Showing posts with label two network cards. Show all posts
Showing posts with label two network cards. Show all posts

16 September 2012

237. Briefly: Packet corrupt during ssh sessions

Whenever I'm using two subnets for my cluster I seems to be having problems with:
Corrupted MAC on input.
Disconnecting: Packet corrupt
It particularly happens when there's a lot of information being passed to the screen. It's a right killer when you're compiling on a remote system. However, while I've been able to get around that by running a GNU Screen session on the remote box it was time to solve it.\


I googled and found:
https://bugs.launchpad.net/ubuntu/+source/linux-source-2.6.17/+bug/60764


The subnet I'm having issues with is operation across a switch. One of the computers on the network is defined as the gateway and I tend to have problems when connecting from it to the other computers on the network.
The gateway server has two interfaces, eth0 which is connected to a router which is connected to the outside world, and eth1 which is connected to the local subnet I'm having problems with.

The fix has been as as simple as
sudo apt-get install ethtool
sudo ethtool -K eth1 rx off tx off

I only had to run this on the gateway box and so far I've had no issues. Depending on how you're managing your network interfaces (i.e. wicd, network-manager, /etc/networking/*) you may want to add it to the post-up section of your /etc/network/interfaces:
post-up ethtool -K eth1 rx off tx off

Links to this post:
http://winscp.net/forum/viewtopic.php?t=12469

09 February 2012

63. Iptables for LAN with one internet connected gateway; sharing internet connection using iptables

Here I show how to share an internet connection with clients on a LAN. It's based in part on the iptables which firestarter generates when setting up connection sharing -- I think one could probably get away with dropping the INBOUND/OUTBOUND sections for the gateway server.

You will probably find that you need to open more ports, depending on your network services. Hopefully it's obvious from the instructions below how to do that. As always, use what you find below as a starting point and expand and correct it as you fool around with it.

While it's easier to use a gui like gufw or firestarter (see previous post), it's easier to get an absolute overview of your firewall configuration if you define each rule using iptables. It's also not that difficult and with a bit of trial and error you can work it out.

The usual caveats apply -- a good 2/3 of my posts are written as I'm teaching myself, while the remainder describe easy, useful, but not always obvious, things and programmes which makes life easier. This lands in the former category.

--- START HERE ---

My network:
One computer has two cards. eth0 is connected to the outside world, eth1 is connected to a switch making up a LAN. Each client is connected to the switch and has static IP (set in /etc/network/interfaces)

The clients are the easiest, so we'll start with them

Client:
create /etc/firewall-rules.sh (e.g. sudo vim /etc/firewall-rules.sh) and put the following in it:

sudo iptables -F #FLUSH

#INPUT
sudo iptables -A INPUT -m conntrack --ctstate ESTABLISHED,RELATED -j ACCEPT #network access
sudo iptables -A INPUT -i lo -j ACCEPT                        #127.0.0.1
sudo iptables -A INPUT -p tcp --dport ssh -j ACCEPT -s 192.168.1.0/24 #ssh
sudo iptables -A INPUT -p tcp --dport www -j ACCEPT -s 192.168.1.0/24 #web server
sudo iptables -A INPUT -p tcp --dport nfs -j ACCEPT -s 192.168.1.0/24 #needed for nfs
sudo iptables -A INPUT -p udp --dport nfs -j ACCEPT -s 192.168.1.0/24 #needed for nfs
sudo iptables -A INPUT -p tcp --dport sunrpc -j ACCEPT -s 192.168.1.0/24 #needed for nfs 
sudo iptables -A INPUT -p udp --dport sunrpc -j ACCEPT -s 192.168.1.0/24 #needed for nfs 
sudo iptables -A INPUT -p icmp --icmp-type 8 -j ACCEPT -s 192.168.1.0/24 #ping
sudo iptables -A INPUT -p udp --dport 60003 -j ACCEPT -s 192.168.1.0/24 #sinfo/d
sudo iptables -A INPUT -m limit --limit 15/minute -j LOG --log-level 7 --log-prefix " Dropped by firewall "
sudo iptables -A INPUT -j DROP                          #drop all else

#OUTPUT
sudo iptables -A OUTPUT -o lo -j ACCEPT #127.0.0.1
sudo iptables -A OUTPUT -j ACCEPT          #all outgoing ok

#FORWARD
sudo iptables -A FORWARD -p icmp --icmp-type 8 -j ACCEPT

#Default behaviour
sudo iptables -P INPUT DROP
sudo iptables -P OUTPUT DROP 
sudo iptables -P FORWARD DROP
Next, change ownership and permission

sudo chown root firewall-rules.sh
sudo chmod 700 firewall-rules.sh

Finally, edit /etc/network/interfaces and put
post-up sh /etc/firewall-rules.sh
as the last line. If you use post-up routing rules as well you can put those before or after.

Done!


The Gateway:
We need to allow the local network access to the services of the gateway, such as apt-cache. We also need to pass through traffic to the outside world.

