man7/netlink.7

   1 '\" t
   2 .\" This man page is Copyright (c) 1998 by Andi Kleen.
   3 .\"
   4 .\" %%%LICENSE_START(GPL_NOVERSION_ONELINE)
   5 .\" Subject to the GPL.
   6 .\" %%%LICENSE_END
   7 .\"
   8 .\" Based on the original comments from Alexey Kuznetsov
   9 .\" Modified 2005-12-27 by Hasso Tepper <hasso@estpak.ee>
  10 .\" $Id: netlink.7,v 1.8 2000/06/22 13:23:00 ak Exp $
  11 .TH NETLINK  7 2016-07-17 "Linux" "Linux Programmer's Manual"
  12 .SH NAME
  13 netlink \- communication between kernel and user space (AF_NETLINK)
  14 .SH SYNOPSIS
  15 .nf
  16 .B #include <asm/types.h>
  17 .B #include <sys/socket.h>
  18 .B #include <linux/netlink.h>
  19
  20 .BI "netlink_socket = socket(AF_NETLINK, " socket_type ", " netlink_family );
  21 .fi
  22 .SH DESCRIPTION
  23 Netlink is used to transfer information between the kernel and
  24 user-space processes.
  25 It consists of a standard sockets-based interface for user space
  26 processes and an internal kernel API for kernel modules.
  27 The internal kernel interface is not documented in this manual page.
  28 There is also an obsolete netlink interface
  29 via netlink character devices; this interface is not documented here
  30 and is provided only for backward compatibility.
  31
  32 Netlink is a datagram-oriented service.
  33 Both
  34 .B SOCK_RAW
  35 and
  36 .B SOCK_DGRAM
  37 are valid values for
  38 .IR socket_type .
  39 However, the netlink protocol does not distinguish between datagram
  40 and raw sockets.
  41
  42 .I netlink_family
  43 selects the kernel module or netlink group to communicate with.
  44 The currently assigned netlink families are:
  45 .TP
  46 .BR NETLINK_ROUTE
  47 Receives routing and link updates and may be used to modify the routing
  48 tables (both IPv4 and IPv6), IP addresses, link parameters,
  49 neighbor setups, queueing disciplines, traffic classes and
  50 packet classifiers (see
  51 .BR rtnetlink (7)).
  52 .TP
  53 .BR NETLINK_W1 " (since Linux 2.6.13)"
  54 Messages from 1-wire subsystem.
  55 .TP
  56 .BR NETLINK_USERSOCK
  57 Reserved for user-mode socket protocols.
  58 .TP
  59 .BR NETLINK_FIREWALL
  60 Transport IPv4 packets from netfilter to user space.
  61 Used by
  62 .I ip_queue
  63 kernel module.
  64 .TP
  65 .BR NETLINK_INET_DIAG " (since Linux 2.6.14)"
  66 .\" FIXME More details on NETLINK_INET_DIAG needed.
  67 INET socket monitoring.
  68 .TP
  69 .BR NETLINK_NFLOG
  70 Netfilter/iptables ULOG.
  71 .TP
  72 .BR NETLINK_XFRM
  73 .\" FIXME More details on NETLINK_XFRM needed.
  74 IPsec.
  75 .TP
  76 .BR NETLINK_SELINUX " (since Linux 2.6.4)"
  77 SELinux event notifications.
  78 .TP
  79 .BR NETLINK_ISCSI " (since Linux 2.6.15)"
  80 .\" FIXME More details on NETLINK_ISCSI needed.
  81 Open-iSCSI.
  82 .TP
  83 .BR NETLINK_AUDIT " (since Linux 2.6.6)"
  84 .\" FIXME More details on NETLINK_AUDIT needed.
  85 Auditing.
  86 .TP
  87 .BR NETLINK_FIB_LOOKUP " (since Linux 2.6.13)"
  88 .\" FIXME More details on NETLINK_FIB_LOOKUP needed.
  89 Access to FIB lookup from user space.
  90 .TP
  91 .BR NETLINK_CONNECTOR " (since Linux 2.6.14)"
  92 Kernel connector.
  93 See
  94 .I Documentation/connector/*
  95 in the Linux kernel source tree for further information.
  96 .TP
  97 .B NETLINK_NETFILTER " (since Linux 2.6.14)"
  98 .\" FIXME More details on NETLINK_NETFILTER needed.
