Quick guide for Netlink

A means to exchange networking information between kernel and userspace:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
The Netlink socket family is a Linux kernel interface used for inter-process communication (IPC) 
between both the kernel and userspace processes, and between different userspace processes, 
in a way similar to the Unix domain sockets. Similarly to the Unix domain sockets, and unlike
INET sockets, Netlink communication cannot traverse host boundaries. However, while the Unix domain
sockets use the file system namespace, Netlink processes are usually addressed by process identifiers (PIDs). [3]

Netlink is designed and used for transferring miscellaneous networking information between the 
kernel space and userspace processes. Networking utilities, such as the iproute2 family and the utilities
used for configuring mac80211-based wireless drivers, use Netlink to communicate with the Linux kernel
from userspace. Netlink provides a standard socket-based interface for userspace processes, and a 
kernel-side API for internal use by kernel modules. Originally, Netlink used the AF_NETLINK socket family. 
https://en.wikipedia.org/wiki/Netlink

Netlink is defined and thoroughly described in RFC3549

The idea is to keep code that's critical for performance in the kernel and to move anything related to control and management in user-space (https://www.linuxjournal.com/article/7356).

1
2
3
4
5
Netlink socket is a special IPC used for transferring information between kernel and user-space processes. 
It provides a full-duplex communication link between the two by way of standard socket APIs for user-space 
processes and a special kernel API for kernel modules. Netlink socket uses the address family AF_NETLINK, as 
compared to AF_INET used by TCP/IP socket. Each netlink socket feature defines its own protocol type in the 
kernel header file include/linux/netlink.h. 
https://www.linuxjournal.com/article/7356

Just a few examples: * iproute2

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
[root@dell-r430-30 ~]# strace -e trace=network /usr/sbin/ip route
socket(AF_NETLINK, SOCK_RAW|SOCK_CLOEXEC, NETLINK_ROUTE) = 3
setsockopt(3, SOL_SOCKET, SO_SNDBUF, [32768], 4) = 0
setsockopt(3, SOL_SOCKET, SO_RCVBUF, [1048576], 4) = 0
bind(3, {sa_family=AF_NETLINK, pid=0, groups=00000000}, 12) = 0
getsockname(3, {sa_family=AF_NETLINK, pid=161345, groups=00000000}, [12]) = 0
sendto(3, "(\0\0\0\32\0\1\3\330A\367[\0\0\0\0\2\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 40, 0, NULL, 0) = 40
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"4\0\0\0\30\0\2\0\330A\367[Av\2\0\2\0\0\0\376\3\0\1\0\0\0\0\10\0\17\0"..., 32768}], msg_controllen=0, msg_flags=0}, 0) = 1312
socket(AF_LOCAL, SOCK_DGRAM|SOCK_CLOEXEC, 0) = 4
default via 10.10.99.254 dev br-redhat 
socket(AF_LOCAL, SOCK_DGRAM|SOCK_CLOEXEC, 0) = 4
10.10.96.0/22 dev br-redhat proto kernel scope link src 10.10.96.194 
socket(AF_LOCAL, SOCK_DGRAM|SOCK_CLOEXEC, 0) = 4
169.254.0.0/16 dev br-redhat scope link metric 1008 
socket(AF_LOCAL, SOCK_DGRAM|SOCK_CLOEXEC, 0) = 4
169.254.0.0/16 dev br-provis scope link metric 1009 
socket(AF_LOCAL, SOCK_DGRAM|SOCK_CLOEXEC, 0) = 4
169.254.0.0/16 dev br-trunk1 scope link metric 1010 
socket(AF_LOCAL, SOCK_DGRAM|SOCK_CLOEXEC, 0) = 4
169.254.0.0/16 dev br-stonith scope link metric 1011 
socket(AF_LOCAL, SOCK_DGRAM|SOCK_CLOEXEC, 0) = 4
169.254.0.0/16 dev br-trunk2 scope link metric 1012 
socket(AF_LOCAL, SOCK_DGRAM|SOCK_CLOEXEC, 0) = 4
192.168.111.0/24 dev br-stonith proto kernel scope link src 192.168.111.1 
socket(AF_LOCAL, SOCK_DGRAM|SOCK_CLOEXEC, 0) = 4
192.168.122.0/24 dev virbr0 proto kernel scope link src 192.168.122.1 
recvmsg(3, {msg_name(12)={sa_family=AF_NETLINK, pid=0, groups=00000000}, msg_iov(1)=[{"\24\0\0\0\3\0\2\0\330A\367[Av\2\0\0\0\0\0", 32768}], msg_controllen=0, msg_flags=0}, 0) = 20
+++ exited with 0 +++

  • OVS

Socket programming

Let's do a little detour towards socket programming, as this will teach the basics for Netlink programming.

C

Resources

The Linux Programming Reference is an excellent book for C programming with Linux.

Program

Let's start with a generic example for IPC via sockets in C. The following contains an example for a UDP socket: https://github.com/andreaskaris/blog/blob/master/messenger.c

Compile:

1
2
[akaris@wks-akaris c]$ gcc messenger.c -o messenger
[akaris@wks-akaris c]$ 

Then run, in one CLI the receiver:

1
2
3
4
5
6
7
[akaris@wks-akaris c]$ ./messenger receiver
This node is a receiver
Starting receiver
Received datagram from 127.0.0.1
Datagram content: MSG
Received datagram from 127.0.0.1
Datagram content: MSG

In the other CLI the sender:

1
2
3
4
5
6
[akaris@wks-akaris c]$ ./messenger sender
This node is a sender
Starting sender
[akaris@wks-akaris c]$ ./messenger sender
This node is a sender
Starting sender

