Comment 15 for bug 1894942

------- Comment From <email address hidden> 2020-10-14 21:19 EDT-------
(In reply to comment #75)
> (In reply to comment #72)
> > Checking SRU compatibility:
> >
> > #1 Source
> > Focal: some (minor) backport noise
> > Bionic: the same backport of Focal applies with offsets.
> > Xenial: some more nois (tracepoint missing), but still ok
> >
> > Not as-is, but pretty close - should be ok
> >
> > #2 Builds
> > I have started builds of those in
> > https://launchpad.net/~ci-train-ppa-service/+archive/ubuntu/4270
> >
> > #3 testing
> > The builds in the PPA succeeded now.
> > I'd ask IBM to verify with their tests to check if they can
> > a) trigger the issue on those releases
> > and
> > b) if the PPA then is fixing for them as expected.
>
> Both focal and xenial look good: the issue can be triggered with an up to
> date distro qemu and can not be triggered with the PPA qemu.
>
> I've ran ran into some unrelated issues with bionic, I will try to test
> there as well after the impediments are out of the way.
>

I've managed to test bionic as well. It is pretty much same as xenial. That means while the source code is broken, in a sense that the compiler is allowed to generate broken code, the generated code does not exhibit the issue in question.

As with xenial I do see intermittent connection problems with iperf3, but the recv queue does not get stuck for good (like in focal, because of the lost interrupt).

The presumably fixes the code source I can not see any positive or negative impact with regards to the intermittent and recoverable connectivity problems.