Skip to contentSkip to conversation

Packet duplication

DISCUSSION Tiaan Wessels 4 participants 18 messages Oct 19, 2007 – Oct 25, 2007 2 messages not in the archive

lore
Share & tools
lore
#0 lore raw

Hi,
Not sure there is any life on this list but in case someone picks me up on netfilter user SETI here goes:
How do I go about duplicating a UDP packet arriving at a machine. Essentially I want to have it go to its original recipient but to another new one also. -j ROUTE --tee seems not to be supported anymore.
Thanks

#1 lore raw

Tiaan Wessels írta:

Quoted 2 lines

Hi,
Not sure there is any life on this list but in case someone picks me up on netfilter user SETI here goes:

There is life !!! :D

Quoted 2 lines

How do I go about duplicating a UDP packet arriving at a machine. Essentially I want to have it go to its original recipient but to another new one also. -j ROUTE --tee seems not to be supported anymore.

False... It is supported but you need the patch-o-matic(-ng) stuff...

--tee Make a copy of the packet, and route that copy to the
given destination. For the original, uncopied packet, behave like a non-terminating tar-
get and continue traversing the rules. Not valid in
combination with --iif' or --continue'

Quoted 1 line

Thanks

Swifty

#2 lore raw
Quoted 6 lines · Gáspár Lajos

Tiaan Wessels írta:

Quoted 2 lines

Hi,
Not sure there is any life on this list but in case someone picks me up on netfilter user SETI here goes:

There is life !!! :D

Quoted 2 lines

How do I go about duplicating a UDP packet arriving at a machine. Essentially I want to have it go to its original recipient but to another new one also. -j ROUTE --tee seems not to be supported anymore.

Quoted 8 lines

False... It is supported but you need the patch-o-matic(-ng) stuff...

--tee Make a copy of the packet, and route that copy to the
given destination. For the original, uncopied packet, behave like a non-terminating tar-
get and continue traversing the rules. Not valid in
combination with --iif' or --continue'

Getting this to work for a novice like me seems to be impossible.
Doing a man on my FC5 system shows --tee to be there under the ROUTE extension and I quote from the man page 'iptables can use extended target modules: the following are included in the standard distribution' however using iptables results in

[root@nst2 ~]# /sbin/iptables -A PREROUTING -t mangle -p udp -d 192.168.3.77 --dport 9090 -j ROUTE --tee iptables v1.3.5: Unknown arg `--tee'

locate libipt_ROUTE.so yields nothing which makes me believe the man page was talking bollocks when claiming the extensions to be part of the 'standard distribution'

After further reading I also came to the conclusion patch-o-matic needed to be used to install the ROUTE module. going to netfilter extensions HOWTO I see I have to get the latest update from CVS like in so

cvs -d :pserver:[email protected]:/cvspublic login

but when I do this with password cvs I get

[root@nst2 ~]# cvs -d :pserver:[email protected]:/cvspublic login
Logging in to :pserver:[email protected]:2401/cvspublic
CVS password:
cvs [login aborted]: connect to pserver.netfilter.org(213.95.27.115):2401 failed: Connection refused

any ideas ?

#5 lore raw
In reply to a message not in the archive

Tiaan Wessels írta:

Quoted 14 lines
Quoted 13 lines · Gáspár Lajos

Tiaan Wessels írta:

Quoted 2 lines

Hi,
Not sure there is any life on this list but in case someone picks me up on netfilter user SETI here goes:

There is life !!! :D

Quoted 2 lines

How do I go about duplicating a UDP packet arriving at a machine. Essentially I want to have it go to its original recipient but to another new one also. -j ROUTE --tee seems not to be supported anymore.

False... It is supported but you need the patch-o-matic(-ng) stuff...

--tee Make a copy of the packet, and route that copy to the
given destination. For the original, uncopied packet, behave like a non-terminating tar-
get and continue traversing the rules. Not valid in
combination with --iif' or --continue'

Quoted 7 lines
Quoted 6 lines
Quoted 1 line

Thanks

Swifty

Getting this to work for a novice like me seems to be impossible.

