diff mbox series

[2/3] drm/i915: Store engine mask in intel_gt

Message ID 20191016122843.9429-2-tvrtko.ursulin@linux.intel.com (mailing list archive)
State New, archived
Headers show
Series [1/3] drm/i915: Make i915_gem_load_power_context take struct intel_gt | expand

Commit Message

Tvrtko Ursulin Oct. 16, 2019, 12:28 p.m. UTC
From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

Medium term goal is to eliminate the i915->engine[] array and to get there
we have recently introduced equivalent array in intel_gt. Now we need to
migrate the code further towards this state.

To allow for_each_engine_masked call sites to pass in gt instead of i915
we need to have a copy of INTEL_INFO()->engine_mask in intel_gt.

For this to work we also need to use engine->id as index when populating
the gt->engine[] array.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/gt/intel_engine_user.c | 4 ++--
 drivers/gpu/drm/i915/gt/intel_gt.c          | 5 +++++
 drivers/gpu/drm/i915/gt/intel_gt.h          | 1 +
 drivers/gpu/drm/i915/gt/intel_gt_types.h    | 4 ++++
 drivers/gpu/drm/i915/i915_drv.c             | 2 ++
 5 files changed, 14 insertions(+), 2 deletions(-)

Comments

Chris Wilson Oct. 16, 2019, 1:03 p.m. UTC | #1
Quoting Tvrtko Ursulin (2019-10-16 13:28:42)
> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> 
> Medium term goal is to eliminate the i915->engine[] array and to get there
> we have recently introduced equivalent array in intel_gt. Now we need to
> migrate the code further towards this state.
> 
> To allow for_each_engine_masked call sites to pass in gt instead of i915
> we need to have a copy of INTEL_INFO()->engine_mask in intel_gt.
> 
> For this to work we also need to use engine->id as index when populating
> the gt->engine[] array.
> 
> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> Cc: Chris Wilson <chris@chris-wilson.co.uk>
> ---
>  drivers/gpu/drm/i915/gt/intel_engine_user.c | 4 ++--
>  drivers/gpu/drm/i915/gt/intel_gt.c          | 5 +++++
>  drivers/gpu/drm/i915/gt/intel_gt.h          | 1 +
>  drivers/gpu/drm/i915/gt/intel_gt_types.h    | 4 ++++
>  drivers/gpu/drm/i915/i915_drv.c             | 2 ++
>  5 files changed, 14 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/gt/intel_engine_user.c b/drivers/gpu/drm/i915/gt/intel_engine_user.c
> index 77cd5de83930..099abae860ec 100644
> --- a/drivers/gpu/drm/i915/gt/intel_engine_user.c
> +++ b/drivers/gpu/drm/i915/gt/intel_engine_user.c
> @@ -183,8 +183,8 @@ static void add_legacy_ring(struct legacy_ring *ring,
>         if (unlikely(idx == -1))
>                 return;
>  
> -       GEM_BUG_ON(idx >= ARRAY_SIZE(ring->gt->engine));
> -       ring->gt->engine[idx] = engine;

