connlimit patch crashes 2.6.11 kernel

connlimit patch crashes 2.6.11 kernel

[Mateusz]

Hi

I experience the problem with connlimit patch included in
patch-o-matic-ng
snapshot from 2005.05.04 and earlier.

I have tested it with 2.6.11.7 kernel and iptables-1.3.1, when on
NATed machine
a go to any website or i send any packet from this machine to
internet via my
server - kernel crashes immediately. When any machines don't send
any
packet (I have also tested it with only my machine connect to
switch with server)
kernel is working fine.

Problem does not occur when I don't apply connlimit patch to
kernel.

I rewrite oops dump here.


[<c03b42aa>] ipt_do_table+0x1fa/0x3b0
[<c037b740>] ip_forward_finish+0x0/0x80
[<c03b7b37>] ipt_hook+0x37/0x40
[<c0362651>] nf_iterate+0x37/0x40
[<c037b740>] ip_forward_finish+0x0/0x80
[<c037b740>] ip_forward_finish+0x0/0x80
[<c0362a55>] nf_hook_slow+0x75/0x120
[<c037b740>] ip_forward_finish+0x0/0x80
[<c037b6ee>] ip_forward+0x1ee/0x240
[<c037b740>] ip_forward_finish+0x0/0x80
[<c037a356>] ip_rcv_finish+0x126/0x290
[<c037a230>] ip_rcv_finish+0x0/0x290
[<c0362d1f>] nf_reinject+0x21f/0x260
[<c037a230>] ip_rcv_finish+0x0/0x290
[<c031869b>] imq_dev_xmit+0x4b/0x60
[<c0363c4b>] qdisc_restart+0x8b/0x200
[<c0318777>] imq_nf_queue+0xc7/0x1a0
[<c03628d3>] nf_queue+0x173/0x280
[<c037a230>] ip_rcv_finish+0x0/0x290
[<c037a230>] ip_rcv_finish+0x0/0x290
[<c037a230>] ip_rcv_finish+0x0/0x290
[<c0362a8c>] nf_hook_slow+0xac/0x120
[<c037a230>] ip_rcv_finish+0x0/0x290
[<c037a007>] ip_rcv+0x427/0x4f0
[<c037a230>] ip_rcv_finish+0x0/0x290
[<c0135d9a>] do_gettimeofday+0x1a/0xb0
[<c0357446>] netif_receive_skb+0x126/0x1b0
[<c03509a7>] alloc_skb+0x47/0xe0
[<c031a1f2>] rtl8139_rx+0x1a2/0x360
[<c031a5a6>] rtl8139_poll+0x56/0x100
[<c035763a>] net_rx_action+0x6a/0xf0
[<c0149b3b>] __do_softirq+0x7b/0x90
[<c01337a1>] do_softirq+0x41/0x50
=======================
[<c0149c25>] irq_exit+0x35/0x40
[<c0133693>] do_IRQ+0x53/0x80
[<c012f5b0>] default_idle+0x0/0x30
[<c0131b5a>] common_interfupt+0x1a/0x20
[<c012f5b0>] default_idle+0x0/0x30
[<c012f5d4>] default_idle+0x24/0x30
[<c012f66c>] cpu_idle+0x4c/0x60
[<c010f880>] start_kernel+0x150/0x170
[<c010f430>] unknown_bootoption+0x0/0x1e0
[<c012f1b1>] L6+0x0/0x2
Code: 43 17 89 d9 8d 04 40 c1 e0 03 29 c1 89 cd 81 ed a8 00 00 00
74 0d 80 7d 54 07 0f 44 54 24 1c 89 54 24 1c 85 db 0f 84 47 01 00
00 <80> 7d 54 07 0f 84 08 01 00 00 8b 54 24 0c 8b 44 24 4c 23 42
08
<0>Kernel panic - not syncing: Fatal exception in interrupt


I could answer at any question about environement.
Thanks in advance for any reply.

Best regards, Mateusz Pluta

PS: sorry for my poor english

Re: connlimit patch crashes 2.6.11 kernel

[Patrick McHardy]

Mateusz wrote:
> Problem does not occur when I don't apply connlimit patch to
> kernel.
> 
> [<c0318777>] imq_nf_queue+0xc7/0x1a0

Does the problem occur without the imq patch?

