]> git.proxmox.com Git - mirror_frr.git/blob - zebra/kernel_netlink.c
Merge pull request #837 from qlyoung/fix-eigrp-routemap-cli
[mirror_frr.git] / zebra / kernel_netlink.c
1 /* Kernel communication using netlink interface.
2 * Copyright (C) 1999 Kunihiro Ishiguro
3 *
4 * This file is part of GNU Zebra.
5 *
6 * GNU Zebra is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License as published by the
8 * Free Software Foundation; either version 2, or (at your option) any
9 * later version.
10 *
11 * GNU Zebra is distributed in the hope that it will be useful, but
12 * WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License along
17 * with this program; see the file COPYING; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 #include <zebra.h>
22
23 #include "linklist.h"
24 #include "if.h"
25 #include "log.h"
26 #include "prefix.h"
27 #include "connected.h"
28 #include "table.h"
29 #include "memory.h"
30 #include "zebra_memory.h"
31 #include "rib.h"
32 #include "thread.h"
33 #include "privs.h"
34 #include "nexthop.h"
35 #include "vrf.h"
36 #include "mpls.h"
37
38 #include "zebra/zserv.h"
39 #include "zebra/zebra_ns.h"
40 #include "zebra/zebra_vrf.h"
41 #include "zebra/debug.h"
42 #include "zebra/kernel_netlink.h"
43 #include "zebra/rt_netlink.h"
44 #include "zebra/if_netlink.h"
45
46 #ifndef SO_RCVBUFFORCE
47 #define SO_RCVBUFFORCE (33)
48 #endif
49
50 /* Hack for GNU libc version 2. */
51 #ifndef MSG_TRUNC
52 #define MSG_TRUNC 0x20
53 #endif /* MSG_TRUNC */
54
55 #ifndef NLMSG_TAIL
56 #define NLMSG_TAIL(nmsg) \
57 ((struct rtattr *)(((u_char *)(nmsg)) + NLMSG_ALIGN((nmsg)->nlmsg_len)))
58 #endif
59
60 #ifndef RTA_TAIL
61 #define RTA_TAIL(rta) \
62 ((struct rtattr *)(((u_char *)(rta)) + RTA_ALIGN((rta)->rta_len)))
63 #endif
64
65 #ifndef RTNL_FAMILY_IP6MR
66 #define RTNL_FAMILY_IP6MR 129
67 #endif
68
69 #ifndef RTPROT_MROUTED
70 #define RTPROT_MROUTED 17
71 #endif
72
73 static const struct message nlmsg_str[] = {{RTM_NEWROUTE, "RTM_NEWROUTE"},
74 {RTM_DELROUTE, "RTM_DELROUTE"},
75 {RTM_GETROUTE, "RTM_GETROUTE"},
76 {RTM_NEWLINK, "RTM_NEWLINK"},
77 {RTM_DELLINK, "RTM_DELLINK"},
78 {RTM_GETLINK, "RTM_GETLINK"},
79 {RTM_NEWADDR, "RTM_NEWADDR"},
80 {RTM_DELADDR, "RTM_DELADDR"},
81 {RTM_GETADDR, "RTM_GETADDR"},
82 {RTM_NEWNEIGH, "RTM_NEWNEIGH"},
83 {RTM_DELNEIGH, "RTM_DELNEIGH"},
84 {RTM_GETNEIGH, "RTM_GETNEIGH"},
85 {0}};
86
87 static const struct message rtproto_str[] = {
88 {RTPROT_REDIRECT, "redirect"},
89 {RTPROT_KERNEL, "kernel"},
90 {RTPROT_BOOT, "boot"},
91 {RTPROT_STATIC, "static"},
92 {RTPROT_GATED, "GateD"},
93 {RTPROT_RA, "router advertisement"},
94 {RTPROT_MRT, "MRT"},
95 {RTPROT_ZEBRA, "Zebra"},
96 #ifdef RTPROT_BIRD
97 {RTPROT_BIRD, "BIRD"},
