46a55bebc54f329a935151a24001f414ef26b77e
[dragonfly.git] / sys / netinet / ip_carp.c
1 /*
2  * Copyright (c) 2002 Michael Shalayeff. All rights reserved.
3  * Copyright (c) 2003 Ryan McBride. All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
15  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
16  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
17  * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT,
18  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
19  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
20  * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
22  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
23  * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
24  * THE POSSIBILITY OF SUCH DAMAGE.
25  */
26 /*
27  * $FreeBSD: src/sys/netinet/ip_carp.c,v 1.48 2007/02/02 09:39:09 glebius Exp $
28  * $DragonFly: src/sys/netinet/ip_carp.c,v 1.10 2008/07/27 10:06:57 sephe Exp $
29  */
30
31 #include "opt_carp.h"
32 #include "opt_inet.h"
33 #include "opt_inet6.h"
34
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/kernel.h>
38 #include <sys/in_cksum.h>
39 #include <sys/limits.h>
40 #include <sys/malloc.h>
41 #include <sys/mbuf.h>
42 #include <sys/time.h>
43 #include <sys/proc.h>
44 #include <sys/priv.h>
45 #include <sys/sockio.h>
46 #include <sys/socket.h>
47 #include <sys/sysctl.h>
48 #include <sys/syslog.h>
49
50 #include <machine/stdarg.h>
51 #include <crypto/sha1.h>
52
53 #include <net/bpf.h>
54 #include <net/ethernet.h>
55 #include <net/if.h>
56 #include <net/if_dl.h>
57 #include <net/if_types.h>
58 #include <net/route.h>
59 #include <net/if_clone.h>
60
61 #ifdef INET
62 #include <netinet/in.h>
63 #include <netinet/in_var.h>
64 #include <netinet/in_systm.h>
65 #include <netinet/ip.h>
66 #include <netinet/ip_var.h>
67 #include <netinet/if_ether.h>
68 #endif
69
70 #ifdef INET6
71 #include <netinet/icmp6.h>
72 #include <netinet/ip6.h>
73 #include <netinet6/ip6_var.h>
74 #include <netinet6/scope6_var.h>
75 #include <netinet6/nd6.h>
76 #endif
77
78 #include <netinet/ip_carp.h>
79
80 #define CARP_IFNAME             "carp"
81 #define CARP_IS_RUNNING(ifp)    \
82         (((ifp)->if_flags & (IFF_UP | IFF_RUNNING)) == (IFF_UP | IFF_RUNNING))
83
84 struct carp_vhaddr {
85         uint32_t                vha_flags;      /* CARP_VHAF_ */
86         const struct in_ifaddr  *vha_ia;        /* carp address */
87         const struct in_ifaddr  *vha_iaback;    /* backing address */
88         TAILQ_ENTRY(carp_vhaddr) vha_link;
89 };
90 TAILQ_HEAD(carp_vhaddr_list, carp_vhaddr);
91
92 struct carp_softc {
93         struct ifnet             sc_if;
94         struct ifnet            *sc_carpdev;    /* parent interface */
95         struct carp_vhaddr_list  sc_vha_list;   /* virtual addr list */
96
97         const struct in_ifaddr  *sc_ia;         /* primary iface address v4 */
98         struct ip_moptions       sc_imo;
99
100 #ifdef INET6
101         struct in6_ifaddr       *sc_ia6;        /* primary iface address v6 */
102         struct ip6_moptions      sc_im6o;
103 #endif /* INET6 */
104         TAILQ_ENTRY(carp_softc)  sc_list;
105
106         enum { INIT = 0, BACKUP, MASTER }
107                                  sc_state;
108         int                      sc_dead;
109
110         int                      sc_suppress;
111
112         int                      sc_sendad_errors;
113 #define CARP_SENDAD_MAX_ERRORS  3
114         int                      sc_sendad_success;
115 #define CARP_SENDAD_MIN_SUCCESS 3
116
117         int                      sc_vhid;
118         int                      sc_advskew;
119         int                      sc_naddrs;     /* actually used IPv4 vha */
120         int                      sc_naddrs6;
121         int                      sc_advbase;    /* seconds */
122         int                      sc_init_counter;
123         uint64_t                 sc_counter;
124
125         /* authentication */
126 #define CARP_HMAC_PAD   64
127         unsigned char            sc_key[CARP_KEY_LEN];
128         unsigned char            sc_pad[CARP_HMAC_PAD];
129         SHA1_CTX                 sc_sha1;
130
131         struct callout           sc_ad_tmo;     /* advertisement timeout */
132         struct callout           sc_md_tmo;     /* master down timeout */
133         struct callout           sc_md6_tmo;    /* master down timeout */
134
135         LIST_ENTRY(carp_softc)   sc_next;       /* Interface clue */
136 };
137
138 struct carp_if {
139         TAILQ_HEAD(, carp_softc) vhif_vrs;
140 };
141
142 enum    { CARP_COUNT_MASTER, CARP_COUNT_RUNNING };
143
144 SYSCTL_DECL(_net_inet_carp);
145
146 static int carp_opts[CARPCTL_MAXID] = { 0, 1, 0, 1, 0, 0 }; /* XXX for now */
147 SYSCTL_INT(_net_inet_carp, CARPCTL_ALLOW, allow, CTLFLAG_RW,
148     &carp_opts[CARPCTL_ALLOW], 0, "Accept incoming CARP packets");
149 SYSCTL_INT(_net_inet_carp, CARPCTL_PREEMPT, preempt, CTLFLAG_RW,
150     &carp_opts[CARPCTL_PREEMPT], 0, "high-priority backup preemption mode");
151 SYSCTL_INT(_net_inet_carp, CARPCTL_LOG, log, CTLFLAG_RW,
152     &carp_opts[CARPCTL_LOG], 0, "log bad carp packets");
153 SYSCTL_INT(_net_inet_carp, CARPCTL_ARPBALANCE, arpbalance, CTLFLAG_RW,
154     &carp_opts[CARPCTL_ARPBALANCE], 0, "balance arp responses");
155
156 static int carp_suppress_preempt = 0;
157 SYSCTL_INT(_net_inet_carp, OID_AUTO, suppress_preempt, CTLFLAG_RD,
158     &carp_suppress_preempt, 0, "Preemption is suppressed");
159
160 static struct carpstats carpstats;
161 SYSCTL_STRUCT(_net_inet_carp, CARPCTL_STATS, stats, CTLFLAG_RW,
162     &carpstats, carpstats,
163     "CARP statistics (struct carpstats, netinet/ip_carp.h)");
164
165 #define CARP_LOG(...)   do {                            \
166         if (carp_opts[CARPCTL_LOG] > 0)                 \
167                 log(LOG_INFO, __VA_ARGS__);             \
168 } while (0)
169
170 #define CARP_DEBUG(...) do {                            \
171         if (carp_opts[CARPCTL_LOG] > 1)                 \
172                 log(LOG_DEBUG, __VA_ARGS__);            \
173 } while (0)
174
175 static void     carp_hmac_prepare(struct carp_softc *);
176 static void     carp_hmac_generate(struct carp_softc *, uint32_t *,
177                     unsigned char *);
178 static int      carp_hmac_verify(struct carp_softc *, uint32_t *,
179                     unsigned char *);
180 static void     carp_setroute(struct carp_softc *, int);
181 static void     carp_input_c(struct mbuf *, struct carp_header *, sa_family_t);
182 static int      carp_clone_create(struct if_clone *, int, caddr_t);
183 static void     carp_clone_destroy(struct ifnet *);
184 static void     carp_detach(struct carp_softc *, int);
185 static int      carp_prepare_ad(struct mbuf *, struct carp_softc *,
186                     struct carp_header *);
187 static void     carp_send_ad_all(void);
188 static void     carp_send_ad_timeout(void *);
189 static void     carp_send_ad(struct carp_softc *);
190 static void     carp_send_arp(struct carp_softc *);
191 static void     carp_master_down_timeout(void *);
192 static void     carp_master_down(struct carp_softc *);
193 static int      carp_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *);
194 static int      carp_looutput(struct ifnet *, struct mbuf *, struct sockaddr *,
195                     struct rtentry *);
196 static void     carp_start(struct ifnet *);
197 static void     carp_setrun(struct carp_softc *, sa_family_t);
198 static void     carp_set_state(struct carp_softc *, int);
199
200 static void     carp_multicast_cleanup(struct carp_softc *);
201 static void     carp_add_addr(struct carp_softc *, struct ifaddr *);
202 static void     carp_del_addr(struct carp_softc *, struct ifaddr *);
203 static void     carp_config_addr(struct carp_softc *, struct ifaddr *);
204 static void     carp_link_addrs(struct carp_softc *, struct ifnet *,
