]> git.proxmox.com Git - mirror_frr.git/blob - zebra/kernel_socket.c
eigrpd: eigrp usage of uint32_t to struct in_addr for router_id data
[mirror_frr.git] / zebra / kernel_socket.c
1 /* Kernel communication using routing socket.
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 #ifndef HAVE_NETLINK
24
25 #include <net/if_types.h>
26 #ifdef __OpenBSD__
27 #include <netmpls/mpls.h>
28 #endif
29
30 #include "if.h"
31 #include "prefix.h"
32 #include "sockunion.h"
33 #include "connected.h"
34 #include "memory.h"
35 #include "zebra_memory.h"
36 #include "ioctl.h"
37 #include "log.h"
38 #include "table.h"
39 #include "rib.h"
40 #include "privs.h"
41 #include "vrf.h"
42 #include "lib_errors.h"
43
44 #include "zebra/rt.h"
45 #include "zebra/interface.h"
46 #include "zebra/zserv.h"
47 #include "zebra/debug.h"
48 #include "zebra/kernel_socket.h"
49 #include "zebra/rib.h"
50 #include "zebra/zebra_errors.h"
51 #include "zebra/zebra_ptm.h"
52
53 extern struct zebra_privs_t zserv_privs;
54
55 /*
56 * Historically, the BSD routing socket has aligned data following a
57 * struct sockaddr to sizeof(long), which was 4 bytes on some
58 * platforms, and 8 bytes on others. NetBSD 6 changed the routing
59 * socket to align to sizeof(uint64_t), which is 8 bytes. OS X
60 * appears to align to sizeof(int), which is 4 bytes.
61 *
62 * Alignment of zero-sized sockaddrs is nonsensical, but historically
63 * BSD defines RT_ROUNDUP(0) to be the alignment interval (rather than
64 * 0). We follow this practice without questioning it, but it is a
65 * bug if quagga calls ROUNDUP with 0.
66 */
67 #ifdef __APPLE__
68 #define ROUNDUP_TYPE int
69 #else
70 #define ROUNDUP_TYPE long
71 #endif
72
73 /*
74 * Because of these varying conventions, the only sane approach is for
75 * the <net/route.h> header to define some flavor of ROUNDUP macro.
76 */
77
78 /* OS X (Xcode as of 2014-12) is known not to define RT_ROUNDUP */
79 #if defined(RT_ROUNDUP)
80 #define ROUNDUP(a) RT_ROUNDUP(a)
81 #endif /* defined(RT_ROUNDUP) */
82
83 #if defined(SUNOS_5)
84 /* Solaris has struct sockaddr_in[6] definitions at 16 / 32 bytes size,
85 * so the whole concept doesn't really apply. */
86 #define ROUNDUP(a) (a)
87 #endif
88
89 /*
90 * If ROUNDUP has not yet been defined in terms of platform-provided
91 * defines, attempt to cope with heuristics.
92 */
93 #if !defined(ROUNDUP)
94
95 /*
96 * If you're porting to a platform that changed RT_ROUNDUP but doesn't
97 * have it in its headers, this will break rather obviously and you'll
98 * have to fix it here.
99 */
100 #define ROUNDUP(a) \
101 ((a) > 0 ? (1 + (((a)-1) | (sizeof(ROUNDUP_TYPE) - 1))) \
102 : sizeof(ROUNDUP_TYPE))
103
104 #endif /* defined(ROUNDUP) */
105
106
107 #if defined(SA_SIZE)
108 /* SAROUNDUP is the only thing we need, and SA_SIZE provides that */
109 #define SAROUNDUP(a) SA_SIZE(a)
110 #else /* !SA_SIZE */
111 /*
112 * Given a pointer (sockaddr or void *), return the number of bytes
113 * taken up by the sockaddr and any padding needed for alignment.
114 */
115 #if defined(HAVE_STRUCT_SOCKADDR_SA_LEN)
116 #define SAROUNDUP(X) ROUNDUP(((struct sockaddr *)(X))->sa_len)
117 #else
118 /*
119 * One would hope all fixed-size structure definitions are aligned,
120 * but round them up nonetheless.
121 */
122 #define SAROUNDUP(X) \
123 (((struct sockaddr *)(X))->sa_family == AF_INET \
124 ? ROUNDUP(sizeof(struct sockaddr_in)) \
125 : (((struct sockaddr *)(X))->sa_family == AF_INET6 \
126 ? ROUNDUP(sizeof(struct sockaddr_in6)) \
127 : (((struct sockaddr *)(X))->sa_family == AF_LINK \
128 ? ROUNDUP(sizeof(struct sockaddr_dl)) \
129 : sizeof(struct sockaddr))))
130 #endif /* HAVE_STRUCT_SOCKADDR_SA_LEN */
131
132 #endif /* !SA_SIZE */
133
134 /* Routing socket message types. */
135 const struct message rtm_type_str[] = {{RTM_ADD, "RTM_ADD"},
136 {RTM_DELETE, "RTM_DELETE"},
137 {RTM_CHANGE, "RTM_CHANGE"},
138 {RTM_GET, "RTM_GET"},
139 {RTM_LOSING, "RTM_LOSING"},
140 {RTM_REDIRECT, "RTM_REDIRECT"},
141 {RTM_MISS, "RTM_MISS"},
142 #ifdef RTM_LOCK
143 {RTM_LOCK, "RTM_LOCK"},
144 #endif /* RTM_LOCK */
145 #ifdef OLDADD
146 {RTM_OLDADD, "RTM_OLDADD"},
147 #endif /* RTM_OLDADD */
148 #ifdef RTM_OLDDEL
149 {RTM_OLDDEL, "RTM_OLDDEL"},
150 #endif /* RTM_OLDDEL */
151 {RTM_RESOLVE, "RTM_RESOLVE"},
152 {RTM_NEWADDR, "RTM_NEWADDR"},
153 {RTM_DELADDR, "RTM_DELADDR"},
154 {RTM_IFINFO, "RTM_IFINFO"},
155 #ifdef RTM_OIFINFO
156 {RTM_OIFINFO, "RTM_OIFINFO"},
157 #endif /* RTM_OIFINFO */
158 #ifdef RTM_NEWMADDR
159 {RTM_NEWMADDR, "RTM_NEWMADDR"},
160 #endif /* RTM_NEWMADDR */
161 #ifdef RTM_DELMADDR
162 {RTM_DELMADDR, "RTM_DELMADDR"},
163 #endif /* RTM_DELMADDR */
164 #ifdef RTM_IFANNOUNCE
165 {RTM_IFANNOUNCE, "RTM_IFANNOUNCE"},
166 #endif /* RTM_IFANNOUNCE */
167 {0}};
168
