://bugs.freedesktop.org/show_bug.cgi?id=109441
[i915#118]: https://gitlab.freedesktop.org/drm/intel/issues/118
[i915#1185]: https://gitlab.freedesktop.org/drm/intel/issues/1185
[i915#1188]: https://gitlab.freedesktop.org/drm/intel/issues/1188
[i915#128]: https://gitlab.freedesktop.org/drm/intel/issues/128
[i915#1519]: https://gitlab.freedesktop.org/drm/intel/issues/1519
[i915#1542]: https://gitlab.freedesktop.org/drm/intel/issues/1542
[i915#1635]: https://gitlab.freedesktop.org/drm/intel/issues/1635
[i915#1732]: https://gitlab.freedesktop.org/drm/intel/issues/1732
[i915#180]: https://gitlab.freedesktop.org/drm/intel/issues/180
[i915#1958]: https://gitlab.freedesktop.org/drm/intel/issues/1958
[i915#1959]: https://gitlab.freedesktop.org/drm/intel/issues/1959
[i915#198]: https://gitlab.freedesktop.org/drm/intel/issues/198
[i915#1982]: https://gitlab.freedesktop.org/drm/intel/issues/1982
[i915#2055]: https://gitlab.freedesktop.org/drm/intel/issues/2055
[i915#2122]: https://gitlab.freedesktop.org/drm/intel/issues/2122
[i915#2190]: https://gitlab.freedesktop.org/drm/intel/issues/2190
[i915#2389]: https://gitlab.freedesktop.org/drm/intel/issues/2389
[i915#2424]: https://gitlab.freedesktop.org/drm/intel/issues/2424
[i915#250]: https://gitlab.freedesktop.org/drm/intel/issues/250
[i915#265]: https://gitlab.freedesktop.org/drm/intel/issues/265
[i915#49]: https://gitlab.freedesktop.org/drm/intel/issues/49
[i915#699]: https://gitlab.freedesktop.org/drm/intel/issues/699
[i915#95]: https://gitlab.freedesktop.org/drm/intel/issues/95
[i915#96]: https://gitlab.freedesktop.org/drm/intel/issues/96

Participating hosts (11 -> 11)
------------------------------

No changes in participating hosts

Build changes
-------------

* Linux: CI_DRM_8973 -> Patchwork_18450

CI-20190529: 20190529
CI_DRM_8973: df57d413475ccb334a5d964f5b830731ecdfc994 @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_5779: f52bf19b5f02d52fc3e201c6467ec3f511227fba @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
Patchwork_18450: b9e9ccc05038546dd23317803456f94d0b9c3640 @ git://anongit.freedesktop.org/gfx-ci/linux
piglit_4509: fdc5a4ca11124ab8413c7988896eec4c97336694 @ git://anongit.freedesktop.org/piglit

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_18450/index.html

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

#5 lore raw
In reply to Zhi Wang

I tested this patch on the suspend/resume case with vGPU created (no need really activate), can still observer the system freeze issue as mentioned in another patch I sent. So I suppose we still need decouple context pin/unpin with submission setup/clean, but move to workload create/destroy?

After made similar changes based on this one, plus the suspend/resume support patch, below tests can pass:

- Create vGPU then suspend/resume.

- Run VM w/ vGPU then suspend/resume.

https://lists.freedesktop.org/archives/intel-gvt-dev/2020-September/007061.html

Quoted 136 lines · Zhi Wang

To support ww locking and per-object implemented in i915, GVT scheduler needs to be refined. Most of the changes are located in shadow batch buffer, shadow wa context in GVT-g, where use quite a lot of i915 gem object APIs.

Cc Maarten Lankhorst <[email protected]>
Cc Joonas Lahtinen <[email protected]>
Cc Zhenyu Wang <[email protected]>
---
  drivers/gpu/drm/i915/gvt/scheduler.c | 68 ++++++++++++++++++++++++++++++------
  1 file changed, 57 insertions(+), 11 deletions(-)
drivers/gpu/drm/i915/gvt/scheduler.c +44 -8
index 1570eb8..fe7ee10 100644
@@ -396,7 +396,9 @@ static void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
  	if (!wa_ctx->indirect_ctx.obj)
  		return;
 
+	i915_gem_object_lock(wa_ctx->indirect_ctx.obj, NULL);
  	i915_gem_object_unpin_map(wa_ctx->indirect_ctx.obj);
+	i915_gem_object_unlock(wa_ctx->indirect_ctx.obj);
  	i915_gem_object_put(wa_ctx->indirect_ctx.obj);
 
  	wa_ctx->indirect_ctx.obj = NULL;
@@ -504,6 +506,7 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
  	struct intel_gvt *gvt = workload->vgpu->gvt;
  	const int gmadr_bytes = gvt->device_info.gmadr_bytes_in_cmd;
  	struct intel_vgpu_shadow_bb *bb;
+	struct i915_gem_ww_ctx ww;
  	int ret;
 
  	list_for_each_entry(bb, &workload->shadow_bb, list) {
@@ -528,10 +531,19 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
  		 * directly
  		 */
  		if (!bb->ppgtt) {
-			bb->vma = i915_gem_object_ggtt_pin(bb->obj,
-							   NULL, 0, 0, 0);
+			i915_gem_ww_ctx_init(&ww, false);
+retry:
+			i915_gem_object_lock(bb->obj, &ww);
+
+			bb->vma = i915_gem_object_ggtt_pin_ww(bb->obj, &ww,
+							      NULL, 0, 0, 0);
  			if (IS_ERR(bb->vma)) {
  				ret = PTR_ERR(bb->vma);
+				if (ret == -EDEADLK) {
+					ret = i915_gem_ww_ctx_backoff(&ww);
+					if (!ret)
+						goto retry;
+				}
  				goto err;
  			}
 
@@ -545,13 +557,18 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
  						      0);
  			if (ret)
  				goto err;
+
+			/* No one is going to touch shadow bb from now on. */
+			i915_gem_object_flush_map(bb->obj);
+
+			i915_gem_object_unlock(bb->obj);
+			i915_gem_ww_ctx_fini(&ww);
  		}
 
-		/* No one is going to touch shadow bb from now on. */
-		i915_gem_object_flush_map(bb->obj);
  	}
  	return 0;
  err:
+	i915_gem_ww_ctx_fini(&ww);
  	release_shadow_batch_buffer(workload);
  	return ret;
  }
@@ -578,14 +595,30 @@ static int prepare_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
  	unsigned char *per_ctx_va =
  		(unsigned char *)wa_ctx->indirect_ctx.shadow_va +
  		wa_ctx->indirect_ctx.size;
+	struct i915_gem_ww_ctx ww;
+	int ret;
 
  	if (wa_ctx->indirect_ctx.size == 0)
  		return 0;
 
-	vma = i915_gem_object_ggtt_pin(wa_ctx->indirect_ctx.obj, NULL,
-				       0, CACHELINE_BYTES, 0);
-	if (IS_ERR(vma))
-		return PTR_ERR(vma);
+	i915_gem_ww_ctx_init(&ww, false);
+retry:
+	i915_gem_object_lock(wa_ctx->indirect_ctx.obj, &ww);
+
+	vma = i915_gem_object_ggtt_pin_ww(wa_ctx->indirect_ctx.obj, &ww, NULL,
+					  0, CACHELINE_BYTES, 0);
+	if (IS_ERR(vma)) {
+		ret = PTR_ERR(vma);
+		if (ret == -EDEADLK) {
+			ret = i915_gem_ww_ctx_backoff(&ww);
+			if (!ret)
+				goto retry;
+		}
+		return ret;
+	}
+
+	i915_gem_object_unlock(wa_ctx->indirect_ctx.obj);
+	i915_gem_ww_ctx_fini(&ww);
 
  	/* FIXME: we are not tracking our pinned VMA leaving it
  	 * up to the core to fix up the stray pin_count upon
@@ -619,12 +652,14 @@ static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload)
 
  	list_for_each_entry_safe(bb, pos, &workload->shadow_bb, list) {
  		if (bb->obj) {
+			i915_gem_object_lock(bb->obj, NULL);
  			if (bb->va && !IS_ERR(bb->va))
  				i915_gem_object_unpin_map(bb->obj);
 
  			if (bb->vma && !IS_ERR(bb->vma))
  				i915_vma_unpin(bb->vma);
 
+			i915_gem_object_unlock(bb->obj);
  			i915_gem_object_put(bb->obj);
  		}
  		list_del(&bb->list);
@@ -1337,6 +1372,7 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
  	struct intel_vgpu_submission *s = &vgpu->submission;
  	struct intel_engine_cs *engine;
  	struct i915_ppgtt *ppgtt;
+	struct i915_gem_ww_ctx ww;
  	enum intel_engine_id i;
  	int ret;
 

drivers/gpu/drm/i915/gvt/scheduler.c int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) · line 1406

