Skip to contentSkip to conversation

drm/i915: make context status notifier head be per engine

RFC PATCH In review [email protected] 4 participants 12 messages Mar 2, 2017 – Mar 17, 2017

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

RFC PATCH 1/1 drm/i915: make context status notifier head be per engine

From: Changbin Du <[email protected]>

hi, Daniel, Chris and All,
As you know, GVTg introduced the context status notifier to schedule the GVTg workload. At that time, the notifier is bound to GVTg context only, so GVTg is not aware of host workloads.

Now we are going to improve GVTg's guest workload scheduler policy, and add Guc emulation support for new Gen graphics. Both these two features require acknowledgment for all contexts running on hardware. (But will not alter host workload.) So here try to make some change.

The change is easy:
1. Move the context status notifier head from i915_gem_context to intel_engine_cs. Which means there is a notifier head per engine instead of per context. Execlist driver still call notifier for each context sched-in/out events of current engine.
2. At GVTg side, it binds a notifier_block for each physical engine at GVTg initialization period. Then GVTg can hear all context status events.

In this RFC patch, GVTg do nothing for host context event, but later will add a function there. But in any case, the notifier callback is a noop if this is no active vGPU.

Since intel_gvt_init() is called at early initialization stage and require the status notifier head has been initiated, I initiate it in intel_engine_setup(). I am not sure if this is the right place, please correct me if wrong. Thank you! Look forward to your feedback. :)

---
 drivers/gpu/drm/i915/gvt/gvt.h          |  2 +-
 drivers/gpu/drm/i915/gvt/scheduler.c    | 46 ++++++++++++++-------------------
 drivers/gpu/drm/i915/i915_gem_context.c |  1 -
 drivers/gpu/drm/i915/i915_gem_context.h |  3 ---
 drivers/gpu/drm/i915/intel_engine_cs.c  |  2 ++
 drivers/gpu/drm/i915/intel_lrc.c        |  3 ++-
 drivers/gpu/drm/i915/intel_ringbuffer.h |  3 +++
 7 files changed, 28 insertions(+), 32 deletions(-)
drivers/gpu/drm/i915/gvt/gvt.h +1 -1
index a41b322..6c1ec7d6 100644
@@ -162,7 +162,6 @@ struct intel_vgpu {
 	atomic_t running_workload_num;
 	DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES);
 	struct i915_gem_context *shadow_ctx;
-	struct notifier_block shadow_ctx_notifier_block;
 
 #if IS_ENABLED(CONFIG_DRM_I915_GVT_KVMGT)
 	struct {
@@ -233,6 +232,7 @@ struct intel_gvt {
 	struct intel_gvt_gtt gtt;
 	struct intel_gvt_opregion opregion;
 	struct intel_gvt_workload_scheduler scheduler;
+	struct notifier_block shadow_ctx_notifier_block[I915_NUM_ENGINES];
 	DECLARE_HASHTABLE(cmd_table, GVT_CMD_HASH_BITS);
 	struct intel_vgpu_type *types;
 	unsigned int num_types;
drivers/gpu/drm/i915/gvt/scheduler.c +20 -26

93 lines not on this page · raw message

drivers/gpu/drm/i915/i915_gem_context.c +0 -1
index 2624520..6e202fb 100644
@@ -335,7 +335,6 @@ __create_hw_context(struct drm_i915_private *dev_priv,
 	ctx->ring_size = 4 * PAGE_SIZE;
 	ctx->desc_template =
 		default_desc_template(dev_priv, dev_priv->mm.aliasing_ppgtt);
-	ATOMIC_INIT_NOTIFIER_HEAD(&ctx->status_notifier);
 
 	/* GuC requires the ring to be placed above GUC_WOPCM_TOP. If GuC is not
 	 * present or not in use we still need a small bias as ring wraparound
drivers/gpu/drm/i915/i915_gem_context.h +0 -3
index 0ac750b..e9c008f 100644
@@ -160,9 +160,6 @@ struct i915_gem_context {
 	/** desc_template: invariant fields for the HW context descriptor */
 	u32 desc_template;
 
-	/** status_notifier: list of callbacks for context-switch changes */
-	struct atomic_notifier_head status_notifier;
-
 	/** guilty_count: How many times this context has caused a GPU hang. */
 	unsigned int guilty_count;
 	/**
drivers/gpu/drm/i915/intel_engine_cs.c +2 -0
index 46c9474..9695daf 100644
@@ -105,6 +105,8 @@ intel_engine_setup(struct drm_i915_private *dev_priv,
 	/* Nothing to do here, execute in order of dependencies */
 	engine->schedule = NULL;
 