98 #endif /* RTPROT_BIRD */
99 {RTPROT_MROUTED, "mroute"},
100 {RTPROT_BGP, "BGP"},
101 {RTPROT_OSPF, "OSPF"},
102 {RTPROT_ISIS, "IS-IS"},
103 {RTPROT_RIP, "RIP"},
104 {RTPROT_RIPNG, "RIPNG"},
105 {0}};
106
107 static const struct message family_str[] = {{AF_INET, "ipv4"},
108 {AF_INET6, "ipv6"},
109 {AF_BRIDGE, "bridge"},
110 {RTNL_FAMILY_IPMR, "ipv4MR"},
111 {RTNL_FAMILY_IP6MR, "ipv6MR"},
112 {0}};
113
114 static const struct message rttype_str[] = {{RTN_UNICAST, "unicast"},
115 {RTN_MULTICAST, "multicast"},
116 {0}};
117
118 extern struct thread_master *master;
119 extern u_int32_t nl_rcvbufsize;
120
121 extern struct zebra_privs_t zserv_privs;
122
123 int netlink_talk_filter(struct sockaddr_nl *snl, struct nlmsghdr *h,
124 ns_id_t ns_id, int startup)
125 {
126 zlog_warn("netlink_talk: ignoring message type 0x%04x NS %u",
127 h->nlmsg_type, ns_id);
128 return 0;
129 }
130
131 static int netlink_recvbuf(struct nlsock *nl, uint32_t newsize)
132 {
133 u_int32_t oldsize;
134 socklen_t newlen = sizeof(newsize);
135 socklen_t oldlen = sizeof(oldsize);
136 int ret;
137
138 ret = getsockopt(nl->sock, SOL_SOCKET, SO_RCVBUF, &oldsize, &oldlen);
139 if (ret < 0) {
140 zlog_err("Can't get %s receive buffer size: %s", nl->name,
141 safe_strerror(errno));
142 return -1;
143 }
144
145 /* Try force option (linux >= 2.6.14) and fall back to normal set */
146 if (zserv_privs.change(ZPRIVS_RAISE))
147 zlog_err("routing_socket: Can't raise privileges");
148 ret = setsockopt(nl->sock, SOL_SOCKET, SO_RCVBUFFORCE, &nl_rcvbufsize,
149 sizeof(nl_rcvbufsize));
150 if (zserv_privs.change(ZPRIVS_LOWER))
151 zlog_err("routing_socket: Can't lower privileges");
152 if (ret < 0)
153 ret = setsockopt(nl->sock, SOL_SOCKET, SO_RCVBUF,
154 &nl_rcvbufsize, sizeof(nl_rcvbufsize));
155 if (ret < 0) {
156 zlog_err("Can't set %s receive buffer size: %s", nl->name,
157 safe_strerror(errno));
158 return -1;
159 }
160
161 ret = getsockopt(nl->sock, SOL_SOCKET, SO_RCVBUF, &newsize, &newlen);
162 if (ret < 0) {
163 zlog_err("Can't get %s receive buffer size: %s", nl->name,
164 safe_strerror(errno));
165 return -1;
166 }
167
168 zlog_info("Setting netlink socket receive buffer size: %u -> %u",
169 oldsize, newsize);
170 return 0;
171 }
172
173 /* Make socket for Linux netlink interface. */
174 static int netlink_socket(struct nlsock *nl, unsigned long groups,
175 ns_id_t ns_id)
176 {
177 int ret;
178 struct sockaddr_nl snl;
179 int sock;
180 int namelen;
181 int save_errno;
182
183 if (zserv_privs.change(ZPRIVS_RAISE)) {
184 zlog_err("Can't raise privileges");
185 return -1;
186 }
187
188 sock = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
189 if (sock < 0) {
190 zlog_err("Can't open %s socket: %s", nl->name,
