Subject: Re: kern/12187: nfs_timer: ignoring error 55- what's it for and why
To: Manuel Bouyer <bouyer@antioche.lip6.fr>
From: Matthew Jacob <mjacob@feral.com>
List: current-users
Date: 04/08/2001 09:29:59
On Sun, 8 Apr 2001, Manuel Bouyer wrote:

> On Sat, Apr 07, 2001 at 10:17:14AM -0700, Matthew Jacob wrote:
> > [...]
> > 
> > tulip.
> 
> What driver ? tlp or de ?

de

> 
> > 
> > My point here is that an out of the box system performing normal operations on
> > an uncongested network should not lock up printing out obscure error
> > messages. I knew it was ENOBUFS.
> 
> It may be normal to get ENOBUFS from time to time. I'm not sure we shouldn't
> print it, it's a diagnostic info after all.

But it's the constant printing of the diagnostic that's really the problem.

> Note that the error can have 2 (or maybe more) sources: we're really
> out of mbuf; or the transmit queue of the adapter is full.
> 
> But this shouldn't lock up. I've got this message from time to time but
> it was never fatal. Maybe the bug is in the adapter driver ?

That's possible.  You'd think you'd get a diagnostic of out of MBCLUSTERs
first though.

I *have* seen this on different hardware, but not so bad. Same h/w running
FreeBSD's got no troubles whatsoever.

-matt