drm/amdgpu/ras: only check bad page for address-based UMC injection

Browse source

UMC error injection on MI300 series is dispatched by the RAS TA using
the injection method; only the "coherent" methods are address based,
the single-shot/persistent/ac-parity ones ignore the address.

The debugfs control path validated the injection address against the
bad page list for every UMC injection. On uniras (SMU v13+) devices the
address is now validated by the ras_mgr inject handler, so the legacy
debugfs bad page check only runs on the legacy RAS path; other ASICs
keep injecting by address.

In the ras_mgr handler an injection is treated as non address-based only
when userspace passes the U64_MAX sentinel address and the method is a
non-address method. In that case the address is cleared to 0 and the bad
page / range validation is skipped; otherwise the injection address is
validated as before.

Signed-off-by: Stanley.Yang <Stanley.Yang@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

...

This commit is contained in:

Stanley.Yang 2026-06-26 20:20:15 +08:00 • committed by Alex Deucher

parent a8880289fd

commit bf65973e5f

2 changed files with 73 additions and 20 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download patch file Download diff file Expand all files Collapse all files

10

drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c

View file

81

drivers/gpu/drm/amd/ras/ras_mgr/amdgpu_ras_cmd.c

View file