Re: [PATCH v4 12/13] drm/xe/xe_ras: Query errors from system controller on probe

["Tauro, Riana" <riana.tauro@intel.com>]

On 4/28/2026 5:16 PM, Raag Jadav wrote:
> On Fri, Apr 17, 2026 at 02:28:24PM +0530, Riana Tauro wrote:
>> Reorder soc remapper and system controller initialization to
>> early probe to allow querying errors on module load.
> ...
>
>> diff --git a/drivers/gpu/drm/xe/xe_ras.c b/drivers/gpu/drm/xe/xe_ras.c
>> index 42ec27c05e9a..7598eeb796f0 100644
>> --- a/drivers/gpu/drm/xe/xe_ras.c
>> +++ b/drivers/gpu/drm/xe/xe_ras.c
>> @@ -479,4 +479,11 @@ void xe_ras_init(struct xe_device *xe)
>>   	/* Get any pages that need to be offlined from firmware and reserve them */
>>   	get_offlined_list(xe);
>>   	get_queued_pages(xe);
> I know it's yet to be merged but should we also add get_pending_event()?

We could add that.

>
>> +	/*
>> +	 * On init, process and log any errors detected by firmware before driver init.
>> +	 * Critical errors are handled in xe_pcode_probe_early(), which enters survivability mode
>> +	 * if required.
>> +	 */
>> +	xe_ras_process_errors(xe);
> What about wedging? Should we continue driver load after declaring wedged?

Wedging is only on CSC and Punit errors. Such errors will be caught 
during boot survivability and appropriate
action will be taken. The rest of the errors require only logging.

Thanks
Riana

>
> Raag
>
>>   }
>> -- 
>> 2.47.1
>>