[Intel-gfx] [PATCH] drm/i915: move intel_hws_csb_write_index() out of i915_drv.h

Wed Feb 9 10:00:00 UTC 2022

On Wed, 09 Feb 2022, Tvrtko Ursulin <tvrtko.ursulin at linux.intel.com> wrote:
> On 08/02/2022 17:16, Jani Nikula wrote:
>> Underscore prefix the index macros, and place
>> INTEL_HWS_CSB_WRITE_INDEX() as a macro next to them, to declutter
>> i915_drv.h.
>> 
>> Signed-off-by: Jani Nikula <jani.nikula at intel.com>
>> ---
>>   drivers/gpu/drm/i915/gt/intel_engine.h               | 6 ++++--
>>   drivers/gpu/drm/i915/gt/intel_execlists_submission.c | 2 +-
>>   drivers/gpu/drm/i915/gvt/execlist.c                  | 2 +-
>>   drivers/gpu/drm/i915/i915_drv.h                      | 8 --------
>>   4 files changed, 6 insertions(+), 12 deletions(-)
>> 
>> diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h
>> index 0e353d8c2bc8..faf26ed37d01 100644
>> --- a/drivers/gpu/drm/i915/gt/intel_engine.h
>> +++ b/drivers/gpu/drm/i915/gt/intel_engine.h
>> @@ -180,8 +180,10 @@ intel_write_status_page(struct intel_engine_cs *engine, int reg, u32 value)
>>   #define I915_GEM_HWS_SCRATCH		0x80
>>   
>>   #define I915_HWS_CSB_BUF0_INDEX		0x10
>> -#define I915_HWS_CSB_WRITE_INDEX	0x1f
>> -#define ICL_HWS_CSB_WRITE_INDEX		0x2f
>> +#define _I915_HWS_CSB_WRITE_INDEX	0x1f
>> +#define _ICL_HWS_CSB_WRITE_INDEX		0x2f
>
> I don't quite get why would these two be the only ones which need 
> underscore prefix?

The others are used directly, these two should only be used via
INTEL_HWS_CSB_WRITE_INDEX(), like they are. That's the hint with the
underscores. Matches what's done in i915_reg.h for example for register
instances and choosing the right register instance.

>
>> +#define INTEL_HWS_CSB_WRITE_INDEX(__i915) \
>> +	(GRAPHICS_VER(__i915) >= 11 ? _ICL_HWS_CSB_WRITE_INDEX : _I915_HWS_CSB_WRITE_INDEX)
>
> Secondly, on the point of the best new home for it, it is better than 
> i915_drv.h that is for sure. But is it the best I am not sure.

>From my POV this one's pretty clear. The index macros are here, and this
is a wrapper to choose the correct index macro. They should be together.

As to moving all of them somewhere in gt/gem, I think it's for follow-up
(by someone else).

BR,
Jani.

>
> CSB in general seems to have identity crisis with some bits being in 
> intel_engine.h, but some also in intel_lrc.h. Neither seems completely 
> right to me. It should all really be in intel_execlists_submission.h. 
> Unless someone would then make an argument the latter is about the i915 
> execlists scheduler backend and shouldn't strictly contain execlists 
> hardware definitions like the CSB layout. Still, at the moment for me it 
> feels like a better place than either intel_engine.h or intel_lrc.h.
>
> And probably all three execlists_ prefix functions from intel_engine.h 
> should go to intel_execlists_submission.h as well.
>
> We can leave the ramblings for later and for now just explain why 
> underscores please, depending on which r-b or not.
>
> Regards,
>
> Tvrtko
>
>>   
>>   void intel_engine_stop(struct intel_engine_cs *engine);
>>   void intel_engine_cleanup(struct intel_engine_cs *engine);
>> diff --git a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
>> index 9bb7c863172f..961d795220a3 100644
>> --- a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
>> +++ b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
>> @@ -3503,7 +3503,7 @@ int intel_execlists_submission_setup(struct intel_engine_cs *engine)
>>   		(u64 *)&engine->status_page.addr[I915_HWS_CSB_BUF0_INDEX];
>>   
>>   	execlists->csb_write =
>> -		&engine->status_page.addr[intel_hws_csb_write_index(i915)];
>> +		&engine->status_page.addr[INTEL_HWS_CSB_WRITE_INDEX(i915)];
>>   
>>   	if (GRAPHICS_VER(i915) < 11)
>>   		execlists->csb_size = GEN8_CSB_ENTRIES;
>> diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c
>> index c8dcda6d4f0d..66d354c4195b 100644
>> --- a/drivers/gpu/drm/i915/gvt/execlist.c
>> +++ b/drivers/gpu/drm/i915/gvt/execlist.c
>> @@ -163,7 +163,7 @@ static void emulate_csb_update(struct intel_vgpu_execlist *execlist,
>>   					       hwsp_gpa + I915_HWS_CSB_BUF0_INDEX * 4 + write_pointer * 8,
>>   					       status, 8);
>>   		intel_gvt_hypervisor_write_gpa(vgpu,
>> -					       hwsp_gpa + intel_hws_csb_write_index(execlist->engine->i915) * 4,
>> +					       hwsp_gpa + INTEL_HWS_CSB_WRITE_INDEX(execlist->engine->i915) * 4,
>>   					       &write_pointer, 4);
>>   	}
>>   
>> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
>> index 8c1706fd81f9..05656cc738d1 100644
>> --- a/drivers/gpu/drm/i915/i915_drv.h
>> +++ b/drivers/gpu/drm/i915/i915_drv.h
>> @@ -1739,14 +1739,6 @@ mkwrite_device_info(struct drm_i915_private *dev_priv)
>>   	return (struct intel_device_info *)INTEL_INFO(dev_priv);
>>   }
>>   
>> -static inline int intel_hws_csb_write_index(struct drm_i915_private *i915)
>> -{
>> -	if (GRAPHICS_VER(i915) >= 11)
>> -		return ICL_HWS_CSB_WRITE_INDEX;
>> -	else
>> -		return I915_HWS_CSB_WRITE_INDEX;
>> -}
>> -
>>   static inline enum i915_map_type
>>   i915_coherent_map_type(struct drm_i915_private *i915,
>>   		       struct drm_i915_gem_object *obj, bool always_coherent)

-- 
Jani Nikula, Intel Open Source Graphics Center