1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause
3 *
4 * Copyright (c) 2022 Alexander V. Chernikov <melifaro@FreeBSD.org>
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25 * SUCH DAMAGE.
26 */
27
28 #include <sys/cdefs.h>
29 #include "opt_inet.h"
30 #include "opt_inet6.h"
31 #include <sys/types.h>
32 #include <sys/eventhandler.h>
33 #include <sys/kernel.h>
34 #include <sys/jail.h>
35 #include <sys/malloc.h>
36 #include <sys/socket.h>
37 #include <sys/sockio.h>
38 #include <sys/syslog.h>
39
40 #include <net/if.h>
41 #include <net/if_dl.h>
42 #include <net/if_media.h>
43 #include <net/if_var.h>
44 #include <net/if_clone.h>
45 #include <net/route.h>
46 #include <net/route/nhop.h>
47 #include <net/route/route_ctl.h>
48 #include <netinet/in_var.h>
49 #include <netinet6/in6_var.h>
50 #include <netinet6/scope6_var.h> /* scope deembedding */
51 #include <netlink/netlink.h>
52 #include <netlink/netlink_ctl.h>
53 #include <netlink/netlink_route.h>
54 #include <netlink/route/route_var.h>
55
56 #define DEBUG_MOD_NAME nl_iface
57 #define DEBUG_MAX_LEVEL LOG_DEBUG3
58 #include <netlink/netlink_debug.h>
59 _DECLARE_DEBUG(LOG_INFO);
60
61 struct netlink_walkargs {
62 struct nl_writer *nw;
63 struct nlmsghdr hdr;
64 struct nlpcb *so;
65 struct ucred *cred;
66 uint32_t fibnum;
67 int family;
68 int error;
69 int count;
70 int dumped;
71 };
72
73 static eventhandler_tag ifdetach_event, ifattach_event, iflink_event, ifaddr_event;
74
75 static SLIST_HEAD(, nl_cloner) nl_cloners = SLIST_HEAD_INITIALIZER(nl_cloners);
76
77 static struct sx rtnl_cloner_lock;
78 SX_SYSINIT(rtnl_cloner_lock, &rtnl_cloner_lock, "rtnl cloner lock");
79
80 /* These are external hooks for CARP. */
81 extern int (*carp_get_vhid_p)(struct ifaddr *);
82
83 /*
84 * RTM_GETLINK request
85 * sendto(3, {{len=32, type=RTM_GETLINK, flags=NLM_F_REQUEST|NLM_F_DUMP, seq=1641940952, pid=0},
86 * {ifi_family=AF_INET, ifi_type=ARPHRD_NETROM, ifi_index=0, ifi_flags=0, ifi_change=0}}, 32, 0, NULL, 0) = 32
87 *
88 * Reply:
89 * {ifi_family=AF_UNSPEC, ifi_type=ARPHRD_ETHER, ifi_index=if_nametoindex("enp0s31f6"), ifi_flags=IFF_UP|IFF_BROADCAST|IFF_RUNNING|IFF_MULTICAST|IFF_LOWER_UP, ifi_change=0},
90 {{nla_len=10, nla_type=IFLA_ADDRESS}, "\xfe\x54\x00\x52\x3e\x90"}
91
92 [
93 {{nla_len=14, nla_type=IFLA_IFNAME}, "enp0s31f6"},
94 {{nla_len=8, nla_type=IFLA_TXQLEN}, 1000},
95 {{nla_len=5, nla_type=IFLA_OPERSTATE}, 6},
96 {{nla_len=5, nla_type=IFLA_LINKMODE}, 0},
97 {{nla_len=8, nla_type=IFLA_MTU}, 1500},
98 {{nla_len=8, nla_type=IFLA_MIN_MTU}, 68},
99 {{nla_len=8, nla_type=IFLA_MAX_MTU}, 9000},
100 {{nla_len=8, nla_type=IFLA_GROUP}, 0},
101 {{nla_len=8, nla_type=IFLA_PROMISCUITY}, 0},
102 {{nla_len=8, nla_type=IFLA_NUM_TX_QUEUES}, 1},
103 {{nla_len=8, nla_type=IFLA_GSO_MAX_SEGS}, 65535},
104 {{nla_len=8, nla_type=IFLA_GSO_MAX_SIZE}, 65536},
105 {{nla_len=8, nla_type=IFLA_NUM_RX_QUEUES}, 1},
106 {{nla_len=5, nla_type=IFLA_CARRIER}, 1},
107 {{nla_len=13, nla_type=IFLA_QDISC}, "fq_codel"},
108 {{nla_len=8, nla_type=IFLA_CARRIER_CHANGES}, 2},
109 {{nla_len=5, nla_type=IFLA_PROTO_DOWN}, 0},
110 {{nla_len=8, nla_type=IFLA_CARRIER_UP_COUNT}, 1},
111 {{nla_len=8, nla_type=IFLA_CARRIER_DOWN_COUNT}, 1},
112 */
113
114 struct if_state {
115 uint8_t ifla_operstate;
116 uint8_t ifla_carrier;
117 };
118
119 static void
get_operstate_ether(if_t ifp,struct if_state * pstate)120 get_operstate_ether(if_t ifp, struct if_state *pstate)
121 {
122 struct ifmediareq ifmr = {};
123 int error;
124 error = if_ioctl(ifp, SIOCGIFMEDIA, (void *)&ifmr);
125
126 if (error != 0) {
127 NL_LOG(LOG_DEBUG, "error calling SIOCGIFMEDIA on %s: %d",
128 if_name(ifp), error);
129 return;
130 }
131
132 switch (IFM_TYPE(ifmr.ifm_active)) {
133 case IFM_ETHER:
134 if (ifmr.ifm_status & IFM_ACTIVE) {
135 pstate->ifla_carrier = 1;
136 if (if_getflags(ifp) & IFF_MONITOR)
137 pstate->ifla_operstate = IF_OPER_DORMANT;
138 else
139 pstate->ifla_operstate = IF_OPER_UP;
140 } else
141 pstate->ifla_operstate = IF_OPER_DOWN;
142 }
143 }
144
145 static bool
get_stats(struct nl_writer * nw,if_t ifp)146 get_stats(struct nl_writer *nw, if_t ifp)
147 {
148 struct rtnl_link_stats64 *stats;
149
150 int nla_len = sizeof(struct nlattr) + sizeof(*stats);
151 struct nlattr *nla = nlmsg_reserve_data(nw, nla_len, struct nlattr);
152 if (nla == NULL)
153 return (false);
154 nla->nla_type = IFLA_STATS64;
155 nla->nla_len = nla_len;
156 stats = (struct rtnl_link_stats64 *)(nla + 1);
157
158 stats->rx_packets = if_getcounter(ifp, IFCOUNTER_IPACKETS);
159 stats->tx_packets = if_getcounter(ifp, IFCOUNTER_OPACKETS);
160 stats->rx_bytes = if_getcounter(ifp, IFCOUNTER_IBYTES);
161 stats->tx_bytes = if_getcounter(ifp, IFCOUNTER_OBYTES);
162 stats->rx_errors = if_getcounter(ifp, IFCOUNTER_IERRORS);
163 stats->tx_errors = if_getcounter(ifp, IFCOUNTER_OERRORS);
164 stats->rx_dropped = if_getcounter(ifp, IFCOUNTER_IQDROPS);
165 stats->tx_dropped = if_getcounter(ifp, IFCOUNTER_OQDROPS);
166 stats->multicast = if_getcounter(ifp, IFCOUNTER_IMCASTS);
167 stats->rx_nohandler = if_getcounter(ifp, IFCOUNTER_NOPROTO);
168
169 return (true);
170 }
171
172 static void
get_operstate(if_t ifp,struct if_state * pstate)173 get_operstate(if_t ifp, struct if_state *pstate)
174 {
175 pstate->ifla_operstate = IF_OPER_UNKNOWN;
176 pstate->ifla_carrier = 0; /* no carrier */
177
178 switch (if_gettype(ifp)) {
179 case IFT_ETHER:
180 case IFT_L2VLAN:
181 get_operstate_ether(ifp, pstate);
182 break;
183 default:
184 /* Map admin state to the operstate */
185 if (if_getflags(ifp) & IFF_UP) {
186 pstate->ifla_operstate = IF_OPER_UP;
187 pstate->ifla_carrier = 1;
188 } else
189 pstate->ifla_operstate = IF_OPER_DOWN;
190 break;
191 }
192 }
193
194 static void
get_hwaddr(struct nl_writer * nw,if_t ifp)195 get_hwaddr(struct nl_writer *nw, if_t ifp)
196 {
197 struct ifreq ifr = {};
198
199 if (if_gethwaddr(ifp, &ifr) == 0) {
200 nlattr_add(nw, IFLAF_ORIG_HWADDR, if_getaddrlen(ifp),
201 ifr.ifr_addr.sa_data);
202 }
203 }
204
205 static unsigned
ifp_flags_to_netlink(const if_t ifp)206 ifp_flags_to_netlink(const if_t ifp)
207 {
208 return (if_getflags(ifp) | if_getdrvflags(ifp));
209 }
210
211 #define LLADDR_CONST(s) ((const void *)((s)->sdl_data + (s)->sdl_nlen))
212 static bool
dump_sa(struct nl_writer * nw,int attr,const struct sockaddr * sa)213 dump_sa(struct nl_writer *nw, int attr, const struct sockaddr *sa)
214 {
215 uint32_t addr_len = 0;
216 const void *addr_data = NULL;
217 #ifdef INET6
218 struct in6_addr addr6;
219 #endif
220
221 if (sa == NULL)
222 return (true);
223
224 switch (sa->sa_family) {
225 #ifdef INET
226 case AF_INET:
227 addr_len = sizeof(struct in_addr);
228 addr_data = &((const struct sockaddr_in *)sa)->sin_addr;