169 static const struct message rtm_flag_str[] = {{RTF_UP, "UP"},
170 {RTF_GATEWAY, "GATEWAY"},
171 {RTF_HOST, "HOST"},
172 {RTF_REJECT, "REJECT"},
173 {RTF_DYNAMIC, "DYNAMIC"},
174 {RTF_MODIFIED, "MODIFIED"},
175 {RTF_DONE, "DONE"},
176 #ifdef RTF_MASK
177 {RTF_MASK, "MASK"},
178 #endif /* RTF_MASK */
179 #ifdef RTF_CLONING
180 {RTF_CLONING, "CLONING"},
181 #endif /* RTF_CLONING */
182 #ifdef RTF_XRESOLVE
183 {RTF_XRESOLVE, "XRESOLVE"},
184 #endif /* RTF_XRESOLVE */
185 #ifdef RTF_LLINFO
186 {RTF_LLINFO, "LLINFO"},
187 #endif /* RTF_LLINFO */
188 {RTF_STATIC, "STATIC"},
189 {RTF_BLACKHOLE, "BLACKHOLE"},
190 #ifdef RTF_PRIVATE
191 {RTF_PRIVATE, "PRIVATE"},
192 #endif /* RTF_PRIVATE */
193 {RTF_PROTO1, "PROTO1"},
194 {RTF_PROTO2, "PROTO2"},
195 #ifdef RTF_PRCLONING
196 {RTF_PRCLONING, "PRCLONING"},
197 #endif /* RTF_PRCLONING */
198 #ifdef RTF_WASCLONED
199 {RTF_WASCLONED, "WASCLONED"},
200 #endif /* RTF_WASCLONED */
201 #ifdef RTF_PROTO3
202 {RTF_PROTO3, "PROTO3"},
203 #endif /* RTF_PROTO3 */
204 #ifdef RTF_PINNED
205 {RTF_PINNED, "PINNED"},
206 #endif /* RTF_PINNED */
207 #ifdef RTF_LOCAL
208 {RTF_LOCAL, "LOCAL"},
209 #endif /* RTF_LOCAL */
210 #ifdef RTF_BROADCAST
211 {RTF_BROADCAST, "BROADCAST"},
212 #endif /* RTF_BROADCAST */
213 #ifdef RTF_MULTICAST
214 {RTF_MULTICAST, "MULTICAST"},
215 #endif /* RTF_MULTICAST */
216 #ifdef RTF_MULTIRT
217 {RTF_MULTIRT, "MULTIRT"},
218 #endif /* RTF_MULTIRT */
219 #ifdef RTF_SETSRC
220 {RTF_SETSRC, "SETSRC"},
221 #endif /* RTF_SETSRC */
222 {0}};
223
224 /* Kernel routing update socket. */
225 int routing_sock = -1;
226
227 /* Kernel dataplane routing update socket, used in the dataplane pthread
228 * context.
229 */
230 int dplane_routing_sock = -1;
231
232 /* Yes I'm checking ugly routing socket behavior. */
233 /* #define DEBUG */
234
235 size_t rta_get(caddr_t sap, void *dest, size_t destlen);
236 size_t rta_getsdlname(caddr_t sap, void *dest, short *destlen);
237
238 /* Supported address family check. */
239 static inline int af_check(int family)
240 {
241 if (family == AF_INET)
242 return 1;
243 if (family == AF_INET6)
244 return 1;
245 return 0;
246 }
247
248 size_t rta_get(caddr_t sap, void *destp, size_t destlen)
249 {
250 struct sockaddr *sa = (struct sockaddr *)sap;
251 uint8_t *dest = destp;
252 size_t tlen, copylen;
253
254 #ifdef HAVE_STRUCT_SOCKADDR_SA_LEN
255 copylen = sa->sa_len;
256 tlen = (copylen == 0) ? sizeof(ROUNDUP_TYPE) : ROUNDUP(copylen);
257 #else /* !HAVE_STRUCT_SOCKADDR_SA_LEN */
258 copylen = tlen = SAROUNDUP(sap);
259 #endif /* !HAVE_STRUCT_SOCKADDR_SA_LEN */
260
261 if (copylen > 0 && dest != NULL && af_check(sa->sa_family)) {
262 if (copylen > destlen) {
263 zlog_warn("%s: destination buffer too small (%lu vs %lu)",
264 __func__, copylen, destlen);
265 memcpy(dest, sap, destlen);
266 } else
267 memcpy(dest, sap, copylen);
268 }
269
270 return tlen;
271 }
272
273 size_t rta_getsdlname(caddr_t sap, void *destp, short *destlen)
274 {
275 struct sockaddr_dl *sdl = (struct sockaddr_dl *)sap;
276 struct sockaddr *sa = (struct sockaddr *)sap;
277 uint8_t *dest = destp;
278 size_t tlen, copylen;
279
280 copylen = sdl->sdl_nlen;
281 #ifdef HAVE_STRUCT_SOCKADDR_SA_LEN
282 tlen = (sa->sa_len == 0) ? sizeof(ROUNDUP_TYPE) : ROUNDUP(sa->sa_len);
283 #else /* !HAVE_STRUCT_SOCKADDR_SA_LEN */
284 tlen = SAROUNDUP(sap);
285 #endif /* !HAVE_STRUCT_SOCKADDR_SA_LEN */
286
287 if (copylen > 0 && dest != NULL && sdl->sdl_family == AF_LINK) {
288 if (copylen > IFNAMSIZ) {
289 zlog_warn("%s: destination buffer too small (%lu vs %d)",
290 __func__, copylen, IFNAMSIZ);
291 memcpy(dest, sdl->sdl_data, IFNAMSIZ);
292 dest[IFNAMSIZ] = 0;
293 *destlen = IFNAMSIZ;
294 } else {
295 memcpy(dest, sdl->sdl_data, copylen);
296 dest[copylen] = 0;
297 *destlen = copylen;
298 }
299 } else
300 *destlen = 0;
301
302 return tlen;
303 }
304
305 /* Dump routing table flag for debug purpose. */
306 static void rtm_flag_dump(int flag)
307 {
308 const struct message *mes;
309 static char buf[BUFSIZ];
310
311 buf[0] = '\0';
312 for (mes = rtm_flag_str; mes->key != 0; mes++) {
313 if (mes->key & flag) {
314 strlcat(buf, mes->str, BUFSIZ);
315 strlcat(buf, " ", BUFSIZ);
316 }
317 }
318 zlog_debug("Kernel: %s", buf);
319 }
320
321 #ifdef RTM_IFANNOUNCE
322 /* Interface adding function */
323 static int ifan_read(struct if_announcemsghdr *ifan)
324 {
325 struct interface *ifp;
326
327 ifp = if_lookup_by_index(ifan->ifan_index, VRF_DEFAULT);
328
329 if (ifp)
330 assert((ifp->ifindex == ifan->ifan_index)
331 || (ifp->ifindex == IFINDEX_INTERNAL));
332
333 if ((ifp == NULL) || ((ifp->ifindex == IFINDEX_INTERNAL)
334 && (ifan->ifan_what == IFAN_ARRIVAL))) {
335 if (IS_ZEBRA_DEBUG_KERNEL)
336 zlog_debug(
337 "%s: creating interface for ifindex %d, name %s",
338 __func__, ifan->ifan_index, ifan->ifan_name);