Quoted 3 lines
drivers/gpu/drm/i915/gvt/scheduler.c +0 -0
@@ -1368,11 +1404,20 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
 
  			ce->ring = __intel_context_ring_size(ring_size);
  		}

Cut here

drivers/gpu/drm/i915/gvt/scheduler.c int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) · line 1417

Quoted 11 lines
drivers/gpu/drm/i915/gvt/scheduler.c +11 -0
+		i915_gem_ww_ctx_init(&ww, false);
+retry:
+		ret = intel_context_pin_ww(ce, &ww);
+		if (ret) {
+			if (ret == -EDEADLK) {
+				ret = i915_gem_ww_ctx_backoff(&ww);
+				if (!ret)
+					goto retry;
+			}
+			goto out_shadow_ctx;
+		}

I move the piece to create_workload. Similar to the change I made in my patch sent.

Quoted 17 lines
drivers/gpu/drm/i915/gvt/scheduler.c +2 -3
-		ret = intel_context_pin(ce);
  		intel_context_put(ce);
-		if (ret)
-			goto out_shadow_ctx;
+		i915_gem_ww_ctx_fini(&ww);
 
  		s->shadow[i] = ce;
  	}
@@ -1400,6 +1445,7 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
  	return 0;
 
  out_shadow_ctx:
+	i915_gem_ww_ctx_fini(&ww);
  	i915_context_ppgtt_root_restore(s, ppgtt);
  	for_each_engine(engine, vgpu->gvt->gt, i) {
  		if (IS_ERR(s->shadow[i]))
Signature · 7 lines
Best Regards,
Colin Xu

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#6 lore raw
Quoted 5 lines · Colin Xu

I tested this patch on the suspend/resume case with vGPU created (no need really activate), can still observer the system freeze issue as mentioned in another patch I sent. So I suppose we still need decouple context pin/unpin with submission setup/clean, but move to workload create/destroy?

yeah, I observed that conflict too. How about merging Zhi's ww lock fix first then next for left suspend/resume fixes? And better if you could pack all fixes in one series.

Thanks

Quoted 169 lines

After made similar changes based on this one, plus the suspend/resume support patch, below tests can pass:

- Create vGPU then suspend/resume.

- Run VM w/ vGPU then suspend/resume.

https://lists.freedesktop.org/archives/intel-gvt-dev/2020-September/007061.html

Quoted 124 lines · Zhi Wang

To support ww locking and per-object implemented in i915, GVT scheduler needs to be refined. Most of the changes are located in shadow batch buffer, shadow wa context in GVT-g, where use quite a lot of i915 gem object APIs.

