Stack backtrace: how can I help?

From: Sven Willenberger (sven_at_dmv.com)
Date: 06/30/04

  • Next message: Niki Denev: "duplicated/bad output from apm(8)"
    To: current@freebsd.org
    Date: Wed, 30 Jun 2004 15:34:05 -0400
    
    

    My abilities to dig into kernel routines, etc is very limited so I am
    asking how I can help those who may be able to fix this recurring
    problem.

    This has been posted by myself and others with utterly no response from
    anyone other than one response saying "it must be a bug".

    Under heavy loads, on 5.2.1-P8 systems, I get a Stack backtrace relating
    to flushing dirty buffers (ffs_fsync).

    the relevant code from ffs_softdep.c ( src/sys/ufs/ffs/ffs_softdep.c,v
    1.149 2003/10/23 21:14:08 jhb )

    getdirtybuf(bpp, mtx, waitfor)
            struct buf **bpp;
            struct mtx *mtx;
            int waitfor;
    {
            struct buf *bp;
            int error;

            /*
             * XXX This code and the code that calls it need to be reviewed
    to
             * verify its use of the vnode interlock.
             */

            for (;;) {
                    if ((bp = *bpp) == NULL)
                            return (0);
                    if (bp->b_vp == NULL)
                            backtrace();
    .....

    It does seem related to the load created by perl (these machines run
    spamassassin through either mimedefang or milter-spamc) and are now
    running 5.8.4; the upgrade to perl made no difference ... still getting
    these backtraces. Each machine handles (filters) roughly 120K email
    messages per day.

    a) what additional information would be of help here
    b) what can I do to help troubleshoot this -- for the most part the
    machines recover after the backtrace (of course they are inoperable
    during the time the trace is generated creating further backlog/work for
    the other machines in the cluster) although occasionally it will cause a
    panic and either reboot or hang at sync.
    c) is it possible to cvsup the latest ffs files and make install those
    without killing the machine?

    _______________________________________________
    freebsd-current@freebsd.org mailing list
    http://lists.freebsd.org/mailman/listinfo/freebsd-current
    To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org"


  • Next message: Niki Denev: "duplicated/bad output from apm(8)"

    Relevant Pages

    • Re: Softupdate/kernel panic ffs_fsync
      ... >> essentially identical configs) I get the following Stack Backtrace ... > becoming frustrating as the machines are bailing under heavy traffic. ... /dev/da0s1e on /tmp (ufs, local, soft-updates) ...
      (freebsd-current)
    • Re: Problem with syslog and perl
      ... > AG> on our development system everything works fine, all log messages from ... > machines), no AG> log mesages at all are written. ... > but it shows no reaction AG> when trying to write messages from perl. ...
      (comp.unix.aix)
    • Re: Can gcc 4.0 be used to build DBI modules if perl is bulit with gcc 3.3?
      ... We have loaded Tiger onto several machines here. ... My production machine still uses 10.39 and I am afraid to upgrade ... because of the problems we are seeing trying to install the DBI and DBD ... had DBI and DBD installed in perl on the 10.39 MacOSX. ...
      (perl.dbi.users)
    • What causes this 32 bit problem??
      ... I've tested your DBD::Oracle 1.20 against both PA-RISC and Itanium ... Machines with the 32 bit versions of 5.8.8 perl. ... Registered Office: 2100 East Grand Avenue, El Segundo California 90245, ...
      (perl.dbi.users)
    • Re: Perl performance difference between Linux distributions
      ... On both machines I used the standard Perl binaries that came with the ... What if you compile your own perl on each machine from the same source ... Maybe the vendor distributions have done something strange to their ...
      (comp.lang.perl.misc)