]>
Commit | Line | Data |
---|---|---|
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 | 53 | extern 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 | 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"}, | |
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 | ||
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"}, | |
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. */ | |
225 | int routing_sock = -1; | |
226 | ||
80776aec MS |
227 | /* Kernel dataplane routing update socket, used in the dataplane pthread |
228 | * context. | |
229 | */ | |
230 | int dplane_routing_sock = -1; | |
231 | ||
718e3744 | 232 | /* Yes I'm checking ugly routing socket behavior. */ |
233 | /* #define DEBUG */ | |
234 | ||
19fb538d | 235 | size_t _rta_get(caddr_t sap, void *destp, size_t destlen, bool checkaf); |
86c57af5 | 236 | size_t rta_get(caddr_t sap, void *dest, size_t destlen); |
19fb538d | 237 | size_t rta_getattr(caddr_t sap, void *destp, size_t destlen); |
86c57af5 RZ |
238 | size_t rta_getsdlname(caddr_t sap, void *dest, short *destlen); |
239 | ||
718e3744 | 240 | /* Supported address family check. */ |
d62a17ae | 241 | static 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 | 250 | size_t _rta_get(caddr_t sap, void *destp, size_t destlen, bool checkaf) |
86c57af5 RZ |
251 | { |
252 | struct sockaddr *sa = (struct sockaddr *)sap; | |
75e710df | 253 | struct sockaddr_dl *sdl; |
86c57af5 RZ |
254 | uint8_t *dest = destp; |
255 | size_t tlen, copylen; | |
256 | ||
257 | #ifdef HAVE_STRUCT_SOCKADDR_SA_LEN | |
258 | copylen = sa->sa_len; | |
259 | tlen = (copylen == 0) ? sizeof(ROUNDUP_TYPE) : ROUNDUP(copylen); | |
260 | #else /* !HAVE_STRUCT_SOCKADDR_SA_LEN */ | |
261 | copylen = tlen = SAROUNDUP(sap); | |
262 | #endif /* !HAVE_STRUCT_SOCKADDR_SA_LEN */ | |
263 | ||
19fb538d RZ |
264 | if (copylen > 0 && dest != NULL) { |
265 | if (checkaf && af_check(sa->sa_family) == 0) | |
266 | return tlen; | |
75e710df RZ |
267 | /* |
268 | * Handle sockaddr_dl corner case: | |
269 | * RTA_NETMASK might be AF_LINK, but it doesn't anything | |
270 | * relevant (e.g. zeroed out fields). Check for this | |
271 | * case and avoid warning log message. | |
272 | */ | |
273 | if (sa->sa_family == AF_LINK) { | |
274 | sdl = (struct sockaddr_dl *)sa; | |
275 | if (sdl->sdl_index == 0 || sdl->sdl_nlen == 0) | |
276 | copylen = sizeof(*sdl) - sizeof(sdl->sdl_data); | |
277 | } | |
19fb538d | 278 | |
86c57af5 RZ |
279 | if (copylen > destlen) { |
280 | zlog_warn("%s: destination buffer too small (%lu vs %lu)", | |
281 | __func__, copylen, destlen); | |
282 | memcpy(dest, sap, destlen); | |
283 | } else | |
284 | memcpy(dest, sap, copylen); | |
285 | } | |
286 | ||
287 | return tlen; | |
288 | } | |
289 | ||
19fb538d RZ |
290 | size_t rta_get(caddr_t sap, void *destp, size_t destlen) |
291 | { | |
292 | return _rta_get(sap, destp, destlen, true); | |
293 | } | |
294 | ||
295 | size_t rta_getattr(caddr_t sap, void *destp, size_t destlen) | |
296 | { | |
297 | return _rta_get(sap, destp, destlen, false); | |
298 | } | |
299 | ||
86c57af5 RZ |
300 | size_t rta_getsdlname(caddr_t sap, void *destp, short *destlen) |
301 | { | |
302 | struct sockaddr_dl *sdl = (struct sockaddr_dl *)sap; | |
303 | struct sockaddr *sa = (struct sockaddr *)sap; | |
304 | uint8_t *dest = destp; | |
305 | size_t tlen, copylen; | |
306 | ||
307 | copylen = sdl->sdl_nlen; | |
308 | #ifdef HAVE_STRUCT_SOCKADDR_SA_LEN | |
309 | tlen = (sa->sa_len == 0) ? sizeof(ROUNDUP_TYPE) : ROUNDUP(sa->sa_len); | |
310 | #else /* !HAVE_STRUCT_SOCKADDR_SA_LEN */ | |
311 | tlen = SAROUNDUP(sap); | |
312 | #endif /* !HAVE_STRUCT_SOCKADDR_SA_LEN */ | |
313 | ||
314 | if (copylen > 0 && dest != NULL && sdl->sdl_family == AF_LINK) { | |
315 | if (copylen > IFNAMSIZ) { | |
316 | zlog_warn("%s: destination buffer too small (%lu vs %d)", | |
317 | __func__, copylen, IFNAMSIZ); | |
318 | memcpy(dest, sdl->sdl_data, IFNAMSIZ); | |
319 | dest[IFNAMSIZ] = 0; | |
320 | *destlen = IFNAMSIZ; | |
321 | } else { | |
322 | memcpy(dest, sdl->sdl_data, copylen); | |
323 | dest[copylen] = 0; | |
324 | *destlen = copylen; | |
325 | } | |
326 | } else | |
327 | *destlen = 0; | |
328 | ||
329 | return tlen; | |
330 | } | |
331 | ||
718e3744 | 332 | /* Dump routing table flag for debug purpose. */ |
d62a17ae | 333 | static void rtm_flag_dump(int flag) |
718e3744 | 334 | { |
d62a17ae | 335 | const struct message *mes; |
336 | static char buf[BUFSIZ]; | |
337 | ||
338 | buf[0] = '\0'; | |
339 | for (mes = rtm_flag_str; mes->key != 0; mes++) { | |
340 | if (mes->key & flag) { | |
341 | strlcat(buf, mes->str, BUFSIZ); | |
342 | strlcat(buf, " ", BUFSIZ); | |
343 | } | |
718e3744 | 344 | } |
d62a17ae | 345 | zlog_debug("Kernel: %s", buf); |
718e3744 | 346 | } |
347 | ||
348 | #ifdef RTM_IFANNOUNCE | |
349 | /* Interface adding function */ | |
d62a17ae | 350 | static int ifan_read(struct if_announcemsghdr *ifan) |
718e3744 | 351 | { |
d62a17ae | 352 | struct interface *ifp; |
353 | ||
354 | ifp = if_lookup_by_index(ifan->ifan_index, VRF_DEFAULT); | |
355 | ||
356 | if (ifp) | |
357 | assert((ifp->ifindex == ifan->ifan_index) | |
358 | || (ifp->ifindex == IFINDEX_INTERNAL)); | |
359 | ||
9d303b37 DL |
360 | if ((ifp == NULL) || ((ifp->ifindex == IFINDEX_INTERNAL) |
361 | && (ifan->ifan_what == IFAN_ARRIVAL))) { | |
d62a17ae | 362 | if (IS_ZEBRA_DEBUG_KERNEL) |
363 | zlog_debug( | |
364 | "%s: creating interface for ifindex %d, name %s", | |
365 | __func__, ifan->ifan_index, ifan->ifan_name); | |
366 | ||
367 | /* Create Interface */ | |
8f90d89b | 368 | ifp = if_get_by_name(ifan->ifan_name, VRF_DEFAULT); |
ff880b78 | 369 | if_set_index(ifp, ifan->ifan_index); |
d62a17ae | 370 | |
371 | if_get_metric(ifp); | |
372 | if_add_update(ifp); | |
373 | } else if (ifp != NULL && ifan->ifan_what == IFAN_DEPARTURE) | |
374 | if_delete_update(ifp); | |
375 | ||
376 | if_get_flags(ifp); | |
377 | if_get_mtu(ifp); | |
378 | if_get_metric(ifp); | |
379 | ||
380 | if (IS_ZEBRA_DEBUG_KERNEL) | |
381 | zlog_debug("%s: interface %s index %d", __func__, | |
382 | ifan->ifan_name, ifan->ifan_index); | |
383 | ||
384 | return 0; | |
718e3744 | 385 | } |
386 | #endif /* RTM_IFANNOUNCE */ | |
387 | ||
9234b382 | 388 | #ifdef HAVE_BSD_IFI_LINK_STATE |
c543a173 | 389 | /* BSD link detect translation */ |
d62a17ae | 390 | static void bsd_linkdetect_translate(struct if_msghdr *ifm) |
c543a173 | 391 | { |
d62a17ae | 392 | if ((ifm->ifm_data.ifi_link_state >= LINK_STATE_UP) |
393 | || (ifm->ifm_data.ifi_link_state == LINK_STATE_UNKNOWN)) | |
394 | SET_FLAG(ifm->ifm_flags, IFF_RUNNING); | |
395 | else | |
396 | UNSET_FLAG(ifm->ifm_flags, IFF_RUNNING); | |
c543a173 | 397 | } |
9234b382 | 398 | #endif /* HAVE_BSD_IFI_LINK_STATE */ |
c543a173 | 399 | |
d62a17ae | 400 | static enum zebra_link_type sdl_to_zebra_link_type(unsigned int sdlt) |
8ccc7e80 | 401 | { |
d62a17ae | 402 | switch (sdlt) { |
403 | case IFT_ETHER: | |
404 | return ZEBRA_LLT_ETHER; | |
405 | case IFT_X25: | |
406 | return ZEBRA_LLT_X25; | |
407 | case IFT_FDDI: | |
408 | return ZEBRA_LLT_FDDI; | |
409 | case IFT_PPP: | |
410 | return ZEBRA_LLT_PPP; | |
411 | case IFT_LOOP: | |
412 | return ZEBRA_LLT_LOOPBACK; | |
413 | case IFT_SLIP: | |
414 | return ZEBRA_LLT_SLIP; | |
415 | case IFT_ARCNET: | |
416 | return ZEBRA_LLT_ARCNET; | |
417 | case IFT_ATM: | |
418 | return ZEBRA_LLT_ATM; | |
419 | case IFT_LOCALTALK: | |
420 | return ZEBRA_LLT_LOCALTLK; | |
421 | case IFT_HIPPI: | |
422 | return ZEBRA_LLT_HIPPI; | |
8ccc7e80 | 423 | #ifdef IFT_IEEE1394 |
d62a17ae | 424 | case IFT_IEEE1394: |
425 | return ZEBRA_LLT_IEEE1394; | |
8ccc7e80 TT |
426 | #endif |
427 | ||
d62a17ae | 428 | default: |
429 | return ZEBRA_LLT_UNKNOWN; | |
430 | } | |
8ccc7e80 TT |
431 | } |
432 | ||
da26e3b6 | 433 | /* |
434 | * Handle struct if_msghdr obtained from reading routing socket or | |
435 | * sysctl (from interface_list). There may or may not be sockaddrs | |
436 | * present after the header. | |
437 | */ | |
d62a17ae | 438 | int ifm_read(struct if_msghdr *ifm) |
718e3744 | 439 | { |
d62a17ae | 440 | struct interface *ifp = NULL; |
441 | struct sockaddr_dl *sdl; | |
442 | char ifname[IFNAMSIZ]; | |
443 | short ifnlen = 0; | |
86c57af5 | 444 | int maskbit; |
d62a17ae | 445 | caddr_t cp; |
446 | ||
447 | /* terminate ifname at head (for strnlen) and tail (for safety) */ | |
448 | ifname[IFNAMSIZ - 1] = '\0'; | |
449 | ||
450 | /* paranoia: sanity check structure */ | |
451 | if (ifm->ifm_msglen < sizeof(struct if_msghdr)) { | |
e914ccbe | 452 | flog_err(EC_ZEBRA_NETLINK_LENGTH_ERROR, |
1c50c1c0 QY |
453 | "ifm_read: ifm->ifm_msglen %d too short\n", |
454 | ifm->ifm_msglen); | |
d62a17ae | 455 | return -1; |
456 | } | |
457 | ||
458 | /* | |
459 | * Check for a sockaddr_dl following the message. First, point to | |
460 | * where a socakddr might be if one follows the message. | |
461 | */ | |
462 | cp = (void *)(ifm + 1); | |
718e3744 | 463 | |
4bfbea8c | 464 | #ifdef SUNOS_5 |
d62a17ae | 465 | /* |
466 | * XXX This behavior should be narrowed to only the kernel versions | |
467 | * for which the structures returned do not match the headers. | |
468 | * | |
469 | * if_msghdr_t on 64 bit kernels in Solaris 9 and earlier versions | |
470 | * is 12 bytes larger than the 32 bit version. | |
471 | */ | |
472 | if (((struct sockaddr *)cp)->sa_family == AF_UNSPEC) | |
473 | cp = cp + 12; | |
4bfbea8c | 474 | #endif |
3e95a074 | 475 | |
86c57af5 RZ |
476 | /* Look up for RTA_IFP and skip others. */ |
477 | for (maskbit = 1; maskbit; maskbit <<= 1) { | |
478 | if ((maskbit & ifm->ifm_addrs) == 0) | |
479 | continue; | |
480 | if (maskbit != RTA_IFP) { | |
481 | cp += rta_get(cp, NULL, 0); | |
482 | continue; | |
483 | } | |
484 | ||
485 | /* Save the pointer to the structure. */ | |
486 | sdl = (struct sockaddr_dl *)cp; | |
487 | cp += rta_getsdlname(cp, ifname, &ifnlen); | |
488 | } | |
d62a17ae | 489 | |
490 | if (IS_ZEBRA_DEBUG_KERNEL) | |
491 | zlog_debug("%s: sdl ifname %s", __func__, | |
492 | (ifnlen ? ifname : "(nil)")); | |
493 | ||
494 | /* | |
495 | * Look up on ifindex first, because ifindices are the primary handle | |
496 | * for | |
497 | * interfaces across the user/kernel boundary, for most systems. (Some | |
498 | * messages, such as up/down status changes on NetBSD, do not include a | |
499 | * sockaddr_dl). | |
500 | */ | |
501 | if ((ifp = if_lookup_by_index(ifm->ifm_index, VRF_DEFAULT)) != NULL) { | |
502 | /* we have an ifp, verify that the name matches as some systems, | |
503 | * eg Solaris, have a 1:many association of ifindex:ifname | |
504 | * if they dont match, we dont have the correct ifp and should | |
505 | * set it back to NULL to let next check do lookup by name | |
506 | */ | |
507 | if (ifnlen && (strncmp(ifp->name, ifname, IFNAMSIZ) != 0)) { | |
508 | if (IS_ZEBRA_DEBUG_KERNEL) | |
509 | zlog_debug( | |
0437e105 | 510 | "%s: ifp name %s doesn't match sdl name %s", |
d62a17ae | 511 | __func__, ifp->name, ifname); |
512 | ifp = NULL; | |
513 | } | |
da26e3b6 | 514 | } |
5c78b3d0 | 515 | |
d62a17ae | 516 | /* |
517 | * If we dont have an ifp, try looking up by name. Particularly as some | |
518 | * systems (Solaris) have a 1:many mapping of ifindex:ifname - the | |
519 | * ifname | |
520 | * is therefore our unique handle to that interface. | |
521 | * | |
522 | * Interfaces specified in the configuration file for which the ifindex | |
523 | * has not been determined will have ifindex == IFINDEX_INTERNAL, and | |
524 | * such | |
525 | * interfaces are found by this search, and then their ifindex values | |
526 | * can | |
527 | * be filled in. | |
528 | */ | |
529 | if ((ifp == NULL) && ifnlen) | |
530 | ifp = if_lookup_by_name(ifname, VRF_DEFAULT); | |
531 | ||
532 | /* | |
533 | * If ifp still does not exist or has an invalid index | |
534 | * (IFINDEX_INTERNAL), | |
535 | * create or fill in an interface. | |
536 | */ | |
537 | if ((ifp == NULL) || (ifp->ifindex == IFINDEX_INTERNAL)) { | |
538 | /* | |
539 | * To create or fill in an interface, a sockaddr_dl (via | |
540 | * RTA_IFP) is required. | |
541 | */ | |
542 | if (!ifnlen) { | |
9df414fe QY |
543 | zlog_debug("Interface index %d (new) missing ifname\n", |
544 | ifm->ifm_index); | |
d62a17ae | 545 | return -1; |
546 | } | |
547 | ||
5c78b3d0 | 548 | #ifndef RTM_IFANNOUNCE |
d62a17ae | 549 | /* Down->Down interface should be ignored here. |
550 | * See further comment below. | |
551 | */ | |
552 | if (!CHECK_FLAG(ifm->ifm_flags, IFF_UP)) | |
553 | return 0; | |
5c78b3d0 | 554 | #endif /* !RTM_IFANNOUNCE */ |
d62a17ae | 555 | |
556 | if (ifp == NULL) { | |
557 | /* Interface that zebra was not previously aware of, so | |
558 | * create. */ | |
bcc24579 | 559 | ifp = if_create(ifname, VRF_DEFAULT); |
d62a17ae | 560 | if (IS_ZEBRA_DEBUG_KERNEL) |
561 | zlog_debug("%s: creating ifp for ifindex %d", | |
562 | __func__, ifm->ifm_index); | |
563 | } | |
564 | ||
565 | if (IS_ZEBRA_DEBUG_KERNEL) | |
566 | zlog_debug( | |
567 | "%s: updated/created ifp, ifname %s, ifindex %d", | |
568 | __func__, ifp->name, ifp->ifindex); | |
569 | /* | |
570 | * Fill in newly created interface structure, or larval | |
571 | * structure with ifindex IFINDEX_INTERNAL. | |
572 | */ | |
ff880b78 | 573 | if_set_index(ifp, ifm->ifm_index); |
d62a17ae | 574 | |
9234b382 | 575 | #ifdef HAVE_BSD_IFI_LINK_STATE /* translate BSD kernel msg for link-state */ |
d62a17ae | 576 | bsd_linkdetect_translate(ifm); |
9234b382 | 577 | #endif /* HAVE_BSD_IFI_LINK_STATE */ |
c543a173 | 578 | |
d62a17ae | 579 | if_flags_update(ifp, ifm->ifm_flags); |
718e3744 | 580 | #if defined(__bsdi__) |
d62a17ae | 581 | if_kvm_get_mtu(ifp); |
718e3744 | 582 | #else |
d62a17ae | 583 | if_get_mtu(ifp); |
718e3744 | 584 | #endif /* __bsdi__ */ |
d62a17ae | 585 | if_get_metric(ifp); |
586 | ||
587 | /* | |
588 | * XXX sockaddr_dl contents can be larger than the structure | |
589 | * definition. There are 2 big families here: | |
590 | * - BSD has sdl_len + sdl_data[16] + overruns sdl_data | |
591 | * we MUST use sdl_len here or we'll truncate data. | |
592 | * - Solaris has no sdl_len, but sdl_data[244] | |
593 | * presumably, it's not going to run past that, so sizeof() | |
594 | * is fine here. | |
86c57af5 | 595 | * a nonzero ifnlen from rta_getsdlname() means sdl is valid |
d62a17ae | 596 | */ |
597 | ifp->ll_type = ZEBRA_LLT_UNKNOWN; | |
598 | ifp->hw_addr_len = 0; | |
599 | if (ifnlen) { | |
ca3ccd87 | 600 | #ifdef HAVE_STRUCT_SOCKADDR_DL_SDL_LEN |
d62a17ae | 601 | memcpy(&((struct zebra_if *)ifp->info)->sdl, sdl, |
602 | sdl->sdl_len); | |
ca3ccd87 | 603 | #else |
d62a17ae | 604 | memcpy(&((struct zebra_if *)ifp->info)->sdl, sdl, |
605 | sizeof(struct sockaddr_dl)); | |
ca3ccd87 | 606 | #endif /* HAVE_STRUCT_SOCKADDR_DL_SDL_LEN */ |
8ccc7e80 | 607 | |
d62a17ae | 608 | ifp->ll_type = sdl_to_zebra_link_type(sdl->sdl_type); |
609 | if (sdl->sdl_alen <= sizeof(ifp->hw_addr)) { | |
610 | memcpy(ifp->hw_addr, LLADDR(sdl), | |
611 | sdl->sdl_alen); | |
612 | ifp->hw_addr_len = sdl->sdl_alen; | |
613 | } | |
614 | } | |
615 | ||
616 | if_add_update(ifp); | |
617 | } else | |
618 | /* | |
619 | * Interface structure exists. Adjust stored flags from | |
620 | * notification. If interface has up->down or down->up | |
621 | * transition, call state change routines (to adjust routes, | |
622 | * notify routing daemons, etc.). (Other flag changes are stored | |
623 | * but apparently do not trigger action.) | |
624 | */ | |
625 | { | |
626 | if (ifp->ifindex != ifm->ifm_index) { | |
9df414fe | 627 | zlog_debug( |
d62a17ae | 628 | "%s: index mismatch, ifname %s, ifp index %d, " |
629 | "ifm index %d", | |
630 | __func__, ifp->name, ifp->ifindex, | |
631 | ifm->ifm_index); | |
632 | return -1; | |
633 | } | |
634 | ||
9234b382 | 635 | #ifdef HAVE_BSD_IFI_LINK_STATE /* translate BSD kernel msg for link-state */ |
d62a17ae | 636 | bsd_linkdetect_translate(ifm); |
9234b382 | 637 | #endif /* HAVE_BSD_IFI_LINK_STATE */ |
c543a173 | 638 | |
d62a17ae | 639 | /* update flags and handle operative->inoperative transition, if |
640 | * any */ | |
641 | if_flags_update(ifp, ifm->ifm_flags); | |
642 | ||
6eb8827d | 643 | #ifndef RTM_IFANNOUNCE |
d62a17ae | 644 | if (!if_is_up(ifp)) { |
645 | /* No RTM_IFANNOUNCE on this platform, so we can never | |
646 | * distinguish between ~IFF_UP and delete. We must | |
647 | * presume | |
648 | * it has been deleted. | |
649 | * Eg, Solaris will not notify us of unplumb. | |
650 | * | |
651 | * XXX: Fixme - this should be runtime detected | |
652 | * So that a binary compiled on a system with IFANNOUNCE | |
653 | * will still behave correctly if run on a platform | |
654 | * without | |
655 | */ | |
656 | if_delete_update(ifp); | |
657 | } | |
6eb8827d | 658 | #endif /* RTM_IFANNOUNCE */ |
d62a17ae | 659 | if (if_is_up(ifp)) { |
1ba27564 | 660 | #if defined(__bsdi__) |
d62a17ae | 661 | if_kvm_get_mtu(ifp); |
1ba27564 | 662 | #else |
d62a17ae | 663 | if_get_mtu(ifp); |
1ba27564 | 664 | #endif /* __bsdi__ */ |
d62a17ae | 665 | if_get_metric(ifp); |
666 | } | |
667 | } | |
5c78b3d0 | 668 | |
718e3744 | 669 | #ifdef HAVE_NET_RT_IFLIST |
d62a17ae | 670 | ifp->stats = ifm->ifm_data; |
718e3744 | 671 | #endif /* HAVE_NET_RT_IFLIST */ |
04427fcb | 672 | ifp->speed = ifm->ifm_data.ifi_baudrate / 1000000; |
718e3744 | 673 | |
d62a17ae | 674 | if (IS_ZEBRA_DEBUG_KERNEL) |
675 | zlog_debug("%s: interface %s index %d", __func__, ifp->name, | |
676 | ifp->ifindex); | |
718e3744 | 677 | |
d62a17ae | 678 | return 0; |
718e3744 | 679 | } |
6b0655a2 | 680 | |
718e3744 | 681 | /* Address read from struct ifa_msghdr. */ |
d62a17ae | 682 | static void ifam_read_mesg(struct ifa_msghdr *ifm, union sockunion *addr, |
683 | union sockunion *mask, union sockunion *brd, | |
684 | char *ifname, short *ifnlen) | |
718e3744 | 685 | { |
d62a17ae | 686 | caddr_t pnt, end; |
687 | union sockunion dst; | |
688 | union sockunion gateway; | |
86c57af5 | 689 | int maskbit; |
d62a17ae | 690 | |
691 | pnt = (caddr_t)(ifm + 1); | |
692 | end = ((caddr_t)ifm) + ifm->ifam_msglen; | |
693 | ||
694 | /* Be sure structure is cleared */ | |
695 | memset(mask, 0, sizeof(union sockunion)); | |
696 | memset(addr, 0, sizeof(union sockunion)); | |
697 | memset(brd, 0, sizeof(union sockunion)); | |
698 | memset(&dst, 0, sizeof(union sockunion)); | |
699 | memset(&gateway, 0, sizeof(union sockunion)); | |
700 | ||
701 | /* We fetch each socket variable into sockunion. */ | |
86c57af5 RZ |
702 | for (maskbit = 1; maskbit; maskbit <<= 1) { |
703 | if ((maskbit & ifm->ifam_addrs) == 0) | |
704 | continue; | |
705 | ||
706 | switch (maskbit) { | |
707 | case RTA_DST: | |
708 | pnt += rta_get(pnt, &dst, sizeof(dst)); | |
709 | break; | |
710 | case RTA_GATEWAY: | |
711 | pnt += rta_get(pnt, &gateway, sizeof(gateway)); | |
712 | break; | |
713 | case RTA_NETMASK: | |
19fb538d | 714 | pnt += rta_getattr(pnt, mask, sizeof(*mask)); |
86c57af5 RZ |
715 | break; |
716 | case RTA_IFP: | |
717 | pnt += rta_getsdlname(pnt, ifname, ifnlen); | |
718 | break; | |
719 | case RTA_IFA: | |
720 | pnt += rta_get(pnt, addr, sizeof(*addr)); | |
721 | break; | |
722 | case RTA_BRD: | |
723 | pnt += rta_get(pnt, brd, sizeof(*brd)); | |
724 | break; | |
725 | ||
726 | default: | |
727 | pnt += rta_get(pnt, NULL, 0); | |
728 | break; | |
729 | } | |
730 | ||
731 | if (pnt > end) { | |
732 | zlog_warn("%s: overflow detected (pnt:%p end:%p)", | |
733 | __func__, pnt, end); | |
734 | break; | |
735 | } | |
736 | } | |
d62a17ae | 737 | |
738 | if (IS_ZEBRA_DEBUG_KERNEL) { | |
7053cae4 | 739 | switch (sockunion_family(addr)) { |
d62a17ae | 740 | case AF_INET: |
741 | case AF_INET6: { | |
742 | char buf[4][INET6_ADDRSTRLEN]; | |
7053cae4 RZ |
743 | int masklen = |
744 | (sockunion_family(addr) == AF_INET) | |
745 | ? ip_masklen(mask->sin.sin_addr) | |
746 | : ip6_masklen(mask->sin6.sin6_addr); | |
d62a17ae | 747 | zlog_debug( |
748 | "%s: ifindex %d, ifname %s, ifam_addrs 0x%x, " | |
749 | "ifam_flags 0x%x, addr %s/%d broad %s dst %s " | |
750 | "gateway %s", | |
751 | __func__, ifm->ifam_index, | |
752 | (ifnlen ? ifname : "(nil)"), ifm->ifam_addrs, | |
753 | ifm->ifam_flags, | |
7053cae4 RZ |
754 | sockunion2str(addr, buf[0], sizeof(buf[0])), |
755 | masklen, | |
756 | sockunion2str(brd, buf[1], sizeof(buf[1])), | |
757 | sockunion2str(&dst, buf[2], sizeof(buf[2])), | |
758 | sockunion2str(&gateway, buf[2], | |
759 | sizeof(buf[2]))); | |
d62a17ae | 760 | } break; |
761 | default: | |
762 | zlog_debug("%s: ifindex %d, ifname %s, ifam_addrs 0x%x", | |
763 | __func__, ifm->ifam_index, | |
764 | (ifnlen ? ifname : "(nil)"), | |
765 | ifm->ifam_addrs); | |
766 | break; | |
767 | } | |
768 | } | |
769 | ||
770 | /* Assert read up end point matches to end point */ | |
86c57af5 | 771 | pnt = (caddr_t)ROUNDUP((size_t)pnt); |
2d100aa5 | 772 | if (pnt != (caddr_t)ROUNDUP((size_t)end)) |
9df414fe | 773 | zlog_debug("ifam_read() doesn't read all socket data"); |
718e3744 | 774 | } |
775 | ||
776 | /* Interface's address information get. */ | |
d62a17ae | 777 | int ifam_read(struct ifa_msghdr *ifam) |
718e3744 | 778 | { |
d62a17ae | 779 | struct interface *ifp = NULL; |
780 | union sockunion addr, mask, brd; | |
781 | char ifname[INTERFACE_NAMSIZ]; | |
782 | short ifnlen = 0; | |
783 | char isalias = 0; | |
784 | int flags = 0; | |
785 | ||
786 | ifname[0] = ifname[INTERFACE_NAMSIZ - 1] = '\0'; | |
787 | ||
788 | /* Allocate and read address information. */ | |
789 | ifam_read_mesg(ifam, &addr, &mask, &brd, ifname, &ifnlen); | |
790 | ||
791 | if ((ifp = if_lookup_by_index(ifam->ifam_index, VRF_DEFAULT)) == NULL) { | |
e914ccbe | 792 | flog_warn(EC_ZEBRA_UNKNOWN_INTERFACE, |
9df414fe | 793 | "%s: no interface for ifname %s, index %d", __func__, |
d62a17ae | 794 | ifname, ifam->ifam_index); |
795 | return -1; | |
796 | } | |
797 | ||
798 | if (ifnlen && strncmp(ifp->name, ifname, INTERFACE_NAMSIZ)) | |
799 | isalias = 1; | |
800 | ||
801 | /* N.B. The info in ifa_msghdr does not tell us whether the RTA_BRD | |
802 | field contains a broadcast address or a peer address, so we are | |
803 | forced to | |
804 | rely upon the interface type. */ | |
805 | if (if_is_pointopoint(ifp)) | |
806 | SET_FLAG(flags, ZEBRA_IFA_PEER); | |
7ab62c53 | 807 | |
6502208c PJ |
808 | #if 0 |
809 | /* it might seem cute to grab the interface metric here, however | |
810 | * we're processing an address update message, and so some systems | |
811 | * (e.g. FBSD) dont bother to fill in ifam_metric. Disabled, but left | |
812 | * in deliberately, as comment. | |
813 | */ | |
d34b8991 | 814 | ifp->metric = ifam->ifam_metric; |
6502208c PJ |
815 | #endif |
816 | ||
d62a17ae | 817 | /* Add connected address. */ |
818 | switch (sockunion_family(&addr)) { | |
819 | case AF_INET: | |
820 | if (ifam->ifam_type == RTM_NEWADDR) | |
821 | connected_add_ipv4(ifp, flags, &addr.sin.sin_addr, | |
822 | ip_masklen(mask.sin.sin_addr), | |
823 | &brd.sin.sin_addr, | |
824 | (isalias ? ifname : NULL)); | |
825 | else | |
826 | connected_delete_ipv4(ifp, flags, &addr.sin.sin_addr, | |
827 | ip_masklen(mask.sin.sin_addr), | |
828 | &brd.sin.sin_addr); | |
829 | break; | |
830 | case AF_INET6: | |
831 | /* Unset interface index from link-local address when IPv6 stack | |
832 | is KAME. */ | |
833 | if (IN6_IS_ADDR_LINKLOCAL(&addr.sin6.sin6_addr)) { | |
834 | SET_IN6_LINKLOCAL_IFINDEX(addr.sin6.sin6_addr, 0); | |
835 | } | |
836 | ||
837 | if (ifam->ifam_type == RTM_NEWADDR) | |
838 | connected_add_ipv6(ifp, flags, &addr.sin6.sin6_addr, | |
60c0687a | 839 | NULL, |
d62a17ae | 840 | ip6_masklen(mask.sin6.sin6_addr), |
d62a17ae | 841 | (isalias ? ifname : NULL)); |
842 | else | |
60c0687a | 843 | connected_delete_ipv6(ifp, &addr.sin6.sin6_addr, NULL, |
608105a7 | 844 | ip6_masklen(mask.sin6.sin6_addr)); |
d62a17ae | 845 | break; |
846 | default: | |
847 | /* Unsupported family silently ignore... */ | |
848 | break; | |
849 | } | |
850 | ||
851 | /* Check interface flag for implicit up of the interface. */ | |
852 | if_refresh(ifp); | |
5c78b3d0 | 853 | |
854 | #ifdef SUNOS_5 | |
d62a17ae | 855 | /* In addition to lacking IFANNOUNCE, on SUNOS IFF_UP is strange. |
856 | * See comments for SUNOS_5 in interface.c::if_flags_mangle. | |
857 | * | |
858 | * Here we take care of case where the real IFF_UP was previously | |
859 | * unset (as kept in struct zebra_if.primary_state) and the mangled | |
860 | * IFF_UP (ie IFF_UP set || listcount(connected) has now transitioned | |
861 | * to unset due to the lost non-primary address having DELADDR'd. | |
862 | * | |
863 | * we must delete the interface, because in between here and next | |
864 | * event for this interface-name the administrator could unplumb | |
865 | * and replumb the interface. | |
866 | */ | |
867 | if (!if_is_up(ifp)) | |
868 | if_delete_update(ifp); | |
5c78b3d0 | 869 | #endif /* SUNOS_5 */ |
d62a17ae | 870 | |
871 | return 0; | |
718e3744 | 872 | } |
6b0655a2 | 873 | |
718e3744 | 874 | /* Interface function for reading kernel routing table information. */ |
d62a17ae | 875 | static int rtm_read_mesg(struct rt_msghdr *rtm, union sockunion *dest, |
876 | union sockunion *mask, union sockunion *gate, | |
877 | char *ifname, short *ifnlen) | |
718e3744 | 878 | { |
d62a17ae | 879 | caddr_t pnt, end; |
86c57af5 | 880 | int maskbit; |
d62a17ae | 881 | |
882 | /* Pnt points out socket data start point. */ | |
883 | pnt = (caddr_t)(rtm + 1); | |
884 | end = ((caddr_t)rtm) + rtm->rtm_msglen; | |
885 | ||
886 | /* rt_msghdr version check. */ | |
887 | if (rtm->rtm_version != RTM_VERSION) | |
e914ccbe | 888 | flog_warn(EC_ZEBRA_RTM_VERSION_MISMATCH, |
9df414fe QY |
889 | "Routing message version different %d should be %d." |
890 | "This may cause problem\n", | |
891 | rtm->rtm_version, RTM_VERSION); | |
d62a17ae | 892 | |
893 | /* Be sure structure is cleared */ | |
894 | memset(dest, 0, sizeof(union sockunion)); | |
895 | memset(gate, 0, sizeof(union sockunion)); | |
896 | memset(mask, 0, sizeof(union sockunion)); | |
897 | ||
898 | /* We fetch each socket variable into sockunion. */ | |
86c57af5 RZ |
899 | /* We fetch each socket variable into sockunion. */ |
900 | for (maskbit = 1; maskbit; maskbit <<= 1) { | |
901 | if ((maskbit & rtm->rtm_addrs) == 0) | |
902 | continue; | |
903 | ||
904 | switch (maskbit) { | |
905 | case RTA_DST: | |
906 | pnt += rta_get(pnt, dest, sizeof(*dest)); | |
907 | break; | |
908 | case RTA_GATEWAY: | |
909 | pnt += rta_get(pnt, gate, sizeof(*gate)); | |
910 | break; | |
911 | case RTA_NETMASK: | |
912 | pnt += rta_get(pnt, mask, sizeof(*mask)); | |
913 | break; | |
914 | case RTA_IFP: | |
915 | pnt += rta_getsdlname(pnt, ifname, ifnlen); | |
916 | break; | |
917 | ||
918 | default: | |
919 | pnt += rta_get(pnt, NULL, 0); | |
920 | break; | |
921 | } | |
922 | ||
923 | if (pnt > end) { | |
924 | zlog_warn("%s: overflow detected (pnt:%p end:%p)", | |
925 | __func__, pnt, end); | |
926 | break; | |
927 | } | |
928 | } | |
d62a17ae | 929 | |
930 | /* If there is netmask information set it's family same as | |
931 | destination family*/ | |
932 | if (rtm->rtm_addrs & RTA_NETMASK) | |
933 | mask->sa.sa_family = dest->sa.sa_family; | |
934 | ||
935 | /* Assert read up to the end of pointer. */ | |
936 | if (pnt != end) | |
9df414fe | 937 | zlog_debug("rtm_read() doesn't read all socket data."); |
d62a17ae | 938 | |
939 | return rtm->rtm_flags; | |
718e3744 | 940 | } |
941 | ||
d62a17ae | 942 | void rtm_read(struct rt_msghdr *rtm) |
718e3744 | 943 | { |
d62a17ae | 944 | int flags; |
d7c0a89a | 945 | uint8_t zebra_flags; |
d62a17ae | 946 | union sockunion dest, mask, gate; |
947 | char ifname[INTERFACE_NAMSIZ + 1]; | |
948 | short ifnlen = 0; | |
fd36be7e | 949 | struct nexthop nh; |
026745a2 DS |
950 | struct prefix p; |
951 | ifindex_t ifindex = 0; | |
952 | afi_t afi; | |
d62a17ae | 953 | |
954 | zebra_flags = 0; | |
955 | ||
956 | /* Read destination and netmask and gateway from rtm message | |
957 | structure. */ | |
958 | flags = rtm_read_mesg(rtm, &dest, &mask, &gate, ifname, &ifnlen); | |
959 | if (!(flags & RTF_DONE)) | |
960 | return; | |
961 | if (IS_ZEBRA_DEBUG_KERNEL) | |
962 | zlog_debug("%s: got rtm of type %d (%s)", __func__, | |
963 | rtm->rtm_type, | |
964 | lookup_msg(rtm_type_str, rtm->rtm_type, NULL)); | |
965 | ||
966 | #ifdef RTF_CLONED /*bsdi, netbsd 1.6*/ | |
967 | if (flags & RTF_CLONED) | |
968 | return; | |
718e3744 | 969 | #endif |
d62a17ae | 970 | #ifdef RTF_WASCLONED /*freebsd*/ |
971 | if (flags & RTF_WASCLONED) | |
972 | return; | |
718e3744 | 973 | #endif |
974 | ||
d62a17ae | 975 | if ((rtm->rtm_type == RTM_ADD || rtm->rtm_type == RTM_CHANGE) |
976 | && !(flags & RTF_UP)) | |
977 | return; | |
978 | ||
979 | /* This is connected route. */ | |
980 | if (!(flags & RTF_GATEWAY)) | |
981 | return; | |
982 | ||
983 | if (flags & RTF_PROTO1) | |
984 | SET_FLAG(zebra_flags, ZEBRA_FLAG_SELFROUTE); | |
985 | ||
a8309422 | 986 | memset(&nh, 0, sizeof(nh)); |
4a7371e9 DS |
987 | |
988 | nh.vrf_id = VRF_DEFAULT; | |
d62a17ae | 989 | /* This is a reject or blackhole route */ |
a8309422 DL |
990 | if (flags & RTF_REJECT) { |
991 | nh.type = NEXTHOP_TYPE_BLACKHOLE; | |
992 | nh.bh_type = BLACKHOLE_REJECT; | |
993 | } else if (flags & RTF_BLACKHOLE) { | |
994 | nh.type = NEXTHOP_TYPE_BLACKHOLE; | |
995 | nh.bh_type = BLACKHOLE_NULL; | |
996 | } | |
d62a17ae | 997 | |
c626d369 DS |
998 | /* |
999 | * Ignore our own messages. | |
1000 | */ | |
1001 | if (rtm->rtm_type != RTM_GET && rtm->rtm_pid == pid) | |
1002 | return; | |
d62a17ae | 1003 | |
d62a17ae | 1004 | if (dest.sa.sa_family == AF_INET) { |
026745a2 | 1005 | afi = AFI_IP; |
d62a17ae | 1006 | p.family = AF_INET; |
1007 | p.u.prefix4 = dest.sin.sin_addr; | |
1008 | if (flags & RTF_HOST) | |
1009 | p.prefixlen = IPV4_MAX_PREFIXLEN; | |
1010 | else | |
1011 | p.prefixlen = ip_masklen(mask.sin.sin_addr); | |
1012 | ||
a8309422 DL |
1013 | if (!nh.type) { |
1014 | nh.type = NEXTHOP_TYPE_IPV4; | |
1015 | nh.gate.ipv4 = gate.sin.sin_addr; | |
1016 | } | |
026745a2 DS |
1017 | } else if (dest.sa.sa_family == AF_INET6) { |
1018 | afi = AFI_IP6; | |
d62a17ae | 1019 | p.family = AF_INET6; |
1020 | p.u.prefix6 = dest.sin6.sin6_addr; | |
1021 | if (flags & RTF_HOST) | |
1022 | p.prefixlen = IPV6_MAX_PREFIXLEN; | |
1023 | else | |
1024 | p.prefixlen = ip6_masklen(mask.sin6.sin6_addr); | |
718e3744 | 1025 | |
1026 | #ifdef KAME | |
d62a17ae | 1027 | if (IN6_IS_ADDR_LINKLOCAL(&gate.sin6.sin6_addr)) { |
1028 | ifindex = IN6_LINKLOCAL_IFINDEX(gate.sin6.sin6_addr); | |
1029 | SET_IN6_LINKLOCAL_IFINDEX(gate.sin6.sin6_addr, 0); | |
1030 | } | |
718e3744 | 1031 | #endif /* KAME */ |
1032 | ||
a8309422 DL |
1033 | if (!nh.type) { |
1034 | nh.type = ifindex ? NEXTHOP_TYPE_IPV6_IFINDEX | |
1035 | : NEXTHOP_TYPE_IPV6; | |
1036 | nh.gate.ipv6 = gate.sin6.sin6_addr; | |
1037 | nh.ifindex = ifindex; | |
1038 | } | |
026745a2 DS |
1039 | } else |
1040 | return; | |
d62a17ae | 1041 | |
026745a2 DS |
1042 | /* |
1043 | * CHANGE: delete the old prefix, we have no further information | |
1044 | * to specify the route really | |
1045 | */ | |
1046 | if (rtm->rtm_type == RTM_CHANGE) | |
1047 | rib_delete(afi, SAFI_UNICAST, VRF_DEFAULT, ZEBRA_ROUTE_KERNEL, | |
1048 | 0, zebra_flags, &p, NULL, NULL, 0, 0, 0, true); | |
1049 | if (rtm->rtm_type == RTM_GET || rtm->rtm_type == RTM_ADD | |
1050 | || rtm->rtm_type == RTM_CHANGE) | |
1051 | rib_add(afi, SAFI_UNICAST, VRF_DEFAULT, ZEBRA_ROUTE_KERNEL, 0, | |
1052 | zebra_flags, &p, NULL, &nh, 0, 0, 0, 0, 0); | |
1053 | else | |
1054 | rib_delete(afi, SAFI_UNICAST, VRF_DEFAULT, ZEBRA_ROUTE_KERNEL, | |
1055 | 0, zebra_flags, &p, NULL, &nh, 0, 0, 0, true); | |
718e3744 | 1056 | } |
1057 | ||
1058 | /* Interface function for the kernel routing table updates. Support | |
6621ca86 | 1059 | * for RTM_CHANGE will be needed. |
1060 | * Exported only for rt_socket.c | |
1061 | */ | |
d62a17ae | 1062 | int rtm_write(int message, union sockunion *dest, union sockunion *mask, |
1063 | union sockunion *gate, union sockunion *mpls, unsigned int index, | |
a8309422 | 1064 | enum blackhole_type bh_type, int metric) |
718e3744 | 1065 | { |
d62a17ae | 1066 | int ret; |
1067 | caddr_t pnt; | |
1068 | struct interface *ifp; | |
1069 | ||
1070 | /* Sequencial number of routing message. */ | |
1071 | static int msg_seq = 0; | |
1072 | ||
1073 | /* Struct of rt_msghdr and buffer for storing socket's data. */ | |
1074 | struct { | |
1075 | struct rt_msghdr rtm; | |
1076 | char buf[512]; | |
1077 | } msg; | |
1078 | ||
80776aec | 1079 | if (dplane_routing_sock < 0) |
d62a17ae | 1080 | return ZEBRA_ERR_EPERM; |
1081 | ||
1082 | /* Clear and set rt_msghdr values */ | |
1083 | memset(&msg, 0, sizeof(struct rt_msghdr)); | |
1084 | msg.rtm.rtm_version = RTM_VERSION; | |
1085 | msg.rtm.rtm_type = message; | |
1086 | msg.rtm.rtm_seq = msg_seq++; | |
1087 | msg.rtm.rtm_addrs = RTA_DST; | |
1088 | msg.rtm.rtm_addrs |= RTA_GATEWAY; | |
1089 | msg.rtm.rtm_flags = RTF_UP; | |
fe6c7157 | 1090 | #ifdef __OpenBSD__ |
d62a17ae | 1091 | msg.rtm.rtm_flags |= RTF_MPATH; |
1092 | msg.rtm.rtm_fmask = RTF_MPLS; | |
d3e2c74a | 1093 | #endif |
d62a17ae | 1094 | msg.rtm.rtm_index = index; |
718e3744 | 1095 | |
d62a17ae | 1096 | if (metric != 0) { |
1097 | msg.rtm.rtm_rmx.rmx_hopcount = metric; | |
1098 | msg.rtm.rtm_inits |= RTV_HOPCOUNT; | |
1099 | } | |
718e3744 | 1100 | |
d62a17ae | 1101 | ifp = if_lookup_by_index(index, VRF_DEFAULT); |
718e3744 | 1102 | |
d62a17ae | 1103 | if (gate && (message == RTM_ADD || message == RTM_CHANGE)) |
1104 | msg.rtm.rtm_flags |= RTF_GATEWAY; | |
718e3744 | 1105 | |
d62a17ae | 1106 | /* When RTF_CLONING is unavailable on BSD, should we set some |
1107 | * other flag instead? | |
1108 | */ | |
e6f148e6 | 1109 | #ifdef RTF_CLONING |
d62a17ae | 1110 | if (!gate && (message == RTM_ADD || message == RTM_CHANGE) && ifp |
1111 | && (ifp->flags & IFF_POINTOPOINT) == 0) | |
1112 | msg.rtm.rtm_flags |= RTF_CLONING; | |
e6f148e6 | 1113 | #endif /* RTF_CLONING */ |
718e3744 | 1114 | |
d62a17ae | 1115 | /* If no protocol specific gateway is specified, use link |
1116 | address for gateway. */ | |
1117 | if (!gate) { | |
1118 | if (!ifp) { | |
1119 | char dest_buf[INET_ADDRSTRLEN] = "NULL", | |
1120 | mask_buf[INET_ADDRSTRLEN] = "255.255.255.255"; | |
1121 | if (dest) | |
1122 | inet_ntop(AF_INET, &dest->sin.sin_addr, | |
1123 | dest_buf, INET_ADDRSTRLEN); | |
1124 | if (mask) | |
1125 | inet_ntop(AF_INET, &mask->sin.sin_addr, | |
1126 | mask_buf, INET_ADDRSTRLEN); | |
9df414fe | 1127 | flog_warn( |
e914ccbe | 1128 | EC_ZEBRA_RTM_NO_GATEWAY, |
d62a17ae | 1129 | "%s: %s/%s: gate == NULL and no gateway found for ifindex %d", |
1130 | __func__, dest_buf, mask_buf, index); | |
1131 | return -1; | |
1132 | } | |
1133 | gate = (union sockunion *)&((struct zebra_if *)ifp->info)->sdl; | |
1134 | } | |
1135 | ||
1136 | if (mask) | |
1137 | msg.rtm.rtm_addrs |= RTA_NETMASK; | |
1138 | else if (message == RTM_ADD || message == RTM_CHANGE) | |
1139 | msg.rtm.rtm_flags |= RTF_HOST; | |
718e3744 | 1140 | |
fe6c7157 | 1141 | #ifdef __OpenBSD__ |
d62a17ae | 1142 | if (mpls) { |
1143 | msg.rtm.rtm_addrs |= RTA_SRC; | |
1144 | msg.rtm.rtm_flags |= RTF_MPLS; | |
1145 | ||
1146 | if (mpls->smpls.smpls_label | |
70e98a7f | 1147 | != htonl(MPLS_LABEL_IMPLICIT_NULL << MPLS_LABEL_OFFSET)) |
d62a17ae | 1148 | msg.rtm.rtm_mpls = MPLS_OP_PUSH; |
1149 | } | |
d3e2c74a RW |
1150 | #endif |
1151 | ||
d62a17ae | 1152 | /* Tagging route with flags */ |
1153 | msg.rtm.rtm_flags |= (RTF_PROTO1); | |
718e3744 | 1154 | |
a8309422 DL |
1155 | switch (bh_type) { |
1156 | case BLACKHOLE_UNSPEC: | |
1157 | break; | |
1158 | case BLACKHOLE_REJECT: | |
d62a17ae | 1159 | msg.rtm.rtm_flags |= RTF_REJECT; |
a8309422 DL |
1160 | break; |
1161 | default: | |
1162 | msg.rtm.rtm_flags |= RTF_BLACKHOLE; | |
1163 | break; | |
1164 | } | |
81dfcaa2 | 1165 | |
718e3744 | 1166 | |
d62a17ae | 1167 | #define SOCKADDRSET(X, R) \ |
1168 | if (msg.rtm.rtm_addrs & (R)) { \ | |
1169 | int len = SAROUNDUP(X); \ | |
1170 | memcpy(pnt, (caddr_t)(X), len); \ | |
1171 | pnt += len; \ | |
1172 | } | |
718e3744 | 1173 | |
d62a17ae | 1174 | pnt = (caddr_t)msg.buf; |
718e3744 | 1175 | |
d62a17ae | 1176 | /* Write each socket data into rtm message buffer */ |
1177 | SOCKADDRSET(dest, RTA_DST); | |
1178 | SOCKADDRSET(gate, RTA_GATEWAY); | |
1179 | SOCKADDRSET(mask, RTA_NETMASK); | |
fe6c7157 | 1180 | #ifdef __OpenBSD__ |
d62a17ae | 1181 | SOCKADDRSET(mpls, RTA_SRC); |
d3e2c74a | 1182 | #endif |
718e3744 | 1183 | |
d62a17ae | 1184 | msg.rtm.rtm_msglen = pnt - (caddr_t)&msg; |
1185 | ||
80776aec | 1186 | ret = write(dplane_routing_sock, &msg, msg.rtm.rtm_msglen); |
d62a17ae | 1187 | |
1188 | if (ret != msg.rtm.rtm_msglen) { | |
1189 | if (errno == EEXIST) | |
1190 | return ZEBRA_ERR_RTEXIST; | |
1191 | if (errno == ENETUNREACH) | |
1192 | return ZEBRA_ERR_RTUNREACH; | |
1193 | if (errno == ESRCH) | |
1194 | return ZEBRA_ERR_RTNOEXIST; | |
1195 | ||
450971aa | 1196 | flog_err_sys(EC_LIB_SOCKET, "%s: write : %s (%d)", __func__, |
9df414fe | 1197 | safe_strerror(errno), errno); |
d62a17ae | 1198 | return ZEBRA_ERR_KERNEL; |
1199 | } | |
1200 | return ZEBRA_ERR_NOERROR; | |
718e3744 | 1201 | } |
1202 | ||
6b0655a2 | 1203 | |
718e3744 | 1204 | #include "thread.h" |
1205 | #include "zebra/zserv.h" | |
1206 | ||
718e3744 | 1207 | /* For debug purpose. */ |
d62a17ae | 1208 | static void rtmsg_debug(struct rt_msghdr *rtm) |
718e3744 | 1209 | { |
d62a17ae | 1210 | zlog_debug("Kernel: Len: %d Type: %s", rtm->rtm_msglen, |
1211 | lookup_msg(rtm_type_str, rtm->rtm_type, NULL)); | |
1212 | rtm_flag_dump(rtm->rtm_flags); | |
1213 | zlog_debug("Kernel: message seq %d", rtm->rtm_seq); | |
1214 | zlog_debug("Kernel: pid %lld, rtm_addrs 0x%x", (long long)rtm->rtm_pid, | |
1215 | rtm->rtm_addrs); | |
718e3744 | 1216 | } |
1217 | ||
1218 | /* This is pretty gross, better suggestions welcome -- mhandler */ | |
1219 | #ifndef RTAX_MAX | |
1220 | #ifdef RTA_NUMBITS | |
1221 | #define RTAX_MAX RTA_NUMBITS | |
1222 | #else | |
1223 | #define RTAX_MAX 8 | |
1224 | #endif /* RTA_NUMBITS */ | |
1225 | #endif /* RTAX_MAX */ | |
1226 | ||
1227 | /* Kernel routing table and interface updates via routing socket. */ | |
d62a17ae | 1228 | static int kernel_read(struct thread *thread) |
718e3744 | 1229 | { |
d62a17ae | 1230 | int sock; |
1231 | int nbytes; | |
1232 | struct rt_msghdr *rtm; | |
1233 | ||
1234 | /* | |
1235 | * This must be big enough for any message the kernel might send. | |
1236 | * Rather than determining how many sockaddrs of what size might be | |
1237 | * in each particular message, just use RTAX_MAX of sockaddr_storage | |
1238 | * for each. Note that the sockaddrs must be after each message | |
1239 | * definition, or rather after whichever happens to be the largest, | |
1240 | * since the buffer needs to be big enough for a message and the | |
1241 | * sockaddrs together. | |
1242 | */ | |
1243 | union { | |
1244 | /* Routing information. */ | |
1245 | struct { | |
1246 | struct rt_msghdr rtm; | |
1247 | struct sockaddr_storage addr[RTAX_MAX]; | |
1248 | } r; | |
1249 | ||
1250 | /* Interface information. */ | |
1251 | struct { | |
1252 | struct if_msghdr ifm; | |
1253 | struct sockaddr_storage addr[RTAX_MAX]; | |
1254 | } im; | |
1255 | ||
1256 | /* Interface address information. */ | |
1257 | struct { | |
1258 | struct ifa_msghdr ifa; | |
1259 | struct sockaddr_storage addr[RTAX_MAX]; | |
1260 | } ia; | |
718e3744 | 1261 | |
1262 | #ifdef RTM_IFANNOUNCE | |
d62a17ae | 1263 | /* Interface arrival/departure */ |
1264 | struct { | |
1265 | struct if_announcemsghdr ifan; | |
1266 | struct sockaddr_storage addr[RTAX_MAX]; | |
1267 | } ian; | |
718e3744 | 1268 | #endif /* RTM_IFANNOUNCE */ |
1269 | ||
d62a17ae | 1270 | } buf; |
718e3744 | 1271 | |
d62a17ae | 1272 | /* Fetch routing socket. */ |
1273 | sock = THREAD_FD(thread); | |
718e3744 | 1274 | |
d62a17ae | 1275 | nbytes = read(sock, &buf, sizeof buf); |
718e3744 | 1276 | |
d62a17ae | 1277 | if (nbytes <= 0) { |
1278 | if (nbytes < 0 && errno != EWOULDBLOCK && errno != EAGAIN) | |
450971aa | 1279 | flog_err_sys(EC_LIB_SOCKET, "routing socket error: %s", |
9df414fe | 1280 | safe_strerror(errno)); |
d62a17ae | 1281 | return 0; |
1282 | } | |
718e3744 | 1283 | |
d62a17ae | 1284 | thread_add_read(zebrad.master, kernel_read, NULL, sock, NULL); |
718e3744 | 1285 | |
d62a17ae | 1286 | if (IS_ZEBRA_DEBUG_KERNEL) |
1287 | rtmsg_debug(&buf.r.rtm); | |
718e3744 | 1288 | |
d62a17ae | 1289 | rtm = &buf.r.rtm; |
718e3744 | 1290 | |
d62a17ae | 1291 | /* |
1292 | * Ensure that we didn't drop any data, so that processing routines | |
1293 | * can assume they have the whole message. | |
1294 | */ | |
1295 | if (rtm->rtm_msglen != nbytes) { | |
9df414fe | 1296 | zlog_debug( |
d62a17ae | 1297 | "kernel_read: rtm->rtm_msglen %d, nbytes %d, type %d\n", |
1298 | rtm->rtm_msglen, nbytes, rtm->rtm_type); | |
1299 | return -1; | |
1300 | } | |
1301 | ||
1302 | switch (rtm->rtm_type) { | |
1303 | case RTM_ADD: | |
1304 | case RTM_DELETE: | |
1305 | case RTM_CHANGE: | |
1306 | rtm_read(rtm); | |
1307 | break; | |
1308 | case RTM_IFINFO: | |
1309 | ifm_read(&buf.im.ifm); | |
1310 | break; | |
1311 | case RTM_NEWADDR: | |
1312 | case RTM_DELADDR: | |
1313 | ifam_read(&buf.ia.ifa); | |
1314 | break; | |
718e3744 | 1315 | #ifdef RTM_IFANNOUNCE |
d62a17ae | 1316 | case RTM_IFANNOUNCE: |
1317 | ifan_read(&buf.ian.ifan); | |
1318 | break; | |
718e3744 | 1319 | #endif /* RTM_IFANNOUNCE */ |
d62a17ae | 1320 | default: |
1321 | if (IS_ZEBRA_DEBUG_KERNEL) | |
1322 | zlog_debug("Unprocessed RTM_type: %d", rtm->rtm_type); | |
1323 | break; | |
1324 | } | |
1325 | return 0; | |
718e3744 | 1326 | } |
1327 | ||
1328 | /* Make routing socket. */ | |
d62a17ae | 1329 | static void routing_socket(struct zebra_ns *zns) |
718e3744 | 1330 | { |
6bb30c2c DL |
1331 | frr_elevate_privs(&zserv_privs) { |
1332 | routing_sock = ns_socket(AF_ROUTE, SOCK_RAW, 0, zns->ns_id); | |
80776aec MS |
1333 | |
1334 | dplane_routing_sock = | |
1335 | ns_socket(AF_ROUTE, SOCK_RAW, 0, zns->ns_id); | |
6bb30c2c | 1336 | } |
d62a17ae | 1337 | |
1338 | if (routing_sock < 0) { | |
1c50c1c0 | 1339 | flog_err_sys(EC_LIB_SOCKET, "Can't init kernel routing socket"); |
d62a17ae | 1340 | return; |
1341 | } | |
1342 | ||
80776aec MS |
1343 | if (dplane_routing_sock < 0) { |
1344 | flog_err_sys(EC_LIB_SOCKET, | |
1345 | "Can't init kernel dataplane routing socket"); | |
1346 | return; | |
1347 | } | |
1348 | ||
d62a17ae | 1349 | /* XXX: Socket should be NONBLOCK, however as we currently |
1350 | * discard failed writes, this will lead to inconsistencies. | |
1351 | * For now, socket must be blocking. | |
1352 | */ | |
1353 | /*if (fcntl (routing_sock, F_SETFL, O_NONBLOCK) < 0) | |
1354 | zlog_warn ("Can't set O_NONBLOCK to routing socket");*/ | |
1355 | ||
d62a17ae | 1356 | /* kernel_read needs rewrite. */ |
1357 | thread_add_read(zebrad.master, kernel_read, NULL, routing_sock, NULL); | |
718e3744 | 1358 | } |
1359 | ||
1360 | /* Exported interface function. This function simply calls | |
1361 | routing_socket (). */ | |
d62a17ae | 1362 | void kernel_init(struct zebra_ns *zns) |
8f7d9fc0 | 1363 | { |
d62a17ae | 1364 | routing_socket(zns); |
8f7d9fc0 FL |
1365 | } |
1366 | ||
62b8bb7a | 1367 | void kernel_terminate(struct zebra_ns *zns, bool complete) |
718e3744 | 1368 | { |
d62a17ae | 1369 | return; |
718e3744 | 1370 | } |
ddfeb486 DL |
1371 | |
1372 | #endif /* !HAVE_NETLINK */ |