  99 Netfilter subsystem.
 100 .TP
 101 .BR NETLINK_IP6_FW
 102 Transport IPv6 packets from netfilter to user space.
 103 Used by
 104 .I ip6_queue
 105 kernel module.
 106 .TP
 107 .B NETLINK_DNRTMSG
 108 DECnet routing messages.
 109 .TP
 110 .BR NETLINK_KOBJECT_UEVENT " (since Linux 2.6.10)"
 111 .\" FIXME More details on NETLINK_KOBJECT_UEVENT needed.
 112 Kernel messages to user space.
 113 .TP
 114 .BR NETLINK_GENERIC " (since Linux 2.6.15)"
 115 Generic netlink family for simplified netlink usage.
 116 .TP
 117 .BR NETLINK_CRYPTO " (since Linux 3.2)"
 118 .\" commit a38f7907b926e4c6c7d389ad96cc38cec2e5a9e9
 119 .\" Author: Steffen Klassert <steffen.klassert@secunet.com>
 120 Netlink interface to request information about ciphers registered
 121 with the kernel crypto API as well as allow configuration of the
 122 kernel crypto API.
 123 .PP
 124 Netlink messages consist of a byte stream with one or multiple
 125 .I nlmsghdr
 126 headers and associated payload.
 127 The byte stream should be accessed only with the standard
 128 .B NLMSG_*
 129 macros.
 130 See
 131 .BR netlink (3)
 132 for further information.
 133
 134 In multipart messages (multiple
 135 .I nlmsghdr
 136 headers with associated payload in one byte stream) the first and all
 137 following headers have the
 138 .B NLM_F_MULTI
 139 flag set, except for the last header which has the type
 140 .BR NLMSG_DONE .
 141
 142 After each
 143 .I nlmsghdr
 144 the payload follows.
 145
 146 .in +4n
 147 .nf
 148 struct nlmsghdr {
 149     __u32 nlmsg_len;    /* Length of message including header */
 150     __u16 nlmsg_type;   /* Type of message content */
 151     __u16 nlmsg_flags;  /* Additional flags */
 152     __u32 nlmsg_seq;    /* Sequence number */
 153     __u32 nlmsg_pid;    /* Sender port ID */
 154 };
 155 .fi
 156 .in
 157
 158 .I nlmsg_type
 159 can be one of the standard message types:
 160 .B NLMSG_NOOP
 161 message is to be ignored,
 162 .B NLMSG_ERROR
 163 message signals an error and the payload contains an
 164 .I nlmsgerr
 165 structure,
 166 .B NLMSG_DONE
 167 message terminates a multipart message.
 168
 169 .in +4n
 170 .nf
 171 struct nlmsgerr {
 172     int error;        /* Negative errno or 0 for acknowledgements */
 173     struct nlmsghdr msg;  /* Message header that caused the error */
 174 };
 175 .fi
 176 .in
 177
 178 A netlink family usually specifies more message types, see the
 179 appropriate manual pages for that, for example,
 180 .BR rtnetlink (7)
 181 for
 182 .BR NETLINK_ROUTE .
 183 .TS
 184 tab(:);
 185 l s
 186 lB l.
 187 Standard flag bits in \fInlmsg_flags\fP
 188 _
 189 NLM_F_REQUEST:Must be set on all request messages.
 190 NLM_F_MULTI:T{
 191 The message is part of a multipart message terminated by
 192 .BR NLMSG_DONE .
 193 T}
 194 NLM_F_ACK:Request for an acknowledgment on success.
 195 NLM_F_ECHO:Echo this request.
 196 .TE
 197 .\" No right adjustment for text blocks in tables
 198 .TS
 199 tab(:);
 200 l s
 201 lB l.
 202 Additional flag bits for GET requests
 203 _
 204 NLM_F_ROOT:Return the complete table instead of a single entry.
 205 NLM_F_MATCH:T{
 206 Return all entries matching criteria passed in message content.
 207 Not implemented yet.
 208 T}
 209 NLM_F_ATOMIC:Return an atomic snapshot of the table.
 210 NLM_F_DUMP:T{
 211 Convenience macro; equivalent to
 212 .br
 213 (NLM_F_ROOT|NLM_F_MATCH).
 214 T}
 215 .TE
 216 .\" FIXME NLM_F_ATOMIC is not used anymore?