339
340 /* Create Interface */
341 ifp = if_get_by_name(ifan->ifan_name, VRF_DEFAULT);
342 if_set_index(ifp, ifan->ifan_index);
343
344 if_get_metric(ifp);
345 if_add_update(ifp);
346 } else if (ifp != NULL && ifan->ifan_what == IFAN_DEPARTURE)
347 if_delete_update(ifp);
348
349 if_get_flags(ifp);
350 if_get_mtu(ifp);
351 if_get_metric(ifp);
352
353 if (IS_ZEBRA_DEBUG_KERNEL)
354 zlog_debug("%s: interface %s index %d", __func__,
355 ifan->ifan_name, ifan->ifan_index);
356
357 return 0;
358 }
359 #endif /* RTM_IFANNOUNCE */
360
361 #ifdef HAVE_BSD_IFI_LINK_STATE
362 /* BSD link detect translation */
363 static void bsd_linkdetect_translate(struct if_msghdr *ifm)
364 {
365 if ((ifm->ifm_data.ifi_link_state >= LINK_STATE_UP)
366 || (ifm->ifm_data.ifi_link_state == LINK_STATE_UNKNOWN))
367 SET_FLAG(ifm->ifm_flags, IFF_RUNNING);
368 else
369 UNSET_FLAG(ifm->ifm_flags, IFF_RUNNING);
370 }
371 #endif /* HAVE_BSD_IFI_LINK_STATE */
372
373 static enum zebra_link_type sdl_to_zebra_link_type(unsigned int sdlt)
374 {
375 switch (sdlt) {
376 case IFT_ETHER:
377 return ZEBRA_LLT_ETHER;
378 case IFT_X25:
379 return ZEBRA_LLT_X25;
380 case IFT_FDDI:
381 return ZEBRA_LLT_FDDI;
382 case IFT_PPP:
383 return ZEBRA_LLT_PPP;
384 case IFT_LOOP:
385 return ZEBRA_LLT_LOOPBACK;
386 case IFT_SLIP:
387 return ZEBRA_LLT_SLIP;
388 case IFT_ARCNET:
389 return ZEBRA_LLT_ARCNET;
390 case IFT_ATM:
391 return ZEBRA_LLT_ATM;
392 case IFT_LOCALTALK:
393 return ZEBRA_LLT_LOCALTLK;
394 case IFT_HIPPI:
395 return ZEBRA_LLT_HIPPI;
396 #ifdef IFT_IEEE1394
397 case IFT_IEEE1394:
398 return ZEBRA_LLT_IEEE1394;
399 #endif
400
401 default:
402 return ZEBRA_LLT_UNKNOWN;
403 }
404 }
405
406 /*
407 * Handle struct if_msghdr obtained from reading routing socket or
408 * sysctl (from interface_list). There may or may not be sockaddrs
409 * present after the header.
410 */
411 int ifm_read(struct if_msghdr *ifm)
412 {
413 struct interface *ifp = NULL;
414 struct sockaddr_dl *sdl;
415 char ifname[IFNAMSIZ];
416 short ifnlen = 0;
417 int maskbit;
418 caddr_t cp;
419
420 /* terminate ifname at head (for strnlen) and tail (for safety) */
421 ifname[IFNAMSIZ - 1] = '\0';
422
423 /* paranoia: sanity check structure */
424 if (ifm->ifm_msglen < sizeof(struct if_msghdr)) {
425 flog_err(EC_ZEBRA_NETLINK_LENGTH_ERROR,
426 "ifm_read: ifm->ifm_msglen %d too short\n",
427 ifm->ifm_msglen);
428 return -1;
429 }
430
431 /*
432 * Check for a sockaddr_dl following the message. First, point to
433 * where a socakddr might be if one follows the message.
434 */
435 cp = (void *)(ifm + 1);
436
437 #ifdef SUNOS_5
438 /*
439 * XXX This behavior should be narrowed to only the kernel versions
440 * for which the structures returned do not match the headers.
441 *
442 * if_msghdr_t on 64 bit kernels in Solaris 9 and earlier versions
443 * is 12 bytes larger than the 32 bit version.
444 */
445 if (((struct sockaddr *)cp)->sa_family == AF_UNSPEC)
446 cp = cp + 12;
447 #endif
448
449 /* Look up for RTA_IFP and skip others. */
450 for (maskbit = 1; maskbit; maskbit <<= 1) {
451 if ((maskbit & ifm->ifm_addrs) == 0)
452 continue;
453 if (maskbit != RTA_IFP) {
454 cp += rta_get(cp, NULL, 0);
455 continue;
456 }
457
458 /* Save the pointer to the structure. */
459 sdl = (struct sockaddr_dl *)cp;
460 cp += rta_getsdlname(cp, ifname, &ifnlen);
461 }
462
463 if (IS_ZEBRA_DEBUG_KERNEL)
464 zlog_debug("%s: sdl ifname %s", __func__,
465 (ifnlen ? ifname : "(nil)"));
466
467 /*
468 * Look up on ifindex first, because ifindices are the primary handle
469 * for
470 * interfaces across the user/kernel boundary, for most systems. (Some
471 * messages, such as up/down status changes on NetBSD, do not include a
472 * sockaddr_dl).
473 */
474 if ((ifp = if_lookup_by_index(ifm->ifm_index, VRF_DEFAULT)) != NULL) {
475 /* we have an ifp, verify that the name matches as some systems,
476 * eg Solaris, have a 1:many association of ifindex:ifname
477 * if they dont match, we dont have the correct ifp and should
478 * set it back to NULL to let next check do lookup by name
479 */
480 if (ifnlen && (strncmp(ifp->name, ifname, IFNAMSIZ) != 0)) {
481 if (IS_ZEBRA_DEBUG_KERNEL)
482 zlog_debug(
483 "%s: ifp name %s doesn't match sdl name %s",
484 __func__, ifp->name, ifname);
485 ifp = NULL;
486 }
487 }
488
489 /*
490 * If we dont have an ifp, try looking up by name. Particularly as some
491 * systems (Solaris) have a 1:many mapping of ifindex:ifname - the
492 * ifname
493 * is therefore our unique handle to that interface.
494 *
495 * Interfaces specified in the configuration file for which the ifindex
496 * has not been determined will have ifindex == IFINDEX_INTERNAL, and
497 * such
498 * interfaces are found by this search, and then their ifindex values
499 * can
500 * be filled in.
