Re: panic: bnx0 Double mbuf allocation failure!

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

On Tue, Jul 01, 2008 at 10:42:40PM +0200, Christoph Egger wrote:
> Fredrik Carlsson wrote:
> >Manuel Bouyer wrote:
> >>On Tue, Jul 01, 2008 at 10:27:27AM +0000, Fredrik Carlsson wrote:
> >> 
> >>>
> >>>We did a desperate upgrade to CURRENT in hope that this would fix the 
> >>>problem ;)
> >>>The machine became much faster but as soon as we put some load on, it 
> >>>panics but
> >>>still responds to ping. There is nothing in the logs after a power 
> >>>cycle.
> >>>
> >>>Is there anything we can do to help troubleshoot this, would a 
> >>>DEBUG-kernel
> >>>help? We can supply ssh access if that would make it easier.
> >>>
> >>>The load that causes this is a MySQL import over the network, a lot 
> >>>of CPU, disk
> >>>access and nfs. The machine panics after 30-60 minutes into the import.
> >>>    
> >>
> >>Do you have the panic message, and eventually the stack trace ?
> >>
> >>  
> >We don't have the panic message (no console access) but we are trying to 
> >create an usable core dump.
> >
> >When we used a MP kernel gdb complained about "Cannot access memory at 
> >address 0xffff80005bfca6c0" so right now we have booted on one CPU and 
> >are trying to produce a panic, but so far it is stable. Before when we 
> >copied files over NFS it paniced several times a few GB into the copy 
> >but now it runs fine, perhaps this has to do with the MP-kernel?
> >
> >Regards
> >Fredrik
> >
> 
> Does this diff help ?
> http://www.netbsd.org/~cegger/if_bnx.diff
> 
> This is a fix ported from FreeBSD via OpenBSD.
> The OpenBSD diffs are:
> http://www.openbsd.org/cgi-bin/cvsweb/src/sys/dev/pci/if_bnx.c.diff?r1=1.63&r2=1.64
> http://www.openbsd.org/cgi-bin/cvsweb/src/sys/dev/pci/if_bnxreg.h.diff?r1=1.25&r2=1.26
> 
> OpenBSD's commit log:
> -------------------------------------------------------
> Fixed a problem that would cause errors (especially when in low memory
> systems) because the RX chain was corrupted when an mbuf was mapped to
> an unexpected number of buffers.

Isn't it what I fixed in rev 1.5 ?

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