It is not so hard... :D

Quoted 6 lines

Doing a man on my FC5 system shows --tee to be there under the ROUTE extension and I quote from the man page 'iptables can use extended target modules: the following are included in the standard distribution' however using iptables results in

[root@nst2 ~]# /sbin/iptables -A PREROUTING -t mangle -p udp -d 192.168.3.77 --dport 9090 -j ROUTE --tee iptables v1.3.5: Unknown arg `--tee'

Note that you may need the --oif option too with tee... (I woukd be glad If anyone could confirm!)

Quoted 2 lines

locate libipt_ROUTE.so yields nothing which makes me believe the man page was talking bollocks when claiming the extensions to be part of the 'standard distribution'

Bad manpage... :D

Quoted 5 lines

After further reading I also came to the conclusion patch-o-matic needed to be used to install the ROUTE module. going to netfilter extensions HOWTO I see I have to get the latest update from CVS like in so

cvs -d :pserver:[email protected]:/cvspublic login

Quoted 8 lines

but when I do this with password cvs I get

[root@nst2 ~]# cvs -d :pserver:[email protected]:/cvspublic login
Logging in to :pserver:[email protected]:2401/cvspublic
CVS password:
cvs [login aborted]: connect to pserver.netfilter.org(213.95.27.115):2401 failed: Connection refused

any ideas ?

Forget cvs... try svn...

Swifty

#6 lore raw
Quoted 41 lines · Gáspár Lajos

Tiaan Wessels írta:

Quoted 14 lines
Quoted 13 lines · Gáspár Lajos

Tiaan Wessels írta:

Quoted 2 lines

Hi,
Not sure there is any life on this list but in case someone picks me up on netfilter user SETI here goes:

There is life !!! :D

Quoted 2 lines

How do I go about duplicating a UDP packet arriving at a machine. Essentially I want to have it go to its original recipient but to another new one also. -j ROUTE --tee seems not to be supported anymore.

False... It is supported but you need the patch-o-matic(-ng) stuff...

--tee Make a copy of the packet, and route that copy to the
given destination. For the original, uncopied packet, behave like a non-terminating tar-
get and continue traversing the rules. Not valid in
combination with --iif' or --continue'

Quoted 7 lines
Quoted 6 lines
Quoted 1 line

Thanks

Swifty

Getting this to work for a novice like me seems to be impossible.

It is not so hard... :D

Quoted 6 lines

Doing a man on my FC5 system shows --tee to be there under the ROUTE extension and I quote from the man page 'iptables can use extended target modules: the following are included in the standard distribution' however using iptables results in

[root@nst2 ~]# /sbin/iptables -A PREROUTING -t mangle -p udp -d 192.168.3.77 --dport 9090 -j ROUTE --tee iptables v1.3.5: Unknown arg `--tee'

Note that you may need the --oif option too with tee... (I woukd be glad If anyone could confirm!)

Quoted 2 lines

locate libipt_ROUTE.so yields nothing which makes me believe the man page was talking bollocks when claiming the extensions to be part of the 'standard distribution'

Bad manpage... :D

Quoted 7 lines

After further reading I also came to the conclusion patch-o-matic needed to be used to install the ROUTE module. going to netfilter extensions HOWTO I see I have to get the latest update from CVS like in so

cvs -d :pserver:[email protected]:/cvspublic login

but when I do this with password cvs I get

Quoted 8 lines
Quoted 6 lines

[root@nst2 ~]# cvs -d :pserver:[email protected]:/cvspublic login
Logging in to :pserver:[email protected]:2401/cvspublic
CVS password:
cvs [login aborted]: connect to pserver.netfilter.org(213.95.27.115):2401 failed: Connection refused

any ideas ?

Forget cvs... try svn...

thanks,
i have located patch-o-matic snapshot for yesterday on netfilter ftp server. so i did a runme extra but it never asks me to apply the ROUTE/--tee patch. is this correct ? must I recompile my kernel in any case ?

#7 lore raw

Tiaan Wessels írta:

Quoted 5 lines

thanks,
i have located patch-o-matic snapshot for yesterday on netfilter ftp server. so i did a runme extra but it never asks me to apply the ROUTE/--tee patch. is this correct ? must I recompile my kernel in any case ?

Yes you have to supply the kernel source-path and compile your own kernel. While you're at it, also download the new iptables source and let POM-ng also patch that source. Compiling the kernel isn't that hard, iptables is easier, just supply the iptables source path and when the kernel is done, compile and install iptables. Remember to remove the existing iptables package from your distribution.