Cc Maarten Lankhorst <[email protected]>
Cc Joonas Lahtinen <[email protected]>
Cc Zhenyu Wang <[email protected]>
---
  drivers/gpu/drm/i915/gvt/scheduler.c | 68 ++++++++++++++++++++++++++++++------
  1 file changed, 57 insertions(+), 11 deletions(-)
drivers/gpu/drm/i915/gvt/scheduler.c +44 -8
index 1570eb8..fe7ee10 100644
@@ -396,7 +396,9 @@ static void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
  	if (!wa_ctx->indirect_ctx.obj)
  		return;
+	i915_gem_object_lock(wa_ctx->indirect_ctx.obj, NULL);
  	i915_gem_object_unpin_map(wa_ctx->indirect_ctx.obj);
+	i915_gem_object_unlock(wa_ctx->indirect_ctx.obj);
  	i915_gem_object_put(wa_ctx->indirect_ctx.obj);
  	wa_ctx->indirect_ctx.obj = NULL;
@@ -504,6 +506,7 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
  	struct intel_gvt *gvt = workload->vgpu->gvt;
  	const int gmadr_bytes = gvt->device_info.gmadr_bytes_in_cmd;
  	struct intel_vgpu_shadow_bb *bb;
+	struct i915_gem_ww_ctx ww;
  	int ret;
  	list_for_each_entry(bb, &workload->shadow_bb, list) {
@@ -528,10 +531,19 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
  		 * directly
  		 */
  		if (!bb->ppgtt) {
-			bb->vma = i915_gem_object_ggtt_pin(bb->obj,
-							   NULL, 0, 0, 0);
+			i915_gem_ww_ctx_init(&ww, false);
+retry:
+			i915_gem_object_lock(bb->obj, &ww);
+
+			bb->vma = i915_gem_object_ggtt_pin_ww(bb->obj, &ww,
+							      NULL, 0, 0, 0);
  			if (IS_ERR(bb->vma)) {
  				ret = PTR_ERR(bb->vma);
+				if (ret == -EDEADLK) {
+					ret = i915_gem_ww_ctx_backoff(&ww);
+					if (!ret)
+						goto retry;
+				}
  				goto err;
  			}
@@ -545,13 +557,18 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
  						      0);
  			if (ret)
  				goto err;
+
+			/* No one is going to touch shadow bb from now on. */
+			i915_gem_object_flush_map(bb->obj);
+
+			i915_gem_object_unlock(bb->obj);
+			i915_gem_ww_ctx_fini(&ww);
  		}
-		/* No one is going to touch shadow bb from now on. */
-		i915_gem_object_flush_map(bb->obj);
  	}
  	return 0;
  err:
+	i915_gem_ww_ctx_fini(&ww);
  	release_shadow_batch_buffer(workload);
  	return ret;
  }
@@ -578,14 +595,30 @@ static int prepare_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
  	unsigned char *per_ctx_va =
  		(unsigned char *)wa_ctx->indirect_ctx.shadow_va +
  		wa_ctx->indirect_ctx.size;
+	struct i915_gem_ww_ctx ww;
+	int ret;
  	if (wa_ctx->indirect_ctx.size == 0)
  		return 0;
-	vma = i915_gem_object_ggtt_pin(wa_ctx->indirect_ctx.obj, NULL,
-				       0, CACHELINE_BYTES, 0);
-	if (IS_ERR(vma))
-		return PTR_ERR(vma);
+	i915_gem_ww_ctx_init(&ww, false);
+retry:
+	i915_gem_object_lock(wa_ctx->indirect_ctx.obj, &ww);
+
+	vma = i915_gem_object_ggtt_pin_ww(wa_ctx->indirect_ctx.obj, &ww, NULL,
+					  0, CACHELINE_BYTES, 0);
+	if (IS_ERR(vma)) {
+		ret = PTR_ERR(vma);
+		if (ret == -EDEADLK) {
+			ret = i915_gem_ww_ctx_backoff(&ww);
+			if (!ret)
+				goto retry;
+		}
+		return ret;
+	}
+
+	i915_gem_object_unlock(wa_ctx->indirect_ctx.obj);
+	i915_gem_ww_ctx_fini(&ww);
  	/* FIXME: we are not tracking our pinned VMA leaving it
  	 * up to the core to fix up the stray pin_count upon
@@ -619,12 +652,14 @@ static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload)
  	list_for_each_entry_safe(bb, pos, &workload->shadow_bb, list) {
  		if (bb->obj) {
+			i915_gem_object_lock(bb->obj, NULL);
  			if (bb->va && !IS_ERR(bb->va))
  				i915_gem_object_unpin_map(bb->obj);
  			if (bb->vma && !IS_ERR(bb->vma))
  				i915_vma_unpin(bb->vma);
+			i915_gem_object_unlock(bb->obj);
  			i915_gem_object_put(bb->obj);
  		}
  		list_del(&bb->list);
@@ -1337,6 +1372,7 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
  	struct intel_vgpu_submission *s = &vgpu->submission;
  	struct intel_engine_cs *engine;
  	struct i915_ppgtt *ppgtt;
+	struct i915_gem_ww_ctx ww;
  	enum intel_engine_id i;
  	int ret;

drivers/gpu/drm/i915/gvt/scheduler.c int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) · line 1405

Quoted 2 lines
drivers/gpu/drm/i915/gvt/scheduler.c +0 -0
@@ -1368,11 +1404,20 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
  			ce->ring = __intel_context_ring_size(ring_size);
  		}

Cut here

drivers/gpu/drm/i915/gvt/scheduler.c int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) · line 1417

Quoted 11 lines
drivers/gpu/drm/i915/gvt/scheduler.c +11 -0
+		i915_gem_ww_ctx_init(&ww, false);
+retry:
+		ret = intel_context_pin_ww(ce, &ww);
+		if (ret) {
+			if (ret == -EDEADLK) {
+				ret = i915_gem_ww_ctx_backoff(&ww);
+				if (!ret)
+					goto retry;
+			}
+			goto out_shadow_ctx;
+		}

I move the piece to create_workload. Similar to the change I made in my patch sent.