501 */
502 if ((ifp == NULL) && ifnlen)
503 ifp = if_lookup_by_name(ifname, VRF_DEFAULT);
504
505 /*
506 * If ifp still does not exist or has an invalid index
507 * (IFINDEX_INTERNAL),
508 * create or fill in an interface.
509 */
510 if ((ifp == NULL) || (ifp->ifindex == IFINDEX_INTERNAL)) {
511 /*
512 * To create or fill in an interface, a sockaddr_dl (via
513 * RTA_IFP) is required.
514 */
515 if (!ifnlen) {
516 zlog_debug("Interface index %d (new) missing ifname\n",
517 ifm->ifm_index);
518 return -1;
519 }
520
521 #ifndef RTM_IFANNOUNCE
522 /* Down->Down interface should be ignored here.
523 * See further comment below.
524 */
525 if (!CHECK_FLAG(ifm->ifm_flags, IFF_UP))
526 return 0;
527 #endif /* !RTM_IFANNOUNCE */
528
529 if (ifp == NULL) {
530 /* Interface that zebra was not previously aware of, so
531 * create. */
532 ifp = if_create(ifname, VRF_DEFAULT);
533 if (IS_ZEBRA_DEBUG_KERNEL)
534 zlog_debug("%s: creating ifp for ifindex %d",
535 __func__, ifm->ifm_index);
536 }
537
538 if (IS_ZEBRA_DEBUG_KERNEL)
539 zlog_debug(
540 "%s: updated/created ifp, ifname %s, ifindex %d",
541 __func__, ifp->name, ifp->ifindex);
542 /*
543 * Fill in newly created interface structure, or larval
544 * structure with ifindex IFINDEX_INTERNAL.
545 */
546 if_set_index(ifp, ifm->ifm_index);
547
548 #ifdef HAVE_BSD_IFI_LINK_STATE /* translate BSD kernel msg for link-state */
549 bsd_linkdetect_translate(ifm);
550 #endif /* HAVE_BSD_IFI_LINK_STATE */
551
552 if_flags_update(ifp, ifm->ifm_flags);
553 #if defined(__bsdi__)
554 if_kvm_get_mtu(ifp);
555 #else
556 if_get_mtu(ifp);
557 #endif /* __bsdi__ */
558 if_get_metric(ifp);
559
560 /*
561 * XXX sockaddr_dl contents can be larger than the structure
562 * definition. There are 2 big families here:
563 * - BSD has sdl_len + sdl_data[16] + overruns sdl_data
564 * we MUST use sdl_len here or we'll truncate data.
565 * - Solaris has no sdl_len, but sdl_data[244]
566 * presumably, it's not going to run past that, so sizeof()
567 * is fine here.
568 * a nonzero ifnlen from rta_getsdlname() means sdl is valid
569 */
570 ifp->ll_type = ZEBRA_LLT_UNKNOWN;
571 ifp->hw_addr_len = 0;
572 if (ifnlen) {
573 #ifdef HAVE_STRUCT_SOCKADDR_DL_SDL_LEN
574 memcpy(&((struct zebra_if *)ifp->info)->sdl, sdl,
575 sdl->sdl_len);
576 #else
577 memcpy(&((struct zebra_if *)ifp->info)->sdl, sdl,
578 sizeof(struct sockaddr_dl));
579 #endif /* HAVE_STRUCT_SOCKADDR_DL_SDL_LEN */
580
581 ifp->ll_type = sdl_to_zebra_link_type(sdl->sdl_type);
582 if (sdl->sdl_alen <= sizeof(ifp->hw_addr)) {
583 memcpy(ifp->hw_addr, LLADDR(sdl),
584 sdl->sdl_alen);
585 ifp->hw_addr_len = sdl->sdl_alen;
586 }
587 }
588
589 if_add_update(ifp);
590 } else
591 /*
592 * Interface structure exists. Adjust stored flags from
593 * notification. If interface has up->down or down->up
594 * transition, call state change routines (to adjust routes,
595 * notify routing daemons, etc.). (Other flag changes are stored
596 * but apparently do not trigger action.)
597 */
598 {
599 if (ifp->ifindex != ifm->ifm_index) {
600 zlog_debug(
601 "%s: index mismatch, ifname %s, ifp index %d, "
602 "ifm index %d",
603 __func__, ifp->name, ifp->ifindex,
604 ifm->ifm_index);
605 return -1;
606 }
607
608 #ifdef HAVE_BSD_IFI_LINK_STATE /* translate BSD kernel msg for link-state */
609 bsd_linkdetect_translate(ifm);
610 #endif /* HAVE_BSD_IFI_LINK_STATE */
611
612 /* update flags and handle operative->inoperative transition, if
613 * any */
614 if_flags_update(ifp, ifm->ifm_flags);
615
616 #ifndef RTM_IFANNOUNCE
617 if (!if_is_up(ifp)) {
618 /* No RTM_IFANNOUNCE on this platform, so we can never
619 * distinguish between ~IFF_UP and delete. We must
620 * presume
621 * it has been deleted.
622 * Eg, Solaris will not notify us of unplumb.