Python

Resources

https://wiki.python.org/moin/UdpCommunication

Program

Let's start with a generic example for IPC via sockets in Python. The following contains an example for a UDP socket: https://github.com/andreaskaris/blog/blob/master/messenger.py

Run the server:

1
2
3
4
5
6
7
[akaris@wks-akaris python]$ ./messenger.py rec
This node is a receiver
Starting receiver
Received datagram from  ('127.0.0.1', 58172)
Datagram content:  b'MSG'
Received datagram from  ('127.0.0.1', 39735)
Datagram content:  b'MSG'

Run the client in another CLI:

1
2
3
4
5
6
7
[akaris@wks-akaris python]$ ./messenger.py sender
This node is a sender
Starting sender
[akaris@wks-akaris python]$ ./messenger.py sender
This node is a sender
Starting sender
[akaris@wks-akaris python]$ 

https://www.linuxjournal.com/article/7356 has a walkthrough for using Netlink in C.

https://www.linuxjournal.com/article/8498 as well.

C

The following C application monitors any route changes that happen in the kernel: https://github.com/andreaskaris/blog/blob/master/netlink.c

Compile the code:

1
gcc netlink.c -o netlink

Start the application and while the application is running, modify the routing table:

1
2
[akaris@wks-akaris python]$ sudo ip r a 4.3.2.0/30 via 127.0.0.1
[akaris@wks-akaris python]$ sudo ip r d 4.3.2.0/30 via 127.0.0.1

The application will yield the following. We can see live when the route is added and when it is deleted:

1
2
3
4
5
6
[akaris@wks-akaris c]$ ./netlink 
Starting netlink receiverd
nlmsg_len: 60, nlmsg_type 24, nlmsg_flags: 1536, nlmsg_seq: 1543015716, nlmsg_pid: 15333
Netmask: 30
nlmsg_len: 60, nlmsg_type 25, nlmsg_flags: 0, nlmsg_seq: 1543015717, nlmsg_pid: 15343
Netmask: 30

Note that a more complete example can be found here: https://gist.github.com/cl4u2/5204374

Python

The following python application monitors any route changes that happen in the kernel: https://github.com/andreaskaris/blog/blob/master/netlink.py

Start the application and while the application is running, modify the routing table:

1
2
[akaris@wks-akaris python]$ sudo ip r a 4.3.2.0/30 via 127.0.0.1
[akaris@wks-akaris python]$ sudo ip r d 4.3.2.0/30 via 127.0.0.1

The application will yield the following. We can see live when the route is added and when it is deleted:

1
2
3
4
5
6
7
8
9
[akaris@wks-akaris python]$ ./netlink.py 
Netlink header:  60 RTM_NEWROUTE 1536 1543012056 13037
Netlink payload:  2 30 0 0 254 3 0 1 0
Netmask: 30
IP:  b'\x04\x03\x02\x00'
Netlink header:  60 RTM_DELROUTE 0 1543012061 13047
Netlink payload:  2 30 0 0 254 3 0 1 0
Netmask: 30
IP:  b'\x04\x03\x02\x00'

strace'ing the application reveals even more about the message structure:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
[akaris@wks-akaris python]$ strace -e trace=network ./netlink.py 
socket(AF_NETLINK, SOCK_RAW|SOCK_CLOEXEC, NETLINK_ROUTE) = 3
bind(3, {sa_family=AF_NETLINK, nl_pid=13307, nl_groups=0x000040}, 12) = 0
recvfrom(3, {{len=60, type=RTM_NEWROUTE, flags=NLM_F_EXCL|NLM_F_CREATE, seq=1543012305, pid=13310}, {rtm_family=AF_INET, rtm_dst_len=30, rtm_src_len=0, rtm_tos=0, rtm_table=RT_TABLE_MAIN, rtm_protocol=RTPROT_BOOT, rtm_scope=RT_SCOPE_UNIVERSE, rtm_type=RTN_UNICAST, rtm_flags=0}, [{{nla_len=8, nla_type=RTA_TABLE}, RT_TABLE_MAIN}, {{nla_len=8, nla_type=RTA_DST}, 4.3.2.0}, {{nla_len=8, nla_type=RTA_GATEWAY}, 127.0.0.1}, {{nla_len=8, nla_type=RTA_OIF}, if_nametoindex("lo")}]}, 65535, 0, NULL, NULL) = 60
Netlink header:  60 RTM_NEWROUTE 1536 1543012305 13310
Netlink payload:  2 30 0 0 254 3 0 1 0
Netmask: 30
IP:  b'\x04\x03\x02\x00'
recvfrom(3, {{len=60, type=RTM_DELROUTE, flags=0, seq=1543012307, pid=13320}, {rtm_family=AF_INET, rtm_dst_len=30, rtm_src_len=0, rtm_tos=0, rtm_table=RT_TABLE_MAIN, rtm_protocol=RTPROT_BOOT, rtm_scope=RT_SCOPE_UNIVERSE, rtm_type=RTN_UNICAST, rtm_flags=0}, [{{nla_len=8, nla_type=RTA_TABLE}, RT_TABLE_MAIN}, {{nla_len=8, nla_type=RTA_DST}, 4.3.2.0}, {{nla_len=8, nla_type=RTA_GATEWAY}, 127.0.0.1}, {{nla_len=8, nla_type=RTA_OIF}, if_nametoindex("lo")}]}, 65535, 0, NULL, NULL) = 60
Netlink header:  60 RTM_DELROUTE 0 1543012307 13320
Netlink payload:  2 30 0 0 254 3 0 1 0
Netmask: 30
IP:  b'\x04\x03\x02\x00'
recvfrom(3, 

Resources

https://en.wikipedia.org/wiki/Netlink

https://tools.ietf.org/html/rfc3549

https://www.linuxjournal.com/article/7356

http://inai.de/documents/Netlink_Protocol.pdf