229 break;
230 #endif
231 #ifdef INET6
232 case AF_INET6:
233 in6_splitscope(&((const struct sockaddr_in6 *)sa)->sin6_addr, &addr6, &addr_len);
234 addr_len = sizeof(struct in6_addr);
235 addr_data = &addr6;
236 break;
237 #endif
238 case AF_LINK:
239 addr_len = ((const struct sockaddr_dl *)sa)->sdl_alen;
240 addr_data = LLADDR_CONST((const struct sockaddr_dl *)sa);
241 break;
242 case AF_UNSPEC:
243 /* Ignore empty SAs without warning */
244 return (true);
245 default:
246 NL_LOG(LOG_DEBUG2, "unsupported family: %d, skipping", sa->sa_family);
247 return (true);
248 }
249
250 return (nlattr_add(nw, attr, addr_len, addr_data));
251 }
252
253 static bool
dump_iface_caps(struct nl_writer * nw,struct ifnet * ifp)254 dump_iface_caps(struct nl_writer *nw, struct ifnet *ifp)
255 {
256 int off = nlattr_add_nested(nw, IFLAF_CAPS);
257 uint32_t active_caps[roundup2(IFCAP_B_SIZE, 32) / 32] = {};
258 uint32_t all_caps[roundup2(IFCAP_B_SIZE, 32) / 32] = {};
259
260 MPASS(sizeof(active_caps) >= 8);
261 MPASS(sizeof(all_caps) >= 8);
262
263 if (off == 0)
264 return (false);
265
266 active_caps[0] = (uint32_t)if_getcapabilities(ifp);
267 all_caps[0] = (uint32_t)if_getcapenable(ifp);
268 active_caps[1] = (uint32_t)if_getcapabilities2(ifp);
269 all_caps[1] = (uint32_t)if_getcapenable2(ifp);
270
271 nlattr_add_u32(nw, NLA_BITSET_SIZE, IFCAP_B_SIZE);
272 nlattr_add(nw, NLA_BITSET_MASK, sizeof(all_caps), all_caps);
273 nlattr_add(nw, NLA_BITSET_VALUE, sizeof(active_caps), active_caps);
274
275 nlattr_set_len(nw, off);
276
277 return (true);
278 }
279
280 /*
281 * Dumps interface state, properties and metrics.
282 * @nw: message writer
283 * @ifp: target interface
284 * @hdr: template header
285 * @if_flags_mask: changed if_[drv]_flags bitmask
286 *
287 * This function is called without epoch and MAY sleep.
288 */
289 static bool
dump_iface(struct nl_writer * nw,if_t ifp,const struct nlmsghdr * hdr,int if_flags_mask)290 dump_iface(struct nl_writer *nw, if_t ifp, const struct nlmsghdr *hdr,
291 int if_flags_mask)
292 {
293 struct epoch_tracker et;
294 struct ifinfomsg *ifinfo;
295
296 NL_LOG(LOG_DEBUG3, "dumping interface %s data", if_name(ifp));
297
298 if (!nlmsg_reply(nw, hdr, sizeof(struct ifinfomsg)))
299 goto enomem;
300
301 ifinfo = nlmsg_reserve_object(nw, struct ifinfomsg);
302 ifinfo->ifi_family = AF_UNSPEC;
303 ifinfo->__ifi_pad = 0;
304 ifinfo->ifi_type = if_gettype(ifp);
305 ifinfo->ifi_index = if_getindex(ifp);
306 ifinfo->ifi_flags = ifp_flags_to_netlink(ifp);
307 ifinfo->ifi_change = if_flags_mask;
308
309 struct if_state ifs = {};
310 get_operstate(ifp, &ifs);
311
312 if (ifs.ifla_operstate == IF_OPER_UP)
313 ifinfo->ifi_flags |= IFF_LOWER_UP;
314
315 nlattr_add_string(nw, IFLA_IFNAME, if_name(ifp));
316 nlattr_add_u8(nw, IFLA_OPERSTATE, ifs.ifla_operstate);
317 nlattr_add_u8(nw, IFLA_CARRIER, ifs.ifla_carrier);
318
319 /*
320 nlattr_add_u8(nw, IFLA_PROTO_DOWN, val);
321 nlattr_add_u8(nw, IFLA_LINKMODE, val);
322 */
323 if (if_getaddrlen(ifp) != 0) {
324 struct ifaddr *ifa;
325
326 NET_EPOCH_ENTER(et);
327 ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
328 if (ifa != NULL)
329 dump_sa(nw, IFLA_ADDRESS, ifa->ifa_addr);
330 NET_EPOCH_EXIT(et);
331 }
332
333 if ((if_getbroadcastaddr(ifp) != NULL)) {
334 nlattr_add(nw, IFLA_BROADCAST, if_getaddrlen(ifp),
335 if_getbroadcastaddr(ifp));
336 }
337
338 nlattr_add_u32(nw, IFLA_MTU, if_getmtu(ifp));
339 /*
340 nlattr_add_u32(nw, IFLA_MIN_MTU, 60);
341 nlattr_add_u32(nw, IFLA_MAX_MTU, 9000);
342 nlattr_add_u32(nw, IFLA_GROUP, 0);
343 */
344
345 if (if_getdescr(ifp) != NULL)
346 nlattr_add_string(nw, IFLA_IFALIAS, if_getdescr(ifp));
347
348 /* Store FreeBSD-specific attributes */
349 int off = nlattr_add_nested(nw, IFLA_FREEBSD);
350 if (off != 0) {
351 get_hwaddr(nw, ifp);
352 dump_iface_caps(nw, ifp);
353
354 nlattr_set_len(nw, off);
355 }
356
357 get_stats(nw, ifp);
358
359 uint32_t val = (if_getflags(ifp) & IFF_PROMISC) != 0;
360 nlattr_add_u32(nw, IFLA_PROMISCUITY, val);
361
362 ifc_dump_ifp_nl(ifp, nw);
363
364 if (nlmsg_end(nw))
365 return (true);
366
367 enomem:
368 NL_LOG(LOG_DEBUG, "unable to dump interface %s state (ENOMEM)", if_name(ifp));
369 nlmsg_abort(nw);
370 return (false);
371 }
372
373 static bool
check_ifmsg(void * hdr,struct nl_pstate * npt)374 check_ifmsg(void *hdr, struct nl_pstate *npt)
375 {
376 struct ifinfomsg *ifm = hdr;
377
378 if (ifm->__ifi_pad != 0 || ifm->ifi_type != 0 ||
379 ifm->ifi_flags != 0 || ifm->ifi_change != 0) {
380 nlmsg_report_err_msg(npt,
381 "strict checking: non-zero values in ifinfomsg header");
382 return (false);
383 }
384
385 return (true);
386 }
387
388 #define _IN(_field) offsetof(struct ifinfomsg, _field)
389 #define _OUT(_field) offsetof(struct nl_parsed_link, _field)
390 static const struct nlfield_parser nlf_p_if[] = {
391 { .off_in = _IN(ifi_type), .off_out = _OUT(ifi_type), .cb = nlf_get_u16 },
392 { .off_in = _IN(ifi_index), .off_out = _OUT(ifi_index), .cb = nlf_get_u32 },
393 { .off_in = _IN(ifi_flags), .off_out = _OUT(ifi_flags), .cb = nlf_get_u32 },
394 { .off_in = _IN(ifi_change), .off_out = _OUT(ifi_change), .cb = nlf_get_u32 },
395 };
396
397 static const struct nlattr_parser nla_p_linfo[] = {
398 { .type = IFLA_INFO_KIND, .off = _OUT(ifla_cloner), .cb = nlattr_get_stringn },
399 { .type = IFLA_INFO_DATA, .off = _OUT(ifla_idata), .cb = nlattr_get_nla },
400 };
401 NL_DECLARE_ATTR_PARSER(linfo_parser, nla_p_linfo);
402
403 static const struct nlattr_parser nla_p_if[] = {
404 { .type = IFLA_IFNAME, .off = _OUT(ifla_ifname), .cb = nlattr_get_string },
405 { .type = IFLA_MTU, .off = _OUT(ifla_mtu), .cb = nlattr_get_uint32 },
406 { .type = IFLA_LINK, .off = _OUT(ifla_link), .cb = nlattr_get_uint32 },
407 { .type = IFLA_LINKINFO, .arg = &linfo_parser, .cb = nlattr_get_nested },
408 { .type = IFLA_IFALIAS, .off = _OUT(ifla_ifalias), .cb = nlattr_get_string },
409 { .type = IFLA_GROUP, .off = _OUT(ifla_group), .cb = nlattr_get_string },
410 { .type = IFLA_ALT_IFNAME, .off = _OUT(ifla_ifname), .cb = nlattr_get_string },
411 };
412 #undef _IN
413 #undef _OUT
414 NL_DECLARE_STRICT_PARSER(ifmsg_parser, struct ifinfomsg, check_ifmsg, nlf_p_if, nla_p_if);
415
416 static bool
match_iface(if_t ifp,void * _arg)417 match_iface(if_t ifp, void *_arg)
418 {
419 struct nl_parsed_link *attrs = (struct nl_parsed_link *)_arg;
420
421 if (attrs->ifi_index != 0 && attrs->ifi_index != if_getindex(ifp))
422 return (false);
423 if (attrs->ifi_type != 0 && attrs->ifi_index != if_gettype(ifp))
424 return (false);
425 if (attrs->ifla_ifname != NULL && strcmp(attrs->ifla_ifname, if_name(ifp)))
426 return (false);
427 /* TODO: add group match */
428
