Re: BNX Double Buffer Allocation

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Fri, Aug 29, 2008 at 12:13:40PM -0700, Aaron J. Grier wrote:
> On Fri, Aug 29, 2008 at 01:28:44PM +0200, Manuel Bouyer wrote:
> > On Thu, Aug 28, 2008 at 03:29:33PM -0700, Aaron J. Grier wrote:
> > > this opposed to the current situation of wedging the network stack
> > > and requiring a reboot to get it back?  the result of killing
> > > connections is the same in both cases!
> > 
> > No reboot needed when I see it, the box recover by itself when the
> > stalled TCP connections time out ... so yes, in my case it's better to
> > let the TCP mechanism time out conections that are really dead,
> > instead of dropping active connections.
> 
> you're lucky.  I've been completely wedged for hours at a time.  a panic
> and reboot would be less disruptive.
> 
> is there any way to find out how many buffers each connection is using?
> a combination of netstat -a and -m?

netstat -a; size of both queues. A mbuf cluster is 2k, so the number
of mbcl is easy to find ...

> 
> this is all still assuming it is an mbuf issue and something else.  I
> still don't know what else draws from mclpool.

Almost anything that needs to store data in mbufs.

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--