623 *
624 * XXX: Fixme - this should be runtime detected
625 * So that a binary compiled on a system with IFANNOUNCE
626 * will still behave correctly if run on a platform
627 * without
628 */
629 if_delete_update(ifp);
630 }
631 #endif /* RTM_IFANNOUNCE */
632 if (if_is_up(ifp)) {
633 #if defined(__bsdi__)
634 if_kvm_get_mtu(ifp);
635 #else
636 if_get_mtu(ifp);
637 #endif /* __bsdi__ */
638 if_get_metric(ifp);
639 }
640 }
641
642 #ifdef HAVE_NET_RT_IFLIST
643 ifp->stats = ifm->ifm_data;
644 #endif /* HAVE_NET_RT_IFLIST */
645 ifp->speed = ifm->ifm_data.ifi_baudrate / 1000000;
646
647 if (IS_ZEBRA_DEBUG_KERNEL)
648 zlog_debug("%s: interface %s index %d", __func__, ifp->name,
649 ifp->ifindex);
650
651 return 0;
652 }
653
654 /* Address read from struct ifa_msghdr. */
655 static void ifam_read_mesg(struct ifa_msghdr *ifm, union sockunion *addr,
656 union sockunion *mask, union sockunion *brd,
657 char *ifname, short *ifnlen)
658 {
659 caddr_t pnt, end;
660 union sockunion dst;
661 union sockunion gateway;
662 int maskbit;
663
664 pnt = (caddr_t)(ifm + 1);
665 end = ((caddr_t)ifm) + ifm->ifam_msglen;
666
667 /* Be sure structure is cleared */
668 memset(mask, 0, sizeof(union sockunion));
669 memset(addr, 0, sizeof(union sockunion));
670 memset(brd, 0, sizeof(union sockunion));
671 memset(&dst, 0, sizeof(union sockunion));
672 memset(&gateway, 0, sizeof(union sockunion));
673
674 /* We fetch each socket variable into sockunion. */
675 for (maskbit = 1; maskbit; maskbit <<= 1) {
676 if ((maskbit & ifm->ifam_addrs) == 0)
677 continue;
678
679 switch (maskbit) {
680 case RTA_DST:
681 pnt += rta_get(pnt, &dst, sizeof(dst));
682 break;
683 case RTA_GATEWAY:
684 pnt += rta_get(pnt, &gateway, sizeof(gateway));
685 break;
686 case RTA_NETMASK:
687 pnt += rta_get(pnt, mask, sizeof(*mask));
688 break;
689 case RTA_IFP:
690 pnt += rta_getsdlname(pnt, ifname, ifnlen);
691 break;
692 case RTA_IFA:
693 pnt += rta_get(pnt, addr, sizeof(*addr));
694 break;
695 case RTA_BRD:
696 pnt += rta_get(pnt, brd, sizeof(*brd));
697 break;
698
699 default:
700 pnt += rta_get(pnt, NULL, 0);
701 break;
702 }
703
704 if (pnt > end) {
705 zlog_warn("%s: overflow detected (pnt:%p end:%p)",
706 __func__, pnt, end);
707 break;
708 }
709 }
710
711 if (IS_ZEBRA_DEBUG_KERNEL) {
712 int family = sockunion_family(addr);
713 switch (family) {
714 case AF_INET:
715 case AF_INET6: {
716 char buf[4][INET6_ADDRSTRLEN];
717 zlog_debug(
718 "%s: ifindex %d, ifname %s, ifam_addrs 0x%x, "
719 "ifam_flags 0x%x, addr %s/%d broad %s dst %s "
720 "gateway %s",
721 __func__, ifm->ifam_index,
722 (ifnlen ? ifname : "(nil)"), ifm->ifam_addrs,
723 ifm->ifam_flags,
724 inet_ntop(family, &addr->sin.sin_addr, buf[0],
725 sizeof(buf[0])),
726 ip_masklen(mask->sin.sin_addr),
727 inet_ntop(family, &brd->sin.sin_addr, buf[1],
728 sizeof(buf[1])),
729 inet_ntop(family, &dst.sin.sin_addr, buf[2],
730 sizeof(buf[2])),
731 inet_ntop(family, &gateway.sin.sin_addr, buf[3],
732 sizeof(buf[3])));
733 } break;
734 default:
735 zlog_debug("%s: ifindex %d, ifname %s, ifam_addrs 0x%x",
736 __func__, ifm->ifam_index,
737 (ifnlen ? ifname : "(nil)"),
738 ifm->ifam_addrs);
739 break;
740 }
741 }
742
743 /* Assert read up end point matches to end point */
744 pnt = (caddr_t)ROUNDUP((size_t)pnt);
745 if (pnt != end)
746 zlog_debug("ifam_read() doesn't read all socket data");
747 }
748
749 /* Interface's address information get. */
750 int ifam_read(struct ifa_msghdr *ifam)
751 {
752 struct interface *ifp = NULL;
753 union sockunion addr, mask, brd;
754 char ifname[INTERFACE_NAMSIZ];
755 short ifnlen = 0;
756 char isalias = 0;
757 int flags = 0;
758
759 ifname[0] = ifname[INTERFACE_NAMSIZ - 1] = '\0';
760
761 /* Allocate and read address information. */
762 ifam_read_mesg(ifam, &addr, &mask, &brd, ifname, &ifnlen);
763
764 if ((ifp = if_lookup_by_index(ifam->ifam_index, VRF_DEFAULT)) == NULL) {
765 flog_warn(EC_ZEBRA_UNKNOWN_INTERFACE,
766 "%s: no interface for ifname %s, index %d", __func__,
767 ifname, ifam->ifam_index);
768 return -1;
769 }
770
771 if (ifnlen && strncmp(ifp->name, ifname, INTERFACE_NAMSIZ))
772 isalias = 1;
773
774 /* N.B. The info in ifa_msghdr does not tell us whether the RTA_BRD
775 field contains a broadcast address or a peer address, so we are
776 forced to
777 rely upon the interface type. */
778 if (if_is_pointopoint(ifp))
779 SET_FLAG(flags, ZEBRA_IFA_PEER);
780
781 #if 0
782 /* it might seem cute to grab the interface metric here, however
783 * we're processing an address update message, and so some systems
784 * (e.g. FBSD) dont bother to fill in ifam_metric. Disabled, but left
785 * in deliberately, as comment.
786 */
787 ifp->metric = ifam->ifam_metric;
788 #endif
789
790 /* Add connected address. */
791 switch (sockunion_family(&addr)) {
792 case AF_INET:
793 if (ifam->ifam_type == RTM_NEWADDR)
794 connected_add_ipv4(ifp, flags, &addr.sin.sin_addr,
795 ip_masklen(mask.sin.sin_addr),
796 &brd.sin.sin_addr,
797 (isalias ? ifname : NULL));
798 else
799 connected_delete_ipv4(ifp, flags, &addr.sin.sin_addr,
800 ip_masklen(mask.sin.sin_addr),
801 &brd.sin.sin_addr);
802 break;
803 case AF_INET6:
804 /* Unset interface index from link-local address when IPv6 stack
805 is KAME. */
806 if (IN6_IS_ADDR_LINKLOCAL(&addr.sin6.sin6_addr)) {
807 SET_IN6_LINKLOCAL_IFINDEX(addr.sin6.sin6_addr, 0);
808 }
809
810 if (ifam->ifam_type == RTM_NEWADDR)
811 connected_add_ipv6(ifp, flags, &addr.sin6.sin6_addr,
812 NULL,
813 ip6_masklen(mask.sin6.sin6_addr),
814 (isalias ? ifname : NULL));
815 else
816 connected_delete_ipv6(ifp, &addr.sin6.sin6_addr, NULL,
817 ip6_masklen(mask.sin6.sin6_addr));
818 break;
819 default:
820 /* Unsupported family silently ignore... */
821 break;
822 }
823
824 /* Check interface flag for implicit up of the interface. */
825 if_refresh(ifp);
826
827 #ifdef SUNOS_5
828 /* In addition to lacking IFANNOUNCE, on SUNOS IFF_UP is strange.
829 * See comments for SUNOS_5 in interface.c::if_flags_mangle.
830 *
831 * Here we take care of case where the real IFF_UP was previously
832 * unset (as kept in struct zebra_if.primary_state) and the mangled
833 * IFF_UP (ie IFF_UP set || listcount(connected) has now transitioned
834 * to unset due to the lost non-primary address having DELADDR'd.
835 *
836 * we must delete the interface, because in between here and next
837 * event for this interface-name the administrator could unplumb
838 * and replumb the interface.