191 safe_strerror(errno));
192 return -1;
193 }
194
195 memset(&snl, 0, sizeof snl);
196 snl.nl_family = AF_NETLINK;
197 snl.nl_groups = groups;
198
199 /* Bind the socket to the netlink structure for anything. */
200 ret = bind(sock, (struct sockaddr *)&snl, sizeof snl);
201 save_errno = errno;
202 if (zserv_privs.change(ZPRIVS_LOWER))
203 zlog_err("Can't lower privileges");
204
205 if (ret < 0) {
206 zlog_err("Can't bind %s socket to group 0x%x: %s", nl->name,
207 snl.nl_groups, safe_strerror(save_errno));
208 close(sock);
209 return -1;
210 }
211
212 /* multiple netlink sockets will have different nl_pid */
213 namelen = sizeof snl;
214 ret = getsockname(sock, (struct sockaddr *)&snl, (socklen_t *)&namelen);
215 if (ret < 0 || namelen != sizeof snl) {
216 zlog_err("Can't get %s socket name: %s", nl->name,
217 safe_strerror(errno));
218 close(sock);
219 return -1;
220 }
221
222 nl->snl = snl;
223 nl->sock = sock;
224 return ret;
225 }
226
227 static int netlink_information_fetch(struct sockaddr_nl *snl,
228 struct nlmsghdr *h, ns_id_t ns_id,
229 int startup)
230 {
231 /* JF: Ignore messages that aren't from the kernel */
232 if (snl->nl_pid != 0) {
233 zlog_err("Ignoring message from pid %u", snl->nl_pid);
234 return 0;
235 }
236
237 switch (h->nlmsg_type) {
238 case RTM_NEWROUTE:
239 return netlink_route_change(snl, h, ns_id, startup);
240 break;
241 case RTM_DELROUTE:
242 return netlink_route_change(snl, h, ns_id, startup);
243 break;
244 case RTM_NEWLINK:
245 return netlink_link_change(snl, h, ns_id, startup);
246 break;
247 case RTM_DELLINK:
248 return netlink_link_change(snl, h, ns_id, startup);
249 break;
250 case RTM_NEWADDR:
251 return netlink_interface_addr(snl, h, ns_id, startup);
252 break;
253 case RTM_DELADDR:
254 return netlink_interface_addr(snl, h, ns_id, startup);
255 break;
256 case RTM_NEWNEIGH:
257 return netlink_neigh_change(snl, h, ns_id);
258 break;
259 case RTM_DELNEIGH:
260 return netlink_neigh_change(snl, h, ns_id);
261 break;
262 default:
263 zlog_warn("Unknown netlink nlmsg_type %d vrf %u\n",
264 h->nlmsg_type, ns_id);
265 break;
266 }
267 return 0;
268 }
269
270 static int kernel_read(struct thread *thread)
271 {
272 struct zebra_ns *zns = (struct zebra_ns *)THREAD_ARG(thread);
273 netlink_parse_info(netlink_information_fetch, &zns->netlink, zns, 5, 0);
274 zns->t_netlink = NULL;
275 thread_add_read(zebrad.master, kernel_read, zns, zns->netlink.sock,
276 &zns->t_netlink);
277
278 return 0;
279 }
280
281 /* Filter out messages from self that occur on listener socket,
282 * caused by our actions on the command socket
283 */
284 static void netlink_install_filter(int sock, __u32 pid)
285 {
286 struct sock_filter filter[] = {
287 /* 0: ldh [4] */