Quoted 14 lines
drivers/gpu/drm/i915/gvt/scheduler.c +2 -3
-		ret = intel_context_pin(ce);
  		intel_context_put(ce);
-		if (ret)
-			goto out_shadow_ctx;
+		i915_gem_ww_ctx_fini(&ww);
  		s->shadow[i] = ce;
  	}
@@ -1400,6 +1445,7 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
  	return 0;
  out_shadow_ctx:
+	i915_gem_ww_ctx_fini(&ww);
  	i915_context_ppgtt_root_restore(s, ppgtt);
  	for_each_engine(engine, vgpu->gvt->gt, i) {
  		if (IS_ERR(s->shadow[i]))
Signature · 2 lines
Best Regards,
Colin Xu
Signature · 1 line
$gpg --keyserver wwwkeys.pgp.net --recv-keys 4D781827

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

#7 lore raw
Quoted 11 lines · Zhenyu Wang
Quoted 5 lines · Colin Xu

I tested this patch on the suspend/resume case with vGPU created (no need really activate), can still observer the system freeze issue as mentioned in another patch I sent. So I suppose we still need decouple context pin/unpin with submission setup/clean, but move to workload create/destroy?

yeah, I observed that conflict too. How about merging Zhi's ww lock fix first then next for left suspend/resume fixes? And better if you could pack all fixes in one series.

Thanks

No problem at all. And I would like to hear more comments if moving the context pin/unpin to workload lifecycle is reasonable.

Quoted 169 lines
Quoted 169 lines

After made similar changes based on this one, plus the suspend/resume support patch, below tests can pass:

- Create vGPU then suspend/resume.

- Run VM w/ vGPU then suspend/resume.

https://lists.freedesktop.org/archives/intel-gvt-dev/2020-September/007061.html

Quoted 124 lines · Zhi Wang

To support ww locking and per-object implemented in i915, GVT scheduler needs to be refined. Most of the changes are located in shadow batch buffer, shadow wa context in GVT-g, where use quite a lot of i915 gem object APIs.

