LinuxLists.cc - Re: 6.7/regression/KASAN: null-ptr-deref in amdgpu_ras_reset_error

2023-11-15 18:16:13

Subject: Re: 6.7/regression/KASAN: null-ptr-deref in amdgpu_ras_reset_error_count+0x2d6

On Wed, Nov 8, 2023 at 4:24 AM Mikhail Gavrilov
<[email protected]> wrote:
>
> Tested-by: Mikhail Gavrilov <[email protected]>
> Thanks, after applying the patch GPU loading meets expectations.
> Games are working so overall all looking good for now.
>

Unfortunately I didn't check this patch on 7900XTX.
And now I see that this patch leads the systems with 7900XTX GPU to a
lot of protection fault messages in the kernel log.

[ 75.211702] gmc_v11_0_process_interrupt: 164 callbacks suppressed
[ 75.211706] amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault
(src_id:0 ring:169 vmid:0 pasid:0, for process pid 0 thread pid 0)
[ 75.211715] amdgpu 0000:03:00.0: amdgpu: in page starting at
address 0x00008087fe5fd000 from client 10
[ 75.211718] amdgpu 0000:03:00.0: amdgpu:
GCVM_L2_PROTECTION_FAULT_STATUS:0x00041D53
[ 75.211720] amdgpu 0000:03:00.0: amdgpu: Faulty UTCL2 client ID: SDMA1 (0xe)
[ 75.211723] amdgpu 0000:03:00.0: amdgpu: MORE_FAULTS: 0x1
[ 75.211725] amdgpu 0000:03:00.0: amdgpu: WALKER_ERROR: 0x1
[ 75.211727] amdgpu 0000:03:00.0: amdgpu: PERMISSION_FAULTS: 0x5
[ 75.211729] amdgpu 0000:03:00.0: amdgpu: MAPPING_ERROR: 0x1
[ 75.211732] amdgpu 0000:03:00.0: amdgpu: RW: 0x1
[ 75.211740] amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault
(src_id:0 ring:169 vmid:0 pasid:0, for process pid 0 thread pid 0)
[ 75.211743] amdgpu 0000:03:00.0: amdgpu: in page starting at
address 0x00008087bffff000 from client 10
[ 75.211746] amdgpu 0000:03:00.0: amdgpu:
GCVM_L2_PROTECTION_FAULT_STATUS:0x00000000
[ 75.211748] amdgpu 0000:03:00.0: amdgpu: Faulty UTCL2 client ID: CB/DB (0x0)
[ 75.211750] amdgpu 0000:03:00.0: amdgpu: MORE_FAULTS: 0x0
[ 75.211752] amdgpu 0000:03:00.0: amdgpu: WALKER_ERROR: 0x0
[ 75.211754] amdgpu 0000:03:00.0: amdgpu: PERMISSION_FAULTS: 0x0
[ 75.211756] amdgpu 0000:03:00.0: amdgpu: MAPPING_ERROR: 0x0
[ 75.211758] amdgpu 0000:03:00.0: amdgpu: RW: 0x0
[ 75.212337] amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault
(src_id:0 ring:169 vmid:0 pasid:0, for process pid 0 thread pid 0)
[ 75.212344] amdgpu 0000:03:00.0: amdgpu: in page starting at
address 0x00008087fe5fd000 from client 10
[ 75.212348] amdgpu 0000:03:00.0: amdgpu:
GCVM_L2_PROTECTION_FAULT_STATUS:0x00041B52
[ 75.212352] amdgpu 0000:03:00.0: amdgpu: Faulty UTCL2 client ID: SDMA0 (0xd)
[ 75.212356] amdgpu 0000:03:00.0: amdgpu: MORE_FAULTS: 0x0
[ 75.212359] amdgpu 0000:03:00.0: amdgpu: WALKER_ERROR: 0x1
[ 75.212363] amdgpu 0000:03:00.0: amdgpu: PERMISSION_FAULTS: 0x5
[ 75.212367] amdgpu 0000:03:00.0: amdgpu: MAPPING_ERROR: 0x1
[ 75.212370] amdgpu 0000:03:00.0: amdgpu: RW: 0x1

The full kernel log attached below.

--
Best Regards,
Mike Gavrilov.

