]> git.proxmox.com Git - mirror_frr.git/blame - zebra/kernel_socket.c
zebra: implement FreeBSD route attr handling
[mirror_frr.git] / zebra / kernel_socket.c
CommitLineData
718e3744 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 *
896014f4
DL
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
718e3744 19 */
20
21#include <zebra.h>
ddfeb486
DL
22
23#ifndef HAVE_NETLINK
24
8ccc7e80 25#include <net/if_types.h>
fe6c7157 26#ifdef __OpenBSD__
d3e2c74a
RW
27#include <netmpls/mpls.h>
28#endif
718e3744 29
30#include "if.h"
31#include "prefix.h"
32#include "sockunion.h"
33#include "connected.h"
34#include "memory.h"
4a1ab8e4 35#include "zebra_memory.h"
718e3744 36#include "ioctl.h"
37#include "log.h"
718e3744 38#include "table.h"
39#include "rib.h"
edd7c245 40#include "privs.h"
78104b9b 41#include "vrf.h"
174482ef 42#include "lib_errors.h"
718e3744 43
a37bd5e0 44#include "zebra/rt.h"
718e3744 45#include "zebra/interface.h"
46#include "zebra/zserv.h"
47#include "zebra/debug.h"
ec1a4283 48#include "zebra/kernel_socket.h"
8f7d9fc0 49#include "zebra/rib.h"
98c15db6 50#include "zebra/zebra_errors.h"
01ce7cba 51#include "zebra/zebra_ptm.h"
718e3744 52
edd7c245 53extern struct zebra_privs_t zserv_privs;
54
4bfbea8c 55/*
cfa0ed09
GT
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.
4bfbea8c 61 *
cfa0ed09
GT
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.
4bfbea8c 66 */
cb800e35
RZ
67#ifdef __APPLE__
68#define ROUNDUP_TYPE int
69#else
70#define ROUNDUP_TYPE long
71#endif
cfa0ed09
GT
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 */
e7f0e649 77
cb800e35 78/* OS X (Xcode as of 2014-12) is known not to define RT_ROUNDUP */
cfa0ed09
GT
79#if defined(RT_ROUNDUP)
80#define ROUNDUP(a) RT_ROUNDUP(a)
81#endif /* defined(RT_ROUNDUP) */
82
f7f9a982
DL
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
cfa0ed09
GT
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/*
adf580f7
DL
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.
cfa0ed09 99 */
d62a17ae 100#define ROUNDUP(a) \
101 ((a) > 0 ? (1 + (((a)-1) | (sizeof(ROUNDUP_TYPE) - 1))) \
102 : sizeof(ROUNDUP_TYPE))
cfa0ed09
GT
103
104#endif /* defined(ROUNDUP) */
105
cb800e35
RZ
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 */
4bfbea8c 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 */
6f0e3f6e 115#if defined(HAVE_STRUCT_SOCKADDR_SA_LEN)
4bfbea8c 116#define SAROUNDUP(X) ROUNDUP(((struct sockaddr *)(X))->sa_len)
56c1f7d8 117#else
4bfbea8c 118/*
119 * One would hope all fixed-size structure definitions are aligned,
120 * but round them up nonetheless.
121 */
d62a17ae 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))))
6f0e3f6e 130#endif /* HAVE_STRUCT_SOCKADDR_SA_LEN */
718e3744 131
e7f0e649
DL
132#endif /* !SA_SIZE */
133
718e3744 134/* Routing socket message types. */
d62a17ae 135const 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"},
7f7651eb 142#ifdef RTM_LOCK
d62a17ae 143 {RTM_LOCK, "RTM_LOCK"},
7f7651eb 144#endif /* RTM_LOCK */
9458b819 145#ifdef OLDADD
d62a17ae 146 {RTM_OLDADD, "RTM_OLDADD"},
9458b819
GT
147#endif /* RTM_OLDADD */
148#ifdef RTM_OLDDEL
d62a17ae 149 {RTM_OLDDEL, "RTM_OLDDEL"},
9458b819 150#endif /* RTM_OLDDEL */
d62a17ae 151 {RTM_RESOLVE, "RTM_RESOLVE"},
152 {RTM_NEWADDR, "RTM_NEWADDR"},
153 {RTM_DELADDR, "RTM_DELADDR"},
154 {RTM_IFINFO, "RTM_IFINFO"},
718e3744 155#ifdef RTM_OIFINFO
d62a17ae 156 {RTM_OIFINFO, "RTM_OIFINFO"},
718e3744 157#endif /* RTM_OIFINFO */
158#ifdef RTM_NEWMADDR
d62a17ae 159 {RTM_NEWMADDR, "RTM_NEWMADDR"},
718e3744 160#endif /* RTM_NEWMADDR */
161#ifdef RTM_DELMADDR
d62a17ae 162 {RTM_DELMADDR, "RTM_DELMADDR"},
718e3744 163#endif /* RTM_DELMADDR */
164#ifdef RTM_IFANNOUNCE
d62a17ae 165 {RTM_IFANNOUNCE, "RTM_IFANNOUNCE"},
718e3744 166#endif /* RTM_IFANNOUNCE */
d62a17ae 167 {0}};
168
169static 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"},
718e3744 176#ifdef RTF_MASK
d62a17ae 177 {RTF_MASK, "MASK"},
718e3744 178#endif /* RTF_MASK */
e6f148e6 179#ifdef RTF_CLONING
d62a17ae 180 {RTF_CLONING, "CLONING"},
e6f148e6 181#endif /* RTF_CLONING */
23b1f400 182#ifdef RTF_XRESOLVE
d62a17ae 183 {RTF_XRESOLVE, "XRESOLVE"},
23b1f400 184#endif /* RTF_XRESOLVE */
c54632ec 185#ifdef RTF_LLINFO
d62a17ae 186 {RTF_LLINFO, "LLINFO"},
c54632ec 187#endif /* RTF_LLINFO */
d62a17ae 188 {RTF_STATIC, "STATIC"},
189 {RTF_BLACKHOLE, "BLACKHOLE"},
6fe70d1b 190#ifdef RTF_PRIVATE
d62a17ae 191 {RTF_PRIVATE, "PRIVATE"},
6fe70d1b 192#endif /* RTF_PRIVATE */
d62a17ae 193 {RTF_PROTO1, "PROTO1"},
194 {RTF_PROTO2, "PROTO2"},
718e3744 195#ifdef RTF_PRCLONING
d62a17ae 196 {RTF_PRCLONING, "PRCLONING"},
718e3744 197#endif /* RTF_PRCLONING */
198#ifdef RTF_WASCLONED
d62a17ae 199 {RTF_WASCLONED, "WASCLONED"},
718e3744 200#endif /* RTF_WASCLONED */
201#ifdef RTF_PROTO3
d62a17ae 202 {RTF_PROTO3, "PROTO3"},
718e3744 203#endif /* RTF_PROTO3 */
204#ifdef RTF_PINNED
d62a17ae 205 {RTF_PINNED, "PINNED"},
718e3744 206#endif /* RTF_PINNED */
207#ifdef RTF_LOCAL
d62a17ae 208 {RTF_LOCAL, "LOCAL"},
718e3744 209#endif /* RTF_LOCAL */
210#ifdef RTF_BROADCAST
d62a17ae 211 {RTF_BROADCAST, "BROADCAST"},
718e3744 212#endif /* RTF_BROADCAST */
213#ifdef RTF_MULTICAST
d62a17ae 214 {RTF_MULTICAST, "MULTICAST"},
718e3744 215#endif /* RTF_MULTICAST */
6fe70d1b 216#ifdef RTF_MULTIRT
d62a17ae 217 {RTF_MULTIRT, "MULTIRT"},
6fe70d1b 218#endif /* RTF_MULTIRT */
219#ifdef RTF_SETSRC
d62a17ae 220 {RTF_SETSRC, "SETSRC"},
6fe70d1b 221#endif /* RTF_SETSRC */
d62a17ae 222 {0}};
718e3744 223
224/* Kernel routing update socket. */
225int routing_sock = -1;
226
80776aec
MS
227/* Kernel dataplane routing update socket, used in the dataplane pthread
228 * context.
