1netlink(7) Miscellaneous Information Manual netlink(7)
2
6 netlink - communication between kernel and user space (AF_NETLINK)
7
9 #include <asm/types.h>
10 #include <sys/socket.h>
11 #include <linux/netlink.h>
12 netlink_socket = socket(AF_NETLINK, socket_type, netlink_family);
14
16 Netlink is used to transfer information between the kernel and user-
17 space processes. It consists of a standard sockets-based interface for
18 user space processes and an internal kernel API for kernel modules.
19 The internal kernel interface is not documented in this manual page.
20 There is also an obsolete netlink interface via netlink character de‐
21 vices; this interface is not documented here and is provided only for
22 backward compatibility.
23 Netlink is a datagram-oriented service. Both SOCK_RAW and SOCK_DGRAM
25 are valid values for socket_type. However, the netlink protocol does
26 not distinguish between datagram and raw sockets.
27 netlink_family selects the kernel module or netlink group to communi‐
29 cate with. The currently assigned netlink families are:
30 NETLINK_ROUTE
32 Receives routing and link updates and may be used to modify the
33 routing tables (both IPv4 and IPv6), IP addresses, link parame‐
34 ters, neighbor setups, queueing disciplines, traffic classes,
35 and packet classifiers (see rtnetlink(7)).
36 NETLINK_W1 (Linux 2.6.13 to Linux 2.16.17)
38 Messages from 1-wire subsystem.
39 NETLINK_USERSOCK
41 Reserved for user-mode socket protocols.
42 NETLINK_FIREWALL (up to and including Linux 3.4)
44 Transport IPv4 packets from netfilter to user space. Used by
45 ip_queue kernel module. After a long period of being declared
46 obsolete (in favor of the more advanced nfnetlink_queue fea‐
47 ture), NETLINK_FIREWALL was removed in Linux 3.5.
48 NETLINK_SOCK_DIAG (since Linux 3.3)
50 Query information about sockets of various protocol families
51 from the kernel (see sock_diag(7)).
52 NETLINK_INET_DIAG (since Linux 2.6.14)
54 An obsolete synonym for NETLINK_SOCK_DIAG.
55 NETLINK_NFLOG (up to and including Linux 3.16)
57 Netfilter/iptables ULOG.
58 NETLINK_XFRM
60 IPsec.
61 NETLINK_SELINUX (since Linux 2.6.4)
63 SELinux event notifications.
64 NETLINK_ISCSI (since Linux 2.6.15)
66 Open-iSCSI.
67 NETLINK_AUDIT (since Linux 2.6.6)
69 Auditing.
70 NETLINK_FIB_LOOKUP (since Linux 2.6.13)
72 Access to FIB lookup from user space.
73 NETLINK_CONNECTOR (since Linux 2.6.14)
75 Kernel connector. See Documentation/driver-api/connector.rst
76 (or /Documentation/connector/connector.* in Linux 5.2 and ear‐
77 lier) in the Linux kernel source tree for further information.
78 NETLINK_NETFILTER (since Linux 2.6.14)
80 Netfilter subsystem.
81 NETLINK_SCSITRANSPORT (since Linux 2.6.19)
83 SCSI Transports.
84 NETLINK_RDMA (since Linux 3.0)
86 Infiniband RDMA.
87 NETLINK_IP6_FW (up to and including Linux 3.4)
89 Transport IPv6 packets from netfilter to user space. Used by
90 ip6_queue kernel module.
91 NETLINK_DNRTMSG
93 DECnet routing messages.
94 NETLINK_KOBJECT_UEVENT (since Linux 2.6.10)
96 Kernel messages to user space.
97 NETLINK_GENERIC (since Linux 2.6.15)
99 Generic netlink family for simplified netlink usage.
100 NETLINK_CRYPTO (since Linux 3.2)
102 Netlink interface to request information about ciphers regis‐
103 tered with the kernel crypto API as well as allow configuration
104 of the kernel crypto API.
105 Netlink messages consist of a byte stream with one or multiple nlmsghdr
107 headers and associated payload. The byte stream should be accessed
108 only with the standard NLMSG_* macros. See netlink(3) for further in‐
109 formation.
110 In multipart messages (multiple nlmsghdr headers with associated pay‐
112 load in one byte stream) the first and all following headers have the
113 NLM_F_MULTI flag set, except for the last header which has the type
114 NLMSG_DONE.
115 After each nlmsghdr the payload follows.
117 struct nlmsghdr {
119 __u32 nlmsg_len; /* Length of message including header */
120 __u16 nlmsg_type; /* Type of message content */
121 __u16 nlmsg_flags; /* Additional flags */
122 __u32 nlmsg_seq; /* Sequence number */
123 __u32 nlmsg_pid; /* Sender port ID */
124 };
125 nlmsg_type can be one of the standard message types: NLMSG_NOOP message
127 is to be ignored, NLMSG_ERROR message signals an error and the payload
128 contains an nlmsgerr structure, NLMSG_DONE message terminates a multi‐
129 part message. Error messages get the original request appended, unless
130 the user requests to cap the error message, and get extra error data if
131 requested.
132 struct nlmsgerr {
134 int error; /* Negative errno or 0 for acknowledgements */
135 struct nlmsghdr msg; /* Message header that caused the error */
136 /*
137 * followed by the message contents
138 * unless NETLINK_CAP_ACK was set
139 * or the ACK indicates success (error == 0).
140 * For example Generic Netlink message with attributes.
141 * message length is aligned with NLMSG_ALIGN()
142 */
143 /*
144 * followed by TLVs defined in enum nlmsgerr_attrs
145 * if NETLINK_EXT_ACK was set
146 */
147 };
148 A netlink family usually specifies more message types, see the appro‐
150 priate manual pages for that, for example, rtnetlink(7) for
151 NETLINK_ROUTE.
152 Standard flag bits in nlmsg_flags
154 ────────────────────────────────────────────────────────────────────────
155 NLM_F_REQUEST Must be set on all request messages.
156 NLM_F_MULTI The message is part of a multipart message ter‐
157 minated by NLMSG_DONE.
158 NLM_F_ACK Request for an acknowledgement on success.
159 NLM_F_ECHO Echo this request.
160 Additional flag bits for GET requests
162 ────────────────────────────────────────────────────────────────────────
163 NLM_F_ROOT Return the complete table instead of a single
164 entry.
165 NLM_F_MATCH Return all entries matching criteria passed
166 in message content. Not implemented yet.
167 NLM_F_ATOMIC Return an atomic snapshot of the table.
168 NLM_F_DUMP Convenience macro; equivalent to
169 (NLM_F_ROOT|NLM_F_MATCH).
170 Note that NLM_F_ATOMIC requires the CAP_NET_ADMIN capability or an ef‐
172 fective UID of 0.
173 Additional flag bits for NEW requests
175 ────────────────────────────────────────────────────────────────────────
176 NLM_F_REPLACE Replace existing matching object.
177 NLM_F_EXCL Don't replace if the object already exists.
178 NLM_F_CREATE Create object if it doesn't already exist.
179 NLM_F_APPEND Add to the end of the object list.
180 nlmsg_seq and nlmsg_pid are used to track messages. nlmsg_pid shows
182 the origin of the message. Note that there isn't a 1:1 relationship
183 between nlmsg_pid and the PID of the process if the message originated
184 from a netlink socket. See the ADDRESS FORMATS section for further in‐
185 formation.
186 Both nlmsg_seq and nlmsg_pid are opaque to netlink core.
188 Netlink is not a reliable protocol. It tries its best to deliver a
190 message to its destination(s), but may drop messages when an out-of-
191 memory condition or other error occurs. For reliable transfer the
192 sender can request an acknowledgement from the receiver by setting the
193 NLM_F_ACK flag. An acknowledgement is an NLMSG_ERROR packet with the
194 error field set to 0. The application must generate acknowledgements
195 for received messages itself. The kernel tries to send an NLMSG_ERROR
196 message for every failed packet. A user process should follow this
197 convention too.
