I'm running a clean Slack 11 install. I run kernel 2.6.18 that was rebuilt to include SMP. My PC has two integrated ethernet cards, Vitesse VSC8201 Gigabit Phy and Marvell 88E8001 Gigabit. They both work under Windows and Linux. My PC connects to a Linksys WRT-54g running OpenWRT which acts as the router/switch/firewall/nat/ap.

The problem: I use rdesktop to connect to a terminal server at work. It runs Windows 2003 Terminal Server. I try to connect "rdesktop 24.123.xxx.xxx" and only part of the login box will show up. I tried passing many different options to rdesktop, but nothing would work. Eventually, rdesktop exits with the response "ERROR: recv: Connection reset by peer"

Later, I tried to access my webmail. It resides behind the same IP, but a different server that has Windows 2003 and Exchange 6.5 handles the connection. I have no trouble logging in to webmail but the connection gets "stalled" and only the first or second message shows up. Mozilla says "transferring data from ..." at the bottom of the browser window, but that is as far as it gets.

I have these problems using Slackware but not Windows, so the problem seems to be with my configuration or a strange bug. The strangest thing is that I can use rdesktop to access a different 2003 server and also a XP machine on my local LAN. It seems that this problem relates to a certain IP. Also, I do not have a firewall set up on my Slack box.

I included some links to screenshots, so you can see where it gets hung up.

snapshot1.png
snapshot2.png
snapshot3.png

Please help. I can provide an wireshark dump if anyone would like to see it.

Thanks!

granth

Well, I would ask first is, what has changed? Did this process work before you installed slack 11? I am guessing you were using 10.2 then? If so, what version of rdesktop did you use with 10.2? Was there a driver version change from old distro to slack 11? Without knowing the issues of latest revisions, those are the questions I would ask first.

Are you using a VPN connection to reach those remote servers? If so, you might be suffering from packet fragmentation due to the fact that the network packets get bigger because of the overhead of the (ipsec) header.
Try lowering the MTU for your interface from the default 1500 to for instance 1492 or even 1360.
Look at the example MTU setting in /etc/rc.d/rc.inet1.conf (search the file for the string MTU).

If it's not a VPN you're using then you still might want to try lowering the MTU since the possibility exists of an intermediate server that fragments your packets.

Eric

I previously had slack-current, kernel 2.6.16.16, rdesktop 1.4.1.

The drivers I previously used were SKGE and FORCEDETH. I used the SKGE interface most of the time.

Now I am using FORCEDETH. SKGE and SK98LIN modules are also loaded. They are two separate drivers for the same piece of hardware. SKGE is supposed to be more stable but I do not know which one it is using. Either way, I'm on the FORCEDETH interface and don't have problems accessing anything else.

I am connecting through a Linksys RV082 VPN Router, but I am not connecting to the VPN. It is a NAT device with port forwarding. I changed the MTU from 1492 to 1500 and back to 1492 again. It also has a working VPN connection to another location.

My WRT-54g has MTU's of 1500 on all of the interfaces. I changed the MTU on my Slack box to 1492, 1360, and back to 1500 (ifconfig eth0 mtu 1500). Is a reboot required?





granth

hmmm. It may not be related, but have you tried using SKGE this time around again?

I wonder if it's possible that two driver modules can conflict? hmm.

As far as if a reboot is required, there could be a daemon that you have to restart, unless I know the daemon, I just do a full reboot. AlienBob or someone else may know a more definitive anwser :-)

I hope I threw around some options for ya. Hope you get 'er up and runnin.

I tried some live linux cd's.

----------
Mandriva one-2007 mona kde-west
----------
kernel 2.6.17-smp
skge
forcedeth

I had the exact same problem using the Mandriva live cd.

----------
SLAX KillBill Edition v 5.1.8
----------
kernel 2.6.16
skge
sk98lin
forcedeth

I did NOT have the problem using this live cd.


---------
Ubuntu 6.10 i386
---------
Could not test due to a bug in the new 'nv' Xorg drivers.


So far, it seems that something in kernel 2.6.17 and up is causing the problem. I dont think it is the ethernet drivers because I've tested both interfaces with the same results. Perhaps it is something with iptables? Anybody have any ideas?


granth

I told ya it was strange.

Any more recommendations?

I upgraded to kernel 2.6.18.2 and I'm still having the same problem. I'm going to build a stripped down kernel to see if I can change anything.

Maybe I should try stepping back to kernel 2.6.16...

From a previous post, it works with sk98lin, but not with skge. rmmod the skge, and see if that helps....

sk98lin driver doesnt seems to work with my chip.

anyway, i have the problem on eth1 using forcedeth and on eth0 using skge.

I'm also having the same problem using my laptop. It has Fedora Core 6 installed, kernel 2.6.18 smp, and sis900 + mii drivers.

Has something gone haywire in the 2.6.17+ kernels?

Hi,

Your problem looks a bit like mine:
Network speed changes dramatically when using 2.6.18.1/2
http://www.linuxquestions.org/questi...d.php?t=500145

After a certain kernel level, without changing anything, there is something in the WAN that slows down or disrupts the connection.

Have you made a network trace to see if some packages get lost?

A possibility is that the higher kernel level use different port for the sending of packages and that the packages from a higher port get 'mangled' or something. But I am not an expert on that.

It is more than just a slow down. In fact, using wget, I tried to download a few files from the problem domain, and I get about 15,5xx bytes and then the connection hangs. It doesn't fail, it simply stops receiving data.

I want to stress that I only have this problem connecting to a specific gateway. I have problems with the web site, remote desktop, and the email server. They are different hosts behind a single gateway device.

I have a BSD server set up at the same location. It is on the same internet connection but has a different real-world IP because it is not behind the gateway device.

This tells me that there is some irregularity with the Linux 2.6.17+ kernel network stack and the Linksys RV082 gateway device.

To make this even more confusing, I can get on a workstation inside the problem domain, ssh into my Slackware box, and even get files from it using scp.

The only time I have trouble, is when my Slackware box initiates the connection.

SOLVED

Wow; I finally found the answer to my problem. It seems the problem is with the Linksys RV082 TCP stack.

http://forums.linksys.com/linksys/bo...essage.id=2396

Apparently, the RV082 does not support TCP scaling. This would cause problems with Windows Vista and Linux kernel 2.6.17+ with enough memory.

You can issue this command to check to check if you are affected:

$ cat /proc/sys/net/ipv4/tcp_wmem
4096 16384 4194304

If the results look like the above, you can issue the following command to turn TCP window scaling off:

echo 0 > /proc/sys/net/ipv4/tcp_window_scaling



This fixed my problem!!! Damn Linksys for not following TCP specs. A fix is rumored to be included with the next firmware release. I am running 1.3.3.5, which is the current release.

Here is a link where you can download the BETA 1.3.3.6 firmware.
http://www.linksysinfo.org/portal/fo...?do=file&id=35

I will try this firmware later to see if it fixes the issue.

Hi I think i have similiar problem with slackware 11 in my notebook
But the solution didn't work for me

When i try to copy large file from my notebook to another notebook/pc
then the network stalled and then i cant connect to the network anymore until restart it
But its always success if I copy file from other pc to the notebook no matter how large the file

The notebook is toshiba satellite A80
network card detected as Marvell Yukon, and hotplug load SKY2 as the driver


Thanks

arifhn