229 */
230int dplane_routing_sock = -1;
231
718e3744 232/* Yes I'm checking ugly routing socket behavior. */
233/* #define DEBUG */
234
19fb538d 235size_t _rta_get(caddr_t sap, void *destp, size_t destlen, bool checkaf);
86c57af5 236size_t rta_get(caddr_t sap, void *dest, size_t destlen);
19fb538d 237size_t rta_getattr(caddr_t sap, void *destp, size_t destlen);
86c57af5
RZ
238size_t rta_getsdlname(caddr_t sap, void *dest, short *destlen);
239
718e3744 240/* Supported address family check. */
d62a17ae 241static inline int af_check(int family)
718e3744 242{
d62a17ae 243 if (family == AF_INET)
244 return 1;
245 if (family == AF_INET6)
246 return 1;
247 return 0;
718e3744 248}
6b0655a2 249
19fb538d 250size_t _rta_get(caddr_t sap, void *destp, size_t destlen, bool checkaf)
86c57af5
RZ
251{
252 struct sockaddr *sa = (struct sockaddr *)sap;
253 uint8_t *dest = destp;
254 size_t tlen, copylen;
255
256#ifdef HAVE_STRUCT_SOCKADDR_SA_LEN
257 copylen = sa->sa_len;
258 tlen = (copylen == 0) ? sizeof(ROUNDUP_TYPE) : ROUNDUP(copylen);
259#else /* !HAVE_STRUCT_SOCKADDR_SA_LEN */
260 copylen = tlen = SAROUNDUP(sap);
261#endif /* !HAVE_STRUCT_SOCKADDR_SA_LEN */
262
19fb538d
RZ
263 if (copylen > 0 && dest != NULL) {
264 if (checkaf && af_check(sa->sa_family) == 0)
265 return tlen;
266
86c57af5
RZ
267 if (copylen > destlen) {
268 zlog_warn("%s: destination buffer too small (%lu vs %lu)",
269 __func__, copylen, destlen);
270 memcpy(dest, sap, destlen);
271 } else
272 memcpy(dest, sap, copylen);
273 }
274
275 return tlen;
276}
277
19fb538d
RZ
278size_t rta_get(caddr_t sap, void *destp, size_t destlen)
279{
280 return _rta_get(sap, destp, destlen, true);
281}
282
283size_t rta_getattr(caddr_t sap, void *destp, size_t destlen)
284{
285 return _rta_get(sap, destp, destlen, false);
286}
287
86c57af5
RZ
288size_t rta_getsdlname(caddr_t sap, void *destp, short *destlen)
289{
290 struct sockaddr_dl *sdl = (struct sockaddr_dl *)sap;
291 struct sockaddr *sa = (struct sockaddr *)sap;
292 uint8_t *dest = destp;
293 size_t tlen, copylen;
294
295 copylen = sdl->sdl_nlen;
296#ifdef HAVE_STRUCT_SOCKADDR_SA_LEN
297 tlen = (sa->sa_len == 0) ? sizeof(ROUNDUP_TYPE) : ROUNDUP(sa->sa_len);
298#else /* !HAVE_STRUCT_SOCKADDR_SA_LEN */
299 tlen = SAROUNDUP(sap);
300#endif /* !HAVE_STRUCT_SOCKADDR_SA_LEN */
301
302 if (copylen > 0 && dest != NULL && sdl->sdl_family == AF_LINK) {
303 if (copylen > IFNAMSIZ) {
304 zlog_warn("%s: destination buffer too small (%lu vs %d)",
305 __func__, copylen, IFNAMSIZ);
306 memcpy(dest, sdl->sdl_data, IFNAMSIZ);
307 dest[IFNAMSIZ] = 0;
308 *destlen = IFNAMSIZ;
309 } else {
310 memcpy(dest, sdl->sdl_data, copylen);
311 dest[copylen] = 0;
312 *destlen = copylen;
313 }
314 } else
315 *destlen = 0;
316
317 return tlen;
318}
319
718e3744 320/* Dump routing table flag for debug purpose. */
d62a17ae 321static void rtm_flag_dump(int flag)
718e3744 322{
d62a17ae 323 const struct message *mes;
324 static char buf[BUFSIZ];
325
326 buf[0] = '\0';
327 for (mes = rtm_flag_str; mes->key != 0; mes++) {
328 if (mes->key & flag) {
329 strlcat(buf, mes->str, BUFSIZ);
330 strlcat(buf, " ", BUFSIZ);
331 }
718e3744 332 }
d62a17ae 333 zlog_debug("Kernel: %s", buf);
718e3744 334}
335
336#ifdef RTM_IFANNOUNCE
337/* Interface adding function */
d62a17ae 338static int ifan_read(struct if_announcemsghdr *ifan)
718e3744 339{
d62a17ae 340 struct interface *ifp;
341
342 ifp = if_lookup_by_index(ifan->ifan_index, VRF_DEFAULT);
343
344 if (ifp)
345 assert((ifp->ifindex == ifan->ifan_index)
346 || (ifp->ifindex == IFINDEX_INTERNAL));
347
9d303b37
DL
348 if ((ifp == NULL) || ((ifp->ifindex == IFINDEX_INTERNAL)
349 && (ifan->ifan_what == IFAN_ARRIVAL))) {
d62a17ae 350 if (IS_ZEBRA_DEBUG_KERNEL)
351 zlog_debug(
352 "%s: creating interface for ifindex %d, name %s",
353 __func__, ifan->ifan_index, ifan->ifan_name);
354
355 /* Create Interface */
8f90d89b 356 ifp = if_get_by_name(ifan->ifan_name, VRF_DEFAULT);
ff880b78 357 if_set_index(ifp, ifan->ifan_index);
d62a17ae 358
359 if_get_metric(ifp);
360 if_add_update(ifp);
361 } else if (ifp != NULL && ifan->ifan_what == IFAN_DEPARTURE)
362 if_delete_update(ifp);
363
364 if_get_flags(ifp);
365 if_get_mtu(ifp);
366 if_get_metric(ifp);
367
368 if (IS_ZEBRA_DEBUG_KERNEL)
369 zlog_debug("%s: interface %s index %d", __func__,
370 ifan->ifan_name, ifan->ifan_index);
371
372 return 0;
718e3744 373}
374#endif /* RTM_IFANNOUNCE */
375
9234b382 376#ifdef HAVE_BSD_IFI_LINK_STATE
c543a173 377/* BSD link detect translation */
d62a17ae 378static void bsd_linkdetect_translate(struct if_msghdr *ifm)
c543a173 379{
d62a17ae 380 if ((ifm->ifm_data.ifi_link_state >= LINK_STATE_UP)
381 || (ifm->ifm_data.ifi_link_state == LINK_STATE_UNKNOWN))
382 SET_FLAG(ifm->ifm_flags, IFF_RUNNING);
383 else
384 UNSET_FLAG(ifm->ifm_flags, IFF_RUNNING);
c543a173 385}
9234b382 386#endif /* HAVE_BSD_IFI_LINK_STATE */
c543a173 387
d62a17ae 388static enum zebra_link_type sdl_to_zebra_link_type(unsigned int sdlt)
8ccc7e80 389{
d62a17ae 390 switch (sdlt) {
391 case IFT_ETHER:
392 return ZEBRA_LLT_ETHER;
393 case IFT_X25:
394 return ZEBRA_LLT_X25;
395 case IFT_FDDI:
396 return ZEBRA_LLT_FDDI;
397 case IFT_PPP:
398 return ZEBRA_LLT_PPP;
399 case IFT_LOOP:
400 return ZEBRA_LLT_LOOPBACK;
401 case IFT_SLIP:
402 return ZEBRA_LLT_SLIP;
403 case IFT_ARCNET:
404 return ZEBRA_LLT_ARCNET;
405 case IFT_ATM:
406 return ZEBRA_LLT_ATM;
407 case IFT_LOCALTALK:
408 return ZEBRA_LLT_LOCALTLK;
409 case IFT_HIPPI:
410 return ZEBRA_LLT_HIPPI;
8ccc7e80 411#ifdef IFT_IEEE1394
d62a17ae 412 case IFT_IEEE1394:
413 return ZEBRA_LLT_IEEE1394;
8ccc7e80
TT
414#endif
415
d62a17ae 416 default:
417 return ZEBRA_LLT_UNKNOWN;
418 }
8ccc7e80
TT
419}
420
da26e3b6 421/*
422 * Handle struct if_msghdr obtained from reading routing socket or
423 * sysctl (from interface_list). There may or may not be sockaddrs
424 * present after the header.