429 return (true);
430 }
431
432 static int
dump_cb(if_t ifp,void * _arg)433 dump_cb(if_t ifp, void *_arg)
434 {
435 struct netlink_walkargs *wa = (struct netlink_walkargs *)_arg;
436 if (!dump_iface(wa->nw, ifp, &wa->hdr, 0))
437 return (ENOMEM);
438 return (0);
439 }
440
441 /*
442 * {nlmsg_len=52, nlmsg_type=RTM_GETLINK, nlmsg_flags=NLM_F_REQUEST, nlmsg_seq=1662842818, nlmsg_pid=0},
443 * {ifi_family=AF_PACKET, ifi_type=ARPHRD_NETROM, ifi_index=0, ifi_flags=0, ifi_change=0},
444 * [
445 * [{nla_len=10, nla_type=IFLA_IFNAME}, "vnet9"],
446 * [{nla_len=8, nla_type=IFLA_EXT_MASK}, RTEXT_FILTER_VF]
447 * ]
448 */
449 static int
rtnl_handle_getlink(struct nlmsghdr * hdr,struct nlpcb * nlp,struct nl_pstate * npt)450 rtnl_handle_getlink(struct nlmsghdr *hdr, struct nlpcb *nlp, struct nl_pstate *npt)
451 {
452 struct epoch_tracker et;
453 if_t ifp;
454 int error = 0;
455
456 struct nl_parsed_link attrs = {};
457 error = nl_parse_nlmsg(hdr, &ifmsg_parser, npt, &attrs);
458 if (error != 0)
459 return (error);
460
461 struct netlink_walkargs wa = {
462 .so = nlp,
463 .nw = npt->nw,
464 .hdr.nlmsg_pid = hdr->nlmsg_pid,
465 .hdr.nlmsg_seq = hdr->nlmsg_seq,
466 .hdr.nlmsg_flags = hdr->nlmsg_flags,
467 .hdr.nlmsg_type = NL_RTM_NEWLINK,
468 };
469
470 /* Fast track for an interface w/ explicit name or index match */
471 if ((attrs.ifi_index != 0) || (attrs.ifla_ifname != NULL)) {
472 if (attrs.ifi_index != 0) {
473 NLP_LOG(LOG_DEBUG3, nlp, "fast track -> searching index %u",
474 attrs.ifi_index);
475 NET_EPOCH_ENTER(et);
476 ifp = ifnet_byindex_ref(attrs.ifi_index);
477 NET_EPOCH_EXIT(et);
478 } else {
479 NLP_LOG(LOG_DEBUG3, nlp, "fast track -> searching name %s",
480 attrs.ifla_ifname);
481 ifp = ifunit_ref(attrs.ifla_ifname);
482 }
483
484 if (ifp != NULL) {
485 if (match_iface(ifp, &attrs)) {
486 if (!dump_iface(wa.nw, ifp, &wa.hdr, 0))
487 error = ENOMEM;
488 } else
489 error = ENODEV;
490 if_rele(ifp);
491 } else
492 error = ENODEV;
493 return (error);
494 }
495
496 /* Always treat non-direct-match as a multipart message */
497 wa.hdr.nlmsg_flags |= NLM_F_MULTI;
498
499 /*
500 * Fetching some link properties require performing ioctl's that may be blocking.
501 * Address it by saving referenced pointers of the matching links,
502 * exiting from epoch and going through the list one-by-one.
503 */
504
505 NL_LOG(LOG_DEBUG2, "Start dump");
506 if_foreach_sleep(match_iface, &attrs, dump_cb, &wa);
507 NL_LOG(LOG_DEBUG2, "End dump, iterated %d dumped %d", wa.count, wa.dumped);
508
509 if (!nlmsg_end_dump(wa.nw, error, &wa.hdr)) {
510 NL_LOG(LOG_DEBUG, "Unable to finalize the dump");
511 return (ENOMEM);
512 }
513
514 return (error);
515 }
516
517 /*
518 * sendmsg(3, {msg_name={sa_family=AF_NETLINK, nl_pid=0, nl_groups=00000000}, msg_namelen=12, msg_iov=[{iov_base=[
519 * {nlmsg_len=60, nlmsg_type=RTM_NEWLINK, nlmsg_flags=NLM_F_REQUEST|NLM_F_ACK|NLM_F_EXCL|NLM_F_CREATE, nlmsg_seq=1662715618, nlmsg_pid=0},
520 * {ifi_family=AF_UNSPEC, ifi_type=ARPHRD_NETROM, ifi_index=0, ifi_flags=0, ifi_change=0},
521 * {nla_len=11, nla_type=IFLA_IFNAME}, "dummy0"],
522 * [
523 * {nla_len=16, nla_type=IFLA_LINKINFO},
524 * [
525 * {nla_len=9, nla_type=IFLA_INFO_KIND}, "dummy"...
526 * ]
527 * ]
528 */
529
530 static int
rtnl_handle_dellink(struct nlmsghdr * hdr,struct nlpcb * nlp,struct nl_pstate * npt)531 rtnl_handle_dellink(struct nlmsghdr *hdr, struct nlpcb *nlp, struct nl_pstate *npt)
532 {
533 struct epoch_tracker et;
534 if_t ifp;
535 int error;
536
537 struct nl_parsed_link attrs = {};
538 error = nl_parse_nlmsg(hdr, &ifmsg_parser, npt, &attrs);
539 if (error != 0)
540 return (error);
541
542 NET_EPOCH_ENTER(et);
543 ifp = ifnet_byindex_ref(attrs.ifi_index);
544 NET_EPOCH_EXIT(et);
545 if (ifp == NULL) {
546 NLP_LOG(LOG_DEBUG, nlp, "unable to find interface %u", attrs.ifi_index);
547 return (ENOENT);
548 }
549 NLP_LOG(LOG_DEBUG3, nlp, "mapped ifindex %u to %s", attrs.ifi_index, if_name(ifp));
550
551 sx_xlock(&ifnet_detach_sxlock);
552 error = if_clone_destroy(if_name(ifp));
553 sx_xunlock(&ifnet_detach_sxlock);
554
555 NLP_LOG(LOG_DEBUG2, nlp, "deleting interface %s returned %d", if_name(ifp), error);
556
557 if_rele(ifp);
558 return (error);
559 }
560
561 /*
562 * New link:
563 * type=RTM_NEWLINK, flags=NLM_F_REQUEST|NLM_F_ACK|NLM_F_EXCL|NLM_F_CREATE, seq=1668185590, pid=0},
564 * {ifi_family=AF_UNSPEC, ifi_type=ARPHRD_NETROM, ifi_index=0, ifi_flags=0, ifi_change=0}
565 * [
566 * {{nla_len=8, nla_type=IFLA_MTU}, 123},
567 * {{nla_len=10, nla_type=IFLA_IFNAME}, "vlan1"},
568 * {{nla_len=24, nla_type=IFLA_LINKINFO},
569 * [
570 * {{nla_len=8, nla_type=IFLA_INFO_KIND}, "vlan"...},
571 * {{nla_len=12, nla_type=IFLA_INFO_DATA}, "\x06\x00\x01\x00\x7b\x00\x00\x00"}]}]}
572 *
573 * Update link:
574 * type=RTM_NEWLINK, flags=NLM_F_REQUEST|NLM_F_ACK, seq=1668185923, pid=0},
575 * {ifi_family=AF_UNSPEC, ifi_type=ARPHRD_NETROM, ifi_index=if_nametoindex("lo"), ifi_flags=0, ifi_change=0},
576 * {{nla_len=8, nla_type=IFLA_MTU}, 123}}
577 *
578 *
579 * Check command availability:
580 * type=RTM_NEWLINK, flags=NLM_F_REQUEST|NLM_F_ACK, seq=0, pid=0},
581 * {ifi_family=AF_UNSPEC, ifi_type=ARPHRD_NETROM, ifi_index=0, ifi_flags=0, ifi_change=0}
582 */
583
584
585 static int
create_link(struct nlmsghdr * hdr,struct nl_parsed_link * lattrs,struct nlattr_bmask * bm,struct nlpcb * nlp,struct nl_pstate * npt)586 create_link(struct nlmsghdr *hdr, struct nl_parsed_link *lattrs,
587 struct nlattr_bmask *bm, struct nlpcb *nlp, struct nl_pstate *npt)
588 {
589 if (lattrs->ifla_ifname == NULL || strlen(lattrs->ifla_ifname) == 0) {
590 NLMSG_REPORT_ERR_MSG(npt, "empty IFLA_IFNAME attribute");
591 return (EINVAL);
592 }
593 if (lattrs->ifla_cloner == NULL || strlen(lattrs->ifla_cloner) == 0) {
594 NLMSG_REPORT_ERR_MSG(npt, "empty IFLA_INFO_KIND attribute");
595 return (EINVAL);
596 }
597
598 struct ifc_data_nl ifd = {
599 .flags = IFC_F_CREATE,
600 .lattrs = lattrs,
601 .bm = bm,
602 .npt = npt,
603 };
604 if (ifc_create_ifp_nl(lattrs->ifla_ifname, &ifd) && ifd.error == 0)
605 nl_store_ifp_cookie(npt, ifd.ifp);
606
607 return (ifd.error);
608 }
609
610 static int
modify_link(struct nlmsghdr * hdr,struct nl_parsed_link * lattrs,struct nlattr_bmask * bm,struct nlpcb * nlp,struct nl_pstate * npt)611 modify_link(struct nlmsghdr *hdr, struct nl_parsed_link *lattrs,
612 struct nlattr_bmask *bm, struct nlpcb *nlp, struct nl_pstate *npt)
613 {
614 if_t ifp = NULL;
615 struct epoch_tracker et;
616
617 if (lattrs->ifi_index == 0 && lattrs->ifla_ifname == NULL) {
618 /*
619 * Applications like ip(8) verify RTM_NEWLINK command
620 * existence by calling it with empty arguments. Always
621 * return "innocent" error in that case.
