RE: Machine hangs(Beta7), only reset button works

From: Robert Watson <rwatson_at_freebsd.org>
Date: Fri, 22 Oct 2004 04:54:35 -0400 (EDT)
On Fri, 22 Oct 2004, Tom Jensen wrote:

> Ok, I managed to break into the with a break over serial console,
> attached is some info from KDB (please note that I have no knowledge
> about using the debugger) 
> 
> I also noticed that I don't have "makeoptions DEBUG=-g" in my kernel
> conf so I have no debug kernel. I rebuild my kernel ASAP so I'm able to
> provide more info sometime tomorrow. 
> 
> Please let me know if there is more info needed. 

Great, into the debugger is good news.  Looking at the thread wait states,
it looks like you might have a vnode deadlock going on.  It would be
useful if you could use the "show lockedvnods" command to show what vnode
locks are held.  If you recompile your kernel with DEBUG_LOCKS (and
compile all modules in, as this produces a kernel that is ABI-incompatible
with most modules), you will get extra debugging information when you use
that command (it will say what locks were acquired where, not just what
locks are held).

It looks like you're using SMBfs could you mention a little about how
that's in use, and whether things like home directories, etc, are mounted
that way?  Is it an element of the system you could remove during testing
to see if the problem goes away?  Also, are you using NFS or other
distributed file systems?

Robert N M Watson             FreeBSD Core Team, TrustedBSD Projects
robert_at_fledge.watson.org      Principal Research Scientist, McAfee Research



> 
> - Tom
> 
> -----Original Message-----
> From: owner-freebsd-current_at_freebsd.org
> [mailto:owner-freebsd-current_at_freebsd.org] On Behalf Of Robert Watson
> Sent: 21. oktober 2004 13:46
> To: Tom Jensen
> Cc: freebsd-current_at_freebsd.org
> Subject: Re: Machine hangs(Beta7), only reset button works
> 
> 
> On Thu, 21 Oct 2004, Tom Jensen wrote:
> 
> > I've been seeing a pretty strange problem lately with my server. 
> > 
> > The box completely freeze typically when it's done running the first 
> > part of my backup script, resulting in no possibility to login on the 
> > console or by SSH, the freeze even happens when I'm sitting in a 
> > terminal and working.
> > 
> > There is no indication in log files etc. about what's causing the 
> > problem and it's not breaking into debugger either :-(
> 
> This should probably be in debugging lore somewhere, but I've observed that
> it's often possible to break into the debugger using a break over serial
> console when it's not possible to break in using syscons.  This is because
> syscons requires the Giant lock, so if the freeze happens because a thread
> is spinning while holding Giant, you can't get in.  This needs to be fixed,
> but hasn't yet been fixed, so in the mean time often useful advice is to use
> a serial console to generate the break.
> 
> If you still can't get into the debugger, you might try some of the various
> watchdog drivers -- some hardware comes with built in watchdog parts, such
> as ichwd(4), or you could try options MP_WATCHDOG on an SMP box if you're
> willing to dedicate a CPU to running as a watchdog for the other cpu(s).
> 
> Robert N M Watson             FreeBSD Core Team, TrustedBSD Projects
> robert_at_fledge.watson.org      Principal Research Scientist, McAfee Research
> 
> 
> > 
> > The backup script is really simple, creating a .tgz file of a given 
> > directory, mounting a windows share (mount_smbfs) and copying the 
> > file. The script is run by cron six times (start at the same time) in 
> > six different directories, this results in the box freezes after the 
> > tar processes finishes.
> > 
> > Attached is the dmesg.boot and the latest top, don't know if it's any 
> > use but it's seems rather strange that a lot of processes are in a 
> > STATE usf (not sure what this means but I don't sees this when the box 
> > is running
> > normally)
> > 
> > The kernel is mostly a generic with the following modifications:
> > 
> > options         IPFIREWALL
> > options         IPFIREWALL_VERBOSE
> > options         IPFIREWALL_VERBOSE_LIMIT=400
> > options         IPDIVERT
> > options         IPSEC
> > options         IPSEC_ESP
> > options         IPSEC_DEBUG
> > device ath
> > device ath_hal
> > options         KDB
> > options         DDB          
> > 
> > bash-2.05b# uname -a
> > FreeBSD bart.motd.dk 5.3-BETA7 FreeBSD 5.3-BETA7 #6: Tue Oct 19 00:36:59
> > CEST 2004     root_at_bart.motd.dk:/usr/obj/usr/src/sys/GW  i386
> > 
> > Any more info needed please let me know.
> > 
> > Best regards
> > 
> > - Tom
> > 
> > 
> 
> _______________________________________________
> freebsd-current_at_freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-current
> To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"
> 
Received on Fri Oct 22 2004 - 06:54:52 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:19 UTC