[PATCH] drm/amdgpu: fix an underflow on non-4KB-page systems

[Christian König]

Am 30.03.21 um 15:23 schrieb Dan Horák:
> On Tue, 30 Mar 2021 21:09:12 +0800
> Xi Ruoyao <xry111 at mengyan1223.wang> wrote:
>
>> On 2021-03-30 21:02 +0800, Xi Ruoyao wrote:
>>> On 2021-03-30 14:55 +0200, Christian König wrote:
>>>> I rather see this as a kernel bug. Can you test if this code fragment
>>>> fixes your issue:
>>>>
>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
>>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
>>>> index 64beb3399604..e1260b517e1b 100644
>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
>>>> @@ -780,7 +780,7 @@ int amdgpu_info_ioctl(struct drm_device *dev, void
>>>> *data, struct drm_file *filp)
>>>>                   }
>>>>                   dev_info->virtual_address_alignment =
>>>> max((int)PAGE_SIZE, AMDGPU_GPU_PAGE_SIZE);
>>>>                   dev_info->pte_fragment_size = (1 <<
>>>> adev->vm_manager.fragment_size) * AMDGPU_GPU_PAGE_SIZE;
>>>> -               dev_info->gart_page_size = AMDGPU_GPU_PAGE_SIZE;
>>>> +               dev_info->gart_page_size =
>>>> dev_info->virtual_address_alignment;
>>>>                   dev_info->cu_active_number = adev->gfx.cu_info.number;
>>>>                   dev_info->cu_ao_mask = adev->gfx.cu_info.ao_cu_mask;
>>>>                   dev_info->ce_ram_size = adev->gfx.ce_ram_size;
>>> It works.  I've seen it at
>>> https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fxen0n%2Flinux%2Fcommit%2F84ada72983838bd7ce54bc32f5d34ac5b5aae191&data=04%7C01%7Cchristian.koenig%40amd.com%7Cf37fddf20a8847edf67808d8f37ef23c%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637527074118791321%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=DZnmee38NGpiWRMX5LmlxOhxAzBMhAusnAWNnCxXTJ0%3D&reserved=0
>>> before (with a common sub-expression, though :).
>> Some comment: on an old version of Fedora ported by Loongson, Xorg just hangs
>> without this commit.  But on the system I built from source, I didn't see any
>> issue before Linux 5.11.  So I misbelieved that it was something already fixed.
>>
>> Dan: you can try it on your PPC 64 with non-4K page as well.
> yup, looks good here as well, ppc64le (Power9) system with 64KB pages

Mhm, as far as I can say this patch never made it to us.

Can you please send it to the mailing list with me on CC?

Thanks,
Christian.

>
>
> 		Dan