205                     struct ifaddr *);
206 static void     carp_unlink_addrs(struct carp_softc *, struct ifnet *,
207                     struct ifaddr *);
208
209 static int      carp_get_vhaddr(struct carp_softc *, struct ifdrv *);
210 static int      carp_config_vhaddr(struct carp_softc *, struct carp_vhaddr *);
211 static int      carp_activate_vhaddr(struct carp_softc *, struct carp_vhaddr *,
212                     struct ifnet *, const struct in_ifaddr *, int);
213 static void     carp_deactivate_vhaddr(struct carp_softc *,
214                     struct carp_vhaddr *);
215
216 static void     carp_sc_state(struct carp_softc *);
217 #ifdef INET6
218 static void     carp_send_na(struct carp_softc *);
219 static int      carp_set_addr6(struct carp_softc *, struct sockaddr_in6 *);
220 static int      carp_del_addr6(struct carp_softc *, struct sockaddr_in6 *);
221 static void     carp_multicast6_cleanup(struct carp_softc *);
222 #endif
223 static void     carp_stop(struct carp_softc *, int);
224 static void     carp_reset(struct carp_softc *, int);
225
226 static void     carp_ifaddr(void *, struct ifnet *, enum ifaddr_event,
227                             struct ifaddr *);
228 static void     carp_ifdetach(void *, struct ifnet *);
229
230 static MALLOC_DEFINE(M_CARP, "CARP", "CARP interfaces");
231
232 static LIST_HEAD(, carp_softc) carpif_list;
233
234 static struct if_clone carp_cloner =
235 IF_CLONE_INITIALIZER(CARP_IFNAME, carp_clone_create, carp_clone_destroy,
236                      0, IF_MAXUNIT);
237
238 static eventhandler_tag carp_ifdetach_event;
239 static eventhandler_tag carp_ifaddr_event;
240
241 static __inline void
242 carp_insert_vhaddr(struct carp_softc *sc, struct carp_vhaddr *vha_new)
243 {
244         struct carp_vhaddr *vha;
245         u_long new_addr, addr;
246
247         KKASSERT((vha_new->vha_flags & CARP_VHAF_ONLIST) == 0);
248
249         /*
250          * Virtual address list is sorted; smaller one first
251          */
252         new_addr = ntohl(vha_new->vha_ia->ia_addr.sin_addr.s_addr);
253
254         TAILQ_FOREACH(vha, &sc->sc_vha_list, vha_link) {
255                 addr = ntohl(vha->vha_ia->ia_addr.sin_addr.s_addr);
256
257                 if (addr > new_addr)
258                         break;
259         }
260         if (vha == NULL)
261                 TAILQ_INSERT_TAIL(&sc->sc_vha_list, vha_new, vha_link);
262         else
263                 TAILQ_INSERT_BEFORE(vha, vha_new, vha_link);
264         vha_new->vha_flags |= CARP_VHAF_ONLIST;
265 }
266
267 static __inline void
268 carp_remove_vhaddr(struct carp_softc *sc, struct carp_vhaddr *vha)
269 {
270         KKASSERT(vha->vha_flags & CARP_VHAF_ONLIST);
271         vha->vha_flags &= ~CARP_VHAF_ONLIST;
272         TAILQ_REMOVE(&sc->sc_vha_list, vha, vha_link);
273 }
274
275 static void
276 carp_hmac_prepare(struct carp_softc *sc)
277 {
278         uint8_t version = CARP_VERSION, type = CARP_ADVERTISEMENT;
279         uint8_t vhid = sc->sc_vhid & 0xff;
280         int i;
281 #ifdef INET6
282         struct ifaddr_container *ifac;
283         struct in6_addr in6;
284 #endif
285 #ifdef INET
286         struct carp_vhaddr *vha;
287 #endif
288
289         /* XXX: possible race here */
290
291         /* compute ipad from key */
292         bzero(sc->sc_pad, sizeof(sc->sc_pad));
293         bcopy(sc->sc_key, sc->sc_pad, sizeof(sc->sc_key));
294         for (i = 0; i < sizeof(sc->sc_pad); i++)
295                 sc->sc_pad[i] ^= 0x36;
296
297         /* precompute first part of inner hash */
298         SHA1Init(&sc->sc_sha1);
299         SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad));
300         SHA1Update(&sc->sc_sha1, (void *)&version, sizeof(version));
301         SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type));
302         SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid));
303 #ifdef INET
304         TAILQ_FOREACH(vha, &sc->sc_vha_list, vha_link) {
305                 SHA1Update(&sc->sc_sha1,
306                     (const uint8_t *)&vha->vha_ia->ia_addr.sin_addr,
307                     sizeof(struct in_addr));
308         }
309 #endif /* INET */
310 #ifdef INET6
311         TAILQ_FOREACH(ifac, &sc->sc_if.if_addrheads[mycpuid], ifa_link) {
312                 struct ifaddr *ifa = ifac->ifa;
313
314                 if (ifa->ifa_addr->sa_family == AF_INET6) {
315                         in6 = ifatoia6(ifa)->ia_addr.sin6_addr;
316                         in6_clearscope(&in6);
317                         SHA1Update(&sc->sc_sha1, (void *)&in6, sizeof(in6));
318                 }
319         }
320 #endif /* INET6 */
321
322         /* convert ipad to opad */
323         for (i = 0; i < sizeof(sc->sc_pad); i++)
324                 sc->sc_pad[i] ^= 0x36 ^ 0x5c;
325 }
326
327 static void
328 carp_hmac_generate(struct carp_softc *sc, uint32_t counter[2],
329     unsigned char md[20])
330 {
331         SHA1_CTX sha1ctx;
332
333         /* fetch first half of inner hash */
334         bcopy(&sc->sc_sha1, &sha1ctx, sizeof(sha1ctx));
335
336         SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter));
337         SHA1Final(md, &sha1ctx);
338
339         /* outer hash */
340         SHA1Init(&sha1ctx);
341         SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad));
342         SHA1Update(&sha1ctx, md, 20);
343         SHA1Final(md, &sha1ctx);
344 }
345
346 static int
347 carp_hmac_verify(struct carp_softc *sc, uint32_t counter[2],
348     unsigned char md[20])
349 {
350         unsigned char md2[20];
351
352         carp_hmac_generate(sc, counter, md2);
353         return (bcmp(md, md2, sizeof(md2)));
354 }
355
356 static void
357 carp_setroute(struct carp_softc *sc, int cmd)
358 {
359 #ifdef INET6
360         struct ifaddr_container *ifac;
361
362         crit_enter();
363         TAILQ_FOREACH(ifac, &sc->sc_if.if_addrheads[mycpuid], ifa_link) {
364                 struct ifaddr *ifa = ifac->ifa;
365
366                 if (ifa->ifa_addr->sa_family == AF_INET6) {
367                         if (cmd == RTM_ADD)
368                                 in6_ifaddloop(ifa);
369                         else
370                                 in6_ifremloop(ifa);
371                 }
372         }
373         crit_exit();
374 #endif /* INET6 */
375 }
376
377 static int
378 carp_clone_create(struct if_clone *ifc, int unit, caddr_t param __unused)
379 {
380         struct carp_softc *sc;
381         struct ifnet *ifp;
382
383         sc = kmalloc(sizeof(*sc), M_CARP, M_WAITOK | M_ZERO);
384         ifp = &sc->sc_if;
385
386         sc->sc_suppress = 0;
387         sc->sc_advbase = CARP_DFLTINTV;
388         sc->sc_vhid = -1;       /* required setting */
389         sc->sc_advskew = 0;
390         sc->sc_init_counter = 1;
391         sc->sc_naddrs = 0;
392         sc->sc_naddrs6 = 0;
393
394         TAILQ_INIT(&sc->sc_vha_list);
395
396 #ifdef INET6
397         sc->sc_im6o.im6o_multicast_hlim = CARP_DFLTTL;
398 #endif
399
400         callout_init(&sc->sc_ad_tmo);
401         callout_init(&sc->sc_md_tmo);
402         callout_init(&sc->sc_md6_tmo);
403
404         ifp->if_softc = sc;
405         if_initname(ifp, CARP_IFNAME, unit);    
406         ifp->if_mtu = ETHERMTU;
407         ifp->if_flags = IFF_LOOPBACK;
408         ifp->if_ioctl = carp_ioctl;
409         ifp->if_output = carp_looutput;
410         ifp->if_start = carp_start;
411         ifp->if_type = IFT_CARP;
412         ifp->if_snd.ifq_maxlen = ifqmaxlen;
413         ifp->if_hdrlen = 0;
414         if_attach(ifp, NULL);
415         bpfattach(ifp, DLT_NULL, sizeof(u_int));
416
417         crit_enter();
418         LIST_INSERT_HEAD(&carpif_list, sc, sc_next);
419         crit_exit();
420
421         return (0);
422 }
423
424 static void
425 carp_clone_destroy(struct ifnet *ifp)
426 {
427         struct carp_softc *sc = ifp->if_softc;
428
429         sc->sc_dead = 1;
430         carp_detach(sc, 1);
431
432         crit_enter();
433         LIST_REMOVE(sc, sc_next);
434         crit_exit();
435         bpfdetach(ifp);
436         if_detach(ifp);
437
438         KASSERT(sc->sc_naddrs == 0, ("certain inet address is still active\n"));
439         kfree(sc, M_CARP);
440 }
441
442 static void
443 carp_detach(struct carp_softc *sc, int detach)
444 {
445         struct carp_if *cif;
446
447         carp_reset(sc, detach);
448
449         carp_multicast_cleanup(sc);
450 #ifdef INET6
451         carp_multicast6_cleanup(sc);
452 #endif
453
454         if (!sc->sc_dead && detach) {
455                 struct carp_vhaddr *vha;
456
457                 TAILQ_FOREACH(vha, &sc->sc_vha_list, vha_link)
458                         carp_deactivate_vhaddr(sc, vha);
459                 KKASSERT(sc->sc_naddrs == 0);
460         }
461
462         if (sc->sc_carpdev != NULL) {
463                 cif = sc->sc_carpdev->if_carp;
464                 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
465                 if (TAILQ_EMPTY(&cif->vhif_vrs)) {
466                         ifpromisc(sc->sc_carpdev, 0);
467                         sc->sc_carpdev->if_carp = NULL;
468                         kfree(cif, M_CARP);
469                 }
470                 sc->sc_carpdev = NULL;
471                 sc->sc_ia = NULL;
472         }
473 }
474
475 /* Detach an interface from the carp. */
476 static void
477 carp_ifdetach(void *arg __unused, struct ifnet *ifp)
478 {
479         struct carp_if *cif = ifp->if_carp;
480         struct carp_softc *sc;
481
482         while (ifp->if_carp &&
483                (sc = TAILQ_FIRST(&cif->vhif_vrs)) != NULL)
484                 carp_detach(sc, 1);
485 }
486
487 /*
488  * process input packet.
489  * we have rearranged checks order compared to the rfc,
490  * but it seems more efficient this way or not possible otherwise.
491  */
492 void
493 carp_input(struct mbuf *m, ...)
494 {
495         struct ip *ip = mtod(m, struct ip *);
496         struct carp_header *ch;
497         int len, iphlen;
498         __va_list ap;
499
500         __va_start(ap, m);
501         iphlen = __va_arg(ap, int);
502         __va_end(ap);
503
504         carpstats.carps_ipackets++;
505
506         if (!carp_opts[CARPCTL_ALLOW]) {
507                 m_freem(m);
508                 return;
509         }
510
511         /* Check if received on a valid carp interface */
512         if (m->m_pkthdr.rcvif->if_carp == NULL) {
513                 carpstats.carps_badif++;
514                 CARP_LOG("carp_input: packet received on non-carp "
515                     "interface: %s\n",
516                     m->m_pkthdr.rcvif->if_xname);
517                 m_freem(m);
518                 return;
519         }
520
521         /* Verify that the IP TTL is CARP_DFLTTL. */
522         if (ip->ip_ttl != CARP_DFLTTL) {
523                 carpstats.carps_badttl++;
524                 CARP_LOG("carp_input: received ttl %d != %d on %s\n",
525                     ip->ip_ttl, CARP_DFLTTL,
526                     m->m_pkthdr.rcvif->if_xname);
527                 m_freem(m);
528                 return;
529         }
530
531         /* Minimal CARP packet size */
532         len = iphlen + sizeof(*ch);
533
534         /*
535          * Verify that the received packet length is
536          * not less than the CARP header
537          */
538         if (m->m_pkthdr.len < len) {
539                 carpstats.carps_badlen++;
540                 CARP_LOG("packet too short %d on %s\n", m->m_pkthdr.len,
541                          m->m_pkthdr.rcvif->if_xname);
542                 m_freem(m);
543                 return;
544         }
545
546         /* Make sure that CARP header is contiguous */
547         if (len > m->m_len) {
548                 m = m_pullup(m, len);
549                 if (m == NULL) {
550                         carpstats.carps_hdrops++;
551                         CARP_LOG("carp_input: m_pullup failed\n");
552                         return;
553                 }
554                 ip = mtod(m, struct ip *);
555         }
556         ch = (struct carp_header *)((uint8_t *)ip + iphlen);
557
558         /* Verify the CARP checksum */
559         if (in_cksum_skip(m, len, iphlen)) {
560                 carpstats.carps_badsum++;
561                 CARP_LOG("carp_input: checksum failed on %s\n",
562                     m->m_pkthdr.rcvif->if_xname);
563                 m_freem(m);
564                 return;
565         }
566         carp_input_c(m, ch, AF_INET);
567 }
568
569 #ifdef INET6
570 int
571 carp6_input(struct mbuf **mp, int *offp, int proto)
572 {
573         struct mbuf *m = *mp;
574         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
575         struct carp_header *ch;
576         u_int len;
577
578         carpstats.carps_ipackets6++;
579
580         if (!carp_opts[CARPCTL_ALLOW]) {
581                 m_freem(m);
582                 return (IPPROTO_DONE);
583         }
584
585         /* check if received on a valid carp interface */
586         if (m->m_pkthdr.rcvif->if_carp == NULL) {
587                 carpstats.carps_badif++;
588                 CARP_LOG("carp6_input: packet received on non-carp "
589                     "interface: %s\n",
590                     m->m_pkthdr.rcvif->if_xname);
591                 m_freem(m);
592                 return (IPPROTO_DONE);
593         }
594
595         /* verify that the IP TTL is 255 */
596         if (ip6->ip6_hlim != CARP_DFLTTL) {
597                 carpstats.carps_badttl++;
598                 CARP_LOG("carp6_input: received ttl %d != 255 on %s\n",
599                     ip6->ip6_hlim,
600                     m->m_pkthdr.rcvif->if_xname);
601                 m_freem(m);
602                 return (IPPROTO_DONE);
603         }
604
605         /* verify that we have a complete carp packet */
606         len = m->m_len;
607         IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch));
608         if (ch == NULL) {
609                 carpstats.carps_badlen++;
610                 CARP_LOG("carp6_input: packet size %u too small\n", len);
611                 return (IPPROTO_DONE);
612         }
613
614         /* verify the CARP checksum */
615         if (in_cksum_range(m, 0, *offp, sizeof(*ch))) {
616                 carpstats.carps_badsum++;
617                 CARP_LOG("carp6_input: checksum failed, on %s\n",
618                     m->m_pkthdr.rcvif->if_xname);
619                 m_freem(m);
620                 return (IPPROTO_DONE);
621         }
622
623         carp_input_c(m, ch, AF_INET6);
624         return (IPPROTO_DONE);
625 }
626 #endif /* INET6 */
627
628 static void
629 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af)
630 {
631         struct ifnet *ifp = m->m_pkthdr.rcvif;
632         struct ifnet *cifp;
633         struct carp_softc *sc;
634         uint64_t tmp_counter;
635         struct timeval sc_tv, ch_tv;
636
637         /* verify that the VHID is valid on the receiving interface */
638         TAILQ_FOREACH(sc, &((struct carp_if *)ifp->if_carp)->vhif_vrs, sc_list)
639                 if (sc->sc_vhid == ch->carp_vhid)
640                         break;
641
642         if (!sc || !CARP_IS_RUNNING(&sc->sc_if)) {
643                 carpstats.carps_badvhid++;
644                 m_freem(m);
645                 return;
646         }
647         cifp = &sc->sc_if;
648
649         getmicrotime(&cifp->if_lastchange);
650         cifp->if_ipackets++;
651         cifp->if_ibytes += m->m_pkthdr.len;
652
653         if (cifp->if_bpf) {
654                 struct ip *ip = mtod(m, struct ip *);
655
656                 /* BPF wants net byte order */
657                 ip->ip_len = htons(ip->ip_len + (ip->ip_hl << 2));
658                 ip->ip_off = htons(ip->ip_off);
