Rob.Montjoy@uc.edu (Robert C. Montjoy) (03/07/90)
Can anyone till me how to decode thes messages. The kernel is panicing about 3 or 4 times a week the addresses in the trace are always indentical. ALso,it always gives the "BAD TRAP" message before dieing. Mar 4 22:52:39 vlsilab vmunix: NFS server babbage.ece.uc.edu ok Mar 4 23:06:54 vlsilab vmunix: BAD TRAP Mar 4 23:06:54 vlsilab vmunix: SIM.vs.7175: Data fault Mar 4 23:06:54 vlsilab vmunix: kernel read fault at addr=0xff640000, pme=0x0 Mar 4 23:06:54 vlsilab vmunix: Sync Error Reg 80<INVALID> Mar 4 23:06:54 vlsilab vmunix: pid=4516, pc=0xf806149c, sp=0xffffecc0, psr=0x10c2, context=0 Mar 4 23:06:54 vlsilab vmunix: g1-g7: 4010e3, 4010e3, f80a97f8, d, 0, f80a9800, f80a9400 Mar 4 23:06:54 vlsilab vmunix: Begin traceback... sp = ffffecc0 Mar 4 23:06:54 vlsilab vmunix: Called from f801fb74, fp=ffffed20, args=ffffc000 ff640000 38a0 ffffec18 f80e6a18 f80c1b20 Mar 4 23:06:54 vlsilab vmunix: Called from f801f988, fp=ffffed98, args=f80e2618 94 1a ffffff6c ffffffff ff09bcb8 Mar 4 23:06:54 vlsilab vmunix: Called from f801f460, fp=ffffedf8, args=1 10 4 f80e3080 f80e0c20 f80e2618 Mar 4 23:06:54 vlsilab vmunix: Called from f801f2b0, fp=ffffee60, args=1 0 0 ff0763ac f80e3080 f80e2618 Mar 4 23:06:54 vlsilab vmunix: Called from f807a26c, fp=ffffeec0, args=ffffefe0 f809e3d0 f809e3d0 0 ffffefb4 ffffefe0 Mar 4 23:06:54 vlsilab vmunix: Called from f8005808, fp=ffffef58, args=8000000 42 0 42 42 210 Mar 4 23:06:54 vlsilab vmunix: Called from 596d8, fp=f7fff4e0, args=e 0 758a0 80000001 e 758a0 Mar 4 23:06:54 vlsilab vmunix: End traceback... Mar 4 23:06:54 vlsilab vmunix: panic: Data fault Mar 4 23:06:54 vlsilab vmunix: zs3: silo overflow Mar 4 23:06:54 vlsilab vmunix: syncing Thanks, Rob Montjoy E-MAIL Rob.Montjoy@UC.EDU uunet!uccba!ucece1!montjoy
lm@sun.com (Larry McVoy) (03/08/90)
In article <5543@brazos.Rice.edu> Rob.Montjoy@uc.edu (Robert C. Montjoy) writes: >Can anyone till me how to decode thes messages. The kernel is panicing >about 3 or 4 times a week the addresses in the trace are always >indentical. ALso,it always gives the "BAD TRAP" message before dieing. > >Mar 4 22:52:39 vlsilab vmunix: NFS server babbage.ece.uc.edu ok >Mar 4 23:06:54 vlsilab vmunix: BAD TRAP > [panic messages deleted] OK, try this. You have to be running the same kernel binary as was running at the time of the crash. If you can't do that, you need a copy of the binary and a core file (usually in /var/crash/`hostname`/vmunix.xxx and .../vmcore.xxx). Say something like # adb /vmunix /dev/kmem or # adb vmunix.0 vmcore.0 The you start walking backwards like so (look at the stack trace above to follow along): 0xf806149c?ia function1+0x100 f801fb74?ia function2+0x80 f801f988?ia function3+0x100 This will give you a symbolic stack trace back. I have a couple of shell scripts that do this automagically for the last panic on a system, these work (or used to work) on sun3/sun4 running 4.1. I'm passing them on for reference purposes only. --- sun4 gettrace ---- #!/bin/sh # $0 [-v] [vmunixfile [messagesfile]] if [ "$1" = "-v" ] then VERBOSE=ON; shift else VERBOSE=OFF fi if [ $# -ge 1 ] then VM=$1; shift else VM=/vmunix fi if [ $# -ge 1 ] then MSGS=$1; shift else MSGS=/var/adm/messages fi sed -e 's/.*vmunix: //' -e 's/pc=0x/PC= /' -e 's/,//' \ < $MSGS \ | egrep 'Called|PC=' > /tmp/trace$$ cat - <<EOF | ed /tmp/trace$$ > /dev/null $ ?PC=? 1,.-1d w q EOF if [ $VERBOSE = ON ] then echo STACK TRACEBACK for $VM on `date` echo '' echo ADDRESSES + ARGS echo '' cat /tmp/trace$$ echo '' echo SYMBOLIC TRACEBACK echo '' fi sed -e 's/Called from \([0-9a-fA-F]*\).*/\1\?ia/p' \ -e 's/PC= \([0-9a-fA-F]*\).*/\1\?ia/p' < /tmp/trace$$ | adb $VM - /bin/rm -f /tmp/trace$$ ---- sun3 gettrace ---- #!/bin/sh if [ "$1" = "-v" ] then VERBOSE=ON; shift else VERBOSE=OFF fi if [ $# -eq 1 ] then VM=$1 else VM=/vmunix fi sed -e 's/.*vmunix: //' -e 's/pc 0x/PC=/' -e 's/,//' \ < /usr/adm/messages \ | egrep 'pid|D0-D7|A0-A7|Called|PC=' > /tmp/trace$$ cat - <<EOF | ed /tmp/trace$$ > /dev/null $ ?PC=? 1,.-2d w q EOF if [ $VERBOSE = ON ] then echo STACK TRACEBACK for $VM on `date` echo '' echo ADDRESSES + ARGS echo '' cat /tmp/trace$$ echo '' echo SYMBOLIC TRACEBACK echo '' fi egrep -v 'pid|D0-D7|A0-A7' < /tmp/trace$$ | \ sed -e 's/Called from \([0-9a-fA-F]*\).*/\1\?ia/p' \ -e 's/.*PC=\([0-9a-fA-F]*\).*/\1\?ia/p' \ | adb $VM - /bin/rm -f /tmp/trace$$ [[Ed's Note: not placed in archives since they were explicitly posted "for reference only" -bdg]] What I say is my opinion. I am not paid to speak for Sun, I'm paid to hack. Besides, I frequently read news when I'm drjhgunghc, err, um, drunk. Larry McVoy, Sun Microsystems (415) 336-7627 ...!sun!lm or lm@sun.com