622 */
623 NLMSG_REPORT_ERR_MSG(npt, "empty ifi_index field");
624 return (EPERM);
625 }
626
627 if (lattrs->ifi_index != 0) {
628 NET_EPOCH_ENTER(et);
629 ifp = ifnet_byindex_ref(lattrs->ifi_index);
630 NET_EPOCH_EXIT(et);
631 if (ifp == NULL) {
632 NLMSG_REPORT_ERR_MSG(npt, "unable to find interface #%u",
633 lattrs->ifi_index);
634 return (ENOENT);
635 }
636 }
637
638 if (ifp == NULL && lattrs->ifla_ifname != NULL) {
639 ifp = ifunit_ref(lattrs->ifla_ifname);
640 if (ifp == NULL) {
641 NLMSG_REPORT_ERR_MSG(npt, "unable to find interface %s",
642 lattrs->ifla_ifname);
643 return (ENOENT);
644 }
645 }
646
647 MPASS(ifp != NULL);
648
649 /*
650 * Modification request can address either
651 * 1) cloned interface, in which case we call the cloner-specific
652 * modification routine
653 * or
654 * 2) non-cloned (e.g. "physical") interface, in which case we call
655 * generic modification routine
656 */
657 struct ifc_data_nl ifd = { .lattrs = lattrs, .bm = bm, .npt = npt };
658 if (!ifc_modify_ifp_nl(ifp, &ifd))
659 ifd.error = nl_modify_ifp_generic(ifp, lattrs, bm, npt);
660
661 if_rele(ifp);
662
663 return (ifd.error);
664 }
665
666
667 static int
rtnl_handle_newlink(struct nlmsghdr * hdr,struct nlpcb * nlp,struct nl_pstate * npt)668 rtnl_handle_newlink(struct nlmsghdr *hdr, struct nlpcb *nlp, struct nl_pstate *npt)
669 {
670 struct nlattr_bmask bm;
671 int error;
672
673 struct nl_parsed_link attrs = {};
674 error = nl_parse_nlmsg(hdr, &ifmsg_parser, npt, &attrs);
675 if (error != 0)
676 return (error);
677 nl_get_attrs_bmask_nlmsg(hdr, &ifmsg_parser, &bm);
678
679 if (hdr->nlmsg_flags & NLM_F_CREATE)
680 return (create_link(hdr, &attrs, &bm, nlp, npt));
681 else
682 return (modify_link(hdr, &attrs, &bm, nlp, npt));
683 }
684
685 static void
set_scope6(struct sockaddr * sa,uint32_t ifindex)686 set_scope6(struct sockaddr *sa, uint32_t ifindex)
687 {
688 #ifdef INET6
689 if (sa != NULL && sa->sa_family == AF_INET6) {
690 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)sa;
691
692 if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr))
693 in6_set_unicast_scopeid(&sa6->sin6_addr, ifindex);
694 }
695 #endif
696 }
697
698 static bool
check_sa_family(const struct sockaddr * sa,int family,const char * attr_name,struct nl_pstate * npt)699 check_sa_family(const struct sockaddr *sa, int family, const char *attr_name,
700 struct nl_pstate *npt)
701 {
702 if (sa == NULL || sa->sa_family == family)
703 return (true);
704
705 nlmsg_report_err_msg(npt, "wrong family for %s attribute: %d != %d",
706 attr_name, family, sa->sa_family);
707 return (false);
708 }
709
710 struct nl_parsed_ifa {
711 uint8_t ifa_family;
712 uint8_t ifa_prefixlen;
713 uint8_t ifa_scope;
714 uint32_t ifa_index;
715 uint32_t ifa_flags;
716 uint32_t ifaf_vhid;
717 uint32_t ifaf_flags;
718 struct sockaddr *ifa_address;
719 struct sockaddr *ifa_local;
720 struct sockaddr *ifa_broadcast;
721 struct ifa_cacheinfo *ifa_cacheinfo;
722 struct sockaddr *f_ifa_addr;
723 struct sockaddr *f_ifa_dst;
724 };
725
726 static int
nlattr_get_cinfo(struct nlattr * nla,struct nl_pstate * npt,const void * arg __unused,void * target)727 nlattr_get_cinfo(struct nlattr *nla, struct nl_pstate *npt,
728 const void *arg __unused, void *target)
729 {
730 if (__predict_false(NLA_DATA_LEN(nla) != sizeof(struct ifa_cacheinfo))) {
731 NLMSG_REPORT_ERR_MSG(npt, "nla type %d size(%u) is not ifa_cacheinfo",
732 nla->nla_type, NLA_DATA_LEN(nla));
733 return (EINVAL);
734 }
735 *((struct ifa_cacheinfo **)target) = (struct ifa_cacheinfo *)NL_RTA_DATA(nla);
736 return (0);
737 }
738
739 #define _IN(_field) offsetof(struct ifaddrmsg, _field)
740 #define _OUT(_field) offsetof(struct nl_parsed_ifa, _field)
741 static const struct nlfield_parser nlf_p_ifa[] = {
742 { .off_in = _IN(ifa_family), .off_out = _OUT(ifa_family), .cb = nlf_get_u8 },
743 { .off_in = _IN(ifa_prefixlen), .off_out = _OUT(ifa_prefixlen), .cb = nlf_get_u8 },
744 { .off_in = _IN(ifa_scope), .off_out = _OUT(ifa_scope), .cb = nlf_get_u8 },
745 { .off_in = _IN(ifa_flags), .off_out = _OUT(ifa_flags), .cb = nlf_get_u8_u32 },
746 { .off_in = _IN(ifa_index), .off_out = _OUT(ifa_index), .cb = nlf_get_u32 },
747 };
748
749 static const struct nlattr_parser nla_p_ifa_fbsd[] = {
750 { .type = IFAF_VHID, .off = _OUT(ifaf_vhid), .cb = nlattr_get_uint32 },
751 { .type = IFAF_FLAGS, .off = _OUT(ifaf_flags), .cb = nlattr_get_uint32 },
752 };
753 NL_DECLARE_ATTR_PARSER(ifa_fbsd_parser, nla_p_ifa_fbsd);
754
755 static const struct nlattr_parser nla_p_ifa[] = {
756 { .type = IFA_ADDRESS, .off = _OUT(ifa_address), .cb = nlattr_get_ip },
757 { .type = IFA_LOCAL, .off = _OUT(ifa_local), .cb = nlattr_get_ip },
758 { .type = IFA_BROADCAST, .off = _OUT(ifa_broadcast), .cb = nlattr_get_ip },
759 { .type = IFA_CACHEINFO, .off = _OUT(ifa_cacheinfo), .cb = nlattr_get_cinfo },
760 { .type = IFA_FLAGS, .off = _OUT(ifa_flags), .cb = nlattr_get_uint32 },
761 { .type = IFA_FREEBSD, .arg = &ifa_fbsd_parser, .cb = nlattr_get_nested },
762 };
763 #undef _IN
764 #undef _OUT
765
766 static bool
post_p_ifa(void * _attrs,struct nl_pstate * npt)767 post_p_ifa(void *_attrs, struct nl_pstate *npt)
768 {
769 struct nl_parsed_ifa *attrs = (struct nl_parsed_ifa *)_attrs;
770
771 if (!check_sa_family(attrs->ifa_address, attrs->ifa_family, "IFA_ADDRESS", npt))
772 return (false);
773 if (!check_sa_family(attrs->ifa_local, attrs->ifa_family, "IFA_LOCAL", npt))
774 return (false);
775 if (!check_sa_family(attrs->ifa_broadcast, attrs->ifa_family, "IFA_BROADADDR", npt))
776 return (false);
777
778 set_scope6(attrs->ifa_address, attrs->ifa_index);
