On 26-10-2026 17:59, Raag Jadav wrote:
> On Fri, Oct 23, 2026 at 10:04:28AM +0530, Badal Nilawar wrote:
>> Add the xe_error_cper tracepoint and emit each assembled CPER record
>> through it. Besides the raw record, the event carries the device name,
>> platform ID, FRU ID, severity and the Intel error section header, so
>> tools can pick up the records from the trace buffer.
> ...
>
>> + * Example trace output::
>> + *
>> + *   xe_error_cper: dev=0000:03:00.0 severity=2
>> + *     platform_id=9046afe5-9041-5124-8614-92550d9e9da6
>> + *     fru_id=91befe38-2f61-3d7b-0000-000000000000
>> + *     intel_err_hdr_raw=
> 'intel_hdr_raw'

The specs uses intel_err_hdr_raw in the example trace output, so I kept 
the same naming here for consistency.

Thanks,

Badal

>
>> + *       01 02 00 18 00 00 00 01 00 00 00 00 64 77 51 14 ...
>> + *     cper_len=1772
>> + *     cper_raw=
>> + *       43 50 45 52 00 01 ff ff ff ff 01 00 02 00 00 00 ...
> ...
>
>> +	    TP_printk("dev=%s severity=%d platform_id=%pUl fru_id=%pUl intel_err_hdr_raw=%s cper_len=%u cper_raw=%s",
> Ditto.
>
> Raag
>
>> +		      __get_str(dev), __entry->sev,
>> +		      __entry->platform_id, __entry->fru_id,
>> +		      __print_hex(__entry->ihdr_raw,
>> +				  sizeof(struct xe_cper_intel_hdr)),
>> +		      __entry->cper_len,
>> +		      __print_hex(__get_dynamic_array(cper),
>> +				  __entry->cper_len))