Perhaps you can find the kernel-config from the distro you're using to import when running "make menuconfig", so you automatically have all options selected that you currently have (unless you choose another kernel version than what you are using now). That way you only have to select the extra's you want.

Grts,
Rob

#8 lore raw
Quoted 16 lines · Rob Sterenborg

Tiaan Wessels írta:

Quoted 6 lines

thanks,
i have located patch-o-matic snapshot for yesterday on netfilter ftp server. so i did a runme extra but it never asks me to apply the ROUTE/--tee patch. is this correct ? must I recompile my kernel in any case ?

Yes you have to supply the kernel source-path and compile your own kernel. While you're at it, also download the new iptables source and let POM-ng also patch that source. Compiling the kernel isn't that hard, iptables is easier, just supply the iptables source path and when the kernel is done, compile and install iptables. Remember to remove the existing iptables package from your distribution.

Quoted 8 lines

Perhaps you can find the kernel-config from the distro you're using to import when running "make menuconfig", so you automatically have all options selected that you currently have (unless you choose another kernel version than what you are using now). That way you only have to select the extra's you want.

i patched the kernel sources with pom and compiled and installed and then compiled and installed iptables 1.3.7 but still i get

nst2:~ # /usr/local/sbin/iptables -A PREROUTE -t mangle -p udp --dport 9090 -j ROUTE
iptables v1.3.7: Couldn't load target `ROUTE':/usr/local/lib/iptables/libipt_ROUTE.so: cannot open shared object file: No such file or directory

#9 lore raw
Quoted 8 lines

i patched the kernel sources with pom and compiled and installed and then compiled and installed iptables 1.3.7 but still i get

nst2:~ # /usr/local/sbin/iptables -A PREROUTE -t mangle -p udp --dport 9090 -j ROUTE
iptables v1.3.7: Couldn't load target `ROUTE':/usr/local/lib/iptables/libipt_ROUTE.so: cannot open shared object file: No such file or directory

So, "locate libipt_ROUTE.so" doesn't return
"/usr/local/lib/iptables/libipt_ROUTE.so".

In that case I suppose there was an error when compiling iptables (perhaps also when compiling the kernel?) that caused libipt_ROUTE.so not to be compiled. When you ran "make", did you see any related errors? Anyway, I don't think I can help you there; I'm not a developer, maybe someone else can.

Grts,
Rob

#10 lore raw
Quoted 10 lines · Rob Sterenborg
Quoted 9 lines

i patched the kernel sources with pom and compiled and installed and then compiled and installed iptables 1.3.7 but still i get

nst2:~ # /usr/local/sbin/iptables -A PREROUTE -t mangle -p udp --dport 9090 -j ROUTE
iptables v1.3.7: Couldn't load target `ROUTE':/usr/local/lib/iptables/libipt_ROUTE.so: cannot open shared object file: No such file or directory

Quoted 10 lines

So, "locate libipt_ROUTE.so" doesn't return
"/usr/local/lib/iptables/libipt_ROUTE.so".

In that case I suppose there was an error when compiling iptables (perhaps also when compiling the kernel?) that caused libipt_ROUTE.so not to be compiled. When you ran "make", did you see any related errors? Anyway, I don't think I can help you there; I'm not a developer, maybe someone else can.

thats correct. locate finds nothing. in an attempt at hacking i added ROUTE to the variable PF_EXT_SLIB in extensions/Makefile and tried to make but then get

nst2:~/iptables-1.3.7 # make
Unable to resolve dependency on linux/netfilter_ipv4/ipt_ROUTE.h. Try 'make clean'.
Extensions found: IPv4:CLUSTERIP IPv4:connbytes IPv4:dccp IPv4:quota IPv4:recent IPv4:statistic IPv4:string IPv6:REJECT IPv6:ah IPv6:esp IPv6:frag IPv6:ipv6header IPv6:hbh IPv6:dst IPv6:rt IPv6:sctp

i could not see any problem in compiling either the kernel or iptables. what did strike me as odd was when i ran the ./runme script in pom it never asks me for the patch on routing so i answer N to all the questions. this right ???