779 set_scope6(attrs->ifa_local, attrs->ifa_index);
780
781 return (true);
782 }
783
784 NL_DECLARE_PARSER_EXT(ifa_parser, struct ifaddrmsg, NULL, nlf_p_ifa, nla_p_ifa, post_p_ifa);
785
786
787 /*
788
789 {ifa_family=AF_INET, ifa_prefixlen=8, ifa_flags=IFA_F_PERMANENT, ifa_scope=RT_SCOPE_HOST, ifa_index=if_nametoindex("lo")},
790 [
791 {{nla_len=8, nla_type=IFA_ADDRESS}, inet_addr("127.0.0.1")},
792 {{nla_len=8, nla_type=IFA_LOCAL}, inet_addr("127.0.0.1")},
793 {{nla_len=7, nla_type=IFA_LABEL}, "lo"},
794 {{nla_len=8, nla_type=IFA_FLAGS}, IFA_F_PERMANENT},
795 {{nla_len=20, nla_type=IFA_CACHEINFO}, {ifa_prefered=4294967295, ifa_valid=4294967295, cstamp=3619, tstamp=3619}}]},
796 ---
797
798 {{len=72, type=RTM_NEWADDR, flags=NLM_F_MULTI, seq=1642191126, pid=566735},
799 {ifa_family=AF_INET6, ifa_prefixlen=96, ifa_flags=IFA_F_PERMANENT, ifa_scope=RT_SCOPE_UNIVERSE, ifa_index=if_nametoindex("virbr0")},
800 [
801 {{nla_len=20, nla_type=IFA_ADDRESS}, inet_pton(AF_INET6, "2a01:4f8:13a:70c:ffff::1")},
802 {{nla_len=20, nla_type=IFA_CACHEINFO}, {ifa_prefered=4294967295, ifa_valid=4294967295, cstamp=4283, tstamp=4283}},
803 {{nla_len=8, nla_type=IFA_FLAGS}, IFA_F_PERMANENT}]},
804 */
805
806 static uint8_t
ifa_get_scope(const struct ifaddr * ifa)807 ifa_get_scope(const struct ifaddr *ifa)
808 {
809 const struct sockaddr *sa;
810 uint8_t addr_scope = RT_SCOPE_UNIVERSE;
811
812 sa = ifa->ifa_addr;
813 switch (sa->sa_family) {
814 #ifdef INET
815 case AF_INET:
816 {
817 struct in_addr addr;
818 addr = ((const struct sockaddr_in *)sa)->sin_addr;
819 if (IN_LOOPBACK(addr.s_addr))
820 addr_scope = RT_SCOPE_HOST;
821 else if (IN_LINKLOCAL(addr.s_addr))
822 addr_scope = RT_SCOPE_LINK;
823 break;
824 }
825 #endif
826 #ifdef INET6
827 case AF_INET6:
828 {
829 const struct in6_addr *addr;
830 addr = &((const struct sockaddr_in6 *)sa)->sin6_addr;
831 if (IN6_IS_ADDR_LOOPBACK(addr))
832 addr_scope = RT_SCOPE_HOST;
833 else if (IN6_IS_ADDR_LINKLOCAL(addr))
834 addr_scope = RT_SCOPE_LINK;
835 break;
836 }
837 #endif
838 }
839
840 return (addr_scope);
841 }
842
843 #ifdef INET6
844 static uint8_t
inet6_get_plen(const struct in6_addr * addr)845 inet6_get_plen(const struct in6_addr *addr)
846 {
847
848 return (bitcount32(addr->s6_addr32[0]) + bitcount32(addr->s6_addr32[1]) +
849 bitcount32(addr->s6_addr32[2]) + bitcount32(addr->s6_addr32[3]));
850 }
851 #endif
852
853 static uint8_t
get_sa_plen(const struct sockaddr * sa)854 get_sa_plen(const struct sockaddr *sa)
855 {
856 #ifdef INET
857 const struct in_addr *paddr;
858 #endif
859 #ifdef INET6
860 const struct in6_addr *paddr6;
861 #endif
862
863 switch (sa->sa_family) {
864 #ifdef INET
865 case AF_INET:
866 paddr = &(((const struct sockaddr_in *)sa)->sin_addr);
867 return bitcount32(paddr->s_addr);
868 #endif
869 #ifdef INET6
870 case AF_INET6:
871 paddr6 = &(((const struct sockaddr_in6 *)sa)->sin6_addr);
872 return inet6_get_plen(paddr6);
873 #endif
874 }
875
876 return (0);
877 }
878
879 #ifdef INET6
880 static uint32_t
in6_flags_to_nl(uint32_t flags)881 in6_flags_to_nl(uint32_t flags)
882 {
883 uint32_t nl_flags = 0;
884
885 if (flags & IN6_IFF_TEMPORARY)
886 nl_flags |= IFA_F_TEMPORARY;
887 if (flags & IN6_IFF_NODAD)
888 nl_flags |= IFA_F_NODAD;
889 if (flags & IN6_IFF_DEPRECATED)
890 nl_flags |= IFA_F_DEPRECATED;
891 if (flags & IN6_IFF_TENTATIVE)
892 nl_flags |= IFA_F_TENTATIVE;
893 if ((flags & (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY)) == 0)
894 flags |= IFA_F_PERMANENT;
895 if (flags & IN6_IFF_DUPLICATED)
896 flags |= IFA_F_DADFAILED;
897 return (nl_flags);
898 }
899
900 static uint32_t
nl_flags_to_in6(uint32_t flags)901 nl_flags_to_in6(uint32_t flags)
902 {
903 uint32_t in6_flags = 0;
904
905 if (flags & IFA_F_TEMPORARY)
906 in6_flags |= IN6_IFF_TEMPORARY;
907 if (flags & IFA_F_NODAD)
908 in6_flags |= IN6_IFF_NODAD;
909 if (flags & IFA_F_DEPRECATED)
910 in6_flags |= IN6_IFF_DEPRECATED;
911 if (flags & IFA_F_TENTATIVE)
912 in6_flags |= IN6_IFF_TENTATIVE;
913 if (flags & IFA_F_DADFAILED)
914 in6_flags |= IN6_IFF_DUPLICATED;
915
916 return (in6_flags);
917 }
918
919 static void
export_cache_info6(struct nl_writer * nw,const struct in6_ifaddr * ia)920 export_cache_info6(struct nl_writer *nw, const struct in6_ifaddr *ia)
921 {
922 struct ifa_cacheinfo ci = {
923 .cstamp = ia->ia6_createtime * 1000,
924 .tstamp = ia->ia6_updatetime * 1000,
925 .ifa_prefered = ia->ia6_lifetime.ia6t_pltime,
926 .ifa_valid = ia->ia6_lifetime.ia6t_vltime,
927 };
928
929 nlattr_add(nw, IFA_CACHEINFO, sizeof(ci), &ci);
930 }
931 #endif
932
933 static void
export_cache_info(struct nl_writer * nw,struct ifaddr * ifa)934 export_cache_info(struct nl_writer *nw, struct ifaddr *ifa)
935 {
936 switch (ifa->ifa_addr->sa_family) {
937 #ifdef INET6
938 case AF_INET6:
939 export_cache_info6(nw, (struct in6_ifaddr *)ifa);
940 break;
941 #endif
942 }
943 }
944
945 /*
946 * {'attrs': [('IFA_ADDRESS', '12.0.0.1'),
947 ('IFA_LOCAL', '12.0.0.1'),
948 ('IFA_LABEL', 'eth10'),
949 ('IFA_FLAGS', 128),
950 ('IFA_CACHEINFO', {'ifa_preferred': 4294967295, 'ifa_valid': 4294967295, 'cstamp': 63745746, 'tstamp': 63745746})],
951 */
952 static bool
dump_iface_addr(struct nl_writer * nw,if_t ifp,struct ifaddr * ifa,const struct nlmsghdr * hdr)953 dump_iface_addr(struct nl_writer *nw, if_t ifp, struct ifaddr *ifa,
954 const struct nlmsghdr *hdr)
955 {
956 struct ifaddrmsg *ifamsg;
957 struct sockaddr *sa = ifa->ifa_addr;
958 struct sockaddr *sa_dst = ifa->ifa_dstaddr;
959
960 NL_LOG(LOG_DEBUG3, "dumping ifa %p type %s(%d) for interface %s",