659                 bpf_mtap(cifp->if_bpf, m);
660         }
661
662         /* verify the CARP version. */
663         if (ch->carp_version != CARP_VERSION) {
664                 carpstats.carps_badver++;
665                 cifp->if_ierrors++;
666                 CARP_LOG("%s; invalid version %d\n", cifp->if_xname,
667                          ch->carp_version);
668                 m_freem(m);
669                 return;
670         }
671
672         /* verify the hash */
673         if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) {
674                 carpstats.carps_badauth++;
675                 cifp->if_ierrors++;
676                 CARP_LOG("%s: incorrect hash\n", cifp->if_xname);
677                 m_freem(m);
678                 return;
679         }
680
681         tmp_counter = ntohl(ch->carp_counter[0]);
682         tmp_counter = tmp_counter<<32;
683         tmp_counter += ntohl(ch->carp_counter[1]);
684
685         /* XXX Replay protection goes here */
686
687         sc->sc_init_counter = 0;
688         sc->sc_counter = tmp_counter;
689
690         sc_tv.tv_sec = sc->sc_advbase;
691         if (carp_suppress_preempt && sc->sc_advskew <  240)
692                 sc_tv.tv_usec = 240 * 1000000 / 256;
693         else
694                 sc_tv.tv_usec = sc->sc_advskew * 1000000 / 256;
695         ch_tv.tv_sec = ch->carp_advbase;
696         ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256;
697
698         switch (sc->sc_state) {
699         case INIT:
700                 break;
701
702         case MASTER:
703                 /*
704                  * If we receive an advertisement from a master who's going to
705                  * be more frequent than us, go into BACKUP state.
706                  */
707                 if (timevalcmp(&sc_tv, &ch_tv, >) ||
708                     timevalcmp(&sc_tv, &ch_tv, ==)) {
709                         callout_stop(&sc->sc_ad_tmo);
710                         CARP_DEBUG("%s: MASTER -> BACKUP "
711                            "(more frequent advertisement received)\n",
712                            cifp->if_xname);
713                         carp_set_state(sc, BACKUP);
714                         carp_setrun(sc, 0);
715                         carp_setroute(sc, RTM_DELETE);
716                 }
717                 break;
718
719         case BACKUP:
720                 /*
721                  * If we're pre-empting masters who advertise slower than us,
722                  * and this one claims to be slower, treat him as down.
723                  */
724                 if (carp_opts[CARPCTL_PREEMPT] &&
725                     timevalcmp(&sc_tv, &ch_tv, <)) {
726                         CARP_DEBUG("%s: BACKUP -> MASTER "
727                             "(preempting a slower master)\n", cifp->if_xname);
728                         carp_master_down(sc);
729                         break;
730                 }
731
732                 /*
733                  *  If the master is going to advertise at such a low frequency
734                  *  that he's guaranteed to time out, we'd might as well just
735                  *  treat him as timed out now.
736                  */
737                 sc_tv.tv_sec = sc->sc_advbase * 3;
738                 if (timevalcmp(&sc_tv, &ch_tv, <)) {
739                         CARP_DEBUG("%s: BACKUP -> MASTER (master timed out)\n",
740                                    cifp->if_xname);
741                         carp_master_down(sc);
742                         break;
743                 }
744
745                 /*
746                  * Otherwise, we reset the counter and wait for the next
747                  * advertisement.
748                  */
749                 carp_setrun(sc, af);
750                 break;
751         }
752         m_freem(m);
753 }
754
755 static int
756 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, struct carp_header *ch)
757 {
758         struct ifnet *cifp = &sc->sc_if;
759         struct m_tag *mtag;
760
761         if (sc->sc_init_counter) {
762                 /* this could also be seconds since unix epoch */
763                 sc->sc_counter = karc4random();
764                 sc->sc_counter = sc->sc_counter << 32;
765                 sc->sc_counter += karc4random();
766         } else {
767                 sc->sc_counter++;
768         }
769
770         ch->carp_counter[0] = htonl((sc->sc_counter >> 32) & 0xffffffff);
771         ch->carp_counter[1] = htonl(sc->sc_counter & 0xffffffff);
772
773         carp_hmac_generate(sc, ch->carp_counter, ch->carp_md);
774
775         /* Tag packet for carp_output */
776         mtag = m_tag_get(PACKET_TAG_CARP, sizeof(struct ifnet *), MB_DONTWAIT);
777         if (mtag == NULL) {
778                 m_freem(m);
779                 cifp->if_oerrors++;
780                 return ENOMEM;
781         }
782         bcopy(&cifp, (caddr_t)(mtag + 1), sizeof(struct ifnet *));
783         m_tag_prepend(m, mtag);
784
785         return 0;
786 }
787
788 static void
789 carp_send_ad_all(void)
790 {
791         struct carp_softc *sc;
792
793         LIST_FOREACH(sc, &carpif_list, sc_next) {
794                 if (sc->sc_carpdev == NULL)
795                         continue;
796
797                 if (CARP_IS_RUNNING(&sc->sc_if) && sc->sc_state == MASTER)
798                         carp_send_ad(sc);
799         }
800 }
801
802 static void
803 carp_send_ad_timeout(void *xsc)
804 {
805         carp_send_ad(xsc);
806 }
807
808 static void
809 carp_send_ad(struct carp_softc *sc)
810 {
811         struct ifnet *cifp = &sc->sc_if;
812         struct carp_header ch;
813         struct timeval tv;
814         struct carp_header *ch_ptr;
815         struct mbuf *m;
816         int len, advbase, advskew;
817
818         if (!CARP_IS_RUNNING(cifp)) {
819                 /* Bow out */
820                 advbase = 255;
821                 advskew = 255;
822         } else {
823                 advbase = sc->sc_advbase;
824                 if (!carp_suppress_preempt || sc->sc_advskew > 240)
825                         advskew = sc->sc_advskew;
826                 else
827                         advskew = 240;
828                 tv.tv_sec = advbase;
829                 tv.tv_usec = advskew * 1000000 / 256;
830         }
831
832         ch.carp_version = CARP_VERSION;
833         ch.carp_type = CARP_ADVERTISEMENT;
834         ch.carp_vhid = sc->sc_vhid;
835         ch.carp_advbase = advbase;
836         ch.carp_advskew = advskew;
837         ch.carp_authlen = 7;    /* XXX DEFINE */
838         ch.carp_pad1 = 0;       /* must be zero */
839         ch.carp_cksum = 0;
840
841 #ifdef INET
842         if (sc->sc_ia != NULL) {
843                 struct ip *ip;
844
845                 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
846                 if (m == NULL) {
847                         cifp->if_oerrors++;
848                         carpstats.carps_onomem++;
849                         /* XXX maybe less ? */
850                         if (advbase != 255 || advskew != 255)
851                                 callout_reset(&sc->sc_ad_tmo, tvtohz_high(&tv),
852                                     carp_send_ad_timeout, sc);
853                         return;
854                 }
855                 len = sizeof(*ip) + sizeof(ch);
856                 m->m_pkthdr.len = len;
857                 m->m_pkthdr.rcvif = NULL;
858                 m->m_len = len;
859                 MH_ALIGN(m, m->m_len);
860                 m->m_flags |= M_MCAST;
861                 ip = mtod(m, struct ip *);
862                 ip->ip_v = IPVERSION;
863                 ip->ip_hl = sizeof(*ip) >> 2;
864                 ip->ip_tos = IPTOS_LOWDELAY;
865                 ip->ip_len = len;
866                 ip->ip_id = ip_newid();
867                 ip->ip_off = IP_DF;
868                 ip->ip_ttl = CARP_DFLTTL;