Attachments:

dmesg-b13.zip (161.94 kB)

2023-11-16 18:57:12

by Alex Deucher

[permalink] [raw]

Subject: Re: 6.7/regression/KASAN: null-ptr-deref in amdgpu_ras_reset_error_count+0x2d6

On Wed, Nov 15, 2023 at 1:14 PM Mikhail Gavrilov
<[email protected]> wrote:
>
> On Wed, Nov 8, 2023 at 4:24 AM Mikhail Gavrilov
> <[email protected]> wrote:
> >
> > Tested-by: Mikhail Gavrilov <[email protected]>
> > Thanks, after applying the patch GPU loading meets expectations.
> > Games are working so overall all looking good for now.
> >
>
> Unfortunately I didn't check this patch on 7900XTX.
> And now I see that this patch leads the systems with 7900XTX GPU to a
> lot of protection fault messages in the kernel log.
>
> [ 75.211702] gmc_v11_0_process_interrupt: 164 callbacks suppressed
> [ 75.211706] amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault
> (src_id:0 ring:169 vmid:0 pasid:0, for process pid 0 thread pid 0)
> [ 75.211715] amdgpu 0000:03:00.0: amdgpu: in page starting at
> address 0x00008087fe5fd000 from client 10
> [ 75.211718] amdgpu 0000:03:00.0: amdgpu:
> GCVM_L2_PROTECTION_FAULT_STATUS:0x00041D53
> [ 75.211720] amdgpu 0000:03:00.0: amdgpu: Faulty UTCL2 client ID: SDMA1 (0xe)
> [ 75.211723] amdgpu 0000:03:00.0: amdgpu: MORE_FAULTS: 0x1
> [ 75.211725] amdgpu 0000:03:00.0: amdgpu: WALKER_ERROR: 0x1
> [ 75.211727] amdgpu 0000:03:00.0: amdgpu: PERMISSION_FAULTS: 0x5
> [ 75.211729] amdgpu 0000:03:00.0: amdgpu: MAPPING_ERROR: 0x1
> [ 75.211732] amdgpu 0000:03:00.0: amdgpu: RW: 0x1
> [ 75.211740] amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault
> (src_id:0 ring:169 vmid:0 pasid:0, for process pid 0 thread pid 0)
> [ 75.211743] amdgpu 0000:03:00.0: amdgpu: in page starting at
> address 0x00008087bffff000 from client 10
> [ 75.211746] amdgpu 0000:03:00.0: amdgpu:
> GCVM_L2_PROTECTION_FAULT_STATUS:0x00000000
> [ 75.211748] amdgpu 0000:03:00.0: amdgpu: Faulty UTCL2 client ID: CB/DB (0x0)
> [ 75.211750] amdgpu 0000:03:00.0: amdgpu: MORE_FAULTS: 0x0
> [ 75.211752] amdgpu 0000:03:00.0: amdgpu: WALKER_ERROR: 0x0
> [ 75.211754] amdgpu 0000:03:00.0: amdgpu: PERMISSION_FAULTS: 0x0
> [ 75.211756] amdgpu 0000:03:00.0: amdgpu: MAPPING_ERROR: 0x0
> [ 75.211758] amdgpu 0000:03:00.0: amdgpu: RW: 0x0
> [ 75.212337] amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault
> (src_id:0 ring:169 vmid:0 pasid:0, for process pid 0 thread pid 0)
> [ 75.212344] amdgpu 0000:03:00.0: amdgpu: in page starting at
> address 0x00008087fe5fd000 from client 10
> [ 75.212348] amdgpu 0000:03:00.0: amdgpu:
> GCVM_L2_PROTECTION_FAULT_STATUS:0x00041B52
> [ 75.212352] amdgpu 0000:03:00.0: amdgpu: Faulty UTCL2 client ID: SDMA0 (0xd)
> [ 75.212356] amdgpu 0000:03:00.0: amdgpu: MORE_FAULTS: 0x0
> [ 75.212359] amdgpu 0000:03:00.0: amdgpu: WALKER_ERROR: 0x1
> [ 75.212363] amdgpu 0000:03:00.0: amdgpu: PERMISSION_FAULTS: 0x5
> [ 75.212367] amdgpu 0000:03:00.0: amdgpu: MAPPING_ERROR: 0x1
> [ 75.212370] amdgpu 0000:03:00.0: amdgpu: RW: 0x1
>
> The full kernel log attached below.

This patch should address the issue:
https://patchwork.freedesktop.org/patch/567101/
If you still see issues, you may also need this series:
https://patchwork.freedesktop.org/series/126220/

Alex

2023-11-17 12:10:21

by Mikhail Gavrilov

[permalink] [raw]

Subject: Re: 6.7/regression/KASAN: null-ptr-deref in amdgpu_ras_reset_error_count+0x2d6

On Thu, Nov 16, 2023 at 11:56 PM Alex Deucher <[email protected]> wrote:
>
> This patch should address the issue:
> https://patchwork.freedesktop.org/patch/567101/
> If you still see issues, you may also need this series:
> https://patchwork.freedesktop.org/series/126220/
>
> Alex

Thanks.
The first one patch is enough.
Tested-on: 7900XTX, 6900XT and 6800M.
Tested-by: Mikhail Gavrilov <[email protected]>

--
Best Regards,
Mike Gavrilov.