961 ifa, rib_print_family(sa->sa_family), sa->sa_family, if_name(ifp));
962
963 if (!nlmsg_reply(nw, hdr, sizeof(struct ifaddrmsg)))
964 goto enomem;
965
966 ifamsg = nlmsg_reserve_object(nw, struct ifaddrmsg);
967 ifamsg->ifa_family = sa->sa_family;
968 ifamsg->ifa_prefixlen = get_sa_plen(ifa->ifa_netmask);
969 ifamsg->ifa_flags = 0; // ifa_flags is useless
970 ifamsg->ifa_scope = ifa_get_scope(ifa);
971 ifamsg->ifa_index = if_getindex(ifp);
972
973 if ((if_getflags(ifp) & IFF_POINTOPOINT) && sa_dst != NULL && sa_dst->sa_family != 0) {
974 /* P2P interface may have IPv6 LL with no dst address */
975 dump_sa(nw, IFA_ADDRESS, sa_dst);
976 dump_sa(nw, IFA_LOCAL, sa);
977 } else {
978 dump_sa(nw, IFA_ADDRESS, sa);
979 #ifdef INET
980 /*
981 * In most cases, IFA_ADDRESS == IFA_LOCAL
982 * Skip IFA_LOCAL for anything except INET
983 */
984 if (sa->sa_family == AF_INET)
985 dump_sa(nw, IFA_LOCAL, sa);
986 #endif
987 }
988 if (if_getflags(ifp) & IFF_BROADCAST)
989 dump_sa(nw, IFA_BROADCAST, ifa->ifa_broadaddr);
990
991 nlattr_add_string(nw, IFA_LABEL, if_name(ifp));
992
993 uint32_t nl_ifa_flags = 0;
994 #ifdef INET6
995 if (sa->sa_family == AF_INET6) {
996 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
997 nl_ifa_flags = in6_flags_to_nl(ia->ia6_flags);
998 }
999 #endif
1000 nlattr_add_u32(nw, IFA_FLAGS, nl_ifa_flags);
1001
1002 export_cache_info(nw, ifa);
1003
1004 /* Store FreeBSD-specific attributes */
1005 int off = nlattr_add_nested(nw, IFA_FREEBSD);
1006 if (off != 0) {
1007 if (ifa->ifa_carp != NULL && carp_get_vhid_p != NULL) {
1008 uint32_t vhid = (uint32_t)(*carp_get_vhid_p)(ifa);
1009 nlattr_add_u32(nw, IFAF_VHID, vhid);
1010 }
1011 #ifdef INET6
1012 if (sa->sa_family == AF_INET6) {
1013 uint32_t ifa_flags = ((struct in6_ifaddr *)ifa)->ia6_flags;
1014
1015 nlattr_add_u32(nw, IFAF_FLAGS, ifa_flags);
1016 }
1017 #endif
1018
1019 nlattr_set_len(nw, off);
1020 }
1021
1022 if (nlmsg_end(nw))
1023 return (true);
1024 enomem:
1025 NL_LOG(LOG_DEBUG, "Failed to dump ifa type %s(%d) for interface %s",
1026 rib_print_family(sa->sa_family), sa->sa_family, if_name(ifp));
1027 nlmsg_abort(nw);
1028 return (false);
1029 }
1030
1031 static int
dump_iface_addrs(struct netlink_walkargs * wa,if_t ifp)1032 dump_iface_addrs(struct netlink_walkargs *wa, if_t ifp)
1033 {
1034 struct ifaddr *ifa;
1035 struct ifa_iter it;
1036 int error = 0;
1037
1038 for (ifa = ifa_iter_start(ifp, &it); ifa != NULL; ifa = ifa_iter_next(&it)) {
1039 if (wa->family != 0 && wa->family != ifa->ifa_addr->sa_family)
1040 continue;
1041 if (ifa->ifa_addr->sa_family == AF_LINK)
1042 continue;
1043 if (prison_if(wa->cred, ifa->ifa_addr) != 0)
1044 continue;
1045 wa->count++;
1046 if (!dump_iface_addr(wa->nw, ifp, ifa, &wa->hdr)) {
1047 error = ENOMEM;
1048 break;
1049 }
1050 wa->dumped++;
1051 }
1052 ifa_iter_finish(&it);
1053
1054 return (error);
1055 }
1056
1057 static int
rtnl_handle_getaddr(struct nlmsghdr * hdr,struct nlpcb * nlp,struct nl_pstate * npt)1058 rtnl_handle_getaddr(struct nlmsghdr *hdr, struct nlpcb *nlp, struct nl_pstate *npt)
1059 {
1060 if_t ifp;
1061 int error = 0;
1062
1063 struct nl_parsed_ifa attrs = {};
1064 error = nl_parse_nlmsg(hdr, &ifa_parser, npt, &attrs);
1065 if (error != 0)
1066 return (error);
1067
1068 struct netlink_walkargs wa = {
1069 .so = nlp,
1070 .nw = npt->nw,
1071 .cred = nlp_get_cred(nlp),
1072 .family = attrs.ifa_family,
1073 .hdr.nlmsg_pid = hdr->nlmsg_pid,
1074 .hdr.nlmsg_seq = hdr->nlmsg_seq,
1075 .hdr.nlmsg_flags = hdr->nlmsg_flags | NLM_F_MULTI,
1076 .hdr.nlmsg_type = NL_RTM_NEWADDR,
1077 };
1078
1079 NL_LOG(LOG_DEBUG2, "Start dump");
1080
1081 if (attrs.ifa_index != 0) {
1082 ifp = ifnet_byindex(attrs.ifa_index);
1083 if (ifp == NULL)
1084 error = ENOENT;
1085 else
1086 error = dump_iface_addrs(&wa, ifp);
1087 } else {
1088 struct if_iter it;
1089
1090 for (ifp = if_iter_start(&it); ifp != NULL; ifp = if_iter_next(&it)) {
1091 error = dump_iface_addrs(&wa, ifp);
1092 if (error != 0)
1093 break;
1094 }
1095 if_iter_finish(&it);
1096 }
1097
1098 NL_LOG(LOG_DEBUG2, "End dump, iterated %d dumped %d", wa.count, wa.dumped);
1099
1100 if (!nlmsg_end_dump(wa.nw, error, &wa.hdr)) {
1101 NL_LOG(LOG_DEBUG, "Unable to finalize the dump");
1102 return (ENOMEM);
1103 }
1104
1105 return (error);
1106 }
1107
1108 #ifdef INET
1109 static int
handle_newaddr_inet(struct nlmsghdr * hdr,struct nl_parsed_ifa * attrs,if_t ifp,struct nlpcb * nlp,struct nl_pstate * npt)1110 handle_newaddr_inet(struct nlmsghdr *hdr, struct nl_parsed_ifa *attrs,
1111 if_t ifp, struct nlpcb *nlp, struct nl_pstate *npt)
1112 {
1113 int plen = attrs->ifa_prefixlen;
1114 int if_flags = if_getflags(ifp);
1115 struct sockaddr_in *addr, *dst;
1116
1117 if (plen > 32) {
1118 nlmsg_report_err_msg(npt, "invalid ifa_prefixlen");
1119 return (EINVAL);
1120 };
1121
1122 if (if_flags & IFF_POINTOPOINT) {
1123 /*
1124 * Only P2P IFAs are allowed by the implementation.
1125 */
1126 if (attrs->ifa_address == NULL || attrs->ifa_local == NULL) {
1127 nlmsg_report_err_msg(npt, "Empty IFA_LOCAL/IFA_ADDRESS");
1128 return (EINVAL);
1129 }
1130 addr = (struct sockaddr_in *)attrs->ifa_local;
1131 dst = (struct sockaddr_in *)attrs->ifa_address;
1132 } else {
1133 /*
1134 * Map the Netlink attributes to FreeBSD ifa layout.
1135 * If only IFA_ADDRESS or IFA_LOCAL is set OR
1136 * both are set to the same value => ifa is not p2p
1137 * and the attribute value contains interface address.
1138 *
1139 * Otherwise (both IFA_ADDRESS and IFA_LOCAL are set and
1140 * different), IFA_LOCAL contains an interface address and
1141 * IFA_ADDRESS contains peer address.
