Failing kernel module tests and test frequency question

Failing kernel module tests and test frequency question

[Patrick Robb]

[-- Attachment #1: Type: text/plain, Size: 882 bytes --]

Hello,

I am seeing our kmods testcase fail on the daily periodic run since June
29: https://dpdkdashboard.iol.unh.edu/results/dashboard/tarballs/25226/

It does coincide with a slew of patches being merged for rc2. But, I
haven't looked at those patches individually, and I also don't know whether
we generally expect breakages for this kernel module compile job to come
from dpdk patches or from patches on the linux kernel project.

I understand that the original/current plan was for running the kernel
module test 1x/day. The question this raises for me is, if we were able to
afford expending the needed compute resources for kmod compile testing on
all incoming patchseries, would that be a valuable addition, or not?

Thanks,
Patrick

-- 

Patrick Robb

Technical Service Manager

UNH InterOperability Laboratory

21 Madbury Rd, Suite 100, Durham, NH 03824

www.iol.unh.edu

[-- Attachment #2: Type: text/html, Size: 3030 bytes --]

Re: Failing kernel module tests and test frequency question

[Ferruh Yigit]

On 7/6/2023 6:50 PM, Patrick Robb wrote:
> Hello,
> 
> I am seeing our kmods testcase fail on the daily periodic run since June
> 29: https://dpdkdashboard.iol.unh.edu/results/dashboard/tarballs/25226/
> <https://dpdkdashboard.iol.unh.edu/results/dashboard/tarballs/25226/>
> 
> It does coincide with a slew of patches being merged for rc2. But, I
> haven't looked at those patches individually, and I also don't know
> whether we generally expect breakages for this kernel module compile job
> to come from dpdk patches or from patches on the linux kernel project.
> 
> I understand that the original/current plan was for running the kernel
> module test 1x/day. The question this raises for me is, if we were able
> to afford expending the needed compute resources for kmod compile
> testing on all incoming patchseries, would that be a valuable addition,
> or not? 
> 
> Thanks,
> Patrick
> 

Hi Patrick,

It is KNI build failure with latest kernel because of an kernel API
change, I am aware of it an have a local patch to fix.

But fix requires kernel version check (to detect correct API) and since
kernel is in the 6.5 merge window and -rc1 is not out, current version
is still 6.4 and version control is not working as expected.
In my local patch I am using 6.4 for version check but that is wrong and
not suitable for an upstream patch.

That is why can you pause the 'dpdk_kmod_compile_torvalds_linux' test
until v6.5-rc1 is out, it shouldn't take long.

Re: Failing kernel module tests and test frequency question

[Ferruh Yigit]

On 7/6/2023 8:42 PM, Ferruh Yigit wrote:
> On 7/6/2023 6:50 PM, Patrick Robb wrote:
>> Hello,
>>
>> I am seeing our kmods testcase fail on the daily periodic run since June
>> 29: https://dpdkdashboard.iol.unh.edu/results/dashboard/tarballs/25226/
>> <https://dpdkdashboard.iol.unh.edu/results/dashboard/tarballs/25226/>
>>
>> It does coincide with a slew of patches being merged for rc2. But, I
>> haven't looked at those patches individually, and I also don't know
>> whether we generally expect breakages for this kernel module compile job
>> to come from dpdk patches or from patches on the linux kernel project.
>>
>> I understand that the original/current plan was for running the kernel
>> module test 1x/day. The question this raises for me is, if we were able
>> to afford expending the needed compute resources for kmod compile
>> testing on all incoming patchseries, would that be a valuable addition,
>> or not? 
>>
>> Thanks,
>> Patrick
>>
> 
> Hi Patrick,
> 
> It is KNI build failure with latest kernel because of an kernel API
> change, I am aware of it an have a local patch to fix.
> 
> But fix requires kernel version check (to detect correct API) and since
> kernel is in the 6.5 merge window and -rc1 is not out, current version
> is still 6.4 and version control is not working as expected.
> In my local patch I am using 6.4 for version check but that is wrong and
> not suitable for an upstream patch.
> 
> That is why can you pause the 'dpdk_kmod_compile_torvalds_linux' test
> until v6.5-rc1 is out, it shouldn't take long.
> 

I have sent following patch, can you please test:
https://patches.dpdk.org/project/dpdk/patch/20230711100941.2082191-1-ferruh.yigit@amd.com/

You can enable test and use patch locally until it is merged to main repo.

