Re: ❌ FAIL (MISSED 2 of 87): Test report for for-kernelci (6.9.0-rc4, arm-next, 6a71d290)

[Jeremy Linton]

Hi,

On 4/23/24 09:42, Donald Zickus wrote:
> (adding Jeremy Linton as the author of the out of tree patch)
> 
> Hi Jeremy,
> 
> 
> On Tuesday, April 23, 2024 at 7:49:26 AM UTC-4 Suzuki K Poulose wrote:
> 
> On 23/04/2024 12:17, James Clark wrote:
>>
>>
>> On 23/04/2024 12:14, James Clark wrote:
>>>
>>>
>>> On 22/04/2024 18:08, Will Deacon wrote:
>>>> [+Suzuki, Mike and James]
>>>>
>>>> On Fri, Apr 19, 2024 at 08:30:09PM -0000, cki-project at redhat.com wrote:
>>>>> Hi, we tested your kernel and here are the results:
>>>>>
>>>>> Overall result: FAILED
>>>>> Merge: OK
>>>>> Compile: OK
>>>>> Test: FAILED
>>>>>
>>>>>
>>>>> Kernel information:
>>>>> Commit message: Merge branch 'for-next/core' into for-kernelci
>>>>>
>>>>> You can find all the details about the test run at
>>>>> https://datawarehouse.cki-project.org/kcidb/checkouts/redhat:1260423326
>>>>>
>>>>> One or more kernel tests failed:
>>>>> Unrecognized or new issues:
>>>>> Boot test
>>>>> aarch64
>>>>> Logs: https://datawarehouse.cki-project.org/kcidb/tests/
> redhat:1260423326-aarch64-kernel_upt_4
>>>>> Non-passing ran subtests:
>>>>> ❌ FAIL distribution/kpkginstall/journalctl-check
>>>>
>>>> I'm not sure if it's the root cause, but the logs here have a tonne of
>>>> coresight ETM splats (I included one at the end of the mail).
>>>>
>>>> https://s3.amazonaws.com/arr-cki-prod-trusted-artifacts/
> trusted-artifacts/1260423326/test_aarch64/6670265232/
> artifacts/run.done.01/job.01/recipes/15985953/tasks/5/
> results/1713555252/logs/journalctl.log
>>>>
>>>> Jonathan has recently done a bunch of work fixing up the ->parent
>>>> pointers for PMU devices, but I don't see anything going near the
>>>> coresight drivers so this is probably unrelated.
>>>>
>>>> Will
>>>>
>>>> --->8
>>>>
>>>> Apr 19 15:33:38 hpe-apollo-cn99xx-03.khw.eng.rdu2.dc.redhat.com kernel:
> CSCFG registered etm103
>>>> Apr 19 15:33:38 hpe-apollo-cn99xx-03.khw.eng.rdu2.dc.redhat.com kernel:
> coresight etm103: CPU103: etm v4.1 initialized
>>>> Apr 19 15:33:38 hpe-apollo-cn99xx-03.khw.eng.rdu2.dc.redhat.com kernel:
> sysfs: cannot create duplicate filename '/devices/system/container/
> ACPI0004:00/ARMHC9FE:00/funnel0/connections/in:0'
>>>
>>> Hi Will,
>>>
>>> It looks like multiple ETMs are described as being connected to the same
>>> port on a funnel. Those connections to build the filename string all
>>> come from the firmware. I don't think there are any changes since 6.8 in
>>> Coresight that would affect device registration.
>>>
>>> I did try to find some previous runs on that machine and I found this
>>> message
>>> (https://s3.amazonaws.com/arr-cki-prod-trusted-artifacts/
> trusted-artifacts/1260423326/test_aarch64/6670265232/
> artifacts/run.done.01/job.01/recipes/15985953/tasks/5/
> results/1713555252/logs/journalctl.log):
>>>
>>> "ETM4 disabled due to firmware bug"
>>>
>>> Although I couldn't find where that string comes from. It seems to me
>>> like there has been a firmware change on that machine recently, or this
>>> is the first time it's been booted with Coresight enabled.
>>>
>>> Also possible is that we removed the kernel part that disabled it due to
>>> whatever that flag is, but I don't think that's happened. Maybe Suzuki
>>> knows about that message and could comment?
>>>
>>> Thanks
>>>
>>> James
>>
>> Ok I found the string, it's from an out of tree patch:
>>
>> https://src.fedoraproject.org/rpms/kernel/c/2db8df88f66eacaa9bec1995431388
> fb3a044566.patch
> 
> 
> Is your patch [1] still relevant in light of the current splat?

Yes. because the firmware doesn't appear to be updated on the test machine.

The downstream patch is quirking/disabling coresight on the machine in 
question because the table is just fundamentally broken. So, yes the 
module could be blacklisted if the test harness could do that on a 
machine by machine basis, or for that matter a corrected table could 
also be injected.




> 
> Cheers,
> Don
> 
> [1]
> - https://gitlab.com/cki-project/kernel-ark/-/commit/ea9bdc9a2dd2fe5aa2de58dcbfa14f64bdd4d716
> 
> 
>>
>>
> 
> 
> Could they not black list the coresight* modules on the system ?
> 
> Suzuki
> 
>