OOPS, 2.4.20, preemption, ptrace, ext3 patches; (sis900?/ncpfs?)

Hello!

I got something like this today (more details below):

<OOPS>
ksymoops 2.4.8 on i686 2.4.20-i. Options used
-V (default)
-k ./ksyms (specified)
-l /proc/modules (default)
-o /lib/modules/2.4.20-i/ (default)
-m ./System.map-2.4.20-i (specified)

OOPS: 0000

CPU: 0  

EIP: 0010: [<c0114f10>] Not tainted
Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010017
eax: c2034000 ebx: 00000000 ecx: 00000001 edx: 00000001
esi: c21234ec edi: 00000001 ebp: c2035f54 esp: c2035f40
ds: 0018 es: 0018 ss: 0018
Process nwsfind (pid: 764, stackpage=c2035000)

Stack: 00000001 00000282 c2aad3e0 c3e9e9e0 fffffffd 00000046 c01c7d4f c2aad3e0
       c01c71f7 c2aad3e0 c3e9e9e0 c01c846a c3e9e9e0 00000000 c01cc9b4 c3e9e9e0
       00000001 c028db68 fffffffd 00000046 c011cfd4 c028db68 c2034000 00000002
Call Trace: [<c01c7d4f>] [<c01c71f7>] [<c01c846a>] [<c01cc9b4>] [<c011cfd4>] [<c0108706>] [[<c010adf3>] 
Code: 8b 53 fc 8b 02 85 c7 75 2f 8b 1b 39 f3 75 f1 b8 00 e0 ff ff

Trace; c01c7d4f <sock_def_write_space+8f/a0>
Trace; c01c71f7 <sock_wfree+37/40>
Trace; c01c846a <__kfree_skb+4a/150>
Trace; c01cc9b4 <net_tx_action+34/c0>
Trace; c011cfd4 <do_softirq+d4/e0>
Trace; c0108706 <do_IRQ+f6/130>
Trace; c010adf3 <call_do_IRQ+5/12>

Code; c0114f10 <__wake_up+30/90>
00000000 <_EIP>:
Code; c0114f10 <__wake_up+30/90> <=====
0: 8b 53 fc mov 0xfffffffc(%ebx),%edx <=====
Code; c0114f13 <__wake_up+33/90>
3: 8b 02 mov (%edx),%eax
Code; c0114f15 <__wake_up+35/90>
5: 85 c7 test %eax,%edi
Code; c0114f17 <__wake_up+37/90>
7: 75 2f jne 38 <_EIP+0x38>
Code; c0114f19 <__wake_up+39/90>
9: 8b 1b mov (%ebx),%ebx
Code; c0114f1b <__wake_up+3b/90>
b: 39 f3 cmp %esi,%ebx
Code; c0114f1d <__wake_up+3d/90>
d: 75 f1 jne 0 <_EIP>
Code; c0114f1f <__wake_up+3f/90>
f: b8 00 e0 ff ff mov $0xffffe000,%eax

Aiee, killing interrupt handler. In interrupt handler: not syncing.

</OOPS>

This station has to have access to ncpfs volumes and for last few days has also been print server. Everything worked fine until motherboard upgrade (new CPU, network card is now onboard,...). Now kernel oopses every time I reboot, after umounting local filesystems.

I believe it may be ncp or sis900 (or both) related (please forgive me if I misunderstood this oops :)

I didn't test neither 2.4.21-rc1, nor 2.4.20 without preemption enabled, nor other NIC, but I will try to do it as soon as possible.

Linux:
-----
Linux indis 2.4.20-i #1 pon mar 17 21:32:14 CET 2003 i686 unknown unknown GNU/Linux
(with ptrace, preempt-2.4.20-1, all 2.4.20 ext3 patches)

Utils:
-----
modutils: 2.4.21
gcc: version 3.2.3 20030309 (Debian prerelease
ncpfs: 2.2.3 (without NDS), as in Debian SID
mount: mount-2.11z
binutils: 2.13.90.0.18

NIC:
---
Silicon Integrated Systems [SiS] SiS900 10/100 Ethernet (rev 82)

- More -
--------

Config:
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.20-i/oops1/config-2.4.20-i

Dmesg:
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.20-i/oops1/dmesg

cpuinfo:
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.20-i/oops1/cpuinfo.txt

lspci:
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.20-i/oops1/lspci.txt
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.20-i/oops1/lspci_-v.txt
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.20-i/oops1/lspci_-v_-v.txt

modules (before reboot):
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.20-i/oops1/modules.txt

oops (one more time)
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.20-i/oops1/oops_decoded.txt

JK

Jacek Kawa, jfk[at]zeus.polsl.gliwice.pl

In addition to previous report, here is an oops from 2.4.21-rc1 (this time without preemption enabled). To reproduce it I only have to try to deconfigure interface (ifdown eth0, where eth0 is SIS 900); once I got it while just pinging another host...