869                 ip->ip_p = IPPROTO_CARP;
870                 ip->ip_sum = 0;
871                 ip->ip_src = sc->sc_ia->ia_addr.sin_addr;
872                 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP);
873
874                 ch_ptr = (struct carp_header *)(&ip[1]);
875                 bcopy(&ch, ch_ptr, sizeof(ch));
876                 if (carp_prepare_ad(m, sc, ch_ptr))
877                         return;
878                 ch_ptr->carp_cksum = in_cksum_skip(m, len, sizeof(*ip));
879
880                 getmicrotime(&cifp->if_lastchange);
881                 cifp->if_opackets++;
882                 cifp->if_obytes += len;
883                 carpstats.carps_opackets++;
884
885                 if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL)) {
886                         cifp->if_oerrors++;
887                         if (sc->sc_sendad_errors < INT_MAX)
888                                 sc->sc_sendad_errors++;
889                         if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
890                                 carp_suppress_preempt++;
891                                 if (carp_suppress_preempt == 1) {
892                                         carp_send_ad_all();
893                                 }
894                         }
895                         sc->sc_sendad_success = 0;
896                 } else {
897                         if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) {
898                                 if (++sc->sc_sendad_success >=
899                                     CARP_SENDAD_MIN_SUCCESS) {
900                                         carp_suppress_preempt--;
901                                         sc->sc_sendad_errors = 0;
902                                 }
903                         } else {
904                                 sc->sc_sendad_errors = 0;
905                         }
906                 }
907         }
908 #endif /* INET */
909 #ifdef INET6
910         if (sc->sc_ia6) {
911                 struct ip6_hdr *ip6;
912
913                 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
914                 if (m == NULL) {
915                         cifp->if_oerrors++;
916                         carpstats.carps_onomem++;
917                         /* XXX maybe less ? */
918                         if (advbase != 255 || advskew != 255)
919                                 callout_reset(&sc->sc_ad_tmo, tvtohz_high(&tv),
920                                     carp_send_ad_timeout, sc);
921                         return;
922                 }
923                 len = sizeof(*ip6) + sizeof(ch);
924                 m->m_pkthdr.len = len;
925                 m->m_pkthdr.rcvif = NULL;
926                 m->m_len = len;
927                 MH_ALIGN(m, m->m_len);
928                 m->m_flags |= M_MCAST;
929                 ip6 = mtod(m, struct ip6_hdr *);
930                 bzero(ip6, sizeof(*ip6));
931                 ip6->ip6_vfc |= IPV6_VERSION;
932                 ip6->ip6_hlim = CARP_DFLTTL;
933                 ip6->ip6_nxt = IPPROTO_CARP;
934                 bcopy(&sc->sc_ia6->ia_addr.sin6_addr, &ip6->ip6_src,
935                     sizeof(struct in6_addr));
936                 /* set the multicast destination */
937
938                 ip6->ip6_dst.s6_addr16[0] = htons(0xff02);
939                 ip6->ip6_dst.s6_addr8[15] = 0x12;
940                 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) {
941                         cifp->if_oerrors++;
942                         m_freem(m);
943                         CARP_LOG("%s: in6_setscope failed\n", __func__);
944                         return;
945                 }
946
947                 ch_ptr = (struct carp_header *)(&ip6[1]);
948                 bcopy(&ch, ch_ptr, sizeof(ch));
949                 if (carp_prepare_ad(m, sc, ch_ptr))
950                         return;
951                 ch_ptr->carp_cksum = in_cksum_skip(m, len, sizeof(*ip6));
952
953                 getmicrotime(&cifp->if_lastchange);
954                 cifp->if_opackets++;
955                 cifp->if_obytes += len;
956                 carpstats.carps_opackets6++;
957
958                 if (ip6_output(m, NULL, NULL, 0, &sc->sc_im6o, NULL, NULL)) {
959                         cifp->if_oerrors++;
960                         if (sc->sc_sendad_errors < INT_MAX)
961                                 sc->sc_sendad_errors++;
962                         if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
963                                 carp_suppress_preempt++;
964                                 if (carp_suppress_preempt == 1) {
965                                         carp_send_ad_all();
966                                 }
967                         }
968                         sc->sc_sendad_success = 0;
969                 } else {
970                         if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) {
971                                 if (++sc->sc_sendad_success >=
972                                     CARP_SENDAD_MIN_SUCCESS) {
973                                         carp_suppress_preempt--;
974                                         sc->sc_sendad_errors = 0;
975                                 }
976                         } else {
977                                 sc->sc_sendad_errors = 0;
978                         }
979                 }
980         }
981 #endif /* INET6 */
982
983         if (advbase != 255 || advskew != 255)
984                 callout_reset(&sc->sc_ad_tmo, tvtohz_high(&tv),
985                     carp_send_ad_timeout, sc);
986 }
987
988 /*
989  * Broadcast a gratuitous ARP request containing
990  * the virtual router MAC address for each IP address
991  * associated with the virtual router.
992  */
993 static void
994 carp_send_arp(struct carp_softc *sc)
995 {
996         const struct carp_vhaddr *vha;
997
998         TAILQ_FOREACH(vha, &sc->sc_vha_list, vha_link) {
999                 if (vha->vha_iaback == NULL)
1000                         continue;
1001
1002                 arp_iainit(sc->sc_carpdev, &vha->vha_ia->ia_addr.sin_addr,
1003                            IF_LLADDR(&sc->sc_if));
1004         }
1005 }
1006
1007 #ifdef INET6
1008 static void
1009 carp_send_na(struct carp_softc *sc)
1010 {
1011         struct ifaddr_container *ifac;
1012         struct in6_addr *in6;
1013         static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
1014
1015         TAILQ_FOREACH(ifac, &sc->sc_if.if_addrheads[mycpuid], ifa_link) {
1016                 struct ifaddr *ifa = ifac->ifa;
1017
1018                 if (ifa->ifa_addr->sa_family != AF_INET6)
1019                         continue;
1020
1021                 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr;
1022                 nd6_na_output(sc->sc_carpdev, &mcast, in6,
1023                     ND_NA_FLAG_OVERRIDE, 1, NULL);
1024                 DELAY(1000);    /* XXX */
1025         }
1026 }
1027 #endif /* INET6 */
1028
1029 static __inline const struct carp_vhaddr *
1030 carp_find_addr(const struct carp_softc *sc, const struct in_addr *addr)
1031 {
1032         struct carp_vhaddr *vha;
1033
1034         TAILQ_FOREACH(vha, &sc->sc_vha_list, vha_link) {
1035                 if (vha->vha_iaback == NULL)
1036                         continue;
1037
1038                 if (vha->vha_ia->ia_addr.sin_addr.s_addr == addr->s_addr)
1039                         return vha;
1040         }
1041         return NULL;
1042 }
1043
1044 static int
1045 carp_iamatch_balance(const struct carp_if *cif, const struct in_addr *itaddr,
1046                      const struct in_addr *isaddr, uint8_t **enaddr)
1047 {
1048         const struct carp_softc *vh;
1049         int index, count = 0;
1050
1051         /*
1052          * XXX proof of concept implementation.