Here's the gateway's /etc/firewall-rules.sh:

sudo iptables -F #FLUSH
# T1 -> eth0 --> inet, eth1 --> LAN (192.168.0/24)

#table nat
sudo iptables -t nat -P PREROUTING ACCEPT
sudo iptables -t nat -P INPUT ACCEPT
sudo iptables -t nat -P OUTPUT ACCEPT
sudo iptables -t nat -P POSTROUTING ACCEPT
sudo iptables -t nat -A POSTROUTING -o eth0 -j MASQUERADE

#table mangle
sudo iptables -t mangle -P PREROUTING ACCEPT
sudo iptables -t mangle -P INPUT ACCEPT
sudo iptables -t mangle -P FORWARD ACCEPT
sudo iptables -t mangle -P OUTPUT ACCEPT
sudo iptables -t mangle -P POSTROUTING ACCEPT 

#main table
sudo iptables -N OUTBOUND
sudo iptables -N INBOUND

#INPUT
sudo iptables -A INPUT -i eth0 -m conntrack --ctstate ESTABLISHED,RELATED -j ACCEPT #allows network access
sudo iptables -A INPUT -i lo -j ACCEPT                                                #127.0.0.1
sudo iptables -A INPUT -i eth1 -p tcp --dport ssh -j ACCEPT -s 192.168.1.0/24 #ssh
sudo iptables -A INPUT -i eth1 -p tcp --dport www -j ACCEPT -s 192.168.1.0/24 #web server
sudo iptables -A INPUT -i eth1 -p tcp --dport nfs -j ACCEPT -s 192.168.1.0/24 #needed for nfs
sudo iptables -A INPUT -i eth1 -p udp --dport nfs -j ACCEPT -s 192.168.1.0/24 #needed for nfs
sudo iptables -A INPUT -i eth1 -p tcp --dport sunrpc -j ACCEPT -s 192.168.1.0/24  #needed for nfs 
sudo iptables -A INPUT -i eth1 -p udp --dport sunrpc -j ACCEPT -s 192.168.1.0/24 #needed for nfs 
sudo iptables -A INPUT -i eth1 -p icmp --icmp-type 8 -j ACCEPT -s 192.168.1.0/24  #ping
sudo iptables -A INPUT -i eth1 -p udp --dport 60003 -j ACCEPT -s 192.168.1.0/24 #sinfo/d
sudo iptables -A INPUT -i eth1 -p tcp --dport 3142 -j ACCEPT -s 192.168.1.0/24             #apt-cache
sudo iptables -A INPUT -i eth1 -d 192.168.1.1 -j INBOUND                                                   #needed for gw -> clients
sudo iptables -A INPUT -i eth0 -s 192.168.1.0/24 -j REJECT 
sudo iptables -A INPUT -m limit --limit 15/minute -j LOG --log-level 7 --log-prefix " Dropped by firewall "
sudo iptables -A INPUT -j DROP                                                       #drop all else


#OUTPUT
sudo iptables -A OUTPUT -o lo -j ACCEPT #localhost 127.0.0.1
sudo iptables -A OUTPUT -o eth0 -j ACCEPT #eth0: all outgoing ok
sudo iptables -A OUTPUT -o eth1 -j ACCEPT                               #eth1: all outgoing ok

#FORWARD
sudo iptables -A FORWARD -p icmp -j ACCEPT 
sudo iptables -A FORWARD -p tcp -s 192.168.1.0/24 -j ACCEPT     #forward everything from local LAN
sudo iptables -A FORWARD -p udp -s 192.168.1.0/24 -j ACCEPT    #forward everything from local LAN
sudo iptables -A FORWARD -i eth0 -j OUTBOUND                           #need both for pass-through
sudo iptables -A FORWARD -i eth1 -j OUTBOUND                           #need both for pass-through


#INBOUND
sudo iptables -A INBOUND -j ACCEPT -m state --state RELATED,ESTABLISHED                              
sudo iptables -A INBOUND -s beryllium -j ACCEPT
sudo iptables -A INBOUND -j ACCEPT -s 192.168.1.0/24

#OUTBOUND
sudo iptables -A OUTBOUND -j ACCEPT

#Default behaviour
sudo iptables -P INPUT DROP
sudo iptables -P OUTPUT DROP 
sudo iptables -P FORWARD DROP

And that's about it.

To check that it loaded do

sudo iptables -L -n -v

The -n is because of this.

Keep on checking what goes into /var/log/firewall.log to see whether you should open more ports or use a more generous (or strict) firewall policy.



Edit: the following was the old way of doing it. The downside is that
1. it gets loaded very late in the boot sequence
2. it doesn't reload on sudo service networking restart

I've migrated away from network-manager, but it might require the method below. Use if the first method doesn't load the firewall rules.

edit /etc/rc.local and put 
sh /etc/firewall-rules.sh
as the second-to-last line to make the rules be added on each boot.

Remember the sudo iptables -m limit --limit 15/minute -j LOG --log-level 7 --log-prefix " Dropped by firewall " line? It doesn't actually do anything yet.

Edit /etc/rsyslog.conf and put
kern.=debug /var/log/firewall.log
anywhere. Restart the service:

sudo service rsyslog restart

There's now a firewall.log in your  /var/log dir.



There is one caveat:

IMPORTANT: for some reason receiving large files via sftp in filezilla FROM a client to the gateway gives

Error: Incorrect MAC received on packet
Error: File transfer failed after transferring 32,768 bytes in 1 second
or
Error: Server sent disconnect message
Error: type 2 (protocol error):
Error: "Packet corrupt"
Error: File transfer failed

Transferring large files TO a client works fine from the gateway and is blazingly fast. Transferring files between clients also works fast and securely.

i.e. on a client I can easily receive files from the gateway. On the gateway I can easily put a file on a client. The opposite directions don't work, whether I do it on the client or on the gateway. It seems like there should be an obvious iptables fix. My network cards are rtl-8169 gigabit pci cards and/or intel e1000 pro

NFS works fine for filetransfer (see this post) but I'm working on figuring out the incorrect MAC problem.

I've already tried with
sudo iptables -t mangle -A FORWARD -p tcp --tcp-flags SYN,RST SYN -j TCPMSS --clamp-mss-to-pmtu

Also, even a firewall consisting of nothing but (apart from flush):

sudo iptables -P INPUT ACCEPTsudo iptables -P OUTPUT ACCEPTsudo iptables -P FORWARD ACCEPT

doesn't solve it

Links to this post:
http://www.debian-srbija.iz.rs/p/kako-da.html