839 */
840 if (!if_is_up(ifp))
841 if_delete_update(ifp);
842 #endif /* SUNOS_5 */
843
844 return 0;
845 }
846
847 /* Interface function for reading kernel routing table information. */
848 static int rtm_read_mesg(struct rt_msghdr *rtm, union sockunion *dest,
849 union sockunion *mask, union sockunion *gate,
850 char *ifname, short *ifnlen)
851 {
852 caddr_t pnt, end;
853 int maskbit;
854
855 /* Pnt points out socket data start point. */
856 pnt = (caddr_t)(rtm + 1);
857 end = ((caddr_t)rtm) + rtm->rtm_msglen;
858
859 /* rt_msghdr version check. */
860 if (rtm->rtm_version != RTM_VERSION)
861 flog_warn(EC_ZEBRA_RTM_VERSION_MISMATCH,
862 "Routing message version different %d should be %d."
863 "This may cause problem\n",
864 rtm->rtm_version, RTM_VERSION);
865
866 /* Be sure structure is cleared */
867 memset(dest, 0, sizeof(union sockunion));
868 memset(gate, 0, sizeof(union sockunion));
869 memset(mask, 0, sizeof(union sockunion));
870
871 /* We fetch each socket variable into sockunion. */
872 /* We fetch each socket variable into sockunion. */
873 for (maskbit = 1; maskbit; maskbit <<= 1) {
874 if ((maskbit & rtm->rtm_addrs) == 0)
875 continue;
876
877 switch (maskbit) {
878 case RTA_DST:
879 pnt += rta_get(pnt, dest, sizeof(*dest));
880 break;
881 case RTA_GATEWAY:
882 pnt += rta_get(pnt, gate, sizeof(*gate));
883 break;
884 case RTA_NETMASK:
885 pnt += rta_get(pnt, mask, sizeof(*mask));
886 break;
887 case RTA_IFP:
888 pnt += rta_getsdlname(pnt, ifname, ifnlen);
889 break;
890
891 default:
892 pnt += rta_get(pnt, NULL, 0);
893 break;
894 }
895
896 if (pnt > end) {
897 zlog_warn("%s: overflow detected (pnt:%p end:%p)",
898 __func__, pnt, end);
899 break;
900 }
901 }
902
903 /* If there is netmask information set it's family same as
904 destination family*/
905 if (rtm->rtm_addrs & RTA_NETMASK)
906 mask->sa.sa_family = dest->sa.sa_family;
907
908 /* Assert read up to the end of pointer. */
909 if (pnt != end)
910 zlog_debug("rtm_read() doesn't read all socket data.");
911
912 return rtm->rtm_flags;
913 }
914
915 void rtm_read(struct rt_msghdr *rtm)
916 {
917 int flags;
918 uint8_t zebra_flags;
919 union sockunion dest, mask, gate;
920 char ifname[INTERFACE_NAMSIZ + 1];
921 short ifnlen = 0;
922 struct nexthop nh;
923 struct prefix p;
924 ifindex_t ifindex = 0;
925 afi_t afi;
926
927 zebra_flags = 0;
928
929 /* Read destination and netmask and gateway from rtm message
930 structure. */
931 flags = rtm_read_mesg(rtm, &dest, &mask, &gate, ifname, &ifnlen);
932 if (!(flags & RTF_DONE))
933 return;
934 if (IS_ZEBRA_DEBUG_KERNEL)
935 zlog_debug("%s: got rtm of type %d (%s)", __func__,
936 rtm->rtm_type,
937 lookup_msg(rtm_type_str, rtm->rtm_type, NULL));
938
939 #ifdef RTF_CLONED /*bsdi, netbsd 1.6*/
940 if (flags & RTF_CLONED)
941 return;
942 #endif
943 #ifdef RTF_WASCLONED /*freebsd*/
944 if (flags & RTF_WASCLONED)
945 return;
946 #endif
947
948 if ((rtm->rtm_type == RTM_ADD || rtm->rtm_type == RTM_CHANGE)
949 && !(flags & RTF_UP))
950 return;
951
952 /* This is connected route. */
953 if (!(flags & RTF_GATEWAY))
954 return;
955
956 if (flags & RTF_PROTO1)
957 SET_FLAG(zebra_flags, ZEBRA_FLAG_SELFROUTE);
958
959 memset(&nh, 0, sizeof(nh));
960
961 nh.vrf_id = VRF_DEFAULT;
962 /* This is a reject or blackhole route */
963 if (flags & RTF_REJECT) {
964 nh.type = NEXTHOP_TYPE_BLACKHOLE;
965 nh.bh_type = BLACKHOLE_REJECT;
966 } else if (flags & RTF_BLACKHOLE) {
967 nh.type = NEXTHOP_TYPE_BLACKHOLE;
968 nh.bh_type = BLACKHOLE_NULL;
969 }
970
971 /*
972 * Ignore our own messages.
973 */
974 if (rtm->rtm_type != RTM_GET && rtm->rtm_pid == pid)
975 return;
976
977 if (dest.sa.sa_family == AF_INET) {
978 afi = AFI_IP;
979 p.family = AF_INET;
980 p.u.prefix4 = dest.sin.sin_addr;
981 if (flags & RTF_HOST)
982 p.prefixlen = IPV4_MAX_PREFIXLEN;
983 else
984 p.prefixlen = ip_masklen(mask.sin.sin_addr);
985
986 if (!nh.type) {
987 nh.type = NEXTHOP_TYPE_IPV4;
988 nh.gate.ipv4 = gate.sin.sin_addr;
989 }
990 } else if (dest.sa.sa_family == AF_INET6) {
991 afi = AFI_IP6;
992 p.family = AF_INET6;
993 p.u.prefix6 = dest.sin6.sin6_addr;
994 if (flags & RTF_HOST)
995 p.prefixlen = IPV6_MAX_PREFIXLEN;
996 else
997 p.prefixlen = ip6_masklen(mask.sin6.sin6_addr);
998
999 #ifdef KAME
1000 if (IN6_IS_ADDR_LINKLOCAL(&gate.sin6.sin6_addr)) {
1001 ifindex = IN6_LINKLOCAL_IFINDEX(gate.sin6.sin6_addr);
1002 SET_IN6_LINKLOCAL_IFINDEX(gate.sin6.sin6_addr, 0);
1003 }
1004 #endif /* KAME */
1005
1006 if (!nh.type) {
1007 nh.type = ifindex ? NEXTHOP_TYPE_IPV6_IFINDEX
1008 : NEXTHOP_TYPE_IPV6;
1009 nh.gate.ipv6 = gate.sin6.sin6_addr;
1010 nh.ifindex = ifindex;
1011 }
1012 } else
1013 return;
1014
1015 /*
1016 * CHANGE: delete the old prefix, we have no further information
1017 * to specify the route really
1018 */
1019 if (rtm->rtm_type == RTM_CHANGE)
1020 rib_delete(afi, SAFI_UNICAST, VRF_DEFAULT, ZEBRA_ROUTE_KERNEL,
1021 0, zebra_flags, &p, NULL, NULL, 0, 0, 0, true);
1022 if (rtm->rtm_type == RTM_GET || rtm->rtm_type == RTM_ADD
1023 || rtm->rtm_type == RTM_CHANGE)
1024 rib_add(afi, SAFI_UNICAST, VRF_DEFAULT, ZEBRA_ROUTE_KERNEL, 0,
1025 zebra_flags, &p, NULL, &nh, 0, 0, 0, 0, 0);
1026 else
1027 rib_delete(afi, SAFI_UNICAST, VRF_DEFAULT, ZEBRA_ROUTE_KERNEL,
1028 0, zebra_flags, &p, NULL, &nh, 0, 0, 0, true);
1029 }
1030
1031 /* Interface function for the kernel routing table updates. Support
1032 * for RTM_CHANGE will be needed.