288 BPF_STMT(BPF_LD | BPF_ABS | BPF_H,
289 offsetof(struct nlmsghdr, nlmsg_type)),
290 /* 1: jeq 0x18 jt 5 jf next */
291 BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htons(RTM_NEWROUTE), 3, 0),
292 /* 2: jeq 0x19 jt 5 jf next */
293 BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htons(RTM_DELROUTE), 2, 0),
294 /* 3: jeq 0x19 jt 5 jf next */
295 BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htons(RTM_NEWNEIGH), 1, 0),
296 /* 4: jeq 0x19 jt 5 jf 8 */
297 BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htons(RTM_DELNEIGH), 0, 3),
298 /* 5: ldw [12] */
299 BPF_STMT(BPF_LD | BPF_ABS | BPF_W,
300 offsetof(struct nlmsghdr, nlmsg_pid)),
301 /* 6: jeq XX jt 7 jf 8 */
302 BPF_JUMP(BPF_JMP | BPF_JEQ | BPF_K, htonl(pid), 0, 1),
303 /* 7: ret 0 (skip) */
304 BPF_STMT(BPF_RET | BPF_K, 0),
305 /* 8: ret 0xffff (keep) */
306 BPF_STMT(BPF_RET | BPF_K, 0xffff),
307 };
308
309 struct sock_fprog prog = {
310 .len = array_size(filter), .filter = filter,
311 };
312
313 if (setsockopt(sock, SOL_SOCKET, SO_ATTACH_FILTER, &prog, sizeof(prog))
314 < 0)
315 zlog_warn("Can't install socket filter: %s\n",
316 safe_strerror(errno));
317 }
318
319 void netlink_parse_rtattr(struct rtattr **tb, int max, struct rtattr *rta,
320 int len)
321 {
322 while (RTA_OK(rta, len)) {
323 if (rta->rta_type <= max)
324 tb[rta->rta_type] = rta;
325 rta = RTA_NEXT(rta, len);
326 }
327 }
328
329 int addattr_l(struct nlmsghdr *n, unsigned int maxlen, int type, void *data,
330 unsigned int alen)
331 {
332 int len;
333 struct rtattr *rta;
334
335 len = RTA_LENGTH(alen);
336
337 if (NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(len) > maxlen)
338 return -1;
339
340 rta = (struct rtattr *)(((char *)n) + NLMSG_ALIGN(n->nlmsg_len));
341 rta->rta_type = type;
342 rta->rta_len = len;
343
344 if (data)
345 memcpy(RTA_DATA(rta), data, alen);
346 else
347 assert(alen == 0);
348
349 n->nlmsg_len = NLMSG_ALIGN(n->nlmsg_len) + RTA_ALIGN(len);
350
351 return 0;
352 }
353
354 int rta_addattr_l(struct rtattr *rta, unsigned int maxlen, int type, void *data,
355 unsigned int alen)
356 {
357 unsigned int len;
358 struct rtattr *subrta;
359
360 len = RTA_LENGTH(alen);
361
362 if (RTA_ALIGN(rta->rta_len) + RTA_ALIGN(len) > maxlen)
363 return -1;
364
365 subrta = (struct rtattr *)(((char *)rta) + RTA_ALIGN(rta->rta_len));
366 subrta->rta_type = type;
367 subrta->rta_len = len;
368
369 if (data)
370 memcpy(RTA_DATA(subrta), data, alen);
371 else
372 assert(alen == 0);
373
374 rta->rta_len = NLMSG_ALIGN(rta->rta_len) + RTA_ALIGN(len);
375
376 return 0;
377 }
378
379 int addattr16(struct nlmsghdr *n, unsigned int maxlen, int type, u_int16_t data)
380 {
381 return addattr_l(n, maxlen, type, &data, sizeof(u_int16_t));
382 }
383
