Linux 1.2.13 general protection: 0000 when crond was envoked

David R VanWagner (davevw@cats.ucsc.edu)
Sun, 3 Sep 1995 02:01:55 -0700 (PDT)


general protection: 0000
EIP: 0010:0012ec14
EFLAGS: 00010206
eax: 00080000 ebx: 57e58999 ecx: 0000002f edx: 00377000
esi: 00377000 edi: bffffb0d ebp: bffffe84 esp: 00b84f6c
ds: 0018 es: 0018 fs: 002b gs: 002b ss: 0018
Process crond (pid: 24022, process nr: 24, stackpage=00b84000)
Stack: 00080000 57e58955 0012303b 00377000 00080000 0000002f 00000000 001230d1
00080000 0011a6fd 0000002f 00377000 0011a7be 00377000 00377000 00377000
00048e00 0011a9de 00000000 00110599 00000000 bffffa38 00000000 00048e00
Call Trace: 0012303b 001230d1 0011a6fd 0011a7be 0011a9de 00110599
Code: 8b 03 85 c0 74 20 90 90 39 70 0c 74 19 89 c3 8b 00 85 c0 75
crond trying to free kernel page-directory: not good

bluelbo:/usr/src/linux$ nm vmlinux | sort | grep 0012ec
0012ec04 T _fcntl_remove_locks
0012ec44 t _copy_flock
0012ece8 t _conflict

I was minding my own business, compiling some C++ code when cron invoked one
of these commands
0 0 * * * mv /tmp/Master.tar.bk1 /tmp/Master.tar.bk2 >/dev/null 2>&1
1 0 * * * mv /tmp/Master.tar /tmp/Master.tar.bk1 >/dev/null 2>&1
2 0 * * * tar cvf /tmp/Master.tar /home/dave/Src/Master >/dev/null 2>&1
and I saw this general protection error. I was *NOT* compiling the
code in the directory that tar was archiving, I was compiling under
Src/mfdb2/.... And the EIP / symbol table cross reference suggests a
problem with file locking?

>From the looks of /var/adm/messages, it was the tar command that was
beginning (or supposed to):
Sep 3 00:02:03 bluelbo kernel: general protection: 0000

I tried to reproduce the problem by staggering all the cron jobs by 3 minutes
and constantly compile the same source (make clean ; make ; ... ) Hasn't
happened again (this is a NEW problem, happenned only once). I will leave
my system grinding to see if it'll happen again. It had been up 8 days
but I rebooted after the "panic."

The 1.2.13 kernel is patched with kerneld's patch-for-1.2.8 (kernel
daemon support) and the fd.c patched to allow reading of Commodore 1581
3.5" floppies, but I don't think those patches are relevant to the
following that happenned on my 386, 12MB RAM, 16MB SWAP, Sound Blaster,
CONFIG_MODVERSIONS=y setup.

Thanks.

-Dave Van Wagner [guy at UCSC talk with drawing of self on shirt]