+	ATOMIC_INIT_NOTIFIER_HEAD(&engine->context_status_notifier);
+
 	dev_priv->engine[id] = engine;
 	return 0;
 }
drivers/gpu/drm/i915/intel_lrc.c +2 -1
index 39329d4..b0daadf 100644
@@ -306,7 +306,8 @@ execlists_context_status_change(struct drm_i915_gem_request *rq,
 	if (!IS_ENABLED(CONFIG_DRM_I915_GVT))
 		return;
 
-	atomic_notifier_call_chain(&rq->ctx->status_notifier, status, rq);
+	atomic_notifier_call_chain(&rq->engine->context_status_notifier,
+				   status, rq);
 }
 
 static void
drivers/gpu/drm/i915/intel_ringbuffer.h +3 -0
index 4091c97..a40f8a7 100644
@@ -408,6 +408,9 @@ struct intel_engine_cs {
 	 */
 	struct i915_gem_context *legacy_active_context;
 
+	/* status_notifier: list of callbacks for context-switch changes */
+	struct atomic_notifier_head context_status_notifier;
+
 	struct intel_engine_hangcheck hangcheck;
 
 	bool needs_cmd_parser;
Signature · 6 lines
2.7.4

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

✗ Fi.CI.BAT: warning for drm/i915: make context status notifier head be per engine

== Series Details ==

== Series Details ==

Series: drm/i915: make context status notifier head be per engine
URL : https://patchwork.freedesktop.org/series/20552/
State : warning

== Summary ==

Series 20552v1 drm/i915: make context status notifier head be per engine
https://patchwork.freedesktop.org/api/1.0/series/20552/revisions/1/mbox/

Test gem_exec_parallel:

        Subgroup basic:
                pass       -> DMESG-WARN (fi-skl-6700hq)
Test pm_rpm:
        Subgroup basic-pci-d3-state:
                incomplete -> PASS       (fi-byt-n2820) fdo#99740

fdo#99740 https://bugs.freedesktop.org/show_bug.cgi?id=99740

fi-bdw-5557u     total:278  pass:267  dwarn:0   dfail:0   fail:0   skip:11 
fi-bsw-n3050     total:278  pass:239  dwarn:0   dfail:0   fail:0   skip:39 
fi-bxt-j4205     total:278  pass:259  dwarn:0   dfail:0   fail:0   skip:19 
fi-bxt-t5700     total:278  pass:258  dwarn:0   dfail:0   fail:0   skip:20 
fi-byt-j1900     total:278  pass:251  dwarn:0   dfail:0   fail:0   skip:27 
fi-byt-n2820     total:278  pass:247  dwarn:0   dfail:0   fail:0   skip:31 
fi-hsw-4770      total:278  pass:262  dwarn:0   dfail:0   fail:0   skip:16 
fi-hsw-4770r     total:278  pass:262  dwarn:0   dfail:0   fail:0   skip:16 
fi-ilk-650       total:278  pass:228  dwarn:0   dfail:0   fail:0   skip:50 
fi-ivb-3520m     total:278  pass:260  dwarn:0   dfail:0   fail:0   skip:18 
fi-ivb-3770      total:278  pass:260  dwarn:0   dfail:0   fail:0   skip:18 
fi-kbl-7500u     total:278  pass:259  dwarn:1   dfail:0   fail:0   skip:18 
fi-skl-6260u     total:278  pass:268  dwarn:0   dfail:0   fail:0   skip:10 
fi-skl-6700hq    total:278  pass:260  dwarn:1   dfail:0   fail:0   skip:17 
fi-skl-6700k     total:278  pass:256  dwarn:4   dfail:0   fail:0   skip:18 
fi-skl-6770hq    total:278  pass:268  dwarn:0   dfail:0   fail:0   skip:10 
fi-snb-2520m     total:278  pass:250  dwarn:0   dfail:0   fail:0   skip:28 
fi-snb-2600      total:278  pass:249  dwarn:0   dfail:0   fail:0   skip:29 

b3c4a574937b2256a716b23037c8b94ffebd7c8a drm-tip: 2017y-03m-02d-08h-59m-08s UTC integration manifest 6fcfeda drm/i915: make context status notifier head be per engine

== Logs ==

For more details see: https://intel-gfx-ci.01.org/CI/Patchwork_4031/
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#2 lore raw
In reply to [email protected]

hi, is there any comments for this? I'd like hear chris and daniel's comments. :)

Thanks,
Changbin

Quoted 25 lines · On Thu, Mar 02, 2017 at 07:23:05PM +0800, [email protected] wrote:

From: Changbin Du <[email protected]>

hi, Daniel, Chris and All,
As you know, GVTg introduced the context status notifier to schedule the GVTg workload. At that time, the notifier is bound to GVTg context only, so GVTg is not aware of host workloads.