425 */
d62a17ae 426int ifm_read(struct if_msghdr *ifm)
718e3744 427{
d62a17ae 428 struct interface *ifp = NULL;
429 struct sockaddr_dl *sdl;
430 char ifname[IFNAMSIZ];
431 short ifnlen = 0;
86c57af5 432 int maskbit;
d62a17ae 433 caddr_t cp;
434
435 /* terminate ifname at head (for strnlen) and tail (for safety) */
436 ifname[IFNAMSIZ - 1] = '\0';
437
438 /* paranoia: sanity check structure */
439 if (ifm->ifm_msglen < sizeof(struct if_msghdr)) {
e914ccbe 440 flog_err(EC_ZEBRA_NETLINK_LENGTH_ERROR,
1c50c1c0
QY
441 "ifm_read: ifm->ifm_msglen %d too short\n",
442 ifm->ifm_msglen);
d62a17ae 443 return -1;
444 }
445
446 /*
447 * Check for a sockaddr_dl following the message. First, point to
448 * where a socakddr might be if one follows the message.
449 */
450 cp = (void *)(ifm + 1);
718e3744 451
4bfbea8c 452#ifdef SUNOS_5
d62a17ae 453 /*
454 * XXX This behavior should be narrowed to only the kernel versions
455 * for which the structures returned do not match the headers.
456 *
457 * if_msghdr_t on 64 bit kernels in Solaris 9 and earlier versions
458 * is 12 bytes larger than the 32 bit version.
459 */
460 if (((struct sockaddr *)cp)->sa_family == AF_UNSPEC)
461 cp = cp + 12;
4bfbea8c 462#endif
3e95a074 463
86c57af5
RZ
464 /* Look up for RTA_IFP and skip others. */
465 for (maskbit = 1; maskbit; maskbit <<= 1) {
466 if ((maskbit & ifm->ifm_addrs) == 0)
467 continue;
468 if (maskbit != RTA_IFP) {
469 cp += rta_get(cp, NULL, 0);
470 continue;
471 }
472
473 /* Save the pointer to the structure. */
474 sdl = (struct sockaddr_dl *)cp;
475 cp += rta_getsdlname(cp, ifname, &ifnlen);
476 }
d62a17ae 477
478 if (IS_ZEBRA_DEBUG_KERNEL)
479 zlog_debug("%s: sdl ifname %s", __func__,
480 (ifnlen ? ifname : "(nil)"));
481
482 /*
483 * Look up on ifindex first, because ifindices are the primary handle
484 * for
485 * interfaces across the user/kernel boundary, for most systems. (Some
486 * messages, such as up/down status changes on NetBSD, do not include a
487 * sockaddr_dl).
488 */
489 if ((ifp = if_lookup_by_index(ifm->ifm_index, VRF_DEFAULT)) != NULL) {
490 /* we have an ifp, verify that the name matches as some systems,
491 * eg Solaris, have a 1:many association of ifindex:ifname
492 * if they dont match, we dont have the correct ifp and should
493 * set it back to NULL to let next check do lookup by name
494 */
495 if (ifnlen && (strncmp(ifp->name, ifname, IFNAMSIZ) != 0)) {
496 if (IS_ZEBRA_DEBUG_KERNEL)
497 zlog_debug(
0437e105 498 "%s: ifp name %s doesn't match sdl name %s",
d62a17ae 499 __func__, ifp->name, ifname);
500 ifp = NULL;
501 }
da26e3b6 502 }
5c78b3d0 503
d62a17ae 504 /*
505 * If we dont have an ifp, try looking up by name. Particularly as some
506 * systems (Solaris) have a 1:many mapping of ifindex:ifname - the
507 * ifname
508 * is therefore our unique handle to that interface.
509 *
510 * Interfaces specified in the configuration file for which the ifindex
511 * has not been determined will have ifindex == IFINDEX_INTERNAL, and
512 * such
513 * interfaces are found by this search, and then their ifindex values
514 * can
515 * be filled in.
516 */
517 if ((ifp == NULL) && ifnlen)
518 ifp = if_lookup_by_name(ifname, VRF_DEFAULT);
519
520 /*
521 * If ifp still does not exist or has an invalid index
522 * (IFINDEX_INTERNAL),
523 * create or fill in an interface.
524 */
525 if ((ifp == NULL) || (ifp->ifindex == IFINDEX_INTERNAL)) {
526 /*
527 * To create or fill in an interface, a sockaddr_dl (via
528 * RTA_IFP) is required.
529 */
530 if (!ifnlen) {
9df414fe
QY
531 zlog_debug("Interface index %d (new) missing ifname\n",
532 ifm->ifm_index);
d62a17ae 533 return -1;
534 }
535
5c78b3d0 536#ifndef RTM_IFANNOUNCE
d62a17ae 537 /* Down->Down interface should be ignored here.
538 * See further comment below.
539 */
540 if (!CHECK_FLAG(ifm->ifm_flags, IFF_UP))
541 return 0;
5c78b3d0 542#endif /* !RTM_IFANNOUNCE */
d62a17ae 543
544 if (ifp == NULL) {
545 /* Interface that zebra was not previously aware of, so
546 * create. */
bcc24579 547 ifp = if_create(ifname, VRF_DEFAULT);
d62a17ae 548 if (IS_ZEBRA_DEBUG_KERNEL)
549 zlog_debug("%s: creating ifp for ifindex %d",
550 __func__, ifm->ifm_index);
551 }
552
553 if (IS_ZEBRA_DEBUG_KERNEL)
554 zlog_debug(
555 "%s: updated/created ifp, ifname %s, ifindex %d",
556 __func__, ifp->name, ifp->ifindex);
557 /*
558 * Fill in newly created interface structure, or larval
559 * structure with ifindex IFINDEX_INTERNAL.