1053          * We use the source ip to decide which virtual host should
1054          * handle the request. If we're master of that virtual host,
1055          * then we respond, otherwise, just drop the arp packet on
1056          * the floor.
1057          */
1058
1059         TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1060                 if (!CARP_IS_RUNNING(&vh->sc_if))
1061                         continue;
1062
1063                 if (carp_find_addr(vh, itaddr) != NULL)
1064                         count++;
1065         }
1066         if (count == 0)
1067                 return 0;
1068
1069         /* this should be a hash, like pf_hash() */
1070         index = ntohl(isaddr->s_addr) % count;
1071         count = 0;
1072
1073         TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1074                 if (!CARP_IS_RUNNING(&vh->sc_if))
1075                         continue;
1076
1077                 if (carp_find_addr(vh, itaddr) == NULL)
1078                         continue;
1079
1080                 if (count == index) {
1081                         if (vh->sc_state == MASTER) {
1082                                 *enaddr = IF_LLADDR(&vh->sc_if);
1083                                 return 1;
1084                         } else {
1085                                 return 0;
1086                         }
1087                 }
1088                 count++;
1089         }
1090         return 0;
1091 }
1092
1093 int
1094 carp_iamatch(const void *v, const struct in_addr *itaddr,
1095              const struct in_addr *isaddr, uint8_t **enaddr)
1096 {
1097         const struct carp_if *cif = v;
1098         const struct carp_softc *vh;
1099
1100         if (carp_opts[CARPCTL_ARPBALANCE])
1101                 return carp_iamatch_balance(cif, itaddr, isaddr, enaddr);
1102
1103         TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1104                 if (!CARP_IS_RUNNING(&vh->sc_if) || vh->sc_state != MASTER)
1105                         continue;
1106
1107                 if (carp_find_addr(vh, itaddr) != NULL) {
1108                         *enaddr = IF_LLADDR(&vh->sc_if);
1109                         return 1;
1110                 }
1111         }
1112         return 0;
1113 }
1114
1115 #ifdef INET6
1116 struct ifaddr *
1117 carp_iamatch6(void *v, struct in6_addr *taddr)
1118 {
1119         struct carp_if *cif = v;
1120         struct carp_softc *vh;
1121
1122         TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1123                 struct ifaddr_container *ifac;
1124
1125                 TAILQ_FOREACH(ifac, &vh->sc_if.if_addrheads[mycpuid],
1126                               ifa_link) {
1127                         struct ifaddr *ifa = ifac->ifa;
1128
1129                         if (IN6_ARE_ADDR_EQUAL(taddr,
1130                             &ifatoia6(ifa)->ia_addr.sin6_addr) &&
1131                             CARP_IS_RUNNING(&vh->sc_if) &&
1132                             vh->sc_state == MASTER) {
1133                                 return (ifa);
1134                         }
1135                 }
1136         }
1137         return (NULL);
1138 }
1139
1140 void *
1141 carp_macmatch6(void *v, struct mbuf *m, const struct in6_addr *taddr)
1142 {
1143         struct m_tag *mtag;
1144         struct carp_if *cif = v;
1145         struct carp_softc *sc;
1146
1147         TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) {
1148                 struct ifaddr_container *ifac;
1149
1150                 TAILQ_FOREACH(ifac, &sc->sc_if.if_addrheads[mycpuid],
1151                               ifa_link) {
1152                         struct ifaddr *ifa = ifac->ifa;
1153
1154                         if (IN6_ARE_ADDR_EQUAL(taddr,
1155                             &ifatoia6(ifa)->ia_addr.sin6_addr) &&
1156                             CARP_IS_RUNNING(&sc->sc_if)) {
1157                                 struct ifnet *ifp = &sc->sc_if;
1158
1159                                 mtag = m_tag_get(PACKET_TAG_CARP,
1160                                     sizeof(struct ifnet *), MB_DONTWAIT);
1161                                 if (mtag == NULL) {
1162                                         /* better a bit than nothing */
1163                                         return (IF_LLADDR(ifp));
1164                                 }
1165                                 bcopy(&ifp, (caddr_t)(mtag + 1),
1166                                     sizeof(struct ifnet *));
1167                                 m_tag_prepend(m, mtag);
1168
1169                                 return (IF_LLADDR(ifp));
1170                         }
1171                 }
1172         }
1173         return (NULL);
1174 }
1175 #endif
1176
1177 int
1178 carp_forus(const void *v, const void *dhost)
1179 {
1180         const struct carp_if *cif = v;
1181         const struct carp_softc *vh;
1182         const uint8_t *ena = dhost;
1183
1184         if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1)
1185                 return 0;
1186
1187         TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1188                 const struct ifnet *cifp = &vh->sc_if;
1189
1190                 if (CARP_IS_RUNNING(cifp) && vh->sc_state == MASTER &&
1191                     !bcmp(dhost, IF_LLADDR(cifp), ETHER_ADDR_LEN))
1192                         return 1;
1193         }
1194         return 0;
1195 }
1196
1197 static void
1198 carp_master_down_timeout(void *xsc)
1199 {
1200         struct carp_softc *sc = xsc;
1201
1202         CARP_DEBUG("%s: BACKUP -> MASTER (master timed out)\n",
1203                    sc->sc_if.if_xname);
1204         carp_master_down(sc);
1205 }
1206
1207 static void
1208 carp_master_down(struct carp_softc *sc)
1209 {
1210         switch (sc->sc_state) {
1211         case INIT:
1212                 kprintf("%s: master_down event in INIT state\n",
1213                         sc->sc_if.if_xname);
1214                 break;
1215
1216         case MASTER:
1217                 break;
1218
1219         case BACKUP:
1220                 carp_set_state(sc, MASTER);
1221                 carp_send_ad(sc);
1222                 carp_send_arp(sc);
1223 #ifdef INET6
1224                 carp_send_na(sc);
1225 #endif /* INET6 */
1226                 carp_setrun(sc, 0);
1227                 carp_setroute(sc, RTM_ADD);
1228                 break;
1229         }
1230 }
1231
1232 /*
1233  * When in backup state, af indicates whether to reset the master down timer
1234  * for v4 or v6. If it's set to zero, reset the ones which are already pending.