Now we are going to improve GVTg's guest workload scheduler policy, and add Guc emulation support for new Gen graphics. Both these two features require acknowledgment for all contexts running on hardware. (But will not alter host workload.) So here try to make some change.

The change is easy:
1. Move the context status notifier head from i915_gem_context to intel_engine_cs. Which means there is a notifier head per engine instead of per context. Execlist driver still call notifier for each context sched-in/out events of current engine.
2. At GVTg side, it binds a notifier_block for each physical engine at GVTg initialization period. Then GVTg can hear all context status events.

In this RFC patch, GVTg do nothing for host context event, but later will add a function there. But in any case, the notifier callback is a noop if this is no active vGPU.

Quoted 6 lines

Since intel_gvt_init() is called at early initialization stage and require the status notifier head has been initiated, I initiate it in intel_engine_setup(). I am not sure if this is the right place, please correct me if wrong. Thank you! Look forward to your feedback. :)

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

#3 lore raw

PATCH 1/1 drm/i915: make context status notifier head be per engine

In reply to [email protected]

From: Changbin Du <[email protected]>

GVTg has introduced the context status notifier to schedule the GVTg workload. At that time, the notifier is bound to GVTg context only, so GVTg is not aware of host workloads.

Now we are going to improve GVTg's guest workload scheduler policy, and add Guc emulation support for new Gen graphics. Both these two features require acknowledgment for all contexts running on hardware. (But will not alter host workload.) So here try to make some change.

The change is simple:
1. Move the context status notifier head from i915_gem_context to intel_engine_cs. Which means there is a notifier head per engine instead of per context. Execlist driver still call notifier for each context sched-in/out events of current engine.
2. At GVTg side, it binds a notifier_block for each physical engine at GVTg initialization period. Then GVTg can hear all context status events.

In this patch, GVTg do nothing for host context event, but later will add a function there. But in any case, the notifier callback is a noop if this is no active vGPU.

Since intel_gvt_init() is called at early initialization stage and require the status notifier head has been initiated, I initiate it in intel_engine_setup().

---
 drivers/gpu/drm/i915/gvt/gvt.h          |  2 +-
 drivers/gpu/drm/i915/gvt/scheduler.c    | 46 ++++++++++++++-------------------
 drivers/gpu/drm/i915/i915_gem_context.c |  1 -
 drivers/gpu/drm/i915/i915_gem_context.h |  3 ---
 drivers/gpu/drm/i915/intel_engine_cs.c  |  2 ++
 drivers/gpu/drm/i915/intel_lrc.c        |  3 ++-
 drivers/gpu/drm/i915/intel_ringbuffer.h |  3 +++
 7 files changed, 28 insertions(+), 32 deletions(-)
drivers/gpu/drm/i915/gvt/gvt.h +1 -1
index 2379192..6dfc48b 100644
@@ -162,7 +162,6 @@ struct intel_vgpu {
 	atomic_t running_workload_num;
 	DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES);
 	struct i915_gem_context *shadow_ctx;
-	struct notifier_block shadow_ctx_notifier_block;
 
 #if IS_ENABLED(CONFIG_DRM_I915_GVT_KVMGT)
 	struct {
@@ -233,6 +232,7 @@ struct intel_gvt {
 	struct intel_gvt_gtt gtt;
 	struct intel_gvt_opregion opregion;
 	struct intel_gvt_workload_scheduler scheduler;
+	struct notifier_block shadow_ctx_notifier_block[I915_NUM_ENGINES];
 	DECLARE_HASHTABLE(cmd_table, GVT_CMD_HASH_BITS);
 	struct intel_vgpu_type *types;
 	unsigned int num_types;
drivers/gpu/drm/i915/gvt/scheduler.c +20 -26

93 lines not on this page · raw message

drivers/gpu/drm/i915/i915_gem_context.c +0 -1
index baceca1..0005085 100644
@@ -318,7 +318,6 @@ __create_hw_context(struct drm_i915_private *dev_priv,
 	ctx->ring_size = 4 * PAGE_SIZE;
 	ctx->desc_template =
 		default_desc_template(dev_priv, dev_priv->mm.aliasing_ppgtt);
-	ATOMIC_INIT_NOTIFIER_HEAD(&ctx->status_notifier);
 
 	/* GuC requires the ring to be placed above GUC_WOPCM_TOP. If GuC is not
 	 * present or not in use we still need a small bias as ring wraparound
drivers/gpu/drm/i915/i915_gem_context.h +0 -3
index 81268c9..4af2ab94 100644
@@ -158,9 +158,6 @@ struct i915_gem_context {
 	/** desc_template: invariant fields for the HW context descriptor */
 	u32 desc_template;
 