560 */
ff880b78 561 if_set_index(ifp, ifm->ifm_index);
d62a17ae 562
9234b382 563#ifdef HAVE_BSD_IFI_LINK_STATE /* translate BSD kernel msg for link-state */
d62a17ae 564 bsd_linkdetect_translate(ifm);
9234b382 565#endif /* HAVE_BSD_IFI_LINK_STATE */
c543a173 566
d62a17ae 567 if_flags_update(ifp, ifm->ifm_flags);
718e3744 568#if defined(__bsdi__)
d62a17ae 569 if_kvm_get_mtu(ifp);
718e3744 570#else
d62a17ae 571 if_get_mtu(ifp);
718e3744 572#endif /* __bsdi__ */
d62a17ae 573 if_get_metric(ifp);
574
575 /*
576 * XXX sockaddr_dl contents can be larger than the structure
577 * definition. There are 2 big families here:
578 * - BSD has sdl_len + sdl_data[16] + overruns sdl_data
579 * we MUST use sdl_len here or we'll truncate data.
580 * - Solaris has no sdl_len, but sdl_data[244]
581 * presumably, it's not going to run past that, so sizeof()
582 * is fine here.
86c57af5 583 * a nonzero ifnlen from rta_getsdlname() means sdl is valid
d62a17ae 584 */
585 ifp->ll_type = ZEBRA_LLT_UNKNOWN;
586 ifp->hw_addr_len = 0;
587 if (ifnlen) {
ca3ccd87 588#ifdef HAVE_STRUCT_SOCKADDR_DL_SDL_LEN
d62a17ae 589 memcpy(&((struct zebra_if *)ifp->info)->sdl, sdl,
590 sdl->sdl_len);
ca3ccd87 591#else
d62a17ae 592 memcpy(&((struct zebra_if *)ifp->info)->sdl, sdl,
593 sizeof(struct sockaddr_dl));
ca3ccd87 594#endif /* HAVE_STRUCT_SOCKADDR_DL_SDL_LEN */
8ccc7e80 595
d62a17ae 596 ifp->ll_type = sdl_to_zebra_link_type(sdl->sdl_type);
597 if (sdl->sdl_alen <= sizeof(ifp->hw_addr)) {
598 memcpy(ifp->hw_addr, LLADDR(sdl),
599 sdl->sdl_alen);
600 ifp->hw_addr_len = sdl->sdl_alen;
601 }
602 }
603
604 if_add_update(ifp);
605 } else
606 /*
607 * Interface structure exists. Adjust stored flags from
608 * notification. If interface has up->down or down->up
609 * transition, call state change routines (to adjust routes,
610 * notify routing daemons, etc.). (Other flag changes are stored
611 * but apparently do not trigger action.)
612 */
613 {
614 if (ifp->ifindex != ifm->ifm_index) {
9df414fe 615 zlog_debug(
d62a17ae 616 "%s: index mismatch, ifname %s, ifp index %d, "
617 "ifm index %d",
618 __func__, ifp->name, ifp->ifindex,
619 ifm->ifm_index);
620 return -1;
621 }
622
9234b382 623#ifdef HAVE_BSD_IFI_LINK_STATE /* translate BSD kernel msg for link-state */
d62a17ae 624 bsd_linkdetect_translate(ifm);
9234b382 625#endif /* HAVE_BSD_IFI_LINK_STATE */
c543a173 626
d62a17ae 627 /* update flags and handle operative->inoperative transition, if
628 * any */
629 if_flags_update(ifp, ifm->ifm_flags);
630
6eb8827d 631#ifndef RTM_IFANNOUNCE
d62a17ae 632 if (!if_is_up(ifp)) {
633 /* No RTM_IFANNOUNCE on this platform, so we can never
634 * distinguish between ~IFF_UP and delete. We must
635 * presume
636 * it has been deleted.
637 * Eg, Solaris will not notify us of unplumb.
638 *
639 * XXX: Fixme - this should be runtime detected
640 * So that a binary compiled on a system with IFANNOUNCE
641 * will still behave correctly if run on a platform
642 * without
643 */
644 if_delete_update(ifp);
645 }
6eb8827d 646#endif /* RTM_IFANNOUNCE */
d62a17ae 647 if (if_is_up(ifp)) {
1ba27564 648#if defined(__bsdi__)
d62a17ae 649 if_kvm_get_mtu(ifp);
1ba27564 650#else
d62a17ae 651 if_get_mtu(ifp);
1ba27564 652#endif /* __bsdi__ */
d62a17ae 653 if_get_metric(ifp);
654 }
655 }
5c78b3d0 656
718e3744 657#ifdef HAVE_NET_RT_IFLIST
d62a17ae 658 ifp->stats = ifm->ifm_data;
718e3744 659#endif /* HAVE_NET_RT_IFLIST */
04427fcb 660 ifp->speed = ifm->ifm_data.ifi_baudrate / 1000000;
718e3744 661
d62a17ae 662 if (IS_ZEBRA_DEBUG_KERNEL)
663 zlog_debug("%s: interface %s index %d", __func__, ifp->name,
664 ifp->ifindex);
718e3744 665
d62a17ae 666 return 0;
718e3744 667}
6b0655a2 668
718e3744 669/* Address read from struct ifa_msghdr. */
d62a17ae 670static void ifam_read_mesg(struct ifa_msghdr *ifm, union sockunion *addr,
671 union sockunion *mask, union sockunion *brd,
672 char *ifname, short *ifnlen)
718e3744 673{
d62a17ae 674 caddr_t pnt, end;
675 union sockunion dst;
676 union sockunion gateway;
86c57af5 677 int maskbit;
d62a17ae 678
679 pnt = (caddr_t)(ifm + 1);
680 end = ((caddr_t)ifm) + ifm->ifam_msglen;
681
682 /* Be sure structure is cleared */
683 memset(mask, 0, sizeof(union sockunion));
684 memset(addr, 0, sizeof(union sockunion));
685 memset(brd, 0, sizeof(union sockunion));
686 memset(&dst, 0, sizeof(union sockunion));
687 memset(&gateway, 0, sizeof(union sockunion));
688
689 /* We fetch each socket variable into sockunion. */
86c57af5
RZ
690 for (maskbit = 1; maskbit; maskbit <<= 1) {
691 if ((maskbit & ifm->ifam_addrs) == 0)
692 continue;
693
694 switch (maskbit) {
695 case RTA_DST:
696 pnt += rta_get(pnt, &dst, sizeof(dst));
697 break;
698 case RTA_GATEWAY:
699 pnt += rta_get(pnt, &gateway, sizeof(gateway));
700 break;
701 case RTA_NETMASK:
19fb538d 702 pnt += rta_getattr(pnt, mask, sizeof(*mask));
86c57af5
RZ
703 break;
704 case RTA_IFP:
705 pnt += rta_getsdlname(pnt, ifname, ifnlen);
706 break;
707 case RTA_IFA:
708 pnt += rta_get(pnt, addr, sizeof(*addr));
709 break;
710 case RTA_BRD:
711 pnt += rta_get(pnt, brd, sizeof(*brd));
712 break;
713
714 default:
715 pnt += rta_get(pnt, NULL, 0);
716 break;
717 }
718
719 if (pnt > end) {
720 zlog_warn("%s: overflow detected (pnt:%p end:%p)",
721 __func__, pnt, end);
722 break;
723 }
724 }
d62a17ae 725
726 if (IS_ZEBRA_DEBUG_KERNEL) {
727 int family = sockunion_family(addr);
728 switch (family) {
729 case AF_INET:
730 case AF_INET6: {
731 char buf[4][INET6_ADDRSTRLEN];
732 zlog_debug(
733 "%s: ifindex %d, ifname %s, ifam_addrs 0x%x, "
734 "ifam_flags 0x%x, addr %s/%d broad %s dst %s "
735 "gateway %s",
736 __func__, ifm->ifam_index,
737 (ifnlen ? ifname : "(nil)"), ifm->ifam_addrs,
738 ifm->ifam_flags,
739 inet_ntop(family, &addr->sin.sin_addr, buf[0],
740 sizeof(buf[0])),
741 ip_masklen(mask->sin.sin_addr),
742 inet_ntop(family, &brd->sin.sin_addr, buf[1],
743 sizeof(buf[1])),
744 inet_ntop(family, &dst.sin.sin_addr, buf[2],
745 sizeof(buf[2])),
746 inet_ntop(family, &gateway.sin.sin_addr, buf[3],
747 sizeof(buf[3])));
748 } break;
749 default:
750 zlog_debug("%s: ifindex %d, ifname %s, ifam_addrs 0x%x",
751 __func__, ifm->ifam_index,
752 (ifnlen ? ifname : "(nil)"),
753 ifm->ifam_addrs);
754 break;
755 }
756 }
757
758 /* Assert read up end point matches to end point */
86c57af5 759 pnt = (caddr_t)ROUNDUP((size_t)pnt);
d62a17ae 760 if (pnt != end)
9df414fe 761 zlog_debug("ifam_read() doesn't read all socket data");
718e3744 762}
763
764/* Interface's address information get. */
d62a17ae 765int ifam_read(struct ifa_msghdr *ifam)
718e3744 766{
d62a17ae 767 struct interface *ifp = NULL;
768 union sockunion addr, mask, brd;
769 char ifname[INTERFACE_NAMSIZ];
770 short ifnlen = 0;
771 char isalias = 0;
772 int flags = 0;
773
774 ifname[0] = ifname[INTERFACE_NAMSIZ - 1] = '\0';
775
776 /* Allocate and read address information. */
777 ifam_read_mesg(ifam, &addr, &mask, &brd, ifname, &ifnlen);
778
779 if ((ifp = if_lookup_by_index(ifam->ifam_index, VRF_DEFAULT)) == NULL) {
e914ccbe 780 flog_warn(EC_ZEBRA_UNKNOWN_INTERFACE,
9df414fe 781 "%s: no interface for ifname %s, index %d", __func__,
d62a17ae 782 ifname, ifam->ifam_index);
783 return -1;
784 }
785
786 if (ifnlen && strncmp(ifp->name, ifname, INTERFACE_NAMSIZ))
787 isalias = 1;
788
789 /* N.B. The info in ifa_msghdr does not tell us whether the RTA_BRD
790 field contains a broadcast address or a peer address, so we are
791 forced to
792 rely upon the interface type. */
793 if (if_is_pointopoint(ifp))
794 SET_FLAG(flags, ZEBRA_IFA_PEER);
7ab62c53 795
6502208c
PJ
796#if 0
797 /* it might seem cute to grab the interface metric here, however
798 * we're processing an address update message, and so some systems
799 * (e.g. FBSD) dont bother to fill in ifam_metric. Disabled, but left
800 * in deliberately, as comment.
801 */
d34b8991 802 ifp->metric = ifam->ifam_metric;
6502208c
PJ
803#endif
804
d62a17ae 805 /* Add connected address. */
806 switch (sockunion_family(&addr)) {
807 case AF_INET:
808 if (ifam->ifam_type == RTM_NEWADDR)
809 connected_add_ipv4(ifp, flags, &addr.sin.sin_addr,
810 ip_masklen(mask.sin.sin_addr),
811 &brd.sin.sin_addr,
812 (isalias ? ifname : NULL));
813 else
814 connected_delete_ipv4(ifp, flags, &addr.sin.sin_addr,
815 ip_masklen(mask.sin.sin_addr),
816 &brd.sin.sin_addr);
817 break;
818 case AF_INET6:
819 /* Unset interface index from link-local address when IPv6 stack
820 is KAME. */
821 if (IN6_IS_ADDR_LINKLOCAL(&addr.sin6.sin6_addr)) {
822 SET_IN6_LINKLOCAL_IFINDEX(addr.sin6.sin6_addr, 0);
823 }
824
825 if (ifam->ifam_type == RTM_NEWADDR)
826 connected_add_ipv6(ifp, flags, &addr.sin6.sin6_addr,
60c0687a 827 NULL,
d62a17ae 828 ip6_masklen(mask.sin6.sin6_addr),
d62a17ae 829 (isalias ? ifname : NULL));
830 else
60c0687a 831 connected_delete_ipv6(ifp, &addr.sin6.sin6_addr, NULL,
608105a7 832 ip6_masklen(mask.sin6.sin6_addr));
d62a17ae 833 break;
834 default:
835 /* Unsupported family silently ignore... */
836 break;
837 }
838
839 /* Check interface flag for implicit up of the interface. */
840 if_refresh(ifp);
5c78b3d0 841
842#ifdef SUNOS_5
d62a17ae 843 /* In addition to lacking IFANNOUNCE, on SUNOS IFF_UP is strange.
844 * See comments for SUNOS_5 in interface.c::if_flags_mangle.
845 *
846 * Here we take care of case where the real IFF_UP was previously
847 * unset (as kept in struct zebra_if.primary_state) and the mangled
848 * IFF_UP (ie IFF_UP set || listcount(connected) has now transitioned
849 * to unset due to the lost non-primary address having DELADDR'd.
850 *
851 * we must delete the interface, because in between here and next
852 * event for this interface-name the administrator could unplumb
853 * and replumb the interface.
854 */
855 if (!if_is_up(ifp))
856 if_delete_update(ifp);
5c78b3d0 857#endif /* SUNOS_5 */
d62a17ae 858
859 return 0;
718e3744 860}
6b0655a2 861
718e3744 862/* Interface function for reading kernel routing table information. */
d62a17ae 863static int rtm_read_mesg(struct rt_msghdr *rtm, union sockunion *dest,
864 union sockunion *mask, union sockunion *gate,
865 char *ifname, short *ifnlen)
718e3744 866{
d62a17ae 867 caddr_t pnt, end;
86c57af5 868 int maskbit;
d62a17ae 869
870 /* Pnt points out socket data start point. */
871 pnt = (caddr_t)(rtm + 1);
872 end = ((caddr_t)rtm) + rtm->rtm_msglen;
873
874 /* rt_msghdr version check. */
875 if (rtm->rtm_version != RTM_VERSION)
e914ccbe 876 flog_warn(EC_ZEBRA_RTM_VERSION_MISMATCH,
9df414fe
QY
877 "Routing message version different %d should be %d."
878 "This may cause problem\n",
879 rtm->rtm_version, RTM_VERSION);
d62a17ae 880
881 /* Be sure structure is cleared */
882 memset(dest, 0, sizeof(union sockunion));
883 memset(gate, 0, sizeof(union sockunion));
884 memset(mask, 0, sizeof(union sockunion));
885
886 /* We fetch each socket variable into sockunion. */
86c57af5
RZ
887 /* We fetch each socket variable into sockunion. */
888 for (maskbit = 1; maskbit; maskbit <<= 1) {
889 if ((maskbit & rtm->rtm_addrs) == 0)
890 continue;
891
892 switch (maskbit) {
893 case RTA_DST:
894 pnt += rta_get(pnt, dest, sizeof(*dest));
895 break;
896 case RTA_GATEWAY:
897 pnt += rta_get(pnt, gate, sizeof(*gate));
898 break;
899 case RTA_NETMASK:
900 pnt += rta_get(pnt, mask, sizeof(*mask));
901 break;
902 case RTA_IFP:
903 pnt += rta_getsdlname(pnt, ifname, ifnlen);
904 break;
905
906 default:
907 pnt += rta_get(pnt, NULL, 0);
908 break;
909 }
910
911 if (pnt > end) {
912 zlog_warn("%s: overflow detected (pnt:%p end:%p)",
913 __func__, pnt, end);
914 break;
915 }
916 }
d62a17ae 917
918 /* If there is netmask information set it's family same as
919 destination family*/
920 if (rtm->rtm_addrs & RTA_NETMASK)
921 mask->sa.sa_family = dest->sa.sa_family;
922
923 /* Assert read up to the end of pointer. */
924 if (pnt != end)
9df414fe 925 zlog_debug("rtm_read() doesn't read all socket data.");
d62a17ae 926
927 return rtm->rtm_flags;
718e3744 928}
929
d62a17ae 930void rtm_read(struct rt_msghdr *rtm)
718e3744 931{
d62a17ae 932 int flags;
d7c0a89a 933 uint8_t zebra_flags;
d62a17ae 934 union sockunion dest, mask, gate;
935 char ifname[INTERFACE_NAMSIZ + 1];
936 short ifnlen = 0;
fd36be7e 937 struct nexthop nh;
026745a2
DS
938 struct prefix p;
939 ifindex_t ifindex = 0;
940 afi_t afi;
d62a17ae 941
942 zebra_flags = 0;
943
944 /* Read destination and netmask and gateway from rtm message
945 structure. */
946 flags = rtm_read_mesg(rtm, &dest, &mask, &gate, ifname, &ifnlen);
947 if (!(flags & RTF_DONE))
948 return;
949 if (IS_ZEBRA_DEBUG_KERNEL)
950 zlog_debug("%s: got rtm of type %d (%s)", __func__,
951 rtm->rtm_type,
952 lookup_msg(rtm_type_str, rtm->rtm_type, NULL));
953
954#ifdef RTF_CLONED /*bsdi, netbsd 1.6*/
955 if (flags & RTF_CLONED)
956 return;
718e3744 957#endif
d62a17ae 958#ifdef RTF_WASCLONED /*freebsd*/
959 if (flags & RTF_WASCLONED)
960 return;
718e3744 961#endif
962
d62a17ae 963 if ((rtm->rtm_type == RTM_ADD || rtm->rtm_type == RTM_CHANGE)
964 && !(flags & RTF_UP))
965 return;
966
967 /* This is connected route. */
968 if (!(flags & RTF_GATEWAY))
969 return;
970
971 if (flags & RTF_PROTO1)
972 SET_FLAG(zebra_flags, ZEBRA_FLAG_SELFROUTE);
973
a8309422 974 memset(&nh, 0, sizeof(nh));
4a7371e9
DS
975
976 nh.vrf_id = VRF_DEFAULT;
d62a17ae 977 /* This is a reject or blackhole route */
a8309422
DL
978 if (flags & RTF_REJECT) {
979 nh.type = NEXTHOP_TYPE_BLACKHOLE;
980 nh.bh_type = BLACKHOLE_REJECT;
981 } else if (flags & RTF_BLACKHOLE) {
982 nh.type = NEXTHOP_TYPE_BLACKHOLE;
983 nh.bh_type = BLACKHOLE_NULL;
984 }
d62a17ae 985
c626d369
DS
986 /*
987 * Ignore our own messages.
988 */
989 if (rtm->rtm_type != RTM_GET && rtm->rtm_pid == pid)
990 return;
d62a17ae 991
d62a17ae 992 if (dest.sa.sa_family == AF_INET) {
026745a2 993 afi = AFI_IP;
d62a17ae 994 p.family = AF_INET;
995 p.u.prefix4 = dest.sin.sin_addr;
996 if (flags & RTF_HOST)
997 p.prefixlen = IPV4_MAX_PREFIXLEN;
998 else
999 p.prefixlen = ip_masklen(mask.sin.sin_addr);
1000
a8309422
DL
1001 if (!nh.type) {
1002 nh.type = NEXTHOP_TYPE_IPV4;
1003 nh.gate.ipv4 = gate.sin.sin_addr;
1004 }
026745a2
DS
1005 } else if (dest.sa.sa_family == AF_INET6) {
1006 afi = AFI_IP6;
d62a17ae 1007 p.family = AF_INET6;
1008 p.u.prefix6 = dest.sin6.sin6_addr;
1009 if (flags & RTF_HOST)
1010 p.prefixlen = IPV6_MAX_PREFIXLEN;
1011 else
1012 p.prefixlen = ip6_masklen(mask.sin6.sin6_addr);
718e3744 1013
1014#ifdef KAME
d62a17ae 1015 if (IN6_IS_ADDR_LINKLOCAL(&gate.sin6.sin6_addr)) {
1016 ifindex = IN6_LINKLOCAL_IFINDEX(gate.sin6.sin6_addr);
1017 SET_IN6_LINKLOCAL_IFINDEX(gate.sin6.sin6_addr, 0);
1018 }
718e3744 1019#endif /* KAME */
1020
a8309422
DL
1021 if (!nh.type) {
1022 nh.type = ifindex ? NEXTHOP_TYPE_IPV6_IFINDEX
1023 : NEXTHOP_TYPE_IPV6;
1024 nh.gate.ipv6 = gate.sin6.sin6_addr;
1025 nh.ifindex = ifindex;
1026 }
026745a2
DS
1027 } else
1028 return;
d62a17ae 1029
026745a2
DS
1030 /*
1031 * CHANGE: delete the old prefix, we have no further information
1032 * to specify the route really
1033 */
1034 if (rtm->rtm_type == RTM_CHANGE)
1035 rib_delete(afi, SAFI_UNICAST, VRF_DEFAULT, ZEBRA_ROUTE_KERNEL,
1036 0, zebra_flags, &p, NULL, NULL, 0, 0, 0, true);
1037 if (rtm->rtm_type == RTM_GET || rtm->rtm_type == RTM_ADD
1038 || rtm->rtm_type == RTM_CHANGE)
1039 rib_add(afi, SAFI_UNICAST, VRF_DEFAULT, ZEBRA_ROUTE_KERNEL, 0,
1040 zebra_flags, &p, NULL, &nh, 0, 0, 0, 0, 0);
1041 else
1042 rib_delete(afi, SAFI_UNICAST, VRF_DEFAULT, ZEBRA_ROUTE_KERNEL,
1043 0, zebra_flags, &p, NULL, &nh, 0, 0, 0, true);
718e3744 1044}
1045
1046/* Interface function for the kernel routing table updates. Support
6621ca86 1047 * for RTM_CHANGE will be needed.
1048 * Exported only for rt_socket.c
1049 */
d62a17ae 1050int rtm_write(int message, union sockunion *dest, union sockunion *mask,
1051 union sockunion *gate, union sockunion *mpls, unsigned int index,
a8309422 1052 enum blackhole_type bh_type, int metric)
718e3744 1053{
d62a17ae 1054 int ret;
1055 caddr_t pnt;
1056 struct interface *ifp;
1057
1058 /* Sequencial number of routing message. */
1059 static int msg_seq = 0;
1060
1061 /* Struct of rt_msghdr and buffer for storing socket's data. */
1062 struct {
1063 struct rt_msghdr rtm;
1064 char buf[512];
1065 } msg;
1066
80776aec 1067 if (dplane_routing_sock < 0)
d62a17ae 1068 return ZEBRA_ERR_EPERM;
1069
1070 /* Clear and set rt_msghdr values */
1071 memset(&msg, 0, sizeof(struct rt_msghdr));
1072 msg.rtm.rtm_version = RTM_VERSION;
1073 msg.rtm.rtm_type = message;
1074 msg.rtm.rtm_seq = msg_seq++;
1075 msg.rtm.rtm_addrs = RTA_DST;
1076 msg.rtm.rtm_addrs |= RTA_GATEWAY;
1077 msg.rtm.rtm_flags = RTF_UP;
fe6c7157 1078#ifdef __OpenBSD__
d62a17ae 1079 msg.rtm.rtm_flags |= RTF_MPATH;
1080 msg.rtm.rtm_fmask = RTF_MPLS;
d3e2c74a 1081#endif
d62a17ae 1082 msg.rtm.rtm_index = index;
718e3744 1083
d62a17ae 1084 if (metric != 0) {
1085 msg.rtm.rtm_rmx.rmx_hopcount = metric;
1086 msg.rtm.rtm_inits |= RTV_HOPCOUNT;
1087 }
718e3744 1088
d62a17ae 1089 ifp = if_lookup_by_index(index, VRF_DEFAULT);
718e3744 1090
d62a17ae 1091 if (gate && (message == RTM_ADD || message == RTM_CHANGE))
1092 msg.rtm.rtm_flags |= RTF_GATEWAY;
718e3744 1093
d62a17ae 1094/* When RTF_CLONING is unavailable on BSD, should we set some
1095 * other flag instead?
1096 */
e6f148e6 1097#ifdef RTF_CLONING
d62a17ae 1098 if (!gate && (message == RTM_ADD || message == RTM_CHANGE) && ifp
1099 && (ifp->flags & IFF_POINTOPOINT) == 0)
1100 msg.rtm.rtm_flags |= RTF_CLONING;
e6f148e6 1101#endif /* RTF_CLONING */
718e3744 1102
d62a17ae 1103 /* If no protocol specific gateway is specified, use link
1104 address for gateway. */
1105 if (!gate) {
1106 if (!ifp) {
1107 char dest_buf[INET_ADDRSTRLEN] = "NULL",
1108 mask_buf[INET_ADDRSTRLEN] = "255.255.255.255";
1109 if (dest)
1110 inet_ntop(AF_INET, &dest->sin.sin_addr,
1111 dest_buf, INET_ADDRSTRLEN);
1112 if (mask)
1113 inet_ntop(AF_INET, &mask->sin.sin_addr,
1114 mask_buf, INET_ADDRSTRLEN);
9df414fe 1115 flog_warn(
e914ccbe 1116 EC_ZEBRA_RTM_NO_GATEWAY,
d62a17ae 1117 "%s: %s/%s: gate == NULL and no gateway found for ifindex %d",
1118 __func__, dest_buf, mask_buf, index);
1119 return -1;
1120 }
1121 gate = (union sockunion *)&((struct zebra_if *)ifp->info)->sdl;
1122 }
1123
1124 if (mask)
1125 msg.rtm.rtm_addrs |= RTA_NETMASK;
1126 else if (message == RTM_ADD || message == RTM_CHANGE)
1127 msg.rtm.rtm_flags |= RTF_HOST;
718e3744 1128
fe6c7157 1129#ifdef __OpenBSD__
d62a17ae 1130 if (mpls) {
1131 msg.rtm.rtm_addrs |= RTA_SRC;
1132 msg.rtm.rtm_flags |= RTF_MPLS;
1133
1134 if (mpls->smpls.smpls_label
70e98a7f 1135 != htonl(MPLS_LABEL_IMPLICIT_NULL << MPLS_LABEL_OFFSET))
d62a17ae 1136 msg.rtm.rtm_mpls = MPLS_OP_PUSH;
1137 }
d3e2c74a
RW
1138#endif
1139
d62a17ae 1140 /* Tagging route with flags */
1141 msg.rtm.rtm_flags |= (RTF_PROTO1);
718e3744 1142
a8309422
DL
1143 switch (bh_type) {
1144 case BLACKHOLE_UNSPEC:
1145 break;
1146 case BLACKHOLE_REJECT:
d62a17ae 1147 msg.rtm.rtm_flags |= RTF_REJECT;
a8309422
DL
1148 break;
1149 default:
1150 msg.rtm.rtm_flags |= RTF_BLACKHOLE;
1151 break;
1152 }
81dfcaa2 1153
718e3744 1154
d62a17ae 1155#define SOCKADDRSET(X, R) \
1156 if (msg.rtm.rtm_addrs & (R)) { \
1157 int len = SAROUNDUP(X); \
1158 memcpy(pnt, (caddr_t)(X), len); \
1159 pnt += len; \
1160 }
718e3744 1161
d62a17ae 1162 pnt = (caddr_t)msg.buf;
718e3744 1163
d62a17ae 1164 /* Write each socket data into rtm message buffer */
1165 SOCKADDRSET(dest, RTA_DST);
1166 SOCKADDRSET(gate, RTA_GATEWAY);
1167 SOCKADDRSET(mask, RTA_NETMASK);
fe6c7157 1168#ifdef __OpenBSD__
d62a17ae 1169 SOCKADDRSET(mpls, RTA_SRC);
d3e2c74a 1170#endif
718e3744 1171
d62a17ae 1172 msg.rtm.rtm_msglen = pnt - (caddr_t)&msg;
1173
80776aec 1174 ret = write(dplane_routing_sock, &msg, msg.rtm.rtm_msglen);
d62a17ae 1175
1176 if (ret != msg.rtm.rtm_msglen) {
1177 if (errno == EEXIST)
1178 return ZEBRA_ERR_RTEXIST;
1179 if (errno == ENETUNREACH)
1180 return ZEBRA_ERR_RTUNREACH;
1181 if (errno == ESRCH)
1182 return ZEBRA_ERR_RTNOEXIST;
1183
450971aa 1184 flog_err_sys(EC_LIB_SOCKET, "%s: write : %s (%d)", __func__,
9df414fe 1185 safe_strerror(errno), errno);
d62a17ae 1186 return ZEBRA_ERR_KERNEL;
1187 }
1188 return ZEBRA_ERR_NOERROR;
718e3744 1189}
1190
6b0655a2 1191
718e3744 1192#include "thread.h"
1193#include "zebra/zserv.h"
1194
718e3744 1195/* For debug purpose. */
d62a17ae 1196static void rtmsg_debug(struct rt_msghdr *rtm)
718e3744 1197{
d62a17ae 1198 zlog_debug("Kernel: Len: %d Type: %s", rtm->rtm_msglen,
1199 lookup_msg(rtm_type_str, rtm->rtm_type, NULL));
1200 rtm_flag_dump(rtm->rtm_flags);
1201 zlog_debug("Kernel: message seq %d", rtm->rtm_seq);
1202 zlog_debug("Kernel: pid %lld, rtm_addrs 0x%x", (long long)rtm->rtm_pid,
1203 rtm->rtm_addrs);
718e3744 1204}
1205
1206/* This is pretty gross, better suggestions welcome -- mhandler */
1207#ifndef RTAX_MAX
1208#ifdef RTA_NUMBITS
1209#define RTAX_MAX RTA_NUMBITS
1210#else
1211#define RTAX_MAX 8
1212#endif /* RTA_NUMBITS */
1213#endif /* RTAX_MAX */
1214
1215/* Kernel routing table and interface updates via routing socket. */
d62a17ae 1216static int kernel_read(struct thread *thread)
718e3744 1217{
d62a17ae 1218 int sock;
1219 int nbytes;
1220 struct rt_msghdr *rtm;
1221
1222 /*
1223 * This must be big enough for any message the kernel might send.
1224 * Rather than determining how many sockaddrs of what size might be
1225 * in each particular message, just use RTAX_MAX of sockaddr_storage
1226 * for each. Note that the sockaddrs must be after each message
1227 * definition, or rather after whichever happens to be the largest,
1228 * since the buffer needs to be big enough for a message and the
1229 * sockaddrs together.
1230 */
1231 union {
1232 /* Routing information. */
1233 struct {
1234 struct rt_msghdr rtm;
1235 struct sockaddr_storage addr[RTAX_MAX];
1236 } r;
1237
1238 /* Interface information. */
1239 struct {
1240 struct if_msghdr ifm;
1241 struct sockaddr_storage addr[RTAX_MAX];
1242 } im;
1243
1244 /* Interface address information. */
1245 struct {
1246 struct ifa_msghdr ifa;
1247 struct sockaddr_storage addr[RTAX_MAX];
1248 } ia;
718e3744 1249
1250#ifdef RTM_IFANNOUNCE
d62a17ae 1251 /* Interface arrival/departure */
1252 struct {
1253 struct if_announcemsghdr ifan;
1254 struct sockaddr_storage addr[RTAX_MAX];
1255 } ian;
718e3744 1256#endif /* RTM_IFANNOUNCE */
1257
d62a17ae 1258 } buf;
718e3744 1259
d62a17ae 1260 /* Fetch routing socket. */
1261 sock = THREAD_FD(thread);
718e3744 1262
d62a17ae 1263 nbytes = read(sock, &buf, sizeof buf);
718e3744 1264
d62a17ae 1265 if (nbytes <= 0) {
1266 if (nbytes < 0 && errno != EWOULDBLOCK && errno != EAGAIN)
450971aa 1267 flog_err_sys(EC_LIB_SOCKET, "routing socket error: %s",
9df414fe 1268 safe_strerror(errno));
d62a17ae 1269 return 0;
1270 }
718e3744 1271
d62a17ae 1272 thread_add_read(zebrad.master, kernel_read, NULL, sock, NULL);
718e3744 1273
d62a17ae 1274 if (IS_ZEBRA_DEBUG_KERNEL)
1275 rtmsg_debug(&buf.r.rtm);
718e3744 1276
d62a17ae 1277 rtm = &buf.r.rtm;
718e3744 1278
d62a17ae 1279 /*
1280 * Ensure that we didn't drop any data, so that processing routines
1281 * can assume they have the whole message.
1282 */
1283 if (rtm->rtm_msglen != nbytes) {
9df414fe 1284 zlog_debug(
d62a17ae 1285 "kernel_read: rtm->rtm_msglen %d, nbytes %d, type %d\n",
1286 rtm->rtm_msglen, nbytes, rtm->rtm_type);
1287 return -1;
1288 }
1289
1290 switch (rtm->rtm_type) {
1291 case RTM_ADD:
1292 case RTM_DELETE:
1293 case RTM_CHANGE:
1294 rtm_read(rtm);
1295 break;
1296 case RTM_IFINFO:
1297 ifm_read(&buf.im.ifm);
1298 break;
1299 case RTM_NEWADDR:
1300 case RTM_DELADDR:
1301 ifam_read(&buf.ia.ifa);
1302 break;
718e3744 1303#ifdef RTM_IFANNOUNCE
d62a17ae 1304 case RTM_IFANNOUNCE:
1305 ifan_read(&buf.ian.ifan);
1306 break;
718e3744 1307#endif /* RTM_IFANNOUNCE */
d62a17ae 1308 default:
1309 if (IS_ZEBRA_DEBUG_KERNEL)
1310 zlog_debug("Unprocessed RTM_type: %d", rtm->rtm_type);
1311 break;
1312 }
1313 return 0;
718e3744 1314}
1315
1316/* Make routing socket. */
d62a17ae 1317static void routing_socket(struct zebra_ns *zns)
718e3744 1318{
6bb30c2c
DL
1319 frr_elevate_privs(&zserv_privs) {
1320 routing_sock = ns_socket(AF_ROUTE, SOCK_RAW, 0, zns->ns_id);
80776aec
MS
1321
1322 dplane_routing_sock =
1323 ns_socket(AF_ROUTE, SOCK_RAW, 0, zns->ns_id);
6bb30c2c 1324 }
d62a17ae 1325
1326 if (routing_sock < 0) {
1c50c1c0 1327 flog_err_sys(EC_LIB_SOCKET, "Can't init kernel routing socket");
d62a17ae 1328 return;
1329 }
1330
80776aec
MS
1331 if (dplane_routing_sock < 0) {
1332 flog_err_sys(EC_LIB_SOCKET,
1333 "Can't init kernel dataplane routing socket");
1334 return;
1335 }
1336
d62a17ae 1337 /* XXX: Socket should be NONBLOCK, however as we currently
1338 * discard failed writes, this will lead to inconsistencies.
1339 * For now, socket must be blocking.
1340 */
1341 /*if (fcntl (routing_sock, F_SETFL, O_NONBLOCK) < 0)
1342 zlog_warn ("Can't set O_NONBLOCK to routing socket");*/
1343
d62a17ae 1344 /* kernel_read needs rewrite. */
1345 thread_add_read(zebrad.master, kernel_read, NULL, routing_sock, NULL);
718e3744 1346}
1347
1348/* Exported interface function. This function simply calls
1349 routing_socket (). */
d62a17ae 1350void kernel_init(struct zebra_ns *zns)
8f7d9fc0 1351{
d62a17ae 1352 routing_socket(zns);
8f7d9fc0
FL
1353}
1354
62b8bb7a 1355void kernel_terminate(struct zebra_ns *zns, bool complete)
718e3744 1356{
d62a17ae 1357 return;
718e3744 1358}
ddfeb486
DL
1359
1360#endif /* !HAVE_NETLINK */