1033 * Exported only for rt_socket.c
1034 */
1035 int rtm_write(int message, union sockunion *dest, union sockunion *mask,
1036 union sockunion *gate, union sockunion *mpls, unsigned int index,
1037 enum blackhole_type bh_type, int metric)
1038 {
1039 int ret;
1040 caddr_t pnt;
1041 struct interface *ifp;
1042
1043 /* Sequencial number of routing message. */
1044 static int msg_seq = 0;
1045
1046 /* Struct of rt_msghdr and buffer for storing socket's data. */
1047 struct {
1048 struct rt_msghdr rtm;
1049 char buf[512];
1050 } msg;
1051
1052 if (dplane_routing_sock < 0)
1053 return ZEBRA_ERR_EPERM;
1054
1055 /* Clear and set rt_msghdr values */
1056 memset(&msg, 0, sizeof(struct rt_msghdr));
1057 msg.rtm.rtm_version = RTM_VERSION;
1058 msg.rtm.rtm_type = message;
1059 msg.rtm.rtm_seq = msg_seq++;
1060 msg.rtm.rtm_addrs = RTA_DST;
1061 msg.rtm.rtm_addrs |= RTA_GATEWAY;
1062 msg.rtm.rtm_flags = RTF_UP;
1063 #ifdef __OpenBSD__
1064 msg.rtm.rtm_flags |= RTF_MPATH;
1065 msg.rtm.rtm_fmask = RTF_MPLS;
1066 #endif
1067 msg.rtm.rtm_index = index;
1068
1069 if (metric != 0) {
1070 msg.rtm.rtm_rmx.rmx_hopcount = metric;
1071 msg.rtm.rtm_inits |= RTV_HOPCOUNT;
1072 }
1073
1074 ifp = if_lookup_by_index(index, VRF_DEFAULT);
1075
1076 if (gate && (message == RTM_ADD || message == RTM_CHANGE))
1077 msg.rtm.rtm_flags |= RTF_GATEWAY;
1078
1079 /* When RTF_CLONING is unavailable on BSD, should we set some
1080 * other flag instead?
1081 */
1082 #ifdef RTF_CLONING
1083 if (!gate && (message == RTM_ADD || message == RTM_CHANGE) && ifp
1084 && (ifp->flags & IFF_POINTOPOINT) == 0)
1085 msg.rtm.rtm_flags |= RTF_CLONING;
1086 #endif /* RTF_CLONING */
1087
1088 /* If no protocol specific gateway is specified, use link
1089 address for gateway. */
1090 if (!gate) {
1091 if (!ifp) {
1092 char dest_buf[INET_ADDRSTRLEN] = "NULL",
1093 mask_buf[INET_ADDRSTRLEN] = "255.255.255.255";
1094 if (dest)
1095 inet_ntop(AF_INET, &dest->sin.sin_addr,
1096 dest_buf, INET_ADDRSTRLEN);
1097 if (mask)
1098 inet_ntop(AF_INET, &mask->sin.sin_addr,
1099 mask_buf, INET_ADDRSTRLEN);
1100 flog_warn(
1101 EC_ZEBRA_RTM_NO_GATEWAY,
1102 "%s: %s/%s: gate == NULL and no gateway found for ifindex %d",
1103 __func__, dest_buf, mask_buf, index);
1104 return -1;
1105 }
1106 gate = (union sockunion *)&((struct zebra_if *)ifp->info)->sdl;
1107 }
1108
1109 if (mask)
1110 msg.rtm.rtm_addrs |= RTA_NETMASK;
1111 else if (message == RTM_ADD || message == RTM_CHANGE)
1112 msg.rtm.rtm_flags |= RTF_HOST;
1113
1114 #ifdef __OpenBSD__
1115 if (mpls) {
1116 msg.rtm.rtm_addrs |= RTA_SRC;
1117 msg.rtm.rtm_flags |= RTF_MPLS;
1118
1119 if (mpls->smpls.smpls_label
1120 != htonl(MPLS_LABEL_IMPLICIT_NULL << MPLS_LABEL_OFFSET))
1121 msg.rtm.rtm_mpls = MPLS_OP_PUSH;
1122 }
1123 #endif
1124
1125 /* Tagging route with flags */
1126 msg.rtm.rtm_flags |= (RTF_PROTO1);
1127
1128 switch (bh_type) {
1129 case BLACKHOLE_UNSPEC:
1130 break;
1131 case BLACKHOLE_REJECT:
1132 msg.rtm.rtm_flags |= RTF_REJECT;
1133 break;
1134 default:
1135 msg.rtm.rtm_flags |= RTF_BLACKHOLE;
1136 break;
1137 }
1138
1139
1140 #define SOCKADDRSET(X, R) \
1141 if (msg.rtm.rtm_addrs & (R)) { \
1142 int len = SAROUNDUP(X); \
1143 memcpy(pnt, (caddr_t)(X), len); \
1144 pnt += len; \
1145 }
1146
1147 pnt = (caddr_t)msg.buf;
1148
1149 /* Write each socket data into rtm message buffer */
1150 SOCKADDRSET(dest, RTA_DST);
1151 SOCKADDRSET(gate, RTA_GATEWAY);
1152 SOCKADDRSET(mask, RTA_NETMASK);
1153 #ifdef __OpenBSD__
1154 SOCKADDRSET(mpls, RTA_SRC);
1155 #endif
1156
1157 msg.rtm.rtm_msglen = pnt - (caddr_t)&msg;
1158
1159 ret = write(dplane_routing_sock, &msg, msg.rtm.rtm_msglen);
1160
1161 if (ret != msg.rtm.rtm_msglen) {
1162 if (errno == EEXIST)
1163 return ZEBRA_ERR_RTEXIST;
1164 if (errno == ENETUNREACH)
1165 return ZEBRA_ERR_RTUNREACH;
1166 if (errno == ESRCH)
1167 return ZEBRA_ERR_RTNOEXIST;
1168
1169 flog_err_sys(EC_LIB_SOCKET, "%s: write : %s (%d)", __func__,
1170 safe_strerror(errno), errno);
1171 return ZEBRA_ERR_KERNEL;
1172 }
1173 return ZEBRA_ERR_NOERROR;
1174 }
1175
1176
1177 #include "thread.h"
1178 #include "zebra/zserv.h"
1179
1180 /* For debug purpose. */
1181 static void rtmsg_debug(struct rt_msghdr *rtm)
1182 {
1183 zlog_debug("Kernel: Len: %d Type: %s", rtm->rtm_msglen,
1184 lookup_msg(rtm_type_str, rtm->rtm_type, NULL));
1185 rtm_flag_dump(rtm->rtm_flags);
1186 zlog_debug("Kernel: message seq %d", rtm->rtm_seq);
1187 zlog_debug("Kernel: pid %lld, rtm_addrs 0x%x", (long long)rtm->rtm_pid,
1188 rtm->rtm_addrs);
1189 }
1190
1191 /* This is pretty gross, better suggestions welcome -- mhandler */
1192 #ifndef RTAX_MAX
1193 #ifdef RTA_NUMBITS
1194 #define RTAX_MAX RTA_NUMBITS
1195 #else
1196 #define RTAX_MAX 8
1197 #endif /* RTA_NUMBITS */
1198 #endif /* RTAX_MAX */
1199
1200 /* Kernel routing table and interface updates via routing socket. */
1201 static int kernel_read(struct thread *thread)
1202 {
1203 int sock;
1204 int nbytes;
1205 struct rt_msghdr *rtm;
1206
1207 /*
1208 * This must be big enough for any message the kernel might send.