-	/** status_notifier: list of callbacks for context-switch changes */
-	struct atomic_notifier_head status_notifier;
-
 	/** guilty_count: How many times this context has caused a GPU hang. */
 	unsigned int guilty_count;
 	/**
drivers/gpu/drm/i915/intel_engine_cs.c +2 -0
index 73fe718..695693a 100644
@@ -105,6 +105,8 @@ intel_engine_setup(struct drm_i915_private *dev_priv,
 	/* Nothing to do here, execute in order of dependencies */
 	engine->schedule = NULL;
 
+	ATOMIC_INIT_NOTIFIER_HEAD(&engine->context_status_notifier);
+
 	dev_priv->engine[id] = engine;
 	return 0;
 }
drivers/gpu/drm/i915/intel_lrc.c +2 -1
index 89f38e7..808ad46 100644
@@ -306,7 +306,8 @@ execlists_context_status_change(struct drm_i915_gem_request *rq,
 	if (!IS_ENABLED(CONFIG_DRM_I915_GVT))
 		return;
 
-	atomic_notifier_call_chain(&rq->ctx->status_notifier, status, rq);
+	atomic_notifier_call_chain(&rq->engine->context_status_notifier,
+				   status, rq);
 }
 
 static void
drivers/gpu/drm/i915/intel_ringbuffer.h +3 -0
index 0ef491d..b7d42a5 100644
@@ -408,6 +408,9 @@ struct intel_engine_cs {
 	 */
 	struct i915_gem_context *legacy_active_context;
 
+	/* status_notifier: list of callbacks for context-switch changes */
+	struct atomic_notifier_head context_status_notifier;
+
 	struct intel_engine_hangcheck hangcheck;
 
 	bool needs_cmd_parser;
Signature · 6 lines
2.7.4

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v1 1/1
#4 lore raw
Quoted 24 lines · On Thu, Mar 09, 2017 at 07:27:24PM +0800, [email protected] wrote:

From: Changbin Du <[email protected]>

GVTg has introduced the context status notifier to schedule the GVTg workload. At that time, the notifier is bound to GVTg context only, so GVTg is not aware of host workloads.

Now we are going to improve GVTg's guest workload scheduler policy, and add Guc emulation support for new Gen graphics. Both these two features require acknowledgment for all contexts running on hardware. (But will not alter host workload.) So here try to make some change.

The change is simple:
1. Move the context status notifier head from i915_gem_context to intel_engine_cs. Which means there is a notifier head per engine instead of per context. Execlist driver still call notifier for each context sched-in/out events of current engine.
2. At GVTg side, it binds a notifier_block for each physical engine at GVTg initialization period. Then GVTg can hear all context status events.

In this patch, GVTg do nothing for host context event, but later will add a function there. But in any case, the notifier callback is a noop if this is no active vGPU.

Quoted 5 lines

Since intel_gvt_init() is called at early initialization stage and require the status notifier head has been initiated, I initiate it in intel_engine_setup().

I presume you will apply this via gvt?

drivers/gpu/drm/i915/gvt/scheduler.c static int populate_shadow_context(struct intel_vgpu_workload *workload) · line 130

Quoted 9 lines
drivers/gpu/drm/i915/gvt/scheduler.c +1 -0
index d3a56c9..64875ec 100644
@@ -127,15 +127,14 @@ static int populate_shadow_context(struct intel_vgpu_workload *workload)
 	return 0;
 }
 
+

Bonus newline

Quoted 16 lines
drivers/gpu/drm/i915/gvt/scheduler.c +4 -6
 static int shadow_context_status_change(struct notifier_block *nb,
 		unsigned long action, void *data)
 {
-	struct intel_vgpu *vgpu = container_of(nb,
-			struct intel_vgpu, shadow_ctx_notifier_block);
-	struct drm_i915_gem_request *req =
-		(struct drm_i915_gem_request *)data;
-	struct intel_gvt_workload_scheduler *scheduler =
-		&vgpu->gvt->scheduler;
+	struct drm_i915_gem_request *req = (struct drm_i915_gem_request *)data;
+	struct intel_gvt *gvt = container_of(nb, struct intel_gvt,
+				shadow_ctx_notifier_block[req->engine->id]);
+	struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
 	struct intel_vgpu_workload *workload =
 		scheduler->current_workload[req->engine->id];
 
Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#5 lore raw

hi, chris,

Quoted 25 lines · Chris Wilson
Quoted 24 lines · On Thu, Mar 09, 2017 at 07:27:24PM +0800, [email protected] wrote:

From: Changbin Du <[email protected]>

GVTg has introduced the context status notifier to schedule the GVTg workload. At that time, the notifier is bound to GVTg context only, so GVTg is not aware of host workloads.

