Re: BNX Double Buffer Allocation

["Aaron J. Grier" <agrier%poofygoof.com@localhost>]

On Thu, Aug 28, 2008 at 10:35:19PM +0200, Manuel Bouyer wrote:
> On Thu, Aug 28, 2008 at 01:24:23PM -0500, David Young wrote:
> > On Thu, Aug 28, 2008 at 11:13:12AM -0700, Aaron J. Grier wrote:
> > > I have seen this on our NAT at work as well.  bumping NMBCLUSTERS
> > > has helped, but what I don't understand is why the network gets
> > > completely wedged.  why can't it recover by dropping packets and
> > > reclaiming mbufs?
> > 
> > Drivers for network interfaces reserve many mbufs for receiving
> > packets, and they are not smart enough to shrink their reservation.
> 
> The mbuf can also be in socket queues; if you drop them you're
> effectively killing the connection.

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!

I see PR8129, which was closed in 2002 with the comment "Does not happen
with tlp."  something either broke along the line, or mbufs could be a
red herring.  what else allocates from mclpool?

-- 
  Aaron J. Grier | "Not your ordinary poofy goof." | 
agrier%poofygoof.com@localhost