Re: [PATCH v2 11/11] drm/xe/ras: Report correctable errors using SIGID

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [Medium] Passing raw `enum xe_ras_component` to `xe_log_from()` instead of a properly constructed `enum xe_log_component`, causing incorrect logging format and kernel warnings.
- [Medium] Unmapped `XE_RAS_COMP_RESERVED` causes `U32_MAX` to be passed as a default SIGID, breaking the stable signature ID tracking scheme by logging `SIGID=4294967295`.
--

commit 9b392e724340e7b28bbb985e627346fd4f036a93
Author: Badal Nilawar <badal.nilawar@intel.com>

drm/xe/ras: Report correctable errors using SIGID

This commit replaces the xe_warn() call for correctable errors with the
xe_log_from() helper to report them using SIGID format.

> diff --git a/drivers/gpu/drm/xe/xe_ras.c b/drivers/gpu/drm/xe/xe_ras.c
> index 4cd1d5eb75f4d..0becfadbb4cdf 100644
> --- a/drivers/gpu/drm/xe/xe_ras.c
> +++ b/drivers/gpu/drm/xe/xe_ras.c
> @@ -458,8 +458,11 @@ void xe_ras_counter_threshold_crossed(struct xe_device *xe,
>  		if (!ras_counter_is_valid(xe, &errors[id]))
>  			continue;
>  
> -		xe_warn(xe, "[RAS]: %s %s detected\n",
> -			comp_to_str(component), sev_to_str(severity));
> +		xe_log_from(xe, ras_sev_to_cper_sev(severity),
> +			    ras_comp_to_hw_sigid(component),

[Severity: Medium]
Will this leak U32_MAX as a SIGID to the logging infrastructure?

If ras_counter_is_valid() allows XE_RAS_COMP_RESERVED (3) to proceed,
ras_comp_to_hw_sigid() hits the default case and returns U32_MAX.

U32_MAX is then passed as the sigid to xe_log_from(), and __xe_log_emit()
blindly formats the sigid using %u, logging "SIGID=4294967295". 
Does this break the stable, machine-matchable token format?

> +			    component,

[Severity: Medium]
Does this pass an invalid component type to xe_log_from()?

xe_log_from() expects its component parameter to be a constructed 
enum xe_log_component (via MAKE_XE_LOG_COMPONENT), where the lower byte is
class and the upper byte is type.

Passing the raw enum xe_ras_component directly causes it to be misinterpreted
as CLASS=1 and TYPE=0. This triggers log_unknown_component_prefix() to be
invoked, printing a corrupted "DRIVER? " prefix and triggering a kernel WARN
on every RAS event if CONFIG_DRM_XE_DEBUG is enabled.

> +			    &errors[id], sizeof(errors[id]),
> +			    "error detected\n");
>  
>  		/* Send event once per component */
>  		if (sent & BIT(component))

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260825175916.1103841-13-badal.nilawar@intel.com?part=11