This is used for the legacy uapi mapping. So we need to adjust the loop
in i915_gem_context.c::default_engines() to use engine->legacy_idx
instead of id. And all other users of engine->legacy_idx are now
backwards.
-Chris
Tvrtko Ursulin Oct. 16, 2019, 1:31 p.m. UTC | #2
On 16/10/2019 14:03, Chris Wilson wrote:
> Quoting Tvrtko Ursulin (2019-10-16 13:28:42)
>> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>>
>> Medium term goal is to eliminate the i915->engine[] array and to get there
>> we have recently introduced equivalent array in intel_gt. Now we need to
>> migrate the code further towards this state.
>>
>> To allow for_each_engine_masked call sites to pass in gt instead of i915
>> we need to have a copy of INTEL_INFO()->engine_mask in intel_gt.
>>
>> For this to work we also need to use engine->id as index when populating
>> the gt->engine[] array.
>>
>> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>> Cc: Chris Wilson <chris@chris-wilson.co.uk>
>> ---
>>   drivers/gpu/drm/i915/gt/intel_engine_user.c | 4 ++--
>>   drivers/gpu/drm/i915/gt/intel_gt.c          | 5 +++++
>>   drivers/gpu/drm/i915/gt/intel_gt.h          | 1 +
>>   drivers/gpu/drm/i915/gt/intel_gt_types.h    | 4 ++++
>>   drivers/gpu/drm/i915/i915_drv.c             | 2 ++
>>   5 files changed, 14 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/i915/gt/intel_engine_user.c b/drivers/gpu/drm/i915/gt/intel_engine_user.c
>> index 77cd5de83930..099abae860ec 100644
>> --- a/drivers/gpu/drm/i915/gt/intel_engine_user.c
>> +++ b/drivers/gpu/drm/i915/gt/intel_engine_user.c
>> @@ -183,8 +183,8 @@ static void add_legacy_ring(struct legacy_ring *ring,
>>          if (unlikely(idx == -1))
>>                  return;
>>   
>> -       GEM_BUG_ON(idx >= ARRAY_SIZE(ring->gt->engine));
>> -       ring->gt->engine[idx] = engine;
> 
> This is used for the legacy uapi mapping. So we need to adjust the loop
> in i915_gem_context.c::default_engines() to use engine->legacy_idx
> instead of id. And all other users of engine->legacy_idx are now
> backwards.

Hmmm that's evil.. I mean so well hidden dependency and contrary to 
years of decoupling efforts. But I gather some mess is expected in these 
circumstances. I'll update.

Regards,

Tvrtko
Chris Wilson Oct. 16, 2019, 1:41 p.m. UTC | #3
Quoting Tvrtko Ursulin (2019-10-16 14:31:45)
> 
> On 16/10/2019 14:03, Chris Wilson wrote:
> > Quoting Tvrtko Ursulin (2019-10-16 13:28:42)
> >> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> >>
> >> Medium term goal is to eliminate the i915->engine[] array and to get there
> >> we have recently introduced equivalent array in intel_gt. Now we need to
> >> migrate the code further towards this state.
> >>
> >> To allow for_each_engine_masked call sites to pass in gt instead of i915
> >> we need to have a copy of INTEL_INFO()->engine_mask in intel_gt.
> >>
> >> For this to work we also need to use engine->id as index when populating
> >> the gt->engine[] array.
> >>
> >> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
> >> Cc: Chris Wilson <chris@chris-wilson.co.uk>
> >> ---
> >>   drivers/gpu/drm/i915/gt/intel_engine_user.c | 4 ++--
> >>   drivers/gpu/drm/i915/gt/intel_gt.c          | 5 +++++
> >>   drivers/gpu/drm/i915/gt/intel_gt.h          | 1 +
> >>   drivers/gpu/drm/i915/gt/intel_gt_types.h    | 4 ++++
> >>   drivers/gpu/drm/i915/i915_drv.c             | 2 ++
> >>   5 files changed, 14 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/drivers/gpu/drm/i915/gt/intel_engine_user.c b/drivers/gpu/drm/i915/gt/intel_engine_user.c
> >> index 77cd5de83930..099abae860ec 100644
> >> --- a/drivers/gpu/drm/i915/gt/intel_engine_user.c
> >> +++ b/drivers/gpu/drm/i915/gt/intel_engine_user.c
> >> @@ -183,8 +183,8 @@ static void add_legacy_ring(struct legacy_ring *ring,
> >>          if (unlikely(idx == -1))
> >>                  return;
> >>   
> >> -       GEM_BUG_ON(idx >= ARRAY_SIZE(ring->gt->engine));
> >> -       ring->gt->engine[idx] = engine;
> > 
> > This is used for the legacy uapi mapping. So we need to adjust the loop
> > in i915_gem_context.c::default_engines() to use engine->legacy_idx
> > instead of id. And all other users of engine->legacy_idx are now
> > backwards.
> 
> Hmmm that's evil.. I mean so well hidden dependency and contrary to 
> years of decoupling efforts. But I gather some mess is expected in these 
> circumstances. I'll update.