198 However, reliable transmissions from kernel to user are impossible in
200 any case. The kernel can't send a netlink message if the socket buffer
201 is full: the message will be dropped and the kernel and the user-space
202 process will no longer have the same view of kernel state. It is up to
203 the application to detect when this happens (via the ENOBUFS error re‐
204 turned by recvmsg(2)) and resynchronize.
205 Address formats
207 The sockaddr_nl structure describes a netlink client in user space or
208 in the kernel. A sockaddr_nl can be either unicast (only sent to one
209 peer) or sent to netlink multicast groups (nl_groups not equal 0).
210 struct sockaddr_nl {
212 sa_family_t nl_family; /* AF_NETLINK */
213 unsigned short nl_pad; /* Zero */
214 pid_t nl_pid; /* Port ID */
215 __u32 nl_groups; /* Multicast groups mask */
216 };
217 nl_pid is the unicast address of netlink socket. It's always 0 if the
219 destination is in the kernel. For a user-space process, nl_pid is usu‐
220 ally the PID of the process owning the destination socket. However,
221 nl_pid identifies a netlink socket, not a process. If a process owns
222 several netlink sockets, then nl_pid can be equal to the process ID
223 only for at most one socket. There are two ways to assign nl_pid to a
224 netlink socket. If the application sets nl_pid before calling bind(2),
225 then it is up to the application to make sure that nl_pid is unique.
226 If the application sets it to 0, the kernel takes care of assigning it.
227 The kernel assigns the process ID to the first netlink socket the
228 process opens and assigns a unique nl_pid to every netlink socket that
229 the process subsequently creates.
230 nl_groups is a bit mask with every bit representing a netlink group
232 number. Each netlink family has a set of 32 multicast groups. When
233 bind(2) is called on the socket, the nl_groups field in the sockaddr_nl
234 should be set to a bit mask of the groups which it wishes to listen to.
235 The default value for this field is zero which means that no multicasts
236 will be received. A socket may multicast messages to any of the multi‐
237 cast groups by setting nl_groups to a bit mask of the groups it wishes
238 to send to when it calls sendmsg(2) or does a connect(2). Only pro‐
239 cesses with an effective UID of 0 or the CAP_NET_ADMIN capability may
240 send or listen to a netlink multicast group. Since Linux 2.6.13, mes‐
241 sages can't be broadcast to multiple groups. Any replies to a message
242 received for a multicast group should be sent back to the sending PID
243 and the multicast group. Some Linux kernel subsystems may additionally
244 allow other users to send and/or receive messages. As at Linux 3.0,
245 the NETLINK_KOBJECT_UEVENT, NETLINK_GENERIC, NETLINK_ROUTE, and
246 NETLINK_SELINUX groups allow other users to receive messages. No
247 groups allow other users to send messages.
248 Socket options
250 To set or get a netlink socket option, call getsockopt(2) to read or
251 setsockopt(2) to write the option with the option level argument set to
252 SOL_NETLINK. Unless otherwise noted, optval is a pointer to an int.
253 NETLINK_PKTINFO (since Linux 2.6.14)
255 Enable nl_pktinfo control messages for received packets to get
256 the extended destination group number.
257 NETLINK_ADD_MEMBERSHIP, NETLINK_DROP_MEMBERSHIP (since Linux 2.6.14)
259 Join/leave a group specified by optval.
260 NETLINK_LIST_MEMBERSHIPS (since Linux 4.2)
262 Retrieve all groups a socket is a member of. optval is a
263 pointer to __u32 and optlen is the size of the array. The array
264 is filled with the full membership set of the socket, and the
265 required array size is returned in optlen.
266 NETLINK_BROADCAST_ERROR (since Linux 2.6.30)
268 When not set, netlink_broadcast() only reports ESRCH errors and
269 silently ignore ENOBUFS errors.
270 NETLINK_NO_ENOBUFS (since Linux 2.6.30)
272 This flag can be used by unicast and broadcast listeners to
273 avoid receiving ENOBUFS errors.
274 NETLINK_LISTEN_ALL_NSID (since Linux 4.2)
276 When set, this socket will receive netlink notifications from
277 all network namespaces that have an nsid assigned into the net‐
278 work namespace where the socket has been opened. The nsid is
279 sent to user space via an ancillary data.
280 NETLINK_CAP_ACK (since Linux 4.3)
282 The kernel may fail to allocate the necessary room for the ac‐
283 knowledgement message back to user space. This option trims off
284 the payload of the original netlink message. The netlink mes‐
285 sage header is still included, so the user can guess from the
286 sequence number which message triggered the acknowledgement.
287
289 The socket interface to netlink first appeared Linux 2.2.
290 Linux 2.0 supported a more primitive device-based netlink interface
292 (which is still available as a compatibility option). This obsolete
293 interface is not described here.
294
296 It is often better to use netlink via libnetlink or libnl than via the
297 low-level kernel interface.
298
300 This manual page is not complete.
301
303 The following example creates a NETLINK_ROUTE netlink socket which will
304 listen to the RTMGRP_LINK (network interface create/delete/up/down
305 events) and RTMGRP_IPV4_IFADDR (IPv4 addresses add/delete events) mul‐
306 ticast groups.
307 struct sockaddr_nl sa;
309 memset(&sa, 0, sizeof(sa));
311 sa.nl_family = AF_NETLINK;
312 sa.nl_groups = RTMGRP_LINK | RTMGRP_IPV4_IFADDR;
313 fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
315 bind(fd, (struct sockaddr *) &sa, sizeof(sa));
316 The next example demonstrates how to send a netlink message to the ker‐
318 nel (pid 0). Note that the application must take care of message se‐
319 quence numbers in order to reliably track acknowledgements.
320 struct nlmsghdr *nh; /* The nlmsghdr with payload to send */
322 struct sockaddr_nl sa;
323 struct iovec iov = { nh, nh->nlmsg_len };
324 struct msghdr msg;
325 msg = { &sa, sizeof(sa), &iov, 1, NULL, 0, 0 };
327 memset(&sa, 0, sizeof(sa));
328 sa.nl_family = AF_NETLINK;
329 nh->nlmsg_pid = 0;
330 nh->nlmsg_seq = ++sequence_number;
331 /* Request an ack from kernel by setting NLM_F_ACK */
332 nh->nlmsg_flags |= NLM_F_ACK;
333 sendmsg(fd, &msg, 0);
335 And the last example is about reading netlink message.
337 int len;
339 /* 8192 to avoid message truncation on platforms with
340 page size > 4096 */
341 struct nlmsghdr buf[8192/sizeof(struct nlmsghdr)];
342 struct iovec iov = { buf, sizeof(buf) };
343 struct sockaddr_nl sa;
344 struct msghdr msg;
345 struct nlmsghdr *nh;
346 msg = { &sa, sizeof(sa), &iov, 1, NULL, 0, 0 };
348 len = recvmsg(fd, &msg, 0);
349 for (nh = (struct nlmsghdr *) buf; NLMSG_OK (nh, len);
351 nh = NLMSG_NEXT (nh, len)) {
352 /* The end of multipart message */
353 if (nh->nlmsg_type == NLMSG_DONE)
354 return;
355 if (nh->nlmsg_type == NLMSG_ERROR)
357 /* Do some error handling */
358 ...
359 /* Continue with parsing payload */
361 ...
362 }
363
365 cmsg(3), netlink(3), capabilities(7), rtnetlink(7), sock_diag(7)
366 information about libnetlink ⟨ftp://ftp.inr.ac.ru/ip-routing/iproute2*⟩
368 information about libnl ⟨http://www.infradead.org/~tgr/libnl/⟩
370 RFC 3549 "Linux Netlink as an IP Services Protocol"
372Linux man-pages 6.05 2023-07-30 netlink(7)