1209 * Rather than determining how many sockaddrs of what size might be
1210 * in each particular message, just use RTAX_MAX of sockaddr_storage
1211 * for each. Note that the sockaddrs must be after each message
1212 * definition, or rather after whichever happens to be the largest,
1213 * since the buffer needs to be big enough for a message and the
1214 * sockaddrs together.
1215 */
1216 union {
1217 /* Routing information. */
1218 struct {
1219 struct rt_msghdr rtm;
1220 struct sockaddr_storage addr[RTAX_MAX];
1221 } r;
1222
1223 /* Interface information. */
1224 struct {
1225 struct if_msghdr ifm;
1226 struct sockaddr_storage addr[RTAX_MAX];
1227 } im;
1228
1229 /* Interface address information. */
1230 struct {
1231 struct ifa_msghdr ifa;
1232 struct sockaddr_storage addr[RTAX_MAX];
1233 } ia;
1234
1235 #ifdef RTM_IFANNOUNCE
1236 /* Interface arrival/departure */
1237 struct {
1238 struct if_announcemsghdr ifan;
1239 struct sockaddr_storage addr[RTAX_MAX];
1240 } ian;
1241 #endif /* RTM_IFANNOUNCE */
1242
1243 } buf;
1244
1245 /* Fetch routing socket. */
1246 sock = THREAD_FD(thread);
1247
1248 nbytes = read(sock, &buf, sizeof buf);
1249
1250 if (nbytes <= 0) {
1251 if (nbytes < 0 && errno != EWOULDBLOCK && errno != EAGAIN)
1252 flog_err_sys(EC_LIB_SOCKET, "routing socket error: %s",
1253 safe_strerror(errno));
1254 return 0;
1255 }
1256
1257 thread_add_read(zebrad.master, kernel_read, NULL, sock, NULL);
1258
1259 if (IS_ZEBRA_DEBUG_KERNEL)
1260 rtmsg_debug(&buf.r.rtm);
1261
1262 rtm = &buf.r.rtm;
1263
1264 /*
1265 * Ensure that we didn't drop any data, so that processing routines
1266 * can assume they have the whole message.
1267 */
1268 if (rtm->rtm_msglen != nbytes) {
1269 zlog_debug(
1270 "kernel_read: rtm->rtm_msglen %d, nbytes %d, type %d\n",
1271 rtm->rtm_msglen, nbytes, rtm->rtm_type);
1272 return -1;
1273 }
1274
1275 switch (rtm->rtm_type) {
1276 case RTM_ADD:
1277 case RTM_DELETE:
1278 case RTM_CHANGE:
1279 rtm_read(rtm);
1280 break;
1281 case RTM_IFINFO:
1282 ifm_read(&buf.im.ifm);
1283 break;
1284 case RTM_NEWADDR:
1285 case RTM_DELADDR:
1286 ifam_read(&buf.ia.ifa);
1287 break;
1288 #ifdef RTM_IFANNOUNCE
1289 case RTM_IFANNOUNCE:
1290 ifan_read(&buf.ian.ifan);
1291 break;
1292 #endif /* RTM_IFANNOUNCE */
1293 default:
1294 if (IS_ZEBRA_DEBUG_KERNEL)
1295 zlog_debug("Unprocessed RTM_type: %d", rtm->rtm_type);
1296 break;
1297 }
1298 return 0;
1299 }
1300
1301 /* Make routing socket. */
1302 static void routing_socket(struct zebra_ns *zns)
1303 {
1304 frr_elevate_privs(&zserv_privs) {
1305 routing_sock = ns_socket(AF_ROUTE, SOCK_RAW, 0, zns->ns_id);
1306
1307 dplane_routing_sock =
1308 ns_socket(AF_ROUTE, SOCK_RAW, 0, zns->ns_id);
1309 }
1310
1311 if (routing_sock < 0) {
1312 flog_err_sys(EC_LIB_SOCKET, "Can't init kernel routing socket");
1313 return;
1314 }
1315
1316 if (dplane_routing_sock < 0) {
1317 flog_err_sys(EC_LIB_SOCKET,
1318 "Can't init kernel dataplane routing socket");
1319 return;
1320 }
1321
1322 /* XXX: Socket should be NONBLOCK, however as we currently
1323 * discard failed writes, this will lead to inconsistencies.
1324 * For now, socket must be blocking.
1325 */
1326 /*if (fcntl (routing_sock, F_SETFL, O_NONBLOCK) < 0)
1327 zlog_warn ("Can't set O_NONBLOCK to routing socket");*/
1328
1329 /* kernel_read needs rewrite. */
1330 thread_add_read(zebrad.master, kernel_read, NULL, routing_sock, NULL);
1331 }
1332
1333 /* Exported interface function. This function simply calls
1334 routing_socket (). */
1335 void kernel_init(struct zebra_ns *zns)
1336 {
1337 routing_socket(zns);
1338 }
1339
1340 void kernel_terminate(struct zebra_ns *zns, bool complete)
1341 {
1342 return;
1343 }
1344
1345 #endif /* !HAVE_NETLINK */