[Intel-gfx] [PATCH 3/4] drm/i915: Pass ww ctx to pin_map, v2.

[Maarten Lankhorst]

Op 29-04-2021 om 16:55 schreef Matthew Auld:
> On Thu, 29 Apr 2021 at 11:10, Maarten Lankhorst
> <maarten.lankhorst at linux.intel.com> wrote:
>> This will allow us to explicitly pass the ww to pin_pages,
>> when it starts taking it.
>>
>> This allows us to finally kill off the explicit passing of ww
>> by retrieving it from the obj.
>>
>> Changes since v1:
>> - Rename 'ret' to ptr, fix error handling of return ptr.
>>
>> Signed-off-by: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> <snip>
>
>> --- a/drivers/gpu/drm/i915/gt/intel_gtt.c
>> +++ b/drivers/gpu/drm/i915/gt/intel_gtt.c
>> @@ -65,7 +65,7 @@ int map_pt_dma_locked(struct i915_address_space *vm, struct drm_i915_gem_object
>>         void *vaddr;
>>
>>         type = i915_coherent_map_type(vm->i915, obj, true);
>> -       vaddr = i915_gem_object_pin_map(obj, type);
>> +       vaddr = i915_gem_object_pin_map(obj, NULL, type);
> Don't we need to pass the ww here? For all our page-tables this is
> both our allocate and map for the backing page(lmem or smem), so at
> least for the locked variant here would it not make sense to pass the
> ww from the vma_pin() etc, in case we need to evict stuff?


Actually yeah, seems to be called from i915_vma_pin_ww, so should pass ww along for sure.. I'll respin.