1142 */
1143 addr = (struct sockaddr_in *)attrs->ifa_local;
1144 if (addr == NULL)
1145 addr = (struct sockaddr_in *)attrs->ifa_address;
1146
1147 if (addr == NULL) {
1148 nlmsg_report_err_msg(npt, "Empty IFA_LOCAL/IFA_ADDRESS");
1149 return (EINVAL);
1150 }
1151
1152 /* Generate broadcast address if not set */
1153 if ((if_flags & IFF_BROADCAST) && attrs->ifa_broadcast == NULL) {
1154 uint32_t s_baddr;
1155 struct sockaddr_in *sin_brd;
1156
1157 if (plen == 31)
1158 s_baddr = INADDR_BROADCAST; /* RFC 3021 */
1159 else {
1160 uint32_t s_mask;
1161
1162 s_mask = htonl(plen ? ~((1 << (32 - plen)) - 1) : 0);
1163 s_baddr = addr->sin_addr.s_addr | ~s_mask;
1164 }
1165
1166 sin_brd = (struct sockaddr_in *)npt_alloc(npt, sizeof(*sin_brd));
1167 if (sin_brd == NULL)
1168 return (ENOMEM);
1169 sin_brd->sin_family = AF_INET;
1170 sin_brd->sin_len = sizeof(*sin_brd);
1171 sin_brd->sin_addr.s_addr = s_baddr;
1172 attrs->ifa_broadcast = (struct sockaddr *)sin_brd;
1173 }
1174 dst = (struct sockaddr_in *)attrs->ifa_broadcast;
1175 }
1176
1177 struct sockaddr_in mask = {
1178 .sin_len = sizeof(struct sockaddr_in),
1179 .sin_family = AF_INET,
1180 .sin_addr.s_addr = htonl(plen ? ~((1 << (32 - plen)) - 1) : 0),
1181 };
1182 struct in_aliasreq req = {
1183 .ifra_addr = *addr,
1184 .ifra_mask = mask,
1185 .ifra_vhid = attrs->ifaf_vhid,
1186 };
1187 if (dst != NULL)
1188 req.ifra_dstaddr = *dst;
1189
1190 return (in_control_ioctl(SIOCAIFADDR, &req, ifp, nlp_get_cred(nlp)));
1191 }
1192
1193 static int
handle_deladdr_inet(struct nlmsghdr * hdr,struct nl_parsed_ifa * attrs,if_t ifp,struct nlpcb * nlp,struct nl_pstate * npt)1194 handle_deladdr_inet(struct nlmsghdr *hdr, struct nl_parsed_ifa *attrs,
1195 if_t ifp, struct nlpcb *nlp, struct nl_pstate *npt)
1196 {
1197 struct sockaddr_in *addr = (struct sockaddr_in *)attrs->ifa_local;
1198
1199 if (addr == NULL)
1200 addr = (struct sockaddr_in *)attrs->ifa_address;
1201
1202 if (addr == NULL) {
1203 nlmsg_report_err_msg(npt, "empty IFA_ADDRESS/IFA_LOCAL");
1204 return (EINVAL);
1205 }
1206
1207 struct in_aliasreq req = { .ifra_addr = *addr };
1208
1209 return (in_control_ioctl(SIOCDIFADDR, &req, ifp, nlp_get_cred(nlp)));
1210 }
1211 #endif
1212
1213 #ifdef INET6
1214 static int
handle_newaddr_inet6(struct nlmsghdr * hdr,struct nl_parsed_ifa * attrs,if_t ifp,struct nlpcb * nlp,struct nl_pstate * npt)1215 handle_newaddr_inet6(struct nlmsghdr *hdr, struct nl_parsed_ifa *attrs,
1216 if_t ifp, struct nlpcb *nlp, struct nl_pstate *npt)
1217 {
1218 struct sockaddr_in6 *addr, *dst;
1219
1220 if (attrs->ifa_prefixlen > 128) {
1221 nlmsg_report_err_msg(npt, "invalid ifa_prefixlen");
1222 return (EINVAL);
1223 }
1224
1225 /*
1226 * In IPv6 implementation, adding non-P2P address to the P2P interface
1227 * is allowed.
1228 */
1229 addr = (struct sockaddr_in6 *)(attrs->ifa_local);
1230 dst = (struct sockaddr_in6 *)(attrs->ifa_address);
1231
1232 if (addr == NULL) {
1233 addr = dst;
1234 dst = NULL;
1235 } else if (dst != NULL) {
1236 if (IN6_ARE_ADDR_EQUAL(&addr->sin6_addr, &dst->sin6_addr)) {
1237 /*
1238 * Sometimes Netlink users fills in both attributes
1239 * with the same address. It still means "non-p2p".
1240 */
1241 dst = NULL;
1242 }
1243 }
1244
1245 if (addr == NULL) {
1246 nlmsg_report_err_msg(npt, "Empty IFA_LOCAL/IFA_ADDRESS");
1247 return (EINVAL);
1248 }
1249
1250 uint32_t flags = nl_flags_to_in6(attrs->ifa_flags) | attrs->ifaf_flags;
1251
1252 uint32_t pltime = 0, vltime = 0;
1253 if (attrs->ifa_cacheinfo != 0) {
1254 pltime = attrs->ifa_cacheinfo->ifa_prefered;
1255 vltime = attrs->ifa_cacheinfo->ifa_valid;
1256 }
1257
1258 struct sockaddr_in6 mask = {
1259 .sin6_len = sizeof(struct sockaddr_in6),
1260 .sin6_family = AF_INET6,
1261 };
1262 ip6_writemask(&mask.sin6_addr, attrs->ifa_prefixlen);
1263
1264 struct in6_aliasreq req = {
1265 .ifra_addr = *addr,
1266 .ifra_prefixmask = mask,
1267 .ifra_flags = flags,
1268 .ifra_lifetime = { .ia6t_vltime = vltime, .ia6t_pltime = pltime },
1269 .ifra_vhid = attrs->ifaf_vhid,
1270 };
1271 if (dst != NULL)
1272 req.ifra_dstaddr = *dst;
1273
1274 return (in6_control_ioctl(SIOCAIFADDR_IN6, &req, ifp, nlp_get_cred(nlp)));
1275 }
1276
1277 static int
handle_deladdr_inet6(struct nlmsghdr * hdr,struct nl_parsed_ifa * attrs,if_t ifp,struct nlpcb * nlp,struct nl_pstate * npt)1278 handle_deladdr_inet6(struct nlmsghdr *hdr, struct nl_parsed_ifa *attrs,
1279 if_t ifp, struct nlpcb *nlp, struct nl_pstate *npt)
1280 {
1281 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)attrs->ifa_local;
1282
1283 if (addr == NULL)
1284 addr = (struct sockaddr_in6 *)(attrs->ifa_address);
1285
1286 if (addr == NULL) {
1287 nlmsg_report_err_msg(npt, "Empty IFA_LOCAL/IFA_ADDRESS");
1288 return (EINVAL);
1289 }
1290
1291 struct in6_aliasreq req = { .ifra_addr = *addr };
1292
1293 return (in6_control_ioctl(SIOCDIFADDR_IN6, &req, ifp, nlp_get_cred(nlp)));
1294 }
1295 #endif
1296
1297
1298 static int
rtnl_handle_addr(struct nlmsghdr * hdr,struct nlpcb * nlp,struct nl_pstate * npt)1299 rtnl_handle_addr(struct nlmsghdr *hdr, struct nlpcb *nlp, struct nl_pstate *npt)
1300 {
1301 struct epoch_tracker et;
1302 int error;
1303
1304 struct nl_parsed_ifa attrs = {};
1305 error = nl_parse_nlmsg(hdr, &ifa_parser, npt, &attrs);
1306 if (error != 0)
1307 return (error);
1308
1309 NET_EPOCH_ENTER(et);
1310 if_t ifp = ifnet_byindex_ref(attrs.ifa_index);
1311 NET_EPOCH_EXIT(et);
1312
1313 if (ifp == NULL) {
1314 nlmsg_report_err_msg(npt, "Unable to find interface with index %u",
1315 attrs.ifa_index);
1316 return (ENOENT);
1317 }
1318 int if_flags = if_getflags(ifp);
1319
1320 #if defined(INET) || defined(INET6)
1321 bool new = hdr->nlmsg_type == NL_RTM_NEWADDR;
1322 #endif
1323
1324 /*
1325 * TODO: Properly handle NLM_F_CREATE / NLM_F_EXCL.
1326 * The current ioctl-based KPI always does an implicit create-or-replace.
1327 * It is not possible to specify fine-grained options.
