Crash report: gettimeofday?

Zefram (A.Main@dcs.warwick.ac.uk)
Sun, 30 Jul 1995 01:12:22 +0100 (BST)


I have experienced a crash, with a 1.2.3 kernel. The system locked
up, to the extent that I couldn't switch VCs. There was a little
hard disk activity a few seconds later, and nothing else. The
three-finger-salute had no effect. The following got logged in
/var/adm/messages:

invalid operand: 0000
EIP: 0010:0011d925
EFLAGS: 00010006
eax: 02e2ebee ebx: 003b0fb0 ecx: 57e11add edx: 000012ea
esi: 00000000 edi: 00000216 ebp: bffffbf8 esp: 003b0f90
ds: 0018 es: 0018 fs: 002b gs: 002b ss: 0018
Process linuxsdoom (pid: 11425, process nr: 21, stackpage=003b0000)
Stack: bffffc20 bffffc18 00000000 02e2ebee 0011d994 003b0fb0 007db000 00000000
30192373 000258e7 001106f9 bffffc20 bffffc18 000000bc 00000000 00000000
bffffbf8 ffffffda 0000002b ffff002b 0011002b 0000002b 0000004e 60018049
Call Trace: 0011d994 001106f9 0011002b
Code: ff e9 89 d1 c1 f9 0c 8b 44 24 0c c1 f8 1f 29 c1 01 ce 89 74

Each line was preceded by "Jul 28 17:50:43 isis kernel: ", which is
probably unimportant, except possibly for the time...

My zSystem.map contains the following:

[...]
0010ff44 t gcc2_compiled.
0010ff44 t signal.o
0010ff94 T _sys_sigreturn <-- 0011002b
00110164 T _setup_frame
001102a4 T _do_signal
00110620 T _lcall7
00110620 t entry.o
00110670 t handle_bottom_half
00110690 t reschedule
001106a0 T _system_call <-- 001106f9
00110740 T ret_from_sys_call
001107e0 t signal_return
[...]
0011d5d8 t time.o
0011d7f8 T _sys_time
0011d828 T _sys_stime
0011d8a8 T _do_gettimeofday <-- 0011d925 (EIP)
0011d968 T _sys_gettimeofday <-- 0011d994
0011d9d8 T _sys_settimeofday
0011db98 T _sys_adjtimex
[...]

Upon rebooting, fsck reported "Deleted inode nnnnnn has zero dtime."
(I forget the number). A few hours later, I had a similar crash.
The EIP and call trace were identical. This time fsck reported no
damage.

This was Slackware's "scsi" 1.2.3 kernel, unmodified. (I haven't
got around to compiling a more recent kernel yet.) I wouldn't have
bothered reporting a crash of such an old version, but kernel/time.c
(which appears to be at fault, if anything is) is identical in the
1.2.11 release.

I hope someone can make use of this.

-zefram