384 int addattr32(struct nlmsghdr *n, unsigned int maxlen, int type, int data)
385 {
386 return addattr_l(n, maxlen, type, &data, sizeof(u_int32_t));
387 }
388
389 struct rtattr *addattr_nest(struct nlmsghdr *n, int maxlen, int type)
390 {
391 struct rtattr *nest = NLMSG_TAIL(n);
392
393 addattr_l(n, maxlen, type, NULL, 0);
394 return nest;
395 }
396
397 int addattr_nest_end(struct nlmsghdr *n, struct rtattr *nest)
398 {
399 nest->rta_len = (u_char *)NLMSG_TAIL(n) - (u_char *)nest;
400 return n->nlmsg_len;
401 }
402
403 struct rtattr *rta_nest(struct rtattr *rta, int maxlen, int type)
404 {
405 struct rtattr *nest = RTA_TAIL(rta);
406
407 rta_addattr_l(rta, maxlen, type, NULL, 0);
408 return nest;
409 }
410
411 int rta_nest_end(struct rtattr *rta, struct rtattr *nest)
412 {
413 nest->rta_len = (u_char *)RTA_TAIL(rta) - (u_char *)nest;
414 return rta->rta_len;
415 }
416
417 const char *nl_msg_type_to_str(uint16_t msg_type)
418 {
419 return lookup_msg(nlmsg_str, msg_type, "");
420 }
421
422 const char *nl_rtproto_to_str(u_char rtproto)
423 {
424 return lookup_msg(rtproto_str, rtproto, "");
425 }
426
427 const char *nl_family_to_str(u_char family)
428 {
429 return lookup_msg(family_str, family, "");
430 }
431
432 const char *nl_rttype_to_str(u_char rttype)
433 {
434 return lookup_msg(rttype_str, rttype, "");
435 }
436
437 /*
438 * netlink_parse_info
439 *
440 * Receive message from netlink interface and pass those information
441 * to the given function.
442 *
443 * filter -> Function to call to read the results
444 * nl -> netlink socket information
445 * zns -> The zebra namespace data
446 * count -> How many we should read in, 0 means as much as possible
447 * startup -> Are we reading in under startup conditions? passed to
448 * the filter.
449 */
450 int netlink_parse_info(int (*filter)(struct sockaddr_nl *, struct nlmsghdr *,
451 ns_id_t, int),
452 struct nlsock *nl, struct zebra_ns *zns, int count,
453 int startup)
454 {
455 int status;
456 int ret = 0;
457 int error;
458 int read_in = 0;
459
460 while (1) {
461 char buf[NL_PKT_BUF_SIZE];
462 struct iovec iov = {.iov_base = buf, .iov_len = sizeof buf};
463 struct sockaddr_nl snl;
464 struct msghdr msg = {.msg_name = (void *)&snl,
465 .msg_namelen = sizeof snl,
466 .msg_iov = &iov,
467 .msg_iovlen = 1};
468 struct nlmsghdr *h;
469
470 if (count && read_in >= count)
471 return 0;
472
473 status = recvmsg(nl->sock, &msg, 0);
474 if (status < 0) {
475 if (errno == EINTR)
476 continue;
477 if (errno == EWOULDBLOCK || errno == EAGAIN)
478 break;
479 zlog_err("%s recvmsg overrun: %s", nl->name,
480 safe_strerror(errno));
481 /*
482 * In this case we are screwed.
483 * There is no good way to
484 * recover zebra at this point.