#11 lore raw
Quoted 4 lines

i could not see any problem in compiling either the kernel or iptables. what did strike me as odd was when i ran the ./runme script in pom it never asks me for the patch on routing so i answer N to all the questions. this right ??? -

I suspect you also couldn't select the ROUTE target when configuring the kernel.

Try:
KERNEL_DIR=/path/to/kernel \
IPTABLES_DIR=/path/to/iptables \
./runme --download extra

Grts,
Rob

#12 lore raw
Quoted 6 lines · Rob Sterenborg
Quoted 5 lines

i could not see any problem in compiling either the kernel or iptables. what did strike me as odd was when i ran the ./runme script in pom it never asks me for the patch on routing so i answer N to all the questions. this right ??? -

Quoted 9 lines

I suspect you also couldn't select the ROUTE target when configuring the kernel.

Try:
KERNEL_DIR=/path/to/kernel \
IPTABLES_DIR=/path/to/iptables \
./runme --download extra

looks promising thanks. it said it download ROUTE extra so patched (never asked me to apply ROUTE patch though) and am now compiling kernel again so will let you know in a days time thanks for your philantropy

#13 lore raw
Quoted 4 lines

looks promising thanks. it said it download ROUTE extra so patched (never asked me to apply ROUTE patch though) and am now compiling kernel again so will let you know in a days time thanks for your philantropy

I tried to patch 2.6.23.1 using POM-ng but it didn't ask to patch for ROUTE as you found out. Maybe it's not supported on this kernel so it'll won't work on this kernel.

To see what happens, I patched the kernel manually (copied files, added stuff to Kconfig and Makefile), configured it for ROUTE (ipv4) and it seems to compile. I can't test if it will work, however.

Grts,
Rob

#14 lore raw
Quoted 6 lines · Rob Sterenborg
Quoted 5 lines

looks promising thanks. it said it download ROUTE extra so patched (never asked me to apply ROUTE patch though) and am now compiling kernel again so will let you know in a days time thanks for your philantropy

Quoted 9 lines

I tried to patch 2.6.23.1 using POM-ng but it didn't ask to patch for ROUTE as you found out. Maybe it's not supported on this kernel so it'll won't work on this kernel.

To see what happens, I patched the kernel manually (copied files, added stuff to Kconfig and Makefile), configured it for ROUTE (ipv4) and it seems to compile. I can't test if it will work, however.

i know this is much asked but any chance you could explain to me what you did ? (you can be brief)

#15 lore raw
Quoted 7 lines
Quoted 5 lines

To see what happens, I patched the kernel manually (copied files, added stuff to Kconfig and Makefile), configured it for ROUTE (ipv4) and it seems to compile. I can't test if it will work, however.

i know this is much asked but any chance you could explain to me what you did ? (you can be brief) -

The patch is not hard to include manually. If you look in POM-ng's patchlets directory, you see all available patches including the ROUTE patch. There, you go to the source you need (I guess linux-2.6). You'll see 2 directories: include and net. Copy these into the kernel source using: cp -R * /path/to/kernel.
Next, you go to the /path/to/kernel/net/ipv4/netfilter directory. You'll find these files:
- Kconfig
- Kconfig.ladd
- Makefile
- Makefile.ladd
Checkout the .ladd files (there's not much in there) and add it somewhere to Kconfig or Makefile (I think if you look into these files you'll probably know where). You can delete the .ladd files afterwards. Do the same in /path/to/kernel/net/ipv6/netfilter.

Now you can configure the kernel for ROUTE and compile it. According to Kconfig.ladd, the ipv6 ROUTE does not work as a module so don't do configure it like that for ipv6; ipv4 seems fine as a module.

Again: I got it to compile but that's as much as I can do here. I do not know if it works.

