Re: [PATCH v3 1/2] drm/xe: Use dynamic allocation for tile and device VRAM region structures

[Jani Nikula <jani.nikula@linux.intel.com>]

On Tue, 24 Jun 2025, Michal Wajdeczko <michal.wajdeczko@intel.com> wrote:
> On 24.06.2025 11:22, Piórkowski, Piotr wrote:
>> diff --git a/drivers/gpu/drm/xe/xe_assert.h b/drivers/gpu/drm/xe/xe_assert.h
>> index 68fe70ce2be3..cd63cdd2dc78 100644
>> --- a/drivers/gpu/drm/xe/xe_assert.h
>> +++ b/drivers/gpu/drm/xe/xe_assert.h
>> @@ -145,7 +145,7 @@
>>  	const struct xe_tile *__tile = (tile);							\
>>  	char __buf[10] __maybe_unused;								\
>>  	xe_assert_msg(tile_to_xe(__tile), condition, "tile: %u VRAM %s\n" msg,			\
>> -		      __tile->id, ({ string_get_size(__tile->mem.vram.actual_physical_size, 1,	\
>> +		      __tile->id, ({ string_get_size(__tile->mem.vram->actual_physical_size, 1,	\
>>  				     STRING_UNITS_2, __buf, sizeof(__buf)); __buf; }), ## arg);	\
>>  })
>>  
>
> this assert will cause NPD on !DGFX, likely you need at least this:
>
> __tile->mem.vram ? __tile->mem.vram->actual_physical_size : 0

There'd be other benefits in adding abstractions for this, such as
making vram an opaque pointer and not requiring its definition be
included absolutely everywhere via xe_assert.h.

BR,
Jani.


-- 
Jani Nikula, Intel