so, i don't have a ka660, but would it seem more generally the case that
when small (or small of a specific size) packets are moved the right way
this happens? does it seem to be a bug in the nfs code that only happens
to show up in this specific configuration, or does it seem to be a bug in
the ze code? does it only happen on a ka660? there are ze's on other newer
(KA6[789]*) machines as well. I have a 4k600 i could try to duplicate
this on, though that machine has run 1.6.1 very stably without trouble. i 
recall seeing when i was running 1.5M on it the xmit logid died message,
but it didnt seem to be fatal. 

isildur

On Thu, 28 Oct 2004, Alex White wrote:

> 
> 
> On Thu, 28 Oct 2004, Patrick Finnegan wrote:
> 
> > So, I'm (trying) to run NetBSD 1.6, currently using the 1.6.2 kernel
> > that's available from ftp.netbsd.org, to run a diskless VAX 4000/300.
> > Every so often, I manage to get the machine into an unusable state,
> > where it appears to freeze up an nfs mount, coinciding with this
> > console message repeated several times:
> >
> > ze0: xmit logic died, resetting...
> 
> Good luck, i've seen this problem since ~1.5 and never got around it. I
> tried several switches to nfs to get it to mount in different ways, or
> transmit data in differently-sized chunks. No beans, I just lived with it
> and eventually gave up with NetBSD on KA660.
> 
> alex/melt
>