Regards
Patrick

RE: connlimit patch crashes 2.6.11 kernel

[Mateusz Pluta]

Patrick McHardy wrote:
> Mateusz wrote:
>> Problem does not occur when I don't apply connlimit patch to
kernel.
>> 
>> [<c0318777>] imq_nf_queue+0xc7/0x1a0
> 
> Does the problem occur without the imq patch?
> 
> Regards
> Patrick

Yes, I use imq patch, and in crash time packets problably go
through 
the imq device, because I use it to traffic shaping.

One other person found my post, and report me the same problem.

Best regards, Mateusz Pluta

RE: connlimit patch crashes 2.6.11 kernel

[tomasz.maj]

Hello,

Yes, I've got the same problem as Mariusz, my kernel 2.6.11 crashes when I
use connlimit patch.

Greetings
Tomek


>> Mateusz wrote:
>>> Problem does not occur when I don't apply connlimit patch to
> kernel.
>>>
>>> [<c0318777>] imq_nf_queue+0xc7/0x1a0
>>
>> Does the problem occur without the imq patch?
>>
>> Regards
>> Patrick
>
> Yes, I use imq patch, and in crash time packets problably go
> through
> the imq device, because I use it to traffic shaping.
>
> One other person found my post, and report me the same problem.
>
> Best regards, Mateusz Pluta
>
>

RE: connlimit patch crashes 2.6.11 kernel

[Forte Systems - Iosif Peterfi]

The same problem here 2.6.11. But no imq this time. Connlimit from latest
snapshot. No problem at module insertion. Crashes when there are rules using
the connlimit match. 

This is my other modules:

sch_sfq 6016 6 - Live 0xf88bf000
ipt_tos 2432 1 - Live 0xf889d000
ipt_helper 3200 3 - Live 0xf8898000
ipt_connbytes 3072 18 - Live 0xf8896000
ipt_CLASSIFY 2944 209 - Live 0xf88bb000
ipt_length 2560 8 - Live 0xf88bd000
sch_esfq 6784 6 - Live 0xf889a000
sch_htb 17536 2 - Live 0xf88a6000
ipt_REDIRECT 3072 3 - Live 0xf888d000
iptable_mangle 3328 1 - Live 0xf881c000
iptable_nat 23868 2 ipt_REDIRECT, Live 0xf889f000
ipt_REJECT 6656 8 - Live 0xf8880000
ipt_limit 3072 29 - Live 0xf888b000
ipt_state 2560 4 - Live 0xf881e000
ip_conntrack 50280 3 ipt_helper,iptable_nat,ipt_state, Live 0xf88ad000
ipt_LOG 7680 15 - Live 0xf8893000
ipt_ULOG 8196 12 - Live 0xf888f000
iptable_filter 3328 1 - Live 0xf8879000
ip_tables 24960 14
ipt_tos,ipt_helper,ipt_connbytes,ipt_CLASSIFY,ipt_length,ipt_REDIRECT,iptabl
e_mangle,iptable_nat,ipt_REJECT,ipt_limit,ipt_state,ipt_LOG,ipt_ULOG,iptable
_filter, Live 0xf8883000
ipv6 246304 16 - Live 0xf88c3000
e100 34560 0 - Live 0xf886f000

The oops message is the same: Kernel panic - not syncing: Fatal exception in
interrupt.
My dump is a bit different but I dont know how to capture it.



-----Original Message-----
From: netfilter-devel-bounces@lists.netfilter.org
[mailto:netfilter-devel-bounces@lists.netfilter.org] On Behalf Of
tomasz.maj@teldom.pl
Sent: Wednesday, May 18, 2005 10:50 AM
To: Mateusz Pluta
Cc: netfilter-devel@lists.netfilter.org; 'Patrick McHardy'
Subject: RE: connlimit patch crashes 2.6.11 kernel

Hello,

Yes, I've got the same problem as Mariusz, my kernel 2.6.11 crashes when I
use connlimit patch.

Greetings
Tomek


>> Mateusz wrote:
>>> Problem does not occur when I don't apply connlimit patch to
> kernel.
>>>
>>> [<c0318777>] imq_nf_queue+0xc7/0x1a0
>>
>> Does the problem occur without the imq patch?
>>
>> Regards
>> Patrick
>
> Yes, I use imq patch, and in crash time packets problably go
> through
> the imq device, because I use it to traffic shaping.
>
> One other person found my post, and report me the same problem.
>
> Best regards, Mateusz Pluta
>
>





-- 
This message was scanned for spam and viruses by BitDefender.
For more information please visit http://linux.bitdefender.com/




-- 
This message was scanned for spam and viruses by BitDefender.
For more information please visit http://linux.bitdefender.com/