As a quick workaround I put there another NIC...

<OOPS>
ksymoops 2.4.8 on i686 2.4.21-rc1+p. Options used
-V (default)
-k ./ksyms (specified)
-l ./modules (specified)
-o /lib/modules/2.4.21-rc1-i2/ (specified) -m ./System.map-2.4.21-rc1-i2 (specified)

Unable to handle kernel paging request at virtual address 00003615
c0116a3b
*pde = 00000000
Oops: 0000
CPU:    0
EIP:    0010:[<c0116a3b>]    Not tainted

Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010082
eax: c32b332c ebx: 00003615 ecx: 00000001 edx: 00000001
esi: c32b332c edi: 00000001 ebp: c2b37e9c esp: c2b37e80
ds: 0018 es: 0018 ss: 0018
Process ifconfig (pid: 897, stackpage=c2b37000)

Stack: 00000001 00000001 00000001 00000282 c3e39380 c556c560 00001002 00000000 
       c01f5295 c10f55d0 c4d00800 c5546bc0 c3e39380 c01f4828 c3e39380 c4a639c0 
       c01f5982 c4a639c0 00000000 00000030 00000009 c601d369 c4a639c0 c556c400 
Call Trace:    [<c01f5295>] [<c01f4828>] [<c01f5982>] [<c601d369>] [<c01f96f5>]
  [<c01fa879>] [<c01f91df>] [<c022b340>] [<c01f2500>] [<c01f2526>] [<c0146380>]
  [<c010734f>]
Code: 8b 13 0f 0d 02 39 c3 74 16 8b 4b fc 8b 01 85 c7 75 19 8b 02 

Trace; c01f5295 <sock_def_write_space+75/80>
Trace; c01f4828 <sock_wfree+48/50>
Trace; c01f5982 <__kfree_skb+42/150>
Trace; c601d369 <[sis900]sis900_close+99/c0>
Trace; c01f96f5 <dev_close+c5/d0>
Trace; c01fa879 <dev_change_flags+129/140>
Trace; c01f91df <dev_get+f/20>
Trace; c022b340 <devinet_ioctl+290/610>
Trace; c01f2500 <sock_ioctl+0/30>
Trace; c01f2526 <sock_ioctl+26/30>
Trace; c0146380 <sys_ioctl+b0/1b0>
Trace; c010734f <system_call+33/38>

Code; c0116a3b <__wake_up+1b/70>
00000000 <_EIP>:
Code; c0116a3b <__wake_up+1b/70> <=====
0: 8b 13 mov (%ebx),%edx <=====
Code; c0116a3d <__wake_up+1d/70>
2: 0f 0d 02 prefetch (%edx)
Code; c0116a40 <__wake_up+20/70>
5: 39 c3 cmp %eax,%ebx
Code; c0116a42 <__wake_up+22/70>
7: 74 16 je 1f <_EIP+0x1f>
Code; c0116a44 <__wake_up+24/70>
9: 8b 4b fc mov 0xfffffffc(%ebx),%ecx
Code; c0116a47 <__wake_up+27/70>
c: 8b 01 mov (%ecx),%eax
Code; c0116a49 <__wake_up+29/70>
e: 85 c7 test %eax,%edi
Code; c0116a4b <__wake_up+2b/70>
10: 75 19 jne 2b <_EIP+0x2b>
Code; c0116a4d <__wake_up+2d/70>
12: 8b 02 mov (%edx),%eax

</OOPS>

PS. There is one strange thing - I've got identical motherboard (same model,
same BIOS) running now the very same kernel (network configuration is different
though: eth0 has no ipx address) and everything seems to be ok there.

Additional information:

config:
------
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.21-rc1/oops2/config

lspci:
-----
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.21-rc1/oops2/lspci
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.21-rc1/oops2/lspci_v
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.21-rc1/oops2/lspci_vv

dmesg:
-----
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.21-rc1/oops2/dmesg

cpuinfo:
-------
http://zeus.polsl.gliwice.pl/~jfk/kernel/2.4.21-rc1/oops2/cpuinfo

gcc: 3.2.3
modutils: 2.4.21
ifconfig: 1.42 (2001-04-13)
binutils: 2.13.90.0.18

Any suggestions are welcome.

Jacek Kawa  **Nobody expects the Spanish Inquisition! Our chief weapon
            is suprise... surprise and fear... fear and surprise...**