Now we are going to improve GVTg's guest workload scheduler policy, and add Guc emulation support for new Gen graphics. Both these two features require acknowledgment for all contexts running on hardware. (But will not alter host workload.) So here try to make some change.

The change is simple:
1. Move the context status notifier head from i915_gem_context to intel_engine_cs. Which means there is a notifier head per engine instead of per context. Execlist driver still call notifier for each context sched-in/out events of current engine.
2. At GVTg side, it binds a notifier_block for each physical engine at GVTg initialization period. Then GVTg can hear all context status events.

In this patch, GVTg do nothing for host context event, but later will add a function there. But in any case, the notifier callback is a noop if this is no active vGPU.

Quoted 9 lines
Quoted 5 lines

Since intel_gvt_init() is called at early initialization stage and require the status notifier head has been initiated, I initiate it in intel_engine_setup().

I presume you will apply this via gvt?

Sure, I'll sync with zhenyu about this. Then update patch with below 'bonus newline' fixed. Thanks.

Quoted 31 lines

drivers/gpu/drm/i915/gvt/scheduler.c static int populate_shadow_context(struct intel_vgpu_workload *workload) · line 130

Quoted 9 lines
drivers/gpu/drm/i915/gvt/scheduler.c +1 -0
index d3a56c9..64875ec 100644
@@ -127,15 +127,14 @@ static int populate_shadow_context(struct intel_vgpu_workload *workload)
 	return 0;
 }
 
+

Bonus newline

Quoted 16 lines
drivers/gpu/drm/i915/gvt/scheduler.c +4 -6
 static int shadow_context_status_change(struct notifier_block *nb,
 		unsigned long action, void *data)
 {
-	struct intel_vgpu *vgpu = container_of(nb,
-			struct intel_vgpu, shadow_ctx_notifier_block);
-	struct drm_i915_gem_request *req =
-		(struct drm_i915_gem_request *)data;
-	struct intel_gvt_workload_scheduler *scheduler =
-		&vgpu->gvt->scheduler;
+	struct drm_i915_gem_request *req = (struct drm_i915_gem_request *)data;
+	struct intel_gvt *gvt = container_of(nb, struct intel_gvt,
+				shadow_ctx_notifier_block[req->engine->id]);
+	struct intel_gvt_workload_scheduler *scheduler = &gvt->scheduler;
 	struct intel_vgpu_workload *workload =
 		scheduler->current_workload[req->engine->id];
 
Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
Signature · 2 lines
Thanks,
Changbin Du

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

#6 lore raw

PATCH v2 1/1 drm/i915: make context status notifier head be per engine

In reply to [email protected]

From: Changbin Du <[email protected]>

GVTg has introduced the context status notifier to schedule the GVTg workload. At that time, the notifier is bound to GVTg context only, so GVTg is not aware of host workloads.

Now we are going to improve GVTg's guest workload scheduler policy, and add Guc emulation support for new Gen graphics. Both these two features require acknowledgment for all contexts running on hardware. (But will not alter host workload.) So here try to make some change.

The change is simple:
1. Move the context status notifier head from i915_gem_context to intel_engine_cs. Which means there is a notifier head per engine instead of per context. Execlist driver still call notifier for each context sched-in/out events of current engine.
2. At GVTg side, it binds a notifier_block for each physical engine at GVTg initialization period. Then GVTg can hear all context status events.

In this patch, GVTg do nothing for host context event, but later will add a function there. But in any case, the notifier callback is a noop if this is no active vGPU.

Since intel_gvt_init() is called at early initialization stage and require the status notifier head has been initiated, I initiate it in intel_engine_setup().

v2: remove a redundant newline. (chris)

---
 drivers/gpu/drm/i915/gvt/gvt.h          |  2 +-
 drivers/gpu/drm/i915/gvt/scheduler.c    | 45 ++++++++++++++-------------------
 drivers/gpu/drm/i915/i915_gem_context.c |  1 -
 drivers/gpu/drm/i915/i915_gem_context.h |  3 ---
 drivers/gpu/drm/i915/intel_engine_cs.c  |  2 ++
 drivers/gpu/drm/i915/intel_lrc.c        |  3 ++-
 drivers/gpu/drm/i915/intel_ringbuffer.h |  3 +++
 7 files changed, 27 insertions(+), 32 deletions(-)
drivers/gpu/drm/i915/gvt/gvt.h +1 -1
index 2379192..6dfc48b 100644
@@ -162,7 +162,6 @@ struct intel_vgpu {
 	atomic_t running_workload_num;
 	DECLARE_BITMAP(tlb_handle_pending, I915_NUM_ENGINES);
 	struct i915_gem_context *shadow_ctx;
-	struct notifier_block shadow_ctx_notifier_block;
 