The decoupling effort is on the uAPI side, I do think the effort in
moving all the implicit legacy layout to a subset of the explicit user
controls is paying off nicely. The next goal was to remove gt->engine[],
so assumptions that it existed and engine->id was an index into it were
removed.
-Chris
diff mbox series

Patch

diff --git a/drivers/gpu/drm/i915/gt/intel_engine_user.c b/drivers/gpu/drm/i915/gt/intel_engine_user.c
index 77cd5de83930..099abae860ec 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_user.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_user.c
@@ -183,8 +183,8 @@  static void add_legacy_ring(struct legacy_ring *ring,
 	if (unlikely(idx == -1))
 		return;
 
-	GEM_BUG_ON(idx >= ARRAY_SIZE(ring->gt->engine));
-	ring->gt->engine[idx] = engine;
+	GEM_BUG_ON(engine->id >= ARRAY_SIZE(ring->gt->engine));
+	ring->gt->engine[engine->id] = engine;
 	ring->instance++;
 
 	engine->legacy_idx = idx;
diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c b/drivers/gpu/drm/i915/gt/intel_gt.c
index b3619a2a5d0e..7a3a9925359b 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt.c
+++ b/drivers/gpu/drm/i915/gt/intel_gt.c
@@ -37,6 +37,11 @@  void intel_gt_init_hw_early(struct drm_i915_private *i915)
 	intel_gt_pm_disable(&i915->gt);
 }
 
+void intel_gt_init_mmio(struct intel_gt *gt, struct drm_i915_private *i915)
+{
+	gt->__info.engine_mask = INTEL_INFO(i915)->engine_mask;
+}
+
 static void init_unused_ring(struct intel_gt *gt, u32 base)
 {
 	struct intel_uncore *uncore = gt->uncore;
diff --git a/drivers/gpu/drm/i915/gt/intel_gt.h b/drivers/gpu/drm/i915/gt/intel_gt.h
index e6ab0bff0efb..a141f9208f69 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt.h
+++ b/drivers/gpu/drm/i915/gt/intel_gt.h
@@ -29,6 +29,7 @@  static inline struct intel_gt *huc_to_gt(struct intel_huc *huc)
 
 void intel_gt_init_early(struct intel_gt *gt, struct drm_i915_private *i915);
 void intel_gt_init_hw_early(struct drm_i915_private *i915);
+void intel_gt_init_mmio(struct intel_gt *gt, struct drm_i915_private *i915);
 int __must_check intel_gt_init_hw(struct intel_gt *gt);
 int intel_gt_init(struct intel_gt *gt);
 void intel_gt_driver_register(struct intel_gt *gt);
diff --git a/drivers/gpu/drm/i915/gt/intel_gt_types.h b/drivers/gpu/drm/i915/gt/intel_gt_types.h
index be4b263621c8..bb813e0a8fa3 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_gt_types.h
@@ -94,6 +94,10 @@  struct intel_gt {
 
 	u32 pm_guc_events;
 
+	struct {
+		intel_engine_mask_t engine_mask;
+	} __info; /* Hack to enable poly-morphic for_each_engine_masked. */
+
 	struct intel_engine_cs *engine[I915_NUM_ENGINES];
 	struct intel_engine_cs *engine_class[MAX_ENGINE_CLASS + 1]
 					    [MAX_ENGINE_INSTANCE + 1];
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index f02a34722217..4d8ee8c54abd 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -592,6 +592,8 @@  static int i915_driver_mmio_probe(struct drm_i915_private *dev_priv)
 
 	intel_device_info_init_mmio(dev_priv);
 
+	intel_gt_init_mmio(&dev_priv->gt, dev_priv);
+
 	intel_uncore_prune_mmio_domains(&dev_priv->uncore);
 
 	intel_uc_init_mmio(&dev_priv->gt.uc);