Date: Sun, 11 Aug 2002 09:08:00 -0400 From: Mike Tancsa <mike@sentex.net> To: stable@freebsd.org, Luigi Rizzo <luigi@freebsd.org> Cc: Ian Dowse <iedowse@maths.tcd.ie> Subject: re: fxp problems with latest stable Re: panics after upgrading to -STABLE Aug 9, 2002 (still something up) Message-ID: <5.1.0.14.0.20020811090030.06a7d640@192.168.0.12> In-Reply-To: <20020811024117.B82972@iguana.icir.org> References: <5.1.0.14.0.20020810234921.014d7ec0@192.168.0.12> <Your <5.1.0.14.0.20020810225848.068a4690@192.168.0.12> <200208110434.aa60377@salmon.maths.tcd.ie> <5.1.0.14.0.20020810234921.014d7ec0@192.168.0.12>
next in thread | previous in thread | raw e-mail | index | archive | help
Thanks once again Luigi and Ian! sysctl -w kern.ipc.mcl_pool_max=0 does
fix it for now.
---Mike
At 02:41 AM 8/11/2002 -0700, Luigi Rizzo wrote:
>Hi,
>I am looking at the problem -- I am able to replicate it so i expect
>to have a fix committed tonight. For the time being,
>setting mcl_pool_max=0 in kern/uipc_mbuf.c should cure the problem.
>
>if_fxp.c seems to be innocent. The problem appears to be related
>to the change to m_getcl()/m_freem(), probably i am not (re)initializing
>some mbuf fields when i put the freed buffers in the free pool,
>resulting in some cases in invalid buffers, which are dropped and
>recycled. Eventually these buffers get out of the way because with
>two identical boxes A and B, freshly rebooted, i see the following
>sequence of events:
>
> A# ping -f B --> losses
> B# ping -f A --> no losses
> A# ping -f B --> no losses now!
>
>and reversing A and B produces (after both reboot) the same effect in
>reverse order. Also, setting kern.ipc.mcl_pool_max=0 on the destination
>host seems to remove the losses.
>
>Back to debugging...
>
> cheers
> luigi
>
>On Sat, Aug 10, 2002 at 11:53:13PM -0400, Mike Tancsa wrote:
> > At 04:34 AM 8/11/2002 +0100, Ian Dowse wrote:
> > >In message <5.1.0.14.0.20020810225848.068a4690@192.168.0.12>, Mike Tancsa
> > >write
> > >s:
> > > >Also, the pings less lossy with your patch. Note, these are non busy
> > > >machines (PIII 1000 and PIV1.6). The patch does seem to make the
> problem
> > > >less acute. But, it is still there.
> > >
> > >Thanks for the info. Yet another thing to try:
> > >
> > > sysctl kern.ipc.mcl_pool_max=0
> > >
> > >Ian
> >
> >
> > That does the trick. Everything on my 2 test boxes are exactly what is in
> > the cvsup repository
> >
> > newhespborder# sysctl -w kern.ipc.mcl_pool_max=0
> > kern.ipc.mcl_pool_max: 10 -> 0
> > newhespborder#
> >
> > ping -q -c 400 -i .01 192.168.43.218
> > PING 192.168.43.218 (192.168.43.218): 56 data bytes
> >
> > --- 192.168.43.218 ping statistics ---
> > 400 packets transmitted, 400 packets received, 0% packet loss
> > round-trip min/avg/max/stddev = 0.119/0.126/0.235/0.013 ms
> >
> >
> > And same with the other box.
> >
> > ping -q -c 400 -i .01 192.168.43.91
> > PING 192.168.43.91 (192.168.43.91): 56 data bytes
> >
> > --- 192.168.43.91 ping statistics ---
> > 400 packets transmitted, 400 packets received, 0% packet loss
> > round-trip min/avg/max/stddev = 0.112/0.121/0.223/0.010 ms
> >
> > The one strange thing I do notice is that initially there is a bit of loss
> > and then its OK.
> >
> > ---Mike
> > --------------------------------------------------------------------
> > Mike Tancsa, tel +1 519 651 3400
> > Sentex Communications, mike@sentex.net
> > Providing Internet since 1994 www.sentex.net
> > Cambridge, Ontario Canada www.sentex.net/mike
> >
--------------------------------------------------------------------
Mike Tancsa, tel +1 519 651 3400
Sentex Communications, mike@sentex.net
Providing Internet since 1994 www.sentex.net
Cambridge, Ontario Canada www.sentex.net/mike
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?5.1.0.14.0.20020811090030.06a7d640>