kern/103256: em0: watchdog timeout -- resetting (6.1-STABLE)

[Gleb Smirnoff]

The following reply was made to PR kern/103256; it has been noted by GNATS.

From: Gleb Smirnoff <glebius at FreeBSD.org>
To: Imri Zvik <imriz at co.zahav.net.il>
Cc: freebsd-gnats-submit at FreeBSD.org
Subject: Re: kern/103256: em0: watchdog timeout -- resetting (6.1-STABLE)
Date: Fri, 15 Sep 2006 18:35:59 +0400

 On Thu, Sep 14, 2006 at 08:46:08AM +0000, Imri Zvik wrote:
 I> >Environment:
 I> FreeBSD wintergreen 6.1-STABLE FreeBSD 6.1-STABLE #1: Wed Aug 30 09:54:35 IDT 2006     root at wintergreen:/usr/obj/usr/src/sys/IGLD  i386
 I> >Description:
 I> I've recently upgraded my 6.0-STABLE machine to 6.1-STABLE, and since the upgrade i'm getting "em0: watchdog timeout -- resetting" messages every once in a while (sometimes twice a day, sometimes every couple of days). This causes the machine to be unstable, network-wise (connections being dropped, etc.)
 I> 
 I> Sep 14 06:35:41 wintergreen kernel: em0: watchdog timeout -- resetting
 I> Sep 14 06:35:41 wintergreen kernel: em0: link state changed to DOWN
 I> Sep 14 06:35:42 wintergreen kernel: em0: link state changed to UP
 I> 
 I> 
 I> i've found this discussion on freebsd-stable (http://www.mail-archive.com/freebsd-stable@freebsd.org/msg81865.html) indicating that i am not the only one suffering from this issue.
 
 As I already have mentioned many times on the mailing lists. Before
 revision 1.65.2.14 of if_em.c, the watchdog timeout event's was not
 logged at all.
 
 However, there is definitely a regression in RELENG_6 that leads to
 fake(?) watchdog timeouts on some systems. Unfortunately, I haven't
 yet reproduced it on my systems.
 
 Some people report that this is SMP issue, setting debug.mpsafenet=0
 plumbs the problem.
 
 Not sure whether you have the regression, or you are seeing the watchdog
 events, that were not logged before.
 
 -- 
 Totus tuus, Glebius.
 GLEBIUS-RIPN GLEB-RIPE