1235  */
1236 static void
1237 carp_setrun(struct carp_softc *sc, sa_family_t af)
1238 {
1239         struct ifnet *cifp = &sc->sc_if;
1240         struct timeval tv;
1241
1242         if (sc->sc_carpdev == NULL) {
1243                 carp_set_state(sc, INIT);
1244                 return;
1245         }
1246
1247         if ((cifp->if_flags & IFF_RUNNING) && sc->sc_vhid > 0 &&
1248             (sc->sc_naddrs || sc->sc_naddrs6)) {
1249                 /* Nothing */
1250         } else {
1251                 carp_setroute(sc, RTM_DELETE);
1252                 return;
1253         }
1254
1255         switch (sc->sc_state) {
1256         case INIT:
1257                 if (carp_opts[CARPCTL_PREEMPT] && !carp_suppress_preempt) {
1258                         carp_send_ad(sc);
1259                         carp_send_arp(sc);
1260 #ifdef INET6
1261                         carp_send_na(sc);
1262 #endif /* INET6 */
1263                         CARP_DEBUG("%s: INIT -> MASTER (preempting)\n",
1264                                    cifp->if_xname);
1265                         carp_set_state(sc, MASTER);
1266                         carp_setroute(sc, RTM_ADD);
1267                 } else {
1268                         CARP_DEBUG("%s: INIT -> BACKUP\n", cifp->if_xname);
1269                         carp_set_state(sc, BACKUP);
1270                         carp_setroute(sc, RTM_DELETE);
1271                         carp_setrun(sc, 0);
1272                 }
1273                 break;
1274
1275         case BACKUP:
1276                 callout_stop(&sc->sc_ad_tmo);
1277                 tv.tv_sec = 3 * sc->sc_advbase;
1278                 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1279                 switch (af) {
1280 #ifdef INET
1281                 case AF_INET:
1282                         callout_reset(&sc->sc_md_tmo, tvtohz_high(&tv),
1283                             carp_master_down_timeout, sc);
1284                         break;
1285 #endif /* INET */
1286 #ifdef INET6
1287                 case AF_INET6:
1288                         callout_reset(&sc->sc_md6_tmo, tvtohz_high(&tv),
1289                             carp_master_down_timeout, sc);
1290                         break;
1291 #endif /* INET6 */
1292                 default:
1293                         if (sc->sc_naddrs)
1294                                 callout_reset(&sc->sc_md_tmo, tvtohz_high(&tv),
1295                                     carp_master_down_timeout, sc);
1296                         if (sc->sc_naddrs6)
1297                                 callout_reset(&sc->sc_md6_tmo, tvtohz_high(&tv),
1298                                     carp_master_down_timeout, sc);
1299                         break;
1300                 }
1301                 break;
1302
1303         case MASTER:
1304                 tv.tv_sec = sc->sc_advbase;
1305                 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1306                 callout_reset(&sc->sc_ad_tmo, tvtohz_high(&tv),
1307                     carp_send_ad_timeout, sc);
1308                 break;
1309         }
1310 }
1311
1312 static void
1313 carp_multicast_cleanup(struct carp_softc *sc)
1314 {
1315         struct ip_moptions *imo = &sc->sc_imo;
1316
1317         if (imo->imo_num_memberships == 0)
1318                 return;
1319         KKASSERT(imo->imo_num_memberships == 1);
1320
1321         in_delmulti(imo->imo_membership[0]);
1322         imo->imo_membership[0] = NULL;
1323         imo->imo_num_memberships = 0;
1324         imo->imo_multicast_ifp = NULL;
1325 }
1326
1327 #ifdef INET6
1328 static void
1329 carp_multicast6_cleanup(struct carp_softc *sc)
1330 {
1331         struct ip6_moptions *im6o = &sc->sc_im6o;
1332
1333         while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1334                 struct in6_multi_mship *imm =
1335                     LIST_FIRST(&im6o->im6o_memberships);
1336
1337                 LIST_REMOVE(imm, i6mm_chain);
1338                 in6_leavegroup(imm);
1339         }
1340         im6o->im6o_multicast_ifp = NULL;
1341 }
1342 #endif
1343
1344 static int
1345 carp_get_vhaddr(struct carp_softc *sc, struct ifdrv *ifd)
1346 {
1347         const struct carp_vhaddr *vha;
1348         struct ifcarpvhaddr *carpa, *carpa0;
1349         int count, len, error;
1350
1351         count = 0;
1352         TAILQ_FOREACH(vha, &sc->sc_vha_list, vha_link)
1353                 ++count;
1354
1355         if (ifd->ifd_len == 0) {
1356                 ifd->ifd_len = count * sizeof(*carpa);
1357                 return 0;
1358         } else if (count == 0 || ifd->ifd_len < sizeof(*carpa)) {
1359                 ifd->ifd_len = 0;
1360                 return 0;
1361         }
1362         len = min(ifd->ifd_len, sizeof(*carpa) * count);
1363         KKASSERT(len >= sizeof(*carpa));
1364
1365         carpa0 = carpa = kmalloc(len, M_TEMP, M_WAITOK | M_NULLOK | M_ZERO);
1366         if (carpa == NULL)
1367                 return ENOMEM;
1368
1369         count = 0;
1370         TAILQ_FOREACH(vha, &sc->sc_vha_list, vha_link) {
1371                 if (len < sizeof(*carpa))
1372                         break;
1373
1374                 carpa->carpa_flags = vha->vha_flags;
1375                 carpa->carpa_addr.sin_family = AF_INET;
1376                 carpa->carpa_addr.sin_addr = vha->vha_ia->ia_addr.sin_addr;
1377
1378                 carpa->carpa_baddr.sin_family = AF_INET;
1379                 if (vha->vha_iaback == NULL) {
1380                         carpa->carpa_baddr.sin_addr.s_addr = INADDR_ANY;
1381                 } else {
1382                         carpa->carpa_baddr.sin_addr =
1383                         vha->vha_iaback->ia_addr.sin_addr;
1384                 }
1385
1386                 ++carpa;
1387                 ++count;
1388                 len -= sizeof(*carpa);
1389         }
1390         ifd->ifd_len = sizeof(*carpa) * count;
1391         KKASSERT(ifd->ifd_len > 0);
1392
1393         error = copyout(carpa0, ifd->ifd_data, ifd->ifd_len);
1394         kfree(carpa0, M_TEMP);
1395         return error;
1396 }
1397
1398 static int
1399 carp_config_vhaddr(struct carp_softc *sc, struct carp_vhaddr *vha)
1400 {
1401         struct ifnet *ifp;
1402         struct in_ifaddr *ia_if;
1403         struct in_ifaddr_container *iac;
1404         const struct sockaddr_in *sin;
1405         u_long iaddr;
1406         int own;
1407
1408         KKASSERT(vha->vha_ia != NULL);
1409
1410         sin = &vha->vha_ia->ia_addr;
1411         iaddr = ntohl(sin->sin_addr.s_addr);
1412
1413         ia_if = NULL;
1414         own = 0;
1415         TAILQ_FOREACH(iac, &in_ifaddrheads[mycpuid], ia_link) {
1416                 struct in_ifaddr *ia = iac->ia;
1417
1418                 if ((ia->ia_flags & IFA_ROUTE) == 0)
1419                         continue;
1420
1421                 if (ia->ia_ifp->if_type == IFT_CARP)
1422                         continue;
1423
1424                 /* and, yeah, we need a multicast-capable iface too */
1425                 if ((ia->ia_ifp->if_flags & IFF_MULTICAST) == 0)
1426                         continue;
1427
1428                 if ((iaddr & ia->ia_subnetmask) == ia->ia_subnet) {
1429                         if (sin->sin_addr.s_addr ==
1430                             ia->ia_addr.sin_addr.s_addr)
1431                                 own = 1;
1432                         if (ia_if == NULL)
1433                                 ia_if = ia;
1434                         else if (sc->sc_carpdev != NULL &&
1435                                  sc->sc_carpdev == ia->ia_ifp)
1436                                 ia_if = ia;
1437                 }
1438         }
1439
1440         carp_deactivate_vhaddr(sc, vha);
1441         if (!ia_if)
1442                 return ENOENT;
1443
1444         ifp = ia_if->ia_ifp;
1445
1446         /* XXX Don't allow parent iface to be changed */
1447         if (sc->sc_carpdev != NULL && sc->sc_carpdev != ifp)
1448                 return EEXIST;
1449
1450         return carp_activate_vhaddr(sc, vha, ifp, ia_if, own);
1451 }
1452
1453 static void
1454 carp_add_addr(struct carp_softc *sc, struct ifaddr *carp_ifa)
1455 {
1456         struct carp_vhaddr *vha_new;
1457         struct in_ifaddr *carp_ia;
1458 #ifdef INVARIANTS
1459         struct carp_vhaddr *vha;
1460 #endif
1461
1462         KKASSERT(carp_ifa->ifa_addr->sa_family == AF_INET);
1463         carp_ia = ifatoia(carp_ifa);
1464
1465 #ifdef INVARIANTS
1466         TAILQ_FOREACH(vha, &sc->sc_vha_list, vha_link)
1467                 KKASSERT(vha->vha_ia != NULL && vha->vha_ia != carp_ia);
1468 #endif
1469
1470         vha_new = kmalloc(sizeof(*vha_new), M_CARP, M_WAITOK | M_ZERO);
1471         vha_new->vha_ia = carp_ia;
1472         carp_insert_vhaddr(sc, vha_new);
1473
1474         if (carp_config_vhaddr(sc, vha_new) != 0) {
1475                 /*
1476                  * If the above configuration fails, it may only mean
1477                  * that the new address is problematic.  However, the
1478                  * carp(4) interface may already have several working
1479                  * addresses.  Since the expected behaviour of
1480                  * SIOC[AS]IFADDR is to put the NIC into working state,
1481                  * we try starting the state machine manually here with
1482                  * the hope that the carp(4)'s previously working
1483                  * addresses still could be brought up.
