QCOM IPQ Boards Kernel CI Automation

QCOM IPQ Boards Kernel CI Automation

[Ram Jangir (QUIC)]

Hi,

We are from Qualcomm. We are actively upstreaming support for some of  our IPQ (Router chipsets, IPQ807X, IPQ5018, IPQ9574, IPQ5332). 
Hence we want to run the kernel CI test suites from the kernel trees (linux, linux-next) .  Boards are there in our local board farm inside QCOM.
So wanted to understand the process involved in testing our boards and getting the results published on the kernelCI dashboard.  
Can you please help us here ?

Thanks,
Ram

Re: QCOM IPQ Boards Kernel CI Automation

[Guillaume Tucker]

Hi Ram,

On 25/04/2023 14:44, Ram Jangir (QUIC) wrote:
> Hi,
> 
> We are from Qualcomm. We are actively upstreaming support for some of  our IPQ (Router chipsets, IPQ807X, IPQ5018, IPQ9574, IPQ5332). 
> Hence we want to run the kernel CI test suites from the kernel trees (linux, linux-next) .  Boards are there in our local board farm inside QCOM.
> So wanted to understand the process involved in testing our boards and getting the results published on the kernelCI dashboard.  
> Can you please help us here ?

Thanks for getting in touch.  I believe Steve (on Cc) once had a
lab connected to KernelCI, is this related in any way?

If your lab is a LAVA lab then it's fairly easy to set up,
there's a documentation page for it:

  https://kernelci.org/docs/labs/lava/

If not, then it's possible to send results from any arbitrary lab
but it's a bit more tricky.  There's currently no way for labs to
get notified when new kernel builds are available so some kind
of "glue code" is needed to bridge this gap.

However, we're now working on a new API which provides a Pub/Sub
interface for this kind of scenario.  The new API is lab-neutral,
it doesn't matter if you're running tests in LAVA or anything
else.  You could even run tests by hand locally and get the
results sent directly to the API.  It's not quite ready yet,
we're hoping to have a publicly-hosted instance this Summer for
people to start using it and start deprecating the current
system.

There's a section in the documentation about it, which is also a
work-in-progress.  Here's an overview:

  https://kernelci.org/docs/api/overview/

If you have a LAVA lab, I believe it would be fine to start by
using the current system now.  If not, then it would probably
make more sense to directly look into using the new API with a
bit of help from our core developers.  There's already a staging
instance available to start experimenting with things, and having
contributions from real-world users with particular types of labs
is a great way to accelerate the API development and adoption.

There's a public KernelCI video meeting every Thursday at 11:00
UTC followed by an Open Hours at 12:00 UTC to discuss any topics,
I guess this would be an interesting one to bring up.  Please let
me know if you want to be added to the calendar invitations.

Thanks,
Guillaume

RE: QCOM IPQ Boards Kernel CI Automation

[Hambardzumyan, Minas]

Guillaume, I am also working on submitting TI upstream test results to Kernel CI -- Vince is helping me address some of technical issues of publishing to the staging server. 

Can you send the invite to me also?

Thanks,
Minas.

-----Original Message-----
From: Guillaume Tucker <guillaume.tucker@collabora.com> 
Sent: Thursday, May 04, 2023 2:59 AM
To: Ram Jangir (QUIC) <quic_rjangir@quicinc.com>; kernelci@lists.linux.dev
Cc: Stephen Boyd <sboyd@kernel.org>
Subject: Re: QCOM IPQ Boards Kernel CI Automation

Hi Ram,

On 25/04/2023 14:44, Ram Jangir (QUIC) wrote:
> Hi,
> 
> We are from Qualcomm. We are actively upstreaming support for some of  our IPQ (Router chipsets, IPQ807X, IPQ5018, IPQ9574, IPQ5332). 
> Hence we want to run the kernel CI test suites from the kernel trees (linux, linux-next) .  Boards are there in our local board farm inside QCOM.
> So wanted to understand the process involved in testing our boards and 
> getting the results published on the kernelCI dashboard. Can you please help us here ?

Thanks for getting in touch.  I believe Steve (on Cc) once had a lab connected to KernelCI, is this related in any way?

If your lab is a LAVA lab then it's fairly easy to set up, there's a documentation page for it:

  https://kernelci.org/docs/labs/lava/

If not, then it's possible to send results from any arbitrary lab but it's a bit more tricky.  There's currently no way for labs to get notified when new kernel builds are available so some kind of "glue code" is needed to bridge this gap.

However, we're now working on a new API which provides a Pub/Sub interface for this kind of scenario.  The new API is lab-neutral, it doesn't matter if you're running tests in LAVA or anything else.  You could even run tests by hand locally and get the results sent directly to the API.  It's not quite ready yet, we're hoping to have a publicly-hosted instance this Summer for people to start using it and start deprecating the current system.

There's a section in the documentation about it, which is also a work-in-progress.  Here's an overview:

  https://kernelci.org/docs/api/overview/

If you have a LAVA lab, I believe it would be fine to start by using the current system now.  If not, then it would probably make more sense to directly look into using the new API with a bit of help from our core developers.  There's already a staging instance available to start experimenting with things, and having contributions from real-world users with particular types of labs is a great way to accelerate the API development and adoption.

There's a public KernelCI video meeting every Thursday at 11:00 UTC followed by an Open Hours at 12:00 UTC to discuss any topics, I guess this would be an interesting one to bring up.  Please let me know if you want to be added to the calendar invitations.

Thanks,
Guillaume