 #if IS_ENABLED(CONFIG_DRM_I915_GVT_KVMGT)
 	struct {
@@ -233,6 +232,7 @@ struct intel_gvt {
 	struct intel_gvt_gtt gtt;
 	struct intel_gvt_opregion opregion;
 	struct intel_gvt_workload_scheduler scheduler;
+	struct notifier_block shadow_ctx_notifier_block[I915_NUM_ENGINES];
 	DECLARE_HASHTABLE(cmd_table, GVT_CMD_HASH_BITS);
 	struct intel_vgpu_type *types;
 	unsigned int num_types;
drivers/gpu/drm/i915/gvt/scheduler.c +19 -26

89 lines not on this page · raw message

drivers/gpu/drm/i915/i915_gem_context.c +0 -1
index baceca1..0005085 100644
@@ -318,7 +318,6 @@ __create_hw_context(struct drm_i915_private *dev_priv,
 	ctx->ring_size = 4 * PAGE_SIZE;
 	ctx->desc_template =
 		default_desc_template(dev_priv, dev_priv->mm.aliasing_ppgtt);
-	ATOMIC_INIT_NOTIFIER_HEAD(&ctx->status_notifier);
 
 	/* GuC requires the ring to be placed above GUC_WOPCM_TOP. If GuC is not
 	 * present or not in use we still need a small bias as ring wraparound
drivers/gpu/drm/i915/i915_gem_context.h +0 -3
index 81268c9..4af2ab94 100644
@@ -158,9 +158,6 @@ struct i915_gem_context {
 	/** desc_template: invariant fields for the HW context descriptor */
 	u32 desc_template;
 
-	/** status_notifier: list of callbacks for context-switch changes */
-	struct atomic_notifier_head status_notifier;
-
 	/** guilty_count: How many times this context has caused a GPU hang. */
 	unsigned int guilty_count;
 	/**
drivers/gpu/drm/i915/intel_engine_cs.c +2 -0
index 73fe718..695693a 100644
@@ -105,6 +105,8 @@ intel_engine_setup(struct drm_i915_private *dev_priv,
 	/* Nothing to do here, execute in order of dependencies */
 	engine->schedule = NULL;
 
+	ATOMIC_INIT_NOTIFIER_HEAD(&engine->context_status_notifier);
+
 	dev_priv->engine[id] = engine;
 	return 0;
 }
drivers/gpu/drm/i915/intel_lrc.c +2 -1
index 89f38e7..808ad46 100644
@@ -306,7 +306,8 @@ execlists_context_status_change(struct drm_i915_gem_request *rq,
 	if (!IS_ENABLED(CONFIG_DRM_I915_GVT))
 		return;
 
-	atomic_notifier_call_chain(&rq->ctx->status_notifier, status, rq);
+	atomic_notifier_call_chain(&rq->engine->context_status_notifier,
+				   status, rq);
 }
 
 static void
drivers/gpu/drm/i915/intel_ringbuffer.h +3 -0
index 0ef491d..b7d42a5 100644
@@ -408,6 +408,9 @@ struct intel_engine_cs {
 	 */
 	struct i915_gem_context *legacy_active_context;
 
+	/* status_notifier: list of callbacks for context-switch changes */
+	struct atomic_notifier_head context_status_notifier;
+
 	struct intel_engine_hangcheck hangcheck;
 
 	bool needs_cmd_parser;
Signature · 6 lines
2.7.4

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Applied
#7 lore raw
Quoted 24 lines · On 2017.03.13 10:47:11 +0800, [email protected] wrote:

From: Changbin Du <[email protected]>

GVTg has introduced the context status notifier to schedule the GVTg workload. At that time, the notifier is bound to GVTg context only, so GVTg is not aware of host workloads.

Now we are going to improve GVTg's guest workload scheduler policy, and add Guc emulation support for new Gen graphics. Both these two features require acknowledgment for all contexts running on hardware. (But will not alter host workload.) So here try to make some change.

The change is simple:
1. Move the context status notifier head from i915_gem_context to intel_engine_cs. Which means there is a notifier head per engine instead of per context. Execlist driver still call notifier for each context sched-in/out events of current engine.
2. At GVTg side, it binds a notifier_block for each physical engine at GVTg initialization period. Then GVTg can hear all context status events.

