Re: ✓ CI.BAT: success for Reapply "drm/xe/gsc: define GSC FW for LNL"

[Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>]

On 7/2/2024 11:02 AM, Lucas De Marchi wrote:
> On Tue, Jul 02, 2024 at 09:25:31AM GMT, Daniele Ceraolo Spurio wrote:
>>
>>
>> On 7/2/2024 7:29 AM, Lucas De Marchi wrote:
>>> On Tue, Jul 02, 2024 at 01:01:28AM GMT, Patchwork wrote:
>>>> == Series Details ==
>>>>
>>>> Series: Reapply "drm/xe/gsc: define GSC FW for LNL"
>>>> URL   : https://patchwork.freedesktop.org/series/135623/
>>>> State : success
>>>>
>>>> == Summary ==
>>>>
>>>> CI Bug Log - changes from 
>>>> xe-1542-886eeb6d89b58f914ee5045fcac54b59a73d8299_BAT -> 
>>>> xe-pw-135623v1_BAT
>>>> ====================================================
>>>>
>>>> Summary
>>>> -------
>>>>
>>>>  **SUCCESS**
>>>>
>>>>  No regressions found.
>>>>
>>>>
>>>>
>>>> Participating hosts (5 -> 4)
>>>> ------------------------------
>>>>
>>>>  Missing    (1): bat-lnl-1
>>>
>>> I guess it didn't really work. +Ryszard +Ewelina: Can we promote LNL to
>>> be considered "a reliable machine from the CI POV" so we don't have
>>> "CI.BAT: success" when LNL execution is missing?  Or is there any other
>>> reason why we report success in this case?
>>
>> Damn. I can't repro the issue anymore with this WA applied and even 
>> in CI we weren't seeing it when I sent it for testing before: 
>> https://patchwork.freedesktop.org/series/134099/ . I did 3 runs in 
>> that case and none of them hit the problem.
>>
>> I've triggered another run to see if we get any better logs.
>
> this change should NOT make the machine "go missing" really. Actually no
> change to xe-only should really make a machine not have any log at
> all....
>
> I believe it was a very unfortunate coincidence and there must be a
> network issue or the like... but we can't consider success when we don't
> have a report for LNL. Particularly for this patch since LNL is
> the only affected platform.
>
> "try again"  in patchwork sounds good for now.

It died again, but this time we got some logs. It died during an fbdev 
test, which makes me think this is due to the display side of the WA 
still being missing. I'll try again to repro locally and if I can't I'll 
send a patch to move the FB out of stolen and see what happens.

Daniele

>
> thanks
> Lucas De Marchi
>
>>
>> Daniele
>>
>>>
>>> thanks
>>> Lucas De Marchi
>>