AMD General Reviewed-by: Tao Zhou <[email protected]>
> -----Original Message----- > From: Sun, Ce(Overlord) <[email protected]> > Sent: Wednesday, September 16, 2026 4:28 PM > To: [email protected] > Cc: Zhang, Hawking <[email protected]>; Chai, Thomas > <[email protected]>; Zhou1, Tao <[email protected]>; Yang, Stanley > <[email protected]>; Sun, Ce(Overlord) <[email protected]> > Subject: [PATCH] drm/amdgpu: skip ras error reset in gpu reset only when > uniras is > enabled > > Original logic skips RAS error reset whenever GPU reset/RAS recovery. > Add amdgpu_uniras_enabled() check: skip reset only for uniras enabled; > non-uniras > ASICs still run RAS error reset during GPU reset. > > Commit "drm/amdgpu: retire MCA support" caused old ASICs UE test failure: > multiple CE/UE events across blocks. Legacy RAS relied on set_debug_mode to > decide skip RAS reset, while set_debug_mode is now only for uniras. > Use amdgpu_uniras_enabled() instead for this condition. > > Fixes: b5b6231598dd ("drm/amdgpu: retire MCA support") > Signed-off-by: Ce Sun <[email protected]> > --- > drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c > b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c > index 6b2b8acdbf9b..c65916d84683 100644 > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c > @@ -1513,7 +1513,8 @@ int amdgpu_ras_reset_error_count(struct amdgpu_device > *adev, > return -EOPNOTSUPP; > > /* skip ras error reset in gpu reset */ > - if (amdgpu_in_reset(adev) || amdgpu_ras_in_recovery(adev)) > + if ((amdgpu_in_reset(adev) || amdgpu_ras_in_recovery(adev)) && > + amdgpu_uniras_enabled(adev)) > return -EOPNOTSUPP; > > if (block_obj->hw_ops->reset_ras_error_count) > -- > 2.34.1
