Bug 82109 – consistent oops in linux-2.4.24-openmosix-r10

• Status: RESOLVED WORKSFORME
• Product: Gentoo Linux

[Description Peter Bienstman (RETIRED) 2005-02-15 04:48:35 UTC]

One of the machines in our cluster crashes during boot, or very quickly afterwards, with the following oops:

kernel BUG in skbuff.c:96!
invalid operand:0000
CPU:0
EIP: 0010:[<c02da4bb>] Not tainted
EFLAGS: 00010086
eax: 00000028 ebx: 00000000 ecx: 00000001 edx: 00000000
esi: 00000006 edi: f75c0280 ebp: f75c0280 esp: c03efc94
ds: 0018 es: 0018 ss: 0018
Process swapper: pid 0, stackpage=c03ef00
..
Code: 0f ab 60 00 99 75 38 co 83 c4 14 c3 89 f6 8d bc 27 00 00 00
<0> kernil panic, Aie, killing interrupt handler
In interrupt handler: not syncing

linux-2.4.24-openmosix-r7 did work fine.


Reproducible: Always
Steps to Reproduce:
1.
2.
3.




# 
# openMosix 
# 
CONFIG_MOSIX=y 
# CONFIG_MOSIX_TOPOLOGY is not set 
CONFIG_MOSIX_SECUREPORTS=y 
CONFIG_MOSIX_DISCLOSURE=3 
# CONFIG_MOSIX_FS is not set 
# CONFIG_MOSIX_PIPE_EXCEPTIONS is not set 
# CONFIG_MOSIX_NO_OOM is not set 
# CONFIG_MOSIX_LOADLIMIT is not set

[Comment 1 Konstantin Arkhipov (RETIRED) 2005-02-15 04:52:13 UTC]

can you please provide output of ksymoops for this oops?

[Comment 2 Peter Bienstman (RETIRED) 2005-02-15 05:00:37 UTC]

If I run ksymoops, I get 

Reading Oops report from the terminal

Can you tell me how exactly I should use it? I just copied the oops report by hand, because I couldn't find it in the syslog.

[Comment 3 Konstantin Arkhipov (RETIRED) 2005-02-15 05:08:13 UTC]

you should save whole oops message in separate file, then - just `cat oops.file | ksymoops`
if you're able to reproduce oops, probably best way to catch whole oops message will be looking at `tail -f /var/log/messages` from remote thru ssh session

[Comment 4 Peter Bienstman (RETIRED) 2005-02-15 05:12:34 UTC]

If I paste my partial handcopied oops, I get:

>>EIP; c02da4bb <skb_over_panic+3b/50>   <=====

>>edi; f75c0280 <_end+37142db4/383e3b94>
>>ebp; f75c0280 <_end+37142db4/383e3b94>
>>esp; c03efc94 <init_task_union+1c94/2000>

Code;  c02da4bb <skb_over_panic+3b/50>
00000000 <_EIP>:
Code;  c02da4bb <skb_over_panic+3b/50>   <=====
   0:   0f ab 60 00               bts    %esp,0x0(%eax)   <=====
Code;  c02da4bf <skb_over_panic+3f/50>
   4:   99                        cltd
Code;  c02da4c0 <skb_over_panic+40/50>
   5:   75 38                     jne    3f <_EIP+0x3f>

Is this sufficient info? The oops doesn't show up in /var/log/messages. Should I change something to syslog-ng?

[Comment 5 Konstantin Arkhipov (RETIRED) 2005-02-15 05:22:55 UTC]

you can try it without files/openmosix-sources-af_unix.patch too.

[Comment 6 Konstantin Arkhipov (RETIRED) 2005-02-15 05:23:47 UTC]

no, your syslog is ok. that's normal behaviour under oops/panics.

[Comment 7 Peter Bienstman (RETIRED) 2005-02-15 06:25:53 UTC]

Without openmosix-sources-af_unix.patch I get the same behaviour...

I think I'll try a vanilla 2.4.26-1, because we can't really afford to reboot our production cluster all the time.

[Comment 8 Konstantin Arkhipov (RETIRED) 2005-04-15 10:12:40 UTC]

is it still actual?
any luck with 2.4.30?

[Comment 9 Konstantin Arkhipov (RETIRED) 2005-04-30 08:32:45 UTC]

still no feedback. closing it.

[Comment 10 Peter Bienstman (RETIRED) 2005-04-30 08:37:13 UTC]

Sorry, didn't see the request for feedback. We won't be trying 2.4.30: the vanilla openmosix 2.4.26-1 works fine for us, and since we're running a production cluster, we can't use it to test gentoo patches.