[PATCH 1/2] drm/amdgpu: Don't change preferred domian when fallback GTT v3

zhoucm1 zhoucm1 at amd.com
Tue Mar 20 03:22:46 UTC 2018



On 2018年03月19日 18:03, Christian König wrote:
> Am 19.03.2018 um 11:00 schrieb Chunming Zhou:
>> v2: add sanity checking
>> v3: make code open
>>
>> Change-Id: I2cf672ad36b8b4cc1a6b2e704f786bf6a155d9ce
>> Signed-off-by: Chunming Zhou <david1.zhou at amd.com>
>
> Reviewed-by: Christian König <christian.koenig at amd.com>
With more thinking, this change breaks the order of fallback.
visible to invisible should be first.
So I will make v4 for handle visible to invisible fallback.

Regards,
David Zhou
>
>> ---
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c    |  5 -----
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 16 ++++++++++++----
>>   2 files changed, 12 insertions(+), 9 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c 
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
>> index 6e6570ff9f8b..660f5e44b1a5 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
>> @@ -85,11 +85,6 @@ int amdgpu_gem_object_create(struct amdgpu_device 
>> *adev, unsigned long size,
>>                   flags &= ~AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
>>                   goto retry;
>>               }
>> -
>> -            if (initial_domain == AMDGPU_GEM_DOMAIN_VRAM) {
>> -                initial_domain |= AMDGPU_GEM_DOMAIN_GTT;
>> -                goto retry;
>> -            }
>>               DRM_DEBUG("Failed to allocate GEM object (%ld, %d, %u, 
>> %d)\n",
>>                     size, initial_domain, alignment, r);
>>           }
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c 
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>> index b3310219e0ac..e167e98c746c 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
>> @@ -440,13 +440,21 @@ static int amdgpu_bo_do_create(struct 
>> amdgpu_device *adev,
>>   #endif
>>         bo->tbo.bdev = &adev->mman.bdev;
>> -    amdgpu_ttm_placement_from_domain(bo, domain);
>> -
>> +    amdgpu_ttm_placement_from_domain(bo, bo->preferred_domains);
>>       r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, type,
>>                    &bo->placement, page_align, &ctx, acc_size,
>>                    sg, resv, &amdgpu_ttm_bo_destroy);
>> -    if (unlikely(r != 0))
>> -        return r;
>> +
>> +    if (unlikely(r && r != -ERESTARTSYS) && bo->allowed_domains !=
>> +        bo->preferred_domains) {
>> +        amdgpu_ttm_placement_from_domain(bo, bo->allowed_domains);
>> +        r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, 
>> type,
>> +                     &bo->placement, page_align, &ctx,
>> +                     acc_size, sg, resv,
>> +                     &amdgpu_ttm_bo_destroy);
>> +        if (unlikely(r != 0))
>> +            return r;
>> +    }
>>         if (adev->gmc.visible_vram_size < adev->gmc.real_vram_size &&
>>           bo->tbo.mem.mem_type == TTM_PL_VRAM &&
>



More information about the amd-gfx mailing list