APPLIED: [SRU][Q][PATCH 0/2] GFX1151 kernel fixes
Mario Limonciello
mario.limonciello at amd.com
Fri Jan 23 16:33:05 UTC 2026
On 1/23/2026 10:29 AM, Stefan Bader wrote:
> On 11/01/2026 19:39, Mario Limonciello wrote:
>> From: "Mario Limonciello (AMD)" <superm1 at kernel.org>
>>
>> [ Impact ]
>> GFX1151 has a problem where the context save / restore area is
>> incorrectly
>> sized and this can lead to hangs when used with ROCm.
>>
>> Patches were upstreamed as part of 6.19-rc1 that fix this issue. They're
>> already in the OEM 6.14 and OEM 6.17 kernels, but should also be added to
>> 6.17 HWE kernels because some users don't run the OEM kernels.
>>
>> [ Fix ]
>> Cherry picks from 6.19-rc1
>>
>> [ Test Plan ]
>> Verify that ROCm works properly on GFX1151
>>
>> [ Where problems could occur ]
>> GFX1151 specifically, with ROCm
>>
>> [ Other Info ]
>> Matching code for ROCm:
>> https://github.com/ROCm/rocm-systems/
>> commit/0a2ea9ef55403341a43df4c96dd946c4641a7f79
>>
>> Jonathan Kim (1):
>> drm/amdkfd: bump minimum vgpr size for gfx1151
>>
>> Mario Limonciello (1):
>> drm/amdkfd: Export the cwsr_size and ctl_stack_size to userspace
>>
>> drivers/gpu/drm/amd/amdkfd/kfd_queue.c | 1 +
>> drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 4 ++++
>> 2 files changed, 5 insertions(+)
>>
>
>
> Applied to questing:linux/master-next. Thanks.
>
> -Stefan
>
Thanks! One clarification I would like:
Does this implicitly mean it also will go to the hwe train noble tracks?
Or does that need a separate PR?
Thanks,
More information about the kernel-team
mailing list