Comments on: Getting Linux routing table using netlink

By: Baa

Baa — Sat, 25 Mar 2023 11:41:32 +0000

Hi,
I have change the code to send request IPv4/IPv6 in main, like below

int main()
{
int nl_sock = open_netlink();

if (do_route_dump_requst(nl_sock, AF_INET) < 0) { perror("Failed to perfom request"); close(nl_sock); return -1; }


get_route_dump_response(nl_sock);
if (do_route_dump_requst(nl_sock, AF_INET6) < 0) {

    perror("Failed to perfom request");

    close(nl_sock);

    return -1;

}
get_route_dump_response(nl_sock);
close (nl_sock);

return 0;

}

and change int do_route_dump_requst(int sock) to
int do_route_dump_requst(int sock, int family)
{
…
int do_route_dump_requst(int sock, int family)
…
}

but the IPv6 always show “Wrong message length: No error information”
if I change AF_INET to AF_UNSPEC , it’s will working fine.

Can you give me some comment?
Thanks

By: Ort

Ort — Wed, 26 Jan 2022 08:06:27 +0000

Hi Oleg,
Thanks for the useful information!

Is there any way to know if a specific route exists in the routing table? For example to use RTM_GETROUTE but also to specify the ip address of the route we’re searching? I don’t want all the entries in the routing table, only a specific entry.
Thanks!

By: Oleg Kutkov

Oleg Kutkov — Mon, 09 Nov 2020 21:31:24 +0000

In reply to Sandy.

Hello. It’s no problem.
This code is just adopted for the IPv4, you can easily switch to IPv6.
Just replace everywhere AF_INET to AF_INET6 + remove the address length check at line 119:

if ((r->rtm_dst_len != 24) && (r->rtm_dst_len != 16)) { return; }

This code just skips your IPv6 addresses due to IPv6 rtm_dst_len == 128.
Without this check IPv6 table can be printed without any problems:

$ ./print_route Main routing table IPv6 ::1/128 dev lo fe80::/64 dev enp4s0 ::1/128 dev lo fe80::d448:77d5:68fa:bf6b/128 dev enp4s0 ff00::/8 dev enp4s0

You can completely remove this check or better adopt it to the 128 bit address lengths.

By: Sandy

Sandy — Mon, 09 Nov 2020 20:57:05 +0000

This code is not working for IPV6 if I change family to AF_INET6 as specified above.
nl_request.rtm.rtm_family = AF_INET6

Please help

By: Digvijay

Digvijay — Wed, 23 Sep 2020 12:24:29 +0000

Very useful. Nicely coded.
I think there should be a return statement in the following block otherwise it can crash:

if (h->nlmsg_type == NLMSG_ERROR) { perror("netlink reported error"); free(buf); }

By: Oleg Kutkov

Oleg Kutkov — Tue, 18 Aug 2020 20:05:48 +0000

In reply to Savva. Hello! Can you repeat your tests with command ip r ? This command is also based on netlink. Command "route -n" uses old-style ioctl requests.

By: Savva

Savva — Tue, 18 Aug 2020 02:38:49 +0000

Hey Oleg, thanks for the post! Very useful. I’m using this code on an ad-hoc network (with olsrd). I am seeing at least a 20 second delay between an ip address appearing among the results of the “route -n” command and the same ip address appearing among the outputs of the get_route_dump_response() function. Any insights as to why this might be happening?

By: Modifying Linux network routes using netlink - Oleg Kutkov personal blog

Modifying Linux network routes using netlink - Oleg Kutkov personal blog — Wed, 28 Aug 2019 23:14:21 +0000

[…] time we talked about getting Linux routing table with simple Netlink code. Now it’s time to do more […]

By: Oleg Kutkov

Oleg Kutkov — Mon, 03 Jun 2019 07:59:22 +0000

In reply to Mike B. Thanks! New article is upcoming. Stay tuned.

By: Mike B

Mike B — Sun, 02 Jun 2019 20:14:45 +0000

very useful, waiting for next articles about netlink