485 */
486 exit(-1);
487 continue;
488 }
489
490 if (status == 0) {
491 zlog_err("%s EOF", nl->name);
492 return -1;
493 }
494
495 if (msg.msg_namelen != sizeof snl) {
496 zlog_err("%s sender address length error: length %d",
497 nl->name, msg.msg_namelen);
498 return -1;
499 }
500
501 if (IS_ZEBRA_DEBUG_KERNEL_MSGDUMP_RECV) {
502 zlog_debug("%s: << netlink message dump [recv]",
503 __func__);
504 zlog_hexdump(buf, status);
505 }
506
507 read_in++;
508 for (h = (struct nlmsghdr *)buf;
509 NLMSG_OK(h, (unsigned int)status);
510 h = NLMSG_NEXT(h, status)) {
511 /* Finish of reading. */
512 if (h->nlmsg_type == NLMSG_DONE)
513 return ret;
514
515 /* Error handling. */
516 if (h->nlmsg_type == NLMSG_ERROR) {
517 struct nlmsgerr *err =
518 (struct nlmsgerr *)NLMSG_DATA(h);
519 int errnum = err->error;
520 int msg_type = err->msg.nlmsg_type;
521
522 /* If the error field is zero, then this is an
523 * ACK */
524 if (err->error == 0) {
525 if (IS_ZEBRA_DEBUG_KERNEL) {
526 zlog_debug(
527 "%s: %s ACK: type=%s(%u), seq=%u, pid=%u",
528 __FUNCTION__, nl->name,
529 nl_msg_type_to_str(
530 err->msg.nlmsg_type),
531 err->msg.nlmsg_type,
532 err->msg.nlmsg_seq,
533 err->msg.nlmsg_pid);
534 }
535
536 /* return if not a multipart message,
537 * otherwise continue */
538 if (!(h->nlmsg_flags & NLM_F_MULTI))
539 return 0;
540 continue;
541 }
542
543 if (h->nlmsg_len
544 < NLMSG_LENGTH(sizeof(struct nlmsgerr))) {
545 zlog_err("%s error: message truncated",
546 nl->name);
547 return -1;
548 }
549
550 /* Deal with errors that occur because of races
551 * in link handling */
552 if (nl == &zns->netlink_cmd
553 && ((msg_type == RTM_DELROUTE
554 && (-errnum == ENODEV
555 || -errnum == ESRCH))
556 || (msg_type == RTM_NEWROUTE
557 && (-errnum == ENETDOWN
558 || -errnum == EEXIST)))) {
559 if (IS_ZEBRA_DEBUG_KERNEL)
560 zlog_debug(
561 "%s: error: %s type=%s(%u), seq=%u, pid=%u",
562 nl->name,
563 safe_strerror(-errnum),
564 nl_msg_type_to_str(
565 msg_type),
566 msg_type,
567 err->msg.nlmsg_seq,
568 err->msg.nlmsg_pid);
569 return 0;
570 }
571
572 /* We see RTM_DELNEIGH when shutting down an
573 * interface with an IPv4
574 * link-local. The kernel should have already
575 * deleted the neighbor
576 * so do not log these as an error.
577 */
578 if (msg_type == RTM_DELNEIGH
579 || (nl == &zns->netlink_cmd
580 && msg_type == RTM_NEWROUTE
581 && (-errnum == ESRCH
582 || -errnum == ENETUNREACH))) {
583 /* This is known to happen in some
584 * situations, don't log
585 * as error.
586 */
587 if (IS_ZEBRA_DEBUG_KERNEL)
588 zlog_debug(
589 "%s error: %s, type=%s(%u), seq=%u, pid=%u",
590 nl->name,
591 safe_strerror(-errnum),
592 nl_msg_type_to_str(
593 msg_type),
594 msg_type,
595 err->msg.nlmsg_seq,
596 err->msg.nlmsg_pid);
597 } else
598 zlog_err(
599 "%s error: %s, type=%s(%u), seq=%u, pid=%u",
600 nl->name,
601 safe_strerror(-errnum),
602 nl_msg_type_to_str(msg_type),
603 msg_type, err->msg.nlmsg_seq,
604 err->msg.nlmsg_pid);
605
606 return -1;
607 }
608
609 /* OK we got netlink message. */
610 if (IS_ZEBRA_DEBUG_KERNEL)
611 zlog_debug(
612 "netlink_parse_info: %s type %s(%u), len=%d, seq=%u, pid=%u",
613 nl->name,
614 nl_msg_type_to_str(h->nlmsg_type),
615 h->nlmsg_type, h->nlmsg_len,
616 h->nlmsg_seq, h->nlmsg_pid);
617
618 /* skip unsolicited messages originating from command
619 * socket
620 * linux sets the originators port-id for {NEW|DEL}ADDR
621 * messages,
622 * so this has to be checked here. */
623 if (nl != &zns->netlink_cmd
624 && h->nlmsg_pid == zns->netlink_cmd.snl.nl_pid
625 && (h->nlmsg_type != RTM_NEWADDR
626 && h->nlmsg_type != RTM_DELADDR)) {
627 if (IS_ZEBRA_DEBUG_KERNEL)
628 zlog_debug(
629 "netlink_parse_info: %s packet comes from %s",
630 zns->netlink_cmd.name,
631 nl->name);
632 continue;
633 }
634
635 error = (*filter)(&snl, h, zns->ns_id, startup);
636 if (error < 0) {
637 zlog_err("%s filter function error", nl->name);
638 ret = error;
639 }
640 }
641
642 /* After error care. */
643 if (msg.msg_flags & MSG_TRUNC) {
644 zlog_err("%s error: message truncated", nl->name);
645 continue;
646 }
647 if (status) {
648 zlog_err("%s error: data remnant size %d", nl->name,
649 status);
650 return -1;
651 }
652 }
653 return ret;
654 }
655
656 /*
657 * netlink_talk
658 *
659 * sendmsg() to netlink socket then recvmsg().
