Re: [PATCH 1/2] tests/intel/xe_survivability: update ignored dmesg regex

["Anirban, Sk" <sk.anirban@intel.com>]

Hi

On 14-07-2026 04:40 pm, Kamil Konieczny wrote:
> Hi Anirban,,
> On 2026-07-13 at 22:59:04 +0530, Anirban, Sk wrote:
>> Hi,
>>
>> On 09-07-2026 08:42 pm, Kamil Konieczny wrote:
>>> Hi Sk,
>>> On 2026-07-09 at 12:39:07 +0530, Sk Anirban wrote:
>>>> The kernel now reports -ENOTRECOVERABLE when the device is wedged.
>>>> As runtime-survivability deliberately wedges the device, update
>>>> dmesg regex so igt_runner does not treat the expected error logs
>>>> as failures.
>>>>
>>>> kernel implementation:
>>>> https://patchwork.freedesktop.org/series/168780/
>>>>
>>>> Signed-off-by: Sk Anirban <sk.anirban@intel.com>
>>>> ---
>>>>    tests/intel/xe_survivability.c | 7 ++++---
>>>>    1 file changed, 4 insertions(+), 3 deletions(-)
>>>>
>>>> diff --git a/tests/intel/xe_survivability.c b/tests/intel/xe_survivability.c
>>>> index 74488bbea..cc54b37ad 100644
>>>> --- a/tests/intel/xe_survivability.c
>>>> +++ b/tests/intel/xe_survivability.c
>>>> @@ -43,12 +43,13 @@ static char bus_addr[NAME_MAX];
>>>>    static void ignore_wedged_in_dmesg(void)
>>>>    {
>>>>    	/* this is needed for igt_runner so it will ignore it */
>>>> -	igt_emit_ignore_dmesg_regex("GT[0-9A-Fa-f]*: failed to enable GuC scheduling policies: -ECANCELED"
>>> Please keep old value for some time (a year?).
>> As kernel changes are merged, we may no longer receive the same error code
>> (ECANCELED) in wedge scenarios.
>> Additionally, if there are any gaps in the kernel patches that still result
>> in ECANCELED being returned, we won't be able to detect that issue using IGT
>> if we keep the old error codes here.
> Hmm, up to you but it could make testing for older kernels hard,
> e.g. those which will not backport kernel changes yet. So up to
> you, both this on one below is acceptable.
Got your point, I will fix that in next rev.
>
>>>> +	igt_emit_ignore_dmesg_regex("GT[0-9A-Fa-f]*: failed to enable GuC scheduling policies: -ENOTRECOVERABLE"
>>>>    				    "|CRITICAL: Xe has declared device [0-9A-Fa-f:.]* as wedged"
>>>> -				    "|GT[0-9A-Fa-f]*: reset failed .-ECANCELED"
>>>> +				    "|GT[0-9A-Fa-f]*: reset failed .-ENOTRECOVERABLE"
>>> Same here, keep both.
>>>
>>>>    				    "|GT[0-9A-Fa-f]*: Failed to submit"
>>>>    				    "|Modules linked in:"
>>>> -				    "|__pfx___drm_");
>>>> +				    "|__pfx___drm_"
>>>> +				    "|-ENOTRECOVERABLE");
>>> Do not add such, it could made to ignore errors in unrelated context.
>> ENOTRECOVERABLE is intended to be returned only for wedged-state scenarios
>> by the kernel. I think the test should ignore any failures that return this
>> error code.
> The problem is that this regex is applied to kernel error and
> you do not know if it will be dropped from other contexts,
> outside of your test. So it is better to make sure it was from
> xe driver. As it is now it is too generic.
>
> Regards,
> Kamil

Sure, in that case, I will also add a few error messages that were 
observed recently for the second revision.

Thanks,
Anirban

>
>> Thanks
>> Anirban
>>
>>> Regards,
>>> Kamil
>>>
>>>>    }
>>>>    static bool check_survivability_mode_sysfs(void)
>>>> -- 
>>>> 2.43.0
>>>>