[ovs-dev] 2.8.0: STP - flush the fdb and mdb when topology changed failure on s390x

[Ilya Maximets]

Hi James.
I had same issue while building with '-march=native' on x86 system.
Changing the repeated time warps to a single long warp almost solved
the issue for me, but I think that testcase is not completely right
and it still rarely fails.

That's the patch I prepared last month for that:
* https://mail.openvswitch.org/pipermail/ovs-dev/2017-September/338708.html

Maybe it'll help or my case will give you some hints about the root cause.
Unfortunately I had no much time to investigate the issue deeply.

Best regards, Ilya Maximets.
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Re: [ovs-dev] 2.8.0: STP - flush the fdb and mdb when topology changed failure on s390x

[James Page]

On Wed, 18 Oct 2017 at 11:45 James Page  wrote:

> Hi All
>
> I've been scratching my head about this test failure for the last few
> hours:
>
>   https://bugs.launchpad.net/ubuntu/+source/openvswitch/+bug/1722799
>
> Basically this test only fails on s390x in Ubuntu; this is the only
> big-endian architecture in Ubuntu now so my first suspicion was to look for
> something related to endian-ess but no cigar so far.  The failure is not
> specific to 2.8.x - I also see the same failure against 2.7.1 but not with
> 2.5.x series versions (will be looking at the diff between those revisions
> next).
>

A bit of bisecting later pointed to this commit:


https://github.com/openvswitch/ovs/commit/013ebfe62d4625afb5bc669e3a32b5a1a236a9be

but I think that just uncovers the actual problem, rather than being the
cause of it...
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

[ovs-dev] 2.8.0: STP - flush the fdb and mdb when topology changed failure on s390x

[James Page]

Hi All

I've been scratching my head about this test failure for the last few hours:

  https://bugs.launchpad.net/ubuntu/+source/openvswitch/+bug/1722799

Basically this test only fails on s390x in Ubuntu; this is the only
big-endian architecture in Ubuntu now so my first suspicion was to look for
something related to endian-ess but no cigar so far.  The failure is not
specific to 2.8.x - I also see the same failure against 2.7.1 but not with
2.5.x series versions (will be looking at the diff between those revisions
next).

Comparing log output between x86_64 and s390x I see a missing STP state
change for the remaining port (p1) in br0 - I have a hunch this is where
the problem lies, but I can't figure out why this state change is not being
triggered;   The fdb entry does clear after about 60 seconds of warped
time, but I suspect that is a general cache age flush, rather than an
intended STP state change causing the flush later.

Thoughts or help much appreciated.

Cheers

James
___
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev