Network problems with linux kernel -2.6.8

Discussion topics, Linux related - not requests for help

Moderators: ChriThor, LXF moderators

Network problems with linux kernel -2.6.8

Postby dejijones » Thu May 26, 2005 8:15 am

I have a strange problems with kernels 2.6.8 and above. I am running Fedora core 2 on my system which is on a network. I found out that once I boot with my system with any kernel above 2.6.8, I can't successfully browse sites outside my network. The browsers returns request to url with half filled pages and at times with no result at all, no error or the message "document contains no data". My ftp's also gets time out, the odd thing is that everything works fine withing my network. And before you start blaming my firewall, I'll like you to know I have no problem whenever I boot with the 2.6.5 kernel. Can anybody tell me whats changed between these kernels, networkwise ?. I have tried all the recent kernels after the 2.6.8 and they all have the same problem. Thanks for any insight to resolving this problem.

Posts: 2
Joined: Thu May 26, 2005 7:25 am

RE: Network problems with linux kernel -2.6.8

Postby towy71 » Thu May 26, 2005 11:33 am

dick@debianbox:~$ uname -a
Linux debianbox 2.6.8-2-k7 #1 Mon Jan 24 03:29:52 EST 2005 i686 GNU/Linux

my kernel seems fine :-)
still looking for that door into summer
Posts: 4317
Joined: Wed Apr 06, 2005 2:11 pm
Location: wild West Wales

RE: Network problems with linux kernel -2.6.8

Postby nelz » Thu May 26, 2005 12:17 pm

Is your firewall running on the same computer? If so, you should try turning it off. The firewall will use iptables, which is part of the kernel, changing the kernel may have broken it.
"Insanity: doing the same thing over and over again and expecting different results." (Albert Einstein)
User avatar
Site admin
Posts: 9008
Joined: Mon Apr 04, 2005 11:52 am
Location: Warrington, UK

RE: Network problems with linux kernel -2.6.8

Postby jjmac » Fri May 27, 2005 1:51 am

When you say yourv'e tryed all the kernels after 2.6.8 ..., in what way ?

Source compiles of mainline kernels, and/or with patch sets applied, if so ... which sets. All the distros will use different patches. There may be some FC patch set issue happening that hasn't been satisfied with the later kernels.

I would think you would need to find out just what patches FC is using in their default kernel, then decide on whether you need to follow that line.

I don't think there has been much actual change in the underlying networking support though. Most of the changes have been with regard to memory management and schedualing issues. And just plain code cleanups/bug fixs.

If you were just compiling with your existing saved '.config', that is, not having gone through a reconfig, jumping off with the earlier one ... you may have accidently turned something off, involving whats implemented as a module, and whats built in etc. Possibly that, and the patch issue will fix it.


Code: Select all

          -|-   If the system is the answer, then the question
                            must have been really stupid               -|-

The FVWM wm -=- -=-

Somebody stole my air guitar, It happened just the other day,
But it's ok, 'cause i've got a spare ...
LXF regular
Posts: 1999
Joined: Fri Apr 08, 2005 1:32 am
Location: Sydney, Australia

Postby samb » Fri May 27, 2005 5:48 am

Given the problem description (connection works fine within the LAN, no WAN connectivity) I'd say it was most likely a routing issue, although I don't see how a kernel upgrade could be directly responsible for that.

What could be happening is that you have multiple network interfaces, and the route is benig set with a specific interface name. If, for example, the 2.6.8 kernel included support for IEEE1394 (Firewire) networking, and the 2.6.5 kernel didn't, then you could have the Firewire 'network card' on eth0, and find your ethernet card bumped back to eth1. That could create exactly the problem you're seeing.

Can you paste the output of
Code: Select all
ip route
Code: Select all
ip link
under both 2.6.5 and 2.6.8?
Bashton Ltd Linux Consultancy
Posts: 4
Joined: Thu May 26, 2005 9:32 pm

Postby dejijones » Mon May 30, 2005 2:17 am

Thanks for your responce so far, I need to clarify a few things. Answering Nelz question, My firewall is not runnig on the same computer. Its running off a openbsd box.

Where in answering the question about my kernel, I am using the precompiled kernel from Fedora, and I have been trying all kernel versions released since 2.6.5 both on Fedora core 2 and 3. Looks like I might have to recompile the kernel myself to find out what has changed from the default versions.

The output from my kernel 2.6.5, and 2.6.9 for ip route and link is;
# ip route dev eth0 proto kernel scope link src dev eth0 scope link
default via dev eth0

# ip link
1: lo: <LOOPBACK,UP> mtu 16436 qdisc noqueue
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
3: eth0: <BROADCAST,MULTICAST,UP> mtu 1500 qdisc pfifo_fast qlen 1000
link/ether 00:40:f4:42:03:71 brd ff:ff:ff:ff:ff:ff

Both kernel's give the same output for ip route and ip link.

Thanks for your help.
Posts: 2
Joined: Thu May 26, 2005 7:25 am

Postby samb » Mon May 30, 2005 7:34 am

Are you able to ping other locations, but not access web sites and the like?

Can you paste a traceroute to, say under both kernels?

Also, is TCP ECN disabled? Some firewalls won't allow ECN packets through.
If you do
Code: Select all
cat /proc/sys/net/ipv4/tcp_ecn
it should read 0. If it returns 1, you can do
Code: Select all
echo 0 > /proc/sys/net/ipv4/tcp_ecn

Note that this won't be preserved across reboots - you need to edit /etc/sysctl.conf for that, we'll cross that bridge if we come to it.
Bashton Ltd Linux Consultancy
Posts: 4
Joined: Thu May 26, 2005 9:32 pm

Return to Discussion

Who is online

Users browsing this forum: No registered users and 1 guest