[ofa-general] IPoIB-UD TX timeouts (OFED 1.2)

[akepner at sgi.com]

On Wed, Apr 30, 2008 at 11:00:55PM +0300, Eli Cohen wrote:
> ....
> when it happens please:
> 1. Check the link error counters.

Unfortunately there appear to be things running that periodically 
reset the counters (to avoid hitting the 32 bit limit), so the 
port counters usually come back as all 0.

> 2. Disconnect and reconnect the cable and see if it recovers.

None of the systems where this has been seen are physically 
accessible to me (and even if they were, finding the right cable 
to pull might be tricky :-)

We have some new information, which I'll post now.

-- 
Arthur