660 * Calls netlink_parse_info to parse returned data
661 *
662 * filter -> The filter to read final results from kernel
663 * nlmsghdr -> The data to send to the kernel
664 * nl -> The netlink socket information
665 * zns -> The zebra namespace information
666 * startup -> Are we reading in under startup conditions
667 * This is passed through eventually to filter.
668 */
669 int netlink_talk(int (*filter)(struct sockaddr_nl *, struct nlmsghdr *, ns_id_t,
670 int startup),
671 struct nlmsghdr *n, struct nlsock *nl, struct zebra_ns *zns,
672 int startup)
673 {
674 int status;
675 struct sockaddr_nl snl;
676 struct iovec iov = {.iov_base = (void *)n, .iov_len = n->nlmsg_len};
677 struct msghdr msg = {
678 .msg_name = (void *)&snl,
679 .msg_namelen = sizeof snl,
680 .msg_iov = &iov,
681 .msg_iovlen = 1,
682 };
683 int save_errno;
684
685 memset(&snl, 0, sizeof snl);
686 snl.nl_family = AF_NETLINK;
687
688 n->nlmsg_seq = ++nl->seq;
689 n->nlmsg_pid = nl->snl.nl_pid;
690
691 /* Request an acknowledgement by setting NLM_F_ACK */
692 n->nlmsg_flags |= NLM_F_ACK;
693
694 if (IS_ZEBRA_DEBUG_KERNEL)
695 zlog_debug(
696 "netlink_talk: %s type %s(%u), len=%d seq=%u flags 0x%x",
697 nl->name, nl_msg_type_to_str(n->nlmsg_type),
698 n->nlmsg_type, n->nlmsg_len, n->nlmsg_seq,
699 n->nlmsg_flags);
700
701 /* Send message to netlink interface. */
702 if (zserv_privs.change(ZPRIVS_RAISE))
703 zlog_err("Can't raise privileges");
704 status = sendmsg(nl->sock, &msg, 0);
705 save_errno = errno;
706 if (zserv_privs.change(ZPRIVS_LOWER))
707 zlog_err("Can't lower privileges");
708
709 if (IS_ZEBRA_DEBUG_KERNEL_MSGDUMP_SEND) {
710 zlog_debug("%s: >> netlink message dump [sent]", __func__);
711 zlog_hexdump(n, n->nlmsg_len);
712 }
713
714 if (status < 0) {
715 zlog_err("netlink_talk sendmsg() error: %s",
716 safe_strerror(save_errno));
717 return -1;
718 }
719
720
721 /*
722 * Get reply from netlink socket.
723 * The reply should either be an acknowlegement or an error.
724 */
725 return netlink_parse_info(filter, nl, zns, 0, startup);
726 }
727
728 /* Issue request message to kernel via netlink socket. GET messages
729 * are issued through this interface.