 217 .sp 1
 218 Note that
 219 .B NLM_F_ATOMIC
 220 requires the
 221 .B CAP_NET_ADMIN
 222 capability or an effective UID of 0.
 223 .TS
 224 tab(:);
 225 l s
 226 lB l.
 227 Additional flag bits for NEW requests
 228 _
 229 NLM_F_REPLACE:Replace existing matching object.
 230 NLM_F_EXCL:Don't replace if the object already exists.
 231 NLM_F_CREATE:Create object if it doesn't already exist.
 232 NLM_F_APPEND:Add to the end of the object list.
 233 .TE
 234 .sp 1
 235 .I nlmsg_seq
 236 and
 237 .I nlmsg_pid
 238 are used to track messages.
 239 .I nlmsg_pid
 240 shows the origin of the message.
 241 Note that there isn't a 1:1 relationship between
 242 .I nlmsg_pid
 243 and the PID of the process if the message originated from a netlink
 244 socket.
 245 See the
 246 .B ADDRESS FORMATS
 247 section for further information.
 248
 249 Both
 250 .I nlmsg_seq
 251 and
 252 .I nlmsg_pid
 253 .\" FIXME Explain more about nlmsg_seq and nlmsg_pid.
 254 are opaque to netlink core.
 255
 256 Netlink is not a reliable protocol.
 257 It tries its best to deliver a message to its destination(s),
 258 but may drop messages when an out-of-memory condition or
 259 other error occurs.
 260 For reliable transfer the sender can request an
 261 acknowledgement from the receiver by setting the
 262 .B NLM_F_ACK
 263 flag.
 264 An acknowledgment is an
 265 .B NLMSG_ERROR
 266 packet with the error field set to 0.
 267 The application must generate acknowledgements for
 268 received messages itself.
 269 The kernel tries to send an
 270 .B NLMSG_ERROR
 271 message for every failed packet.
 272 A user process should follow this convention too.
 273
 274 However, reliable transmissions from kernel to user are impossible
 275 in any case.
 276 The kernel can't send a netlink message if the socket buffer is full:
 277 the message will be dropped and the kernel and the user-space process will
 278 no longer have the same view of kernel state.
 279 It is up to the application to detect when this happens (via the
 280 .B ENOBUFS
 281 error returned by
 282 .BR recvmsg (2))
 283 and resynchronize.
 284 .SS Address formats
 285 The
 286 .I sockaddr_nl
 287 structure describes a netlink client in user space or in the kernel.
 288 A
 289 .I sockaddr_nl
 290 can be either unicast (only sent to one peer) or sent to
 291 netlink multicast groups
 292 .RI ( nl_groups
 293 not equal 0).
 294
 295 .in +4n
 296 .nf
 297 struct sockaddr_nl {
 298     sa_family_t     nl_family;  /* AF_NETLINK */
 299     unsigned short  nl_pad;     /* Zero */
 300     pid_t           nl_pid;     /* Port ID */
 301     __u32           nl_groups;  /* Multicast groups mask */
 302 };
 303 .fi
 304 .in
 305
 306 .I nl_pid
 307 is the unicast address of netlink socket.
 308 It's always 0 if the destination is in the kernel.
 309 For a user-space process,
 310 .I nl_pid
 311 is usually the PID of the process owning the destination socket.
 312 However,
 313 .I nl_pid
 314 identifies a netlink socket, not a process.
 315 If a process owns several netlink
 316 sockets, then
 317 .I nl_pid
 318 can be equal to the process ID only for at most one socket.
 319 There are two ways to assign
 320 .I nl_pid
 321 to a netlink socket.
 322 If the application sets
 323 .I nl_pid
 324 before calling
 325 .BR bind (2),
 326 then it is up to the application to make sure that
 327 .I nl_pid
 328 is unique.
 329 If the application sets it to 0, the kernel takes care of assigning it.
 330 The kernel assigns the process ID to the first netlink socket the process
 331 opens and assigns a unique
 332 .I nl_pid
 333 to every netlink socket that the process subsequently creates.
 334
 335 .I nl_groups
 336 is a bit mask with every bit representing a netlink group number.
 337 Each netlink family has a set of 32 multicast groups.
 338 When
 339 .BR bind (2)
 340 is called on the socket, the
 341 .I nl_groups
 342 field in the
 343 .I sockaddr_nl
 344 should be set to a bit mask of the groups which it wishes to listen to.