Cc Maarten Lankhorst <[email protected]>
Cc Joonas Lahtinen <[email protected]>
Cc Zhenyu Wang <[email protected]>
---
   drivers/gpu/drm/i915/gvt/scheduler.c | 68 ++++++++++++++++++++++++++++++------
   1 file changed, 57 insertions(+), 11 deletions(-)
drivers/gpu/drm/i915/gvt/scheduler.c +44 -8
index 1570eb8..fe7ee10 100644
@@ -396,7 +396,9 @@ static void release_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
   	if (!wa_ctx->indirect_ctx.obj)
   		return;
+	i915_gem_object_lock(wa_ctx->indirect_ctx.obj, NULL);
   	i915_gem_object_unpin_map(wa_ctx->indirect_ctx.obj);
+	i915_gem_object_unlock(wa_ctx->indirect_ctx.obj);
   	i915_gem_object_put(wa_ctx->indirect_ctx.obj);
   	wa_ctx->indirect_ctx.obj = NULL;
@@ -504,6 +506,7 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
   	struct intel_gvt *gvt = workload->vgpu->gvt;
   	const int gmadr_bytes = gvt->device_info.gmadr_bytes_in_cmd;
   	struct intel_vgpu_shadow_bb *bb;
+	struct i915_gem_ww_ctx ww;
   	int ret;
   	list_for_each_entry(bb, &workload->shadow_bb, list) {
@@ -528,10 +531,19 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
   		 * directly
   		 */
   		if (!bb->ppgtt) {
-			bb->vma = i915_gem_object_ggtt_pin(bb->obj,
-							   NULL, 0, 0, 0);
+			i915_gem_ww_ctx_init(&ww, false);
+retry:
+			i915_gem_object_lock(bb->obj, &ww);
+
+			bb->vma = i915_gem_object_ggtt_pin_ww(bb->obj, &ww,
+							      NULL, 0, 0, 0);
   			if (IS_ERR(bb->vma)) {
   				ret = PTR_ERR(bb->vma);
+				if (ret == -EDEADLK) {
+					ret = i915_gem_ww_ctx_backoff(&ww);
+					if (!ret)
+						goto retry;
+				}
   				goto err;
   			}
@@ -545,13 +557,18 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
   						      0);
   			if (ret)
   				goto err;
+
+			/* No one is going to touch shadow bb from now on. */
+			i915_gem_object_flush_map(bb->obj);
+
+			i915_gem_object_unlock(bb->obj);
+			i915_gem_ww_ctx_fini(&ww);
   		}
-		/* No one is going to touch shadow bb from now on. */
-		i915_gem_object_flush_map(bb->obj);
   	}
   	return 0;
   err:
+	i915_gem_ww_ctx_fini(&ww);
   	release_shadow_batch_buffer(workload);
   	return ret;
   }
@@ -578,14 +595,30 @@ static int prepare_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
   	unsigned char *per_ctx_va =
   		(unsigned char *)wa_ctx->indirect_ctx.shadow_va +
   		wa_ctx->indirect_ctx.size;
+	struct i915_gem_ww_ctx ww;
+	int ret;
   	if (wa_ctx->indirect_ctx.size == 0)
   		return 0;
-	vma = i915_gem_object_ggtt_pin(wa_ctx->indirect_ctx.obj, NULL,
-				       0, CACHELINE_BYTES, 0);
-	if (IS_ERR(vma))
-		return PTR_ERR(vma);
+	i915_gem_ww_ctx_init(&ww, false);
+retry:
+	i915_gem_object_lock(wa_ctx->indirect_ctx.obj, &ww);
+
+	vma = i915_gem_object_ggtt_pin_ww(wa_ctx->indirect_ctx.obj, &ww, NULL,
+					  0, CACHELINE_BYTES, 0);
+	if (IS_ERR(vma)) {
+		ret = PTR_ERR(vma);
+		if (ret == -EDEADLK) {
+			ret = i915_gem_ww_ctx_backoff(&ww);
+			if (!ret)
+				goto retry;
+		}
+		return ret;
+	}
+
+	i915_gem_object_unlock(wa_ctx->indirect_ctx.obj);
+	i915_gem_ww_ctx_fini(&ww);
   	/* FIXME: we are not tracking our pinned VMA leaving it
   	 * up to the core to fix up the stray pin_count upon
@@ -619,12 +652,14 @@ static void release_shadow_batch_buffer(struct intel_vgpu_workload *workload)
   	list_for_each_entry_safe(bb, pos, &workload->shadow_bb, list) {
   		if (bb->obj) {
+			i915_gem_object_lock(bb->obj, NULL);
   			if (bb->va && !IS_ERR(bb->va))
   				i915_gem_object_unpin_map(bb->obj);
   			if (bb->vma && !IS_ERR(bb->vma))
   				i915_vma_unpin(bb->vma);
+			i915_gem_object_unlock(bb->obj);
   			i915_gem_object_put(bb->obj);
   		}
   		list_del(&bb->list);
@@ -1337,6 +1372,7 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
   	struct intel_vgpu_submission *s = &vgpu->submission;
   	struct intel_engine_cs *engine;
   	struct i915_ppgtt *ppgtt;
+	struct i915_gem_ww_ctx ww;
   	enum intel_engine_id i;
   	int ret;

drivers/gpu/drm/i915/gvt/scheduler.c int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) · line 1405

Quoted 2 lines
drivers/gpu/drm/i915/gvt/scheduler.c +0 -0
@@ -1368,11 +1404,20 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
   			ce->ring = __intel_context_ring_size(ring_size);
   		}

Cut here

drivers/gpu/drm/i915/gvt/scheduler.c int intel_vgpu_setup_submission(struct intel_vgpu *vgpu) · line 1417

Quoted 11 lines
drivers/gpu/drm/i915/gvt/scheduler.c +11 -0
+		i915_gem_ww_ctx_init(&ww, false);
+retry:
+		ret = intel_context_pin_ww(ce, &ww);
+		if (ret) {
+			if (ret == -EDEADLK) {
+				ret = i915_gem_ww_ctx_backoff(&ww);
+				if (!ret)
+					goto retry;
+			}
+			goto out_shadow_ctx;
+		}

I move the piece to create_workload. Similar to the change I made in my patch sent.

Quoted 14 lines
drivers/gpu/drm/i915/gvt/scheduler.c +2 -3
-		ret = intel_context_pin(ce);
   		intel_context_put(ce);
-		if (ret)
-			goto out_shadow_ctx;
+		i915_gem_ww_ctx_fini(&ww);
   		s->shadow[i] = ce;
   	}
@@ -1400,6 +1445,7 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
   	return 0;
   out_shadow_ctx:
+	i915_gem_ww_ctx_fini(&ww);
   	i915_context_ppgtt_root_restore(s, ppgtt);
   	for_each_engine(engine, vgpu->gvt->gt, i) {
   		if (IS_ERR(s->shadow[i]))
Signature · 2 lines
Best Regards,
Colin Xu
Signature · 7 lines
Best Regards,
Colin Xu

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