In this patch, GVTg do nothing for host context event, but later will add a function there. But in any case, the notifier callback is a noop if this is no active vGPU.

Quoted 8 lines

Since intel_gvt_init() is called at early initialization stage and require the status notifier head has been initiated, I initiate it in intel_engine_setup().

v2: remove a redundant newline. (chris)

Applied to gvt-next, thanks!

Signature · 3 lines
Open Source Technology Center, Intel ltd.

$gpg --keyserver wwwkeys.pgp.net --recv-keys 4D781827

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

#8 lore raw
Quoted 29 lines · Zhenyu Wang
Quoted 28 lines · On 2017.03.13 10:47:11 +0800, [email protected] wrote:

From: Changbin Du <[email protected]>

GVTg has introduced the context status notifier to schedule the GVTg workload. At that time, the notifier is bound to GVTg context only, so GVTg is not aware of host workloads.

Now we are going to improve GVTg's guest workload scheduler policy, and add Guc emulation support for new Gen graphics. Both these two features require acknowledgment for all contexts running on hardware. (But will not alter host workload.) So here try to make some change.

The change is simple:
1. Move the context status notifier head from i915_gem_context to intel_engine_cs. Which means there is a notifier head per engine instead of per context. Execlist driver still call notifier for each context sched-in/out events of current engine.
2. At GVTg side, it binds a notifier_block for each physical engine at GVTg initialization period. Then GVTg can hear all context status events.

In this patch, GVTg do nothing for host context event, but later will add a function there. But in any case, the notifier callback is a noop if this is no active vGPU.

Since intel_gvt_init() is called at early initialization stage and require the status notifier head has been initiated, I initiate it in intel_engine_setup().

Quoted 6 lines
Quoted 4 lines

v2: remove a redundant newline. (chris)

Applied to gvt-next, thanks!

It turns out this fixes a use-after-free bug that CI was falling over. Applied to dinq.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100232
-Chris

Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#9 lore raw
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines

Since intel_gvt_init() is called at early initialization stage and require the status notifier head has been initiated, I initiate it in intel_engine_setup().

v2: remove a redundant newline. (chris)

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

Applied to gvt-next, thanks!

It turns out this fixes a use-after-free bug that CI was falling over. Applied to dinq.

wow, that's out of my expectation. lucky!

Quoted 9 lines

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100232
-Chris

Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
intel-gvt-dev mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gvt-dev
Signature · 2 lines
Thanks,
Changbin Du

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

#10 lore raw

✓ Fi.CI.BAT: success for drm/i915: make context status notifier head be per engine (rev2)

In reply to [email protected]
== Series Details ==

== Series Details ==

Series: drm/i915: make context status notifier head be per engine (rev2)
URL : https://patchwork.freedesktop.org/series/20552/
State : success

== Summary ==

Series 20552v2 drm/i915: make context status notifier head be per engine
https://patchwork.freedesktop.org/api/1.0/series/20552/revisions/2/mbox/

Test gem_ctx_create:

        Subgroup basic-files:
                incomplete -> PASS       (fi-skl-6700k) fdo#100130
Test gem_exec_flush:
        Subgroup basic-uc-ro-default:
                incomplete -> PASS       (fi-kbl-7500u) fdo#100130

fdo#100130 https://bugs.freedesktop.org/show_bug.cgi?id=100130

fi-bdw-5557u     total:278  pass:267  dwarn:0   dfail:0   fail:0   skip:11  time: 465s
fi-bsw-n3050     total:278  pass:239  dwarn:0   dfail:0   fail:0   skip:39  time: 613s
fi-bxt-j4205     total:278  pass:259  dwarn:0   dfail:0   fail:0   skip:19  time: 532s
fi-bxt-t5700     total:278  pass:258  dwarn:0   dfail:0   fail:0   skip:20  time: 612s
fi-byt-j1900     total:278  pass:251  dwarn:0   dfail:0   fail:0   skip:27  time: 506s
fi-byt-n2820     total:278  pass:247  dwarn:0   dfail:0   fail:0   skip:31  time: 500s
fi-hsw-4770      total:278  pass:262  dwarn:0   dfail:0   fail:0   skip:16  time: 443s
fi-hsw-4770r     total:278  pass:262  dwarn:0   dfail:0   fail:0   skip:16  time: 432s
fi-ilk-650       total:278  pass:228  dwarn:0   dfail:0   fail:0   skip:50  time: 447s
fi-ivb-3520m     total:278  pass:260  dwarn:0   dfail:0   fail:0   skip:18  time: 505s
fi-ivb-3770      total:278  pass:260  dwarn:0   dfail:0   fail:0   skip:18  time: 491s
fi-kbl-7500u     total:278  pass:260  dwarn:0   dfail:0   fail:0   skip:18  time: 470s
fi-skl-6260u     total:278  pass:268  dwarn:0   dfail:0   fail:0   skip:10  time: 509s
fi-skl-6700hq    total:278  pass:261  dwarn:0   dfail:0   fail:0   skip:17  time: 597s
fi-skl-6700k     total:53   pass:45   dwarn:0   dfail:0   fail:0   skip:7   time: 0s
fi-skl-6770hq    total:278  pass:268  dwarn:0   dfail:0   fail:0   skip:10  time: 547s
fi-snb-2520m     total:278  pass:250  dwarn:0   dfail:0   fail:0   skip:28  time: 558s
fi-snb-2600      total:278  pass:249  dwarn:0   dfail:0   fail:0   skip:29  time: 419s

