Comment # 15 on bug 1201859 from Dario Faggioli

(In reply to Michal Kube������ek from comment #12)
> I'm not really an expert on drivers, I tend to focus on core networking,
>
Yeah, I know. In fact, thanks for answering my call, and giving us your opinion
anyway.

Do you know if we there is anyone else that we can ping, that works more on
these specific aspects (even if not specifically on this driver)?

> but AFAICS the log excerpt in the initial comment rather indicates
> a hardware/firmware issue than a software one. The watchdog timeout with
> the stack trace is just a fallout of the actual problem (kernel tells us
> that a packet passed to NIC to transmit has not been transmitted within
> a time limit (few seconds) which should not happen unless something is
> seriously wrong).
> 
Indeed. And this is what we concluded, when we say the first instance of this
bug (the one I added a reference to).

What's strange here, in my opinion, is that Mathias says he has started to see
this problem after a recent kernel update. I was thinking it could have been a
patch that we've backported... but this is Tumbleweed, and I don't think there
are many backports applied to the TW kernel.

Maybe this requires a certain traffic pattern and, even if Mathias says also
the workload is being the same, some unrelated change makes it now
possible/more probable for such conditions to occur?