[patch] problems with "fix visible VRAM handling during faults"

[Christian König]

Am 08.05.24 um 12:17 schrieb Michel Dänzer:
> On 2024-05-07 18:39, Jeremy Day wrote:
>> This is just to report that I've had usually well-behaved applications
>> sometimes having problems with memory access violations since kernel
>> version 6.9-rc5.  This past weekend I stumbled across a way to reliably
>> reproduce the problem in the form of a Skyrim save file which causes a
>> crash shortly after loading the game on affected kernels.
>>
>> Things go back to running smoothly only if I revert one of the changes
>> in 5th April's "[PATCH] drm/amdgpu: fix visible VRAM handling during
>> faults" as follows.
>>
>> Patch is against v6.9-rc7.  It restores the check for partially
>> visible-to-cpu memory in amdgpu_bo_fault_reserve_notify.  Things
>> seem stable again with this change.
> Does this instead of your patch help by any chance?
>
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> index 109fe557a02b..29c197c00018 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> @@ -427,7 +427,7 @@ bool amdgpu_res_cpu_visible(struct amdgpu_device *adev,
>
>          amdgpu_res_first(res, 0, res->size, &cursor);
>          while (cursor.remaining) {
> -               if ((cursor.start + cursor.size) >= adev->gmc.visible_vram_size)
> +               if ((cursor.start + cursor.size) > adev->gmc.visible_vram_size)

Oh, good catch. Yes that might be it.

Thanks a lot,
Christian.

>                          return false;
>                  amdgpu_res_next(&cursor, cursor.size);
>          }
>
>