 345 The default value for this field is zero which means that no multicasts
 346 will be received.
 347 A socket may multicast messages to any of the multicast groups by setting
 348 .I nl_groups
 349 to a bit mask of the groups it wishes to send to when it calls
 350 .BR sendmsg (2)
 351 or does a
 352 .BR connect (2).
 353 Only processes with an effective UID of 0 or the
 354 .B CAP_NET_ADMIN
 355 capability may send or listen to a netlink multicast group.
 356 Since Linux 2.6.13,
 357 .\" commit d629b836d151d43332492651dd841d32e57ebe3b
 358 messages can't be broadcast to multiple groups.
 359 Any replies to a message received for a multicast group should be
 360 sent back to the sending PID and the multicast group.
 361 Some Linux kernel subsystems may additionally allow other users
 362 to send and/or receive messages.
 363 As at Linux 3.0, the
 364 .BR NETLINK_KOBJECT_UEVENT ,
 365 .BR NETLINK_GENERIC ,
 366 .BR NETLINK_ROUTE ,
 367 and
 368 .BR NETLINK_SELINUX
 369 groups allow other users to receive messages.
 370 No groups allow other users to send messages.
 371
 372 .SS Socket options
 373 To set or get a netlink socket option, call
 374 .BR getsockopt (2)
 375 to read or
 376 .BR setsockopt (2)
 377 to write the option with the option level argument set to
 378 .BR SOL_NETLINK .
 379 Unless otherwise noted,
 380 .I optval
 381 is a pointer to an
 382 .IR int .
 383 .TP
 384 .BR NETLINK_PKTINFO " (since Linux 2.6.14)"
 385 .\"     commit 9a4595bc7e67962f13232ee55a64e063062c3a99
 386 .\"     Author: Patrick McHardy <kaber@trash.net>
 387 Enable
 388 .B nl_pktinfo
 389 control messages for received packets to get the extended
 390 destination group number.
 391 .TP
 392 .BR NETLINK_ADD_MEMBERSHIP ,\  NETLINK_DROP_MEMBERSHIP " (since Linux 2.6.14)"
 393 .\"     commit 9a4595bc7e67962f13232ee55a64e063062c3a99
 394 .\"     Author: Patrick McHardy <kaber@trash.net>
 395 Join/leave a group specified by
 396 .IR optval .
 397 .TP
 398 .BR NETLINK_LIST_MEMBERSHIPS " (since Linux 4.2)"
 399 .\"     commit b42be38b2778eda2237fc759e55e3b698b05b315
 400 .\"     Author: David Herrmann <dh.herrmann@gmail.com>
 401 Retrieve all groups a socket is a member of.
 402 .I optval
 403 is a pointer to
 404 .B __u32
 405 and
 406 .I optlen
 407 is the size of the array.
 408 The array is filled with the full membership set of the
 409 socket, and the required array size is returned in
 410 .I optlen.
 411 .TP
 412 .BR NETLINK_BROADCAST_ERROR " (since Linux 2.6.30)"
 413 .\"     commit be0c22a46cfb79ab2342bb28fde99afa94ef868e
 414 .\"     Author: Pablo Neira Ayuso <pablo@netfilter.org>
 415 When not set,
 416 .B netlink_broadcast()
 417 only reports
 418 .B ESRCH
 419 errors and silently ignore
 420 .B NOBUFS
 421 errors.
 422 .TP
 423 .BR NETLINK_NO_ENOBUFS " (since Linux 2.6.30)"
 424 .\"     commit 38938bfe3489394e2eed5e40c9bb8f66a2ce1405
 425 .\"     Author: Pablo Neira Ayuso <pablo@netfilter.org>
 426 This flag can be used by unicast and broadcast listeners to avoid receiving
 427 .B ENOBUFS
 428 errors.
 429 .TP
 430 .BR NETLINK_LISTEN_ALL_NSID " (since Linux 4.2)"
 431 .\"     commit 59324cf35aba5336b611074028777838a963d03b
 432 .\"     Author: Nicolas Dichtel <nicolas.dichtel@6wind.com>
 433 When set, this socket will receive netlink notifications from
 434 all network namespaces that have an
 435 .I nsid
 436 assigned into the network namespace where the socket has been opened.
 437 The
 438 .I nsid
 439 is sent to user space via an ancillary data.