730 */
731 int netlink_request(struct nlsock *nl, struct nlmsghdr *n)
732 {
733 int ret;
734 struct sockaddr_nl snl;
735 int save_errno;
736
737 /* Check netlink socket. */
738 if (nl->sock < 0) {
739 zlog_err("%s socket isn't active.", nl->name);
740 return -1;
741 }
742
743 /* Fill common fields for all requests. */
744 n->nlmsg_flags = NLM_F_ROOT | NLM_F_MATCH | NLM_F_REQUEST;
745 n->nlmsg_pid = nl->snl.nl_pid;
746 n->nlmsg_seq = ++nl->seq;
747
748 memset(&snl, 0, sizeof snl);
749 snl.nl_family = AF_NETLINK;
750
751 /* Raise capabilities and send message, then lower capabilities. */
752 if (zserv_privs.change(ZPRIVS_RAISE)) {
753 zlog_err("Can't raise privileges");
754 return -1;
755 }
756
757 ret = sendto(nl->sock, (void *)n, n->nlmsg_len, 0,
758 (struct sockaddr *)&snl, sizeof snl);
759 save_errno = errno;
760
761 if (zserv_privs.change(ZPRIVS_LOWER))
762 zlog_err("Can't lower privileges");
763
764 if (ret < 0) {
765 zlog_err("%s sendto failed: %s", nl->name,
766 safe_strerror(save_errno));
767 return -1;
768 }
769
770 return 0;
771 }
772
773 /* Exported interface function. This function simply calls
774 netlink_socket (). */
775 void kernel_init(struct zebra_ns *zns)
776 {
777 unsigned long groups;
778
779 /* Initialize netlink sockets */
780 groups = RTMGRP_LINK | RTMGRP_IPV4_ROUTE | RTMGRP_IPV4_IFADDR
781 | RTMGRP_IPV6_ROUTE | RTMGRP_IPV6_IFADDR | RTMGRP_IPV4_MROUTE
782 | RTMGRP_NEIGH;
783
784 snprintf(zns->netlink.name, sizeof(zns->netlink.name),
785 "netlink-listen (NS %u)", zns->ns_id);
786 zns->netlink.sock = -1;
787 netlink_socket(&zns->netlink, groups, zns->ns_id);
788
789 snprintf(zns->netlink_cmd.name, sizeof(zns->netlink_cmd.name),
790 "netlink-cmd (NS %u)", zns->ns_id);
791 zns->netlink_cmd.sock = -1;
792 netlink_socket(&zns->netlink_cmd, 0, zns->ns_id);
793
794 /* Register kernel socket. */
795 if (zns->netlink.sock > 0) {
796 /* Only want non-blocking on the netlink event socket */
797 if (fcntl(zns->netlink.sock, F_SETFL, O_NONBLOCK) < 0)
798 zlog_err("Can't set %s socket flags: %s",
799 zns->netlink.name, safe_strerror(errno));
800
801 /* Set receive buffer size if it's set from command line */
802 if (nl_rcvbufsize)
803 netlink_recvbuf(&zns->netlink, nl_rcvbufsize);
804
805 netlink_install_filter(zns->netlink.sock,
806 zns->netlink_cmd.snl.nl_pid);
807 zns->t_netlink = NULL;
808 thread_add_read(zebrad.master, kernel_read, zns,
809 zns->netlink.sock, &zns->t_netlink);
810 }
811
812 rt_netlink_init();
813 }
814
815 void kernel_terminate(struct zebra_ns *zns)
816 {
817 THREAD_READ_OFF(zns->t_netlink);
818
819 if (zns->netlink.sock >= 0) {
820 close(zns->netlink.sock);
821 zns->netlink.sock = -1;
822 }
823
824 if (zns->netlink_cmd.sock >= 0) {
825 close(zns->netlink_cmd.sock);
826 zns->netlink_cmd.sock = -1;
827 }
828 }