Re: Failing kernel module tests and test frequency question

[Ferruh Yigit]

On 7/11/2023 11:12 AM, Ferruh Yigit wrote:
> On 7/6/2023 8:42 PM, Ferruh Yigit wrote:
>> On 7/6/2023 6:50 PM, Patrick Robb wrote:
>>> Hello,
>>>
>>> I am seeing our kmods testcase fail on the daily periodic run since June
>>> 29: https://dpdkdashboard.iol.unh.edu/results/dashboard/tarballs/25226/
>>> <https://dpdkdashboard.iol.unh.edu/results/dashboard/tarballs/25226/>
>>>
>>> It does coincide with a slew of patches being merged for rc2. But, I
>>> haven't looked at those patches individually, and I also don't know
>>> whether we generally expect breakages for this kernel module compile job
>>> to come from dpdk patches or from patches on the linux kernel project.
>>>
>>> I understand that the original/current plan was for running the kernel
>>> module test 1x/day. The question this raises for me is, if we were able
>>> to afford expending the needed compute resources for kmod compile
>>> testing on all incoming patchseries, would that be a valuable addition,
>>> or not? 
>>>
>>> Thanks,
>>> Patrick
>>>
>>
>> Hi Patrick,
>>
>> It is KNI build failure with latest kernel because of an kernel API
>> change, I am aware of it an have a local patch to fix.
>>
>> But fix requires kernel version check (to detect correct API) and since
>> kernel is in the 6.5 merge window and -rc1 is not out, current version
>> is still 6.4 and version control is not working as expected.
>> In my local patch I am using 6.4 for version check but that is wrong and
>> not suitable for an upstream patch.
>>
>> That is why can you pause the 'dpdk_kmod_compile_torvalds_linux' test
>> until v6.5-rc1 is out, it shouldn't take long.
>>
> 
> I have sent following patch, can you please test:
> https://patches.dpdk.org/project/dpdk/patch/20230711100941.2082191-1-ferruh.yigit@amd.com/
> 
> You can enable test and use patch locally until it is merged to main repo.
> 

Fix merged for -rc3, so test should be OK now.

Re: Failing kernel module tests and test frequency question

[Patrick Robb]

[-- Attachment #1: Type: text/plain, Size: 2352 bytes --]

Hello I see the test is passing now. Thank you Ferruh!

On Thu, Jul 13, 2023 at 4:56 AM Ferruh Yigit <ferruh.yigit@amd.com> wrote:

> On 7/11/2023 11:12 AM, Ferruh Yigit wrote:
> > On 7/6/2023 8:42 PM, Ferruh Yigit wrote:
> >> On 7/6/2023 6:50 PM, Patrick Robb wrote:
> >>> Hello,
> >>>
> >>> I am seeing our kmods testcase fail on the daily periodic run since
> June
> >>> 29:
> https://dpdkdashboard.iol.unh.edu/results/dashboard/tarballs/25226/
> >>> <https://dpdkdashboard.iol.unh.edu/results/dashboard/tarballs/25226/>
> >>>
> >>> It does coincide with a slew of patches being merged for rc2. But, I
> >>> haven't looked at those patches individually, and I also don't know
> >>> whether we generally expect breakages for this kernel module compile
> job
> >>> to come from dpdk patches or from patches on the linux kernel project.
> >>>
> >>> I understand that the original/current plan was for running the kernel
> >>> module test 1x/day. The question this raises for me is, if we were able
> >>> to afford expending the needed compute resources for kmod compile
> >>> testing on all incoming patchseries, would that be a valuable addition,
> >>> or not?
> >>>
> >>> Thanks,
> >>> Patrick
> >>>
> >>
> >> Hi Patrick,
> >>
> >> It is KNI build failure with latest kernel because of an kernel API
> >> change, I am aware of it an have a local patch to fix.
> >>
> >> But fix requires kernel version check (to detect correct API) and since
> >> kernel is in the 6.5 merge window and -rc1 is not out, current version
> >> is still 6.4 and version control is not working as expected.
> >> In my local patch I am using 6.4 for version check but that is wrong and
> >> not suitable for an upstream patch.
> >>
> >> That is why can you pause the 'dpdk_kmod_compile_torvalds_linux' test
> >> until v6.5-rc1 is out, it shouldn't take long.
> >>
> >
> > I have sent following patch, can you please test:
> >
> https://patches.dpdk.org/project/dpdk/patch/20230711100941.2082191-1-ferruh.yigit@amd.com/
> >
> > You can enable test and use patch locally until it is merged to main
> repo.
> >
>
> Fix merged for -rc3, so test should be OK now.
>
>

-- 

Patrick Robb

Technical Service Manager

UNH InterOperability Laboratory

21 Madbury Rd, Suite 100, Durham, NH 03824

www.iol.unh.edu

[-- Attachment #2: Type: text/html, Size: 5303 bytes --]