Re: pre-2.1.37-7: OOPS (plus NumLock problem)

Carlo E. Prelz (
Sun, 11 May 1997 07:23:42 +0100 (WET DST)

On 10 May 1997, Linus Torvalds wrote:

> Would people try out pre-2.1.37-7 and see what it breaks? The largest
> change since the last pre-patch is simply that I split up the old bit
> operations in two: one for atomically doing a test-and-change operation,
> and one for just doing the change operation.

I did, and was bitten. I booted the kernel, then went to xwindows and
fired lynx. The machine froze solid, and I had to reboot. I rebooted, and
started lynx from text mode. I got the following oops:

Unable to handle kernel NULL pointer dereference at virtual address 00000000
current->tss.cr3 = 01472000, lr3 = 01472000
*pde = 00000000
Oops: 0002
CPU: 0
EIP: 0010:[<c0175f57>]
EFLAGS: 00010283
eax: 00000000 ebx: c0f748a8 ecx: 33743978 edx: c1fb9ca8
esi: c0f7486c edi: c1fb9ce0 ebp: c1fb9df0 esp: c1471f24
ds: 0018 es: 0018 ss: 0018
Process lynx (pid: 144, process nr: 30, stackpage=c1471000)
Stack: c1fb9ce0 c1fb9d58 00000000 00000000 00000286 c1fb9f8c c0172c66 c1fb9ce0
c1fb9ce0 c0499dec 00000000 bfffdb28 c017e54c c1fb9ce0 00000000 c0499dec
c0499d60 c01624d3 c0499dec 00000000 c0499d60 c016292c c0499dec c0de45c0
Call Trace: [<c0172c66>] [<c017e54c>] [<c01624d3>] [<c016292c>] [<c0121acf>] [<c0121b4a>] [<c0121bc6>]
Code: 08 00 00 83 c4 0c 5b 5e 5f 5d 83 c4 08 c3 8d 76 00 83 ec 2c
Error in symbol table input.
Cannot find map file.

As you can see, the symbol table info was not found by the kernel
(Still could not find why...) I compiled ksymoops and it segfaults
(long time since I used it last time). So here are the info from my
system map, extracted by hand:

00000000c0121a98 T sys_creat
00000000c0121ab0 T __fput
00000000c0121afc T close_fp
00000000c0121b80 T sys_close
00000000c0121bd0 T sys_vhangup
00000000c0162410 T sock_alloc
00000000c01624b4 T sock_release
00000000c01624f4 T sock_sendmsg
00000000c01628c8 t sock_poll
00000000c01628f4 t sock_close
00000000c0162930 t sock_fasync
00000000c0172ae4 T tcp_shutdown
00000000c0172b50 T tcp_close
00000000c0172d58 t wait_for_connect
00000000c0175b44 T tcp_do_retransmit
00000000c0175d44 T tcp_send_fin
00000000c0175f68 T tcp_send_synack
00000000c017e4bc t inet_dup
00000000c017e4dc T inet_release
00000000c017e554 t inet_bind

After this oops, I got back to the prompt. But when I started shutdown
(to go back to pre-patch-5 - I did not test pre-patch-6) I got another
oops and the machine froze. I hoped to find this other oops in the log
files but there was no trace of it (I may redo all steps and write
down the oops stuff if needed).


Another topic: I read a couple of messages about having problems with numlock & the like. This is what happens to me:

1) the NumLock led is ON at boot (it was off with 2.1.36, IIRC). 2) if I then boot as root on term 1, and I press the NumLock key once, the NumLock led does NOT go off 3) if I press it a second time, it goes off, and then all goes on more or less OK. 4) If I press numlock while the booting process is going on, or to a term that is not logged in, or otherwise I do not follow steps 2 and 3 above, I find myself in a condition when the machine remains locked for a few seconds. If I connect from another (networked) machine with ssh and run top, I can see that the situation unlocks when top does a refresh of its output. If I type keys, it seems that the unlocking process is delayed. If I wait patiently, when the top output on the other PC changes I can type again on that multiple console, or switch to another. But if I do other things with numlock or capslock keys, I am blocked again for another period.

Am I the only one??

Ciao Carlo

  *                                          ...Ma appena fuori tutto e' gomma,
* K * Carlo E. Prelz -     tutto e' cicca impiastricciata...
  *                                     (Marco Zappa-Niente cicca nella scuola)