1328 */
1329
1330 switch (attrs.ifa_family) {
1331 #ifdef INET
1332 case AF_INET:
1333 if (new)
1334 error = handle_newaddr_inet(hdr, &attrs, ifp, nlp, npt);
1335 else
1336 error = handle_deladdr_inet(hdr, &attrs, ifp, nlp, npt);
1337 break;
1338 #endif
1339 #ifdef INET6
1340 case AF_INET6:
1341 if (new)
1342 error = handle_newaddr_inet6(hdr, &attrs, ifp, nlp, npt);
1343 else
1344 error = handle_deladdr_inet6(hdr, &attrs, ifp, nlp, npt);
1345 break;
1346 #endif
1347 default:
1348 error = EAFNOSUPPORT;
1349 }
1350
1351 if (error == 0 && !(if_flags & IFF_UP) && (if_getflags(ifp) & IFF_UP))
1352 if_up(ifp);
1353
1354 if_rele(ifp);
1355
1356 return (error);
1357 }
1358
1359
1360 static void
rtnl_handle_ifaddr(void * arg __unused,struct ifaddr * ifa,int cmd)1361 rtnl_handle_ifaddr(void *arg __unused, struct ifaddr *ifa, int cmd)
1362 {
1363 struct nlmsghdr hdr = {};
1364 struct nl_writer nw = {};
1365 uint32_t group = 0;
1366
1367 switch (ifa->ifa_addr->sa_family) {
1368 #ifdef INET
1369 case AF_INET:
1370 group = RTNLGRP_IPV4_IFADDR;
1371 break;
1372 #endif
1373 #ifdef INET6
1374 case AF_INET6:
1375 group = RTNLGRP_IPV6_IFADDR;
1376 break;
1377 #endif
1378 default:
1379 NL_LOG(LOG_DEBUG2, "ifa notification for unknown AF: %d",
1380 ifa->ifa_addr->sa_family);
1381 return;
1382 }
1383
1384 if (!nl_has_listeners(NETLINK_ROUTE, group))
1385 return;
1386
1387 if (!nlmsg_get_group_writer(&nw, NLMSG_LARGE, NETLINK_ROUTE, group)) {
1388 NL_LOG(LOG_DEBUG, "error allocating group writer");
1389 return;
1390 }
1391
1392 hdr.nlmsg_type = (cmd == RTM_DELETE) ? NL_RTM_DELADDR : NL_RTM_NEWADDR;
1393
1394 dump_iface_addr(&nw, ifa->ifa_ifp, ifa, &hdr);
1395 nlmsg_flush(&nw);
1396 }
1397
1398 static void
rtnl_handle_ifevent(if_t ifp,int nlmsg_type,int if_flags_mask)1399 rtnl_handle_ifevent(if_t ifp, int nlmsg_type, int if_flags_mask)
1400 {
1401 struct nlmsghdr hdr = { .nlmsg_type = nlmsg_type };
1402 struct nl_writer nw = {};
1403
1404 if (!nl_has_listeners(NETLINK_ROUTE, RTNLGRP_LINK))
1405 return;
1406
1407 if (!nlmsg_get_group_writer(&nw, NLMSG_LARGE, NETLINK_ROUTE, RTNLGRP_LINK)) {
1408 NL_LOG(LOG_DEBUG, "error allocating mbuf");
1409 return;
1410 }
1411 dump_iface(&nw, ifp, &hdr, if_flags_mask);
1412 nlmsg_flush(&nw);
1413 }
1414
1415 static void
rtnl_handle_ifattach(void * arg,if_t ifp)1416 rtnl_handle_ifattach(void *arg, if_t ifp)
1417 {
1418 NL_LOG(LOG_DEBUG2, "ifnet %s", if_name(ifp));
1419 rtnl_handle_ifevent(ifp, NL_RTM_NEWLINK, 0);
1420 }
1421
1422 static void
rtnl_handle_ifdetach(void * arg,if_t ifp)1423 rtnl_handle_ifdetach(void *arg, if_t ifp)
1424 {
1425 NL_LOG(LOG_DEBUG2, "ifnet %s", if_name(ifp));
1426 rtnl_handle_ifevent(ifp, NL_RTM_DELLINK, 0);
1427 }
1428
1429 static void
rtnl_handle_iflink(void * arg,if_t ifp)1430 rtnl_handle_iflink(void *arg, if_t ifp)
1431 {
1432 NL_LOG(LOG_DEBUG2, "ifnet %s", if_name(ifp));
1433 rtnl_handle_ifevent(ifp, NL_RTM_NEWLINK, 0);
1434 }
1435
1436 void
rtnl_handle_ifnet_event(if_t ifp,int if_flags_mask)1437 rtnl_handle_ifnet_event(if_t ifp, int if_flags_mask)
1438 {
1439 NL_LOG(LOG_DEBUG2, "ifnet %s", if_name(ifp));
1440 rtnl_handle_ifevent(ifp, NL_RTM_NEWLINK, if_flags_mask);
1441 }
1442
1443 static const struct rtnl_cmd_handler cmd_handlers[] = {
1444 {
1445 .cmd = NL_RTM_GETLINK,
1446 .name = "RTM_GETLINK",
1447 .cb = &rtnl_handle_getlink,
1448 .flags = RTNL_F_NOEPOCH | RTNL_F_ALLOW_NONVNET_JAIL,
1449 },
1450 {
1451 .cmd = NL_RTM_DELLINK,
1452 .name = "RTM_DELLINK",
1453 .cb = &rtnl_handle_dellink,
1454 .priv = PRIV_NET_IFDESTROY,
1455 .flags = RTNL_F_NOEPOCH,
1456 },
1457 {
1458 .cmd = NL_RTM_NEWLINK,
1459 .name = "RTM_NEWLINK",
1460 .cb = &rtnl_handle_newlink,
1461 .priv = PRIV_NET_IFCREATE,
1462 .flags = RTNL_F_NOEPOCH,
1463 },
1464 {
1465 .cmd = NL_RTM_GETADDR,
1466 .name = "RTM_GETADDR",
1467 .cb = &rtnl_handle_getaddr,
1468 .flags = RTNL_F_ALLOW_NONVNET_JAIL,
1469 },
1470 {
1471 .cmd = NL_RTM_NEWADDR,
1472 .name = "RTM_NEWADDR",
1473 .cb = &rtnl_handle_addr,
1474 .priv = PRIV_NET_ADDIFADDR,
1475 .flags = RTNL_F_NOEPOCH,
1476 },
1477 {
1478 .cmd = NL_RTM_DELADDR,
1479 .name = "RTM_DELADDR",
1480 .cb = &rtnl_handle_addr,
1481 .priv = PRIV_NET_DELIFADDR,
1482 .flags = RTNL_F_NOEPOCH,
1483 },
1484 };
1485
1486 static const struct nlhdr_parser *all_parsers[] = {
1487 &ifmsg_parser, &ifa_parser, &ifa_fbsd_parser,
1488 };
1489
1490 void
rtnl_iface_add_cloner(struct nl_cloner * cloner)1491 rtnl_iface_add_cloner(struct nl_cloner *cloner)
1492 {
1493 sx_xlock(&rtnl_cloner_lock);
1494 SLIST_INSERT_HEAD(&nl_cloners, cloner, next);
1495 sx_xunlock(&rtnl_cloner_lock);
1496 }
1497
1498 void
rtnl_iface_del_cloner(struct nl_cloner * cloner)1499 rtnl_iface_del_cloner(struct nl_cloner *cloner)
1500 {
1501 sx_xlock(&rtnl_cloner_lock);
1502 SLIST_REMOVE(&nl_cloners, cloner, nl_cloner, next);
1503 sx_xunlock(&rtnl_cloner_lock);
1504 }
1505
1506 void
rtnl_ifaces_init(void)1507 rtnl_ifaces_init(void)
1508 {
1509 ifattach_event = EVENTHANDLER_REGISTER(
1510 ifnet_arrival_event, rtnl_handle_ifattach, NULL,
1511 EVENTHANDLER_PRI_ANY);
1512 ifdetach_event = EVENTHANDLER_REGISTER(
1513 ifnet_departure_event, rtnl_handle_ifdetach, NULL,
1514 EVENTHANDLER_PRI_ANY);
1515 ifaddr_event = EVENTHANDLER_REGISTER(
1516 rt_addrmsg, rtnl_handle_ifaddr, NULL,
1517 EVENTHANDLER_PRI_ANY);
1518 iflink_event = EVENTHANDLER_REGISTER(
1519 ifnet_link_event, rtnl_handle_iflink, NULL,
1520 EVENTHANDLER_PRI_ANY);
1521 NL_VERIFY_PARSERS(all_parsers);
1522 rtnl_register_messages(cmd_handlers, NL_ARRAY_LEN(cmd_handlers));
1523 }
1524
1525 void
rtnl_ifaces_destroy(void)1526 rtnl_ifaces_destroy(void)
1527 {
1528 EVENTHANDLER_DEREGISTER(ifnet_arrival_event, ifattach_event);
1529 EVENTHANDLER_DEREGISTER(ifnet_departure_event, ifdetach_event);
1530 EVENTHANDLER_DEREGISTER(rt_addrmsg, ifaddr_event);
1531 EVENTHANDLER_DEREGISTER(ifnet_link_event, iflink_event);
1532 }
1533