789c80247ee92848a46d66a32ca723e16a261913 drm-tip: 2017y-03m-09d-12h-32m-54s UTC integration manifest f793ac6 drm/i915: make context status notifier head be per engine

== Logs ==

For more details see: https://intel-gfx-ci.01.org/CI/Patchwork_4115/
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#11 lore raw

✗ Fi.CI.BAT: failure for drm/i915: make context status notifier head be per engine (rev3)

In reply to [email protected]
== Series Details ==

== Series Details ==

Series: drm/i915: make context status notifier head be per engine (rev3)
URL : https://patchwork.freedesktop.org/series/20552/
State : failure

== Summary ==

Series 20552v3 drm/i915: make context status notifier head be per engine
https://patchwork.freedesktop.org/api/1.0/series/20552/revisions/3/mbox/

Test drv_module_reload:

        Subgroup basic-reload-inject:
                pass       -> INCOMPLETE (fi-hsw-4770r)
Test gem_exec_flush:
        Subgroup basic-batch-kernel-default-uc:
                pass       -> FAIL       (fi-snb-2600) fdo#100007
Test kms_force_connector_basic:
        Subgroup prune-stale-modes:
                skip       -> PASS       (fi-snb-2520m)

fdo#100007 https://bugs.freedesktop.org/show_bug.cgi?id=100007

fi-bdw-5557u     total:278  pass:267  dwarn:0   dfail:0   fail:0   skip:11  time: 454s
fi-bsw-n3050     total:278  pass:239  dwarn:0   dfail:0   fail:0   skip:39  time: 624s
fi-bxt-j4205     total:278  pass:259  dwarn:0   dfail:0   fail:0   skip:19  time: 535s
fi-bxt-t5700     total:278  pass:258  dwarn:0   dfail:0   fail:0   skip:20  time: 579s
fi-byt-j1900     total:278  pass:251  dwarn:0   dfail:0   fail:0   skip:27  time: 505s
fi-byt-n2820     total:278  pass:247  dwarn:0   dfail:0   fail:0   skip:31  time: 504s
fi-hsw-4770      total:278  pass:262  dwarn:0   dfail:0   fail:0   skip:16  time: 441s
fi-hsw-4770r     total:276  pass:260  dwarn:0   dfail:0   fail:0   skip:15  time: 0s
fi-ilk-650       total:278  pass:228  dwarn:0   dfail:0   fail:0   skip:50  time: 445s
fi-ivb-3520m     total:278  pass:260  dwarn:0   dfail:0   fail:0   skip:18  time: 495s
fi-ivb-3770      total:278  pass:260  dwarn:0   dfail:0   fail:0   skip:18  time: 491s
fi-kbl-7500u     total:278  pass:259  dwarn:1   dfail:0   fail:0   skip:18  time: 474s
fi-skl-6260u     total:278  pass:268  dwarn:0   dfail:0   fail:0   skip:10  time: 502s
fi-skl-6700hq    total:278  pass:261  dwarn:0   dfail:0   fail:0   skip:17  time: 585s
fi-skl-6700k     total:278  pass:256  dwarn:4   dfail:0   fail:0   skip:18  time: 496s
fi-skl-6770hq    total:278  pass:268  dwarn:0   dfail:0   fail:0   skip:10  time: 543s
fi-snb-2520m     total:278  pass:250  dwarn:0   dfail:0   fail:0   skip:28  time: 567s
fi-snb-2600      total:278  pass:248  dwarn:0   dfail:0   fail:1   skip:29  time: 419s

2cb12884655eab41d4992b33ccb36c609c4537d3 drm-tip: 2017y-03m-12d-13h-00m-05s UTC integration manifest 2bd4f8a drm/i915: make context status notifier head be per engine

== Logs ==

For more details see: https://intel-gfx-ci.01.org/CI/Patchwork_4147/
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx