LAVA SSH sessions to LNG RT kernels on Arndale fail

[Mike Holmes]

When using LAVA SSH sessions with non RT LNG images everything works
http://validation.linaro.org/scheduler/job/77248/log_file#L_29_414

When using the RT image they fail
http://validation.linaro.org/scheduler/job/77247/log_file#L_29_348

The root appears to be no IP address in the RT kernel case
  udhcpc (v1.21.1) started
 Sending discover...
 [ 46.192401] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
 [ 46.193006] asix 1-3.2.4:1.0 eth0: link up, 100Mbps, full-duplex, lpa 0x41E1
 Sending discover...
 Sending discover...
 No lease, failing

The same issue has also seen with KVM tests that use networking.

[Mike Holmes (mike-holmes)]

See also https://bugs.launchpad.net/linaro-networking/+bug/1234718

[Vincent Hsu (vincent-hsu)]

I run the similar lava tests 3 times today.
fail 1 times
http://validation.linaro.org/scheduler/job/77488/definition
but success 2 times
http://validation.linaro.org/scheduler/job/77599/definition
http://validation.linaro.org/scheduler/job/77573/definition
Furthermore, it won't happen in Kim's local environment.
Condering the udhcp default setting in OE just retried 3 times.
Maybe the bug cause by the unstable networking environment in lava lab.

[Kim Phillips (kim-phillips)]

my environment may have lower net latency than the lab, but that doesn't mean the lab's dhcp server is unstable.

In the job that failed, you can see the failure here:

http://validation.linaro.org/scheduler/job/77488/log_file#L_29_366

but a previous success in the same job, here:

http://validation.linaro.org/scheduler/job/77488/log_file#L_15_368

[Mike Holmes (mike-holmes)]

With Zis patch [1] http://validation.linaro.org/scheduler/job/77750/log_file SSH worked three out of three times, in both the lng and the regular lab.

[1] https://groups.google.com/a/linaro.org/forum/#!topic/linaro-networking/qr1-D6EaTKw

I will add it to the lng image

[Vincent Hsu (vincent-hsu)]

I got the same faild logs with Mike's image which apply zi's patch in both the lng and the regular lab.
In regular lab
     http://validation.linaro.org/scheduler/job/78123/log_file#L_29_353
In lng lab
     http://validation.linaro.org/scheduler/job/78093/log_file#L_35_399
     http://validation.linaro.org/scheduler/job/78126/log_file#L_15_384
I have a successful one, so it may not relate to my json file.
http://validation.linaro.org/scheduler/job/78129/log_file#L_29_422

I can reproduce this issue with non-rt kernel in both the lng and the regular lab..
In regular lab
fail:
     http://validation.linaro.org/scheduler/job/78149/log_file#L_29_372
success:
     http://validation.linaro.org/scheduler/job/78129/log_file#L_29_422
     http://validation.linaro.org/scheduler/job/78148/log_file#L_29_421
In lng lab
fail:
     http://validation.linaro.org/scheduler/job/78150/log_file#L_29_376
success:
     http://validation.linaro.org/scheduler/job/78151/log_file#L_29_448

This task is more interesting one, it didn't start ethernet interface.
     http://validation.linaro.org/scheduler/job/78152/log_file#L_33_17

[Kim Phillips (kim-phillips)]

> I got the same faild logs with Mike's image which apply zi's patch in both the lng and the regular lab.

We've got a better patch than Zi's original one. It's sitting in the LNG kernel as of a few hours ago:

https://groups.google.com/a/linaro.org/forum/#!topic/linaro-networking/h8vFTnjZ-C0

I don't know which version (Zi's or Steve Capper's three patches) is included in "Mike's image", but it may be worth finding out, and re-testing if what was tested was Zi's original patch alone.

[Zi Shen Lim (zlim)]

On Thu, Oct 10, 2013 at 7:43 PM, Kim Phillips <email address hidden>wrote:

> > I got the same faild logs with Mike's image which apply zi's patch in
> both the lng and the regular lab.
>
> We've got a better patch than Zi's original one. It's sitting in the
> LNG kernel as of a few hours ago:
>
> https://groups.google.com/a/linaro.org/forum/#!topic/linaro-
> networking/h8vFTnjZ-C0
>
> I don't know which version (Zi's or Steve Capper's three patches) is
> included in "Mike's image", but it may be worth finding out, and re-
> testing if what was tested was Zi's original patch alone.
>
>
Yes. Please retest with Steve Capper's patches (especially the thp and
fast_gup patches), as they handle the thp split case not covered by my
quick-n-dirty version.

> --
> You received this bug notification because you are a member of Linaro
> Networking Group, which is subscribed to linaro-networking.
> Matching subscriptions: LNG all
> https://bugs.launchpad.net/bugs/1236655
>
> Title:
> LAVA SSH sessions to LNG RT kernels on Arndale fail
>
> Status in Linaro networking Group:
> Fix Committed
>
> Bug description:
> When using LAVA SSH sessions with non RT LNG images everything works
> http://validation.linaro.org/scheduler/job/77248/log_file#L_29_414
>
> When using the RT image they fail
> http://validation.linaro.org/scheduler/job/77247/log_file#L_29_348
>
> The root appears to be no IP address in the RT kernel case
> udhcpc (v1.21.1) started
> Sending discover...
> [ 46.192401] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
> [ 46.193006] asix 1-3.2.4:1.0 eth0: link up, 100Mbps, full-duplex, lpa
> 0x41E1
> Sending discover...
> Sending discover...
> No lease, failing
>
> The same issue has also seen with KVM tests that use networking.
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/linaro-networking/+bug/1236655/+subscriptions
>