Grts,
Rob

#16 lore raw
Quoted 26 lines · Rob Sterenborg
Quoted 9 lines
Quoted 6 lines

To see what happens, I patched the kernel manually (copied files, added stuff to Kconfig and Makefile), configured it for ROUTE (ipv4) and it seems to compile. I can't test if it will work, however.

i know this is much asked but any chance you could explain to me what you did ? (you can be brief) -

The patch is not hard to include manually. If you look in POM-ng's patchlets directory, you see all available patches including the ROUTE patch. There, you go to the source you need (I guess linux-2.6). You'll see 2 directories: include and net. Copy these into the kernel source using: cp -R * /path/to/kernel.
Next, you go to the /path/to/kernel/net/ipv4/netfilter directory. You'll find these files:
- Kconfig
- Kconfig.ladd
- Makefile
- Makefile.ladd
Checkout the .ladd files (there's not much in there) and add it somewhere to Kconfig or Makefile (I think if you look into these files you'll probably know where). You can delete the .ladd files afterwards. Do the same in /path/to/kernel/net/ipv6/netfilter.

Quoted 9 lines

Now you can configure the kernel for ROUTE and compile it. According to Kconfig.ladd, the ipv6 ROUTE does not work as a module so don't do configure it like that for ipv6; ipv4 seems fine as a module.

Again: I got it to compile but that's as much as I can do here. I do not know if it works.

thanks a million Rob,
I got it working at last using your instructions.
cheers

#17 lore raw
Quoted 1 line

thanks a million Rob,

You're welcome..

Quoted 2 lines

I got it working at last using your instructions.
cheers

Does this mean that the ROUTE patch is actually working (I didn't get to test that) and which kernel version did you use?

Grts,
Rob

#18 lore raw
Quoted 6 lines · Rob Sterenborg
Quoted 2 lines

thanks a million Rob,

You're welcome..

Quoted 8 lines
Quoted 3 lines

I got it working at last using your instructions.
cheers

Does this mean that the ROUTE patch is actually working (I didn't get to test that) and which kernel version did you use?

not exactly, it only meant all error messages was gone printed to the tty when i run the iptables command. i'm now trying to create what i originally set out to do. if you'r interested here's my scenario.

i have one machine with two interfaces eth0 192.168.51.151 and eth1 192.168.3.11. i have udp packets coming in on eth1 from 192.168.3.77 destined for 192.168.3.11:9090 . i want to duplicate these packets so they go to their original destination but also go out on eth0 to 192.168.51.46:9090. that's in short what i'm trying. so i have the following rules i'm trying now (hope this is not laughable as i'm a novice)

/usr/local/sbin/iptables -F INPUT
/usr/local/sbin/iptables -F OUTPUT
/usr/local/sbin/iptables -F FORWARD
/usr/local/sbin/iptables -t nat -F POSTROUTING
/usr/local/sbin/iptables -t nat -F PREROUTING
/usr/local/sbin/iptables -t mangle -F POSTROUTING
/usr/local/sbin/iptables -t mangle -F PREROUTING

/usr/local/sbin/iptables -t mangle -A PREROUTING -p udp --dport 9090 -j ROUTE --tee /usr/local/sbin/iptables -t nat -A PREROUTING -p udp --dport 9090 -j DNAT --to-destination 192.168.51.46:9090

but alas once i run these commands the packets just disappear. it doesn't reach either of the two destinations then. any comments ?

#19 lore raw
Quoted 3 lines · [email protected] wrote:
Quoted 1 line · Rob Sterenborg

and which kernel version did you use?

[...]

Quoted 2 lines

/usr/local/sbin/iptables -t mangle -A PREROUTING -p udp --dport 9090 -j ROUTE --tee

Searching for how "-j ROUTE --tee" works, I stumbled on these pages:
http://www.gossamer-threads.com/lists/iptables/devel/68316
http://www.cl.cam.ac.uk/research/dtg/research/wiki/MulticastRouting

From what I read there, --tee is used like this:

iptables -t mangle -A INPUT [matches] \
-j ROUTE --tee --gw <mirror_ip>

iptables -t mangle -A OUTPUT -d <dst_ip_net> -o <if_out> \
-j ROUTE --tee --oif <mirror_if_out>

Quoted 5 lines

/usr/local/sbin/iptables -t nat -A PREROUTING -p udp --dport 9090 -j DNAT --to-destination 192.168.51.46:9090

but alas once i run these commands the packets just disappear. it doesn't reach either of the two destinations then. any comments ? -

AFAICS you forgot to tell iptables to where you want to packet copied.

I also found that there appears to be a TEE target in development.
http://www.gossamer-threads.com/lists/iptables/devel/68781

Grts,
Rob