 440 .TP
 441 .BR NETLINK_CAP_ACK " (since Linux 4.2)"
 442 .\"     commit 0a6a3a23ea6efde079a5b77688541a98bf202721
 443 .\"     Author: Christophe Ricard <christophe.ricard@gmail.com>
 444 The kernel may fail to allocate the necessary room for the acknowledgment
 445 message back to user space.
 446 This option trims off the payload of the original netlink message.
 447 The netlink message header is still included, so the user can guess from the
 448 sequence number which message triggered the acknowledgment.
 449 .SH VERSIONS
 450 The socket interface to netlink first appeared Linux 2.2.
 451
 452 Linux 2.0 supported a more primitive device-based netlink interface
 453 (which is still available as a compatibility option).
 454 This obsolete interface is not described here.
 455 .SH NOTES
 456 It is often better to use netlink via
 457 .I libnetlink
 458 or
 459 .I libnl
 460 than via the low-level kernel interface.
 461 .SH BUGS
 462 This manual page is not complete.
 463 .SH EXAMPLE
 464 The following example creates a
 465 .B NETLINK_ROUTE
 466 netlink socket which will listen to the
 467 .B RTMGRP_LINK
 468 (network interface create/delete/up/down events) and
 469 .B RTMGRP_IPV4_IFADDR
 470 (IPv4 addresses add/delete events) multicast groups.
 471
 472 .in +4n
 473 .nf
 474 struct sockaddr_nl sa;
 475
 476 memset(&sa, 0, sizeof(sa));
 477 sa.nl_family = AF_NETLINK;
 478 sa.nl_groups = RTMGRP_LINK | RTMGRP_IPV4_IFADDR;
 479
 480 fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
 481 bind(fd, (struct sockaddr *) &sa, sizeof(sa));
 482 .fi
 483 .in
 484
 485 The next example demonstrates how to send a netlink message to the
 486 kernel (pid 0).
 487 Note that the application must take care of message sequence numbers
 488 in order to reliably track acknowledgements.
 489
 490 .in +4n
 491 .nf
 492 struct nlmsghdr *nh;    /* The nlmsghdr with payload to send */
 493 struct sockaddr_nl sa;
 494 struct iovec iov = { nh, nh\->nlmsg_len };
 495 struct msghdr msg;
 496
 497 msg = { &sa, sizeof(sa), &iov, 1, NULL, 0, 0 };
 498 memset(&sa, 0, sizeof(sa));
 499 sa.nl_family = AF_NETLINK;
 500 nh\->nlmsg_pid = 0;
 501 nh\->nlmsg_seq = ++sequence_number;
 502 /* Request an ack from kernel by setting NLM_F_ACK */
 503 nh\->nlmsg_flags |= NLM_F_ACK;
 504
 505 sendmsg(fd, &msg, 0);
 506 .fi
 507 .in
 508
 509 And the last example is about reading netlink message.
 510
 511 .in +4n
 512 .nf
 513 int len;
 514 char buf[4096];
 515 struct iovec iov = { buf, sizeof(buf) };
 516 struct sockaddr_nl sa;
 517 struct msghdr msg;
 518 struct nlmsghdr *nh;
 519
 520 msg = { &sa, sizeof(sa), &iov, 1, NULL, 0, 0 };
 521 len = recvmsg(fd, &msg, 0);
 522
 523 for (nh = (struct nlmsghdr *) buf; NLMSG_OK (nh, len);
 524      nh = NLMSG_NEXT (nh, len)) {
 525     /* The end of multipart message */
 526     if (nh\->nlmsg_type == NLMSG_DONE)
 527         return;
 528
 529     if (nh\->nlmsg_type == NLMSG_ERROR)
 530         /* Do some error handling */
 531     ...
 532
 533     /* Continue with parsing payload */
 534     ...
 535 }
 536 .fi
 537 .in
 538 .SH SEE ALSO
 539 .BR cmsg (3),
 540 .BR netlink (3),
 541 .BR capabilities (7),
 542 .BR rtnetlink (7)
 543
 544 .UR ftp://ftp.inr.ac.ru\:/ip-routing\:/iproute2*
 545 information about libnetlink
 546 .UE
 547
 548 .UR http://people.suug.ch\:/~tgr\:/libnl/
 549 information about libnl
 550 .UE
 551
 552 RFC 3549 "Linux Netlink as an IP Services Protocol"