1484                  */
1485                 carp_hmac_prepare(sc);
1486                 carp_set_state(sc, INIT);
1487                 carp_setrun(sc, 0);
1488         }
1489 }
1490
1491 static void
1492 carp_del_addr(struct carp_softc *sc, struct ifaddr *carp_ifa)
1493 {
1494         struct carp_vhaddr *vha;
1495         struct in_ifaddr *carp_ia;
1496
1497         KKASSERT(carp_ifa->ifa_addr->sa_family == AF_INET);
1498         carp_ia = ifatoia(carp_ifa);
1499
1500         TAILQ_FOREACH(vha, &sc->sc_vha_list, vha_link) {
1501                 KKASSERT(vha->vha_ia != NULL);
1502                 if (vha->vha_ia == carp_ia)
1503                         break;
1504         }
1505         KASSERT(vha != NULL, ("no corresponding vhaddr %p\n", carp_ifa));
1506
1507         /*
1508          * Remove the vhaddr from the list before deactivating
1509          * the vhaddr, so that the HMAC could be correctly
1510          * updated in carp_deactivate_vhaddr()
1511          */
1512         carp_remove_vhaddr(sc, vha);
1513
1514         carp_deactivate_vhaddr(sc, vha);
1515         kfree(vha, M_CARP);
1516 }
1517
1518 static void
1519 carp_config_addr(struct carp_softc *sc, struct ifaddr *carp_ifa)
1520 {
1521         struct carp_vhaddr *vha;
1522         struct in_ifaddr *carp_ia;
1523
1524         KKASSERT(carp_ifa->ifa_addr->sa_family == AF_INET);
1525         carp_ia = ifatoia(carp_ifa);
1526
1527         TAILQ_FOREACH(vha, &sc->sc_vha_list, vha_link) {
1528                 KKASSERT(vha->vha_ia != NULL);
1529                 if (vha->vha_ia == carp_ia)
1530                         break;
1531         }
1532         KASSERT(vha != NULL, ("no corresponding vhaddr %p\n", carp_ifa));
1533
1534         /* Remove then reinsert, to keep the vhaddr list sorted */
1535         carp_remove_vhaddr(sc, vha);
1536         carp_insert_vhaddr(sc, vha);
1537
1538         if (carp_config_vhaddr(sc, vha) != 0) {
1539                 /* See the comment in carp_add_addr() */
1540                 carp_hmac_prepare(sc);
1541                 carp_set_state(sc, INIT);
1542                 carp_setrun(sc, 0);
1543         }
1544 }
1545
1546 #ifdef INET6
1547 static int
1548 carp_set_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6)
1549 {
1550         struct ifnet *ifp;
1551         struct carp_if *cif;
1552         struct in6_ifaddr *ia, *ia_if;
1553         struct ip6_moptions *im6o = &sc->sc_im6o;
1554         struct in6_multi_mship *imm;
1555         struct in6_addr in6;
1556         int own, error;
1557
1558         if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1559                 carp_setrun(sc, 0);
1560                 return (0);
1561         }
1562
1563         /* we have to do it by hands to check we won't match on us */
1564         ia_if = NULL; own = 0;
1565         for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1566                 int i;
1567
1568                 for (i = 0; i < 4; i++) {
1569                         if ((sin6->sin6_addr.s6_addr32[i] &
1570                             ia->ia_prefixmask.sin6_addr.s6_addr32[i]) !=
1571                             (ia->ia_addr.sin6_addr.s6_addr32[i] &
1572                             ia->ia_prefixmask.sin6_addr.s6_addr32[i]))
1573                                 break;
1574                 }
1575                 /* and, yeah, we need a multicast-capable iface too */
1576                 if (ia->ia_ifp != &sc->sc_if &&
1577                     (ia->ia_ifp->if_flags & IFF_MULTICAST) &&
1578                     (i == 4)) {
1579                         if (!ia_if)
1580                                 ia_if = ia;
1581                         if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
1582                             &ia->ia_addr.sin6_addr))
1583                                 own++;
1584                 }
1585         }
1586
1587         if (!ia_if)
1588                 return (EADDRNOTAVAIL);
1589         ia = ia_if;
1590         ifp = ia->ia_ifp;
1591
1592         if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 ||
1593             (im6o->im6o_multicast_ifp && im6o->im6o_multicast_ifp != ifp))
1594                 return (EADDRNOTAVAIL);
1595
1596         if (!sc->sc_naddrs6) {
1597                 im6o->im6o_multicast_ifp = ifp;
1598
1599                 /* join CARP multicast address */
1600                 bzero(&in6, sizeof(in6));
1601                 in6.s6_addr16[0] = htons(0xff02);
1602                 in6.s6_addr8[15] = 0x12;
1603                 if (in6_setscope(&in6, ifp, NULL) != 0)
1604                         goto cleanup;
1605                 if ((imm = in6_joingroup(ifp, &in6, &error)) == NULL)
1606                         goto cleanup;
1607                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1608
1609                 /* join solicited multicast address */
1610                 bzero(&in6, sizeof(in6));
1611                 in6.s6_addr16[0] = htons(0xff02);
1612                 in6.s6_addr32[1] = 0;
1613                 in6.s6_addr32[2] = htonl(1);
1614                 in6.s6_addr32[3] = sin6->sin6_addr.s6_addr32[3];
1615                 in6.s6_addr8[12] = 0xff;
1616                 if (in6_setscope(&in6, ifp, NULL) != 0)
1617                         goto cleanup;
1618                 if ((imm = in6_joingroup(ifp, &in6, &error)) == NULL)
1619                         goto cleanup;
1620                 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1621         }
1622
1623         if (!ifp->if_carp) {
1624                 cif = kmalloc(sizeof(*cif), M_CARP, M_WAITOK | M_ZERO);
1625
1626                 if ((error = ifpromisc(ifp, 1))) {
1627                         kfree(cif, M_CARP);
1628                         goto cleanup;
1629                 }
1630
1631                 TAILQ_INIT(&cif->vhif_vrs);
1632                 ifp->if_carp = cif;
1633         } else {
1634                 struct carp_softc *vr;
1635
1636                 cif = ifp->if_carp;
1637                 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
1638                         if (vr != sc && vr->sc_vhid == sc->sc_vhid) {
1639                                 error = EINVAL;
1640                                 goto cleanup;
1641                         }
1642                 }
1643         }
1644         sc->sc_ia6 = ia;
1645         sc->sc_carpdev = ifp;
1646
1647         { /* XXX prevent endless loop if already in queue */
1648         struct carp_softc *vr, *after = NULL;
1649         int myself = 0;
1650         cif = ifp->if_carp;
1651
1652         TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
1653                 if (vr == sc)
1654                         myself = 1;
1655                 if (vr->sc_vhid < sc->sc_vhid)
1656                         after = vr;
1657         }
1658
1659         if (!myself) {
1660                 /* We're trying to keep things in order */
1661                 if (after == NULL)
1662                         TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list);
1663                 else
1664                         TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list);
1665         }
1666         }
1667
1668         sc->sc_naddrs6++;
1669         if (own)
1670                 sc->sc_advskew = 0;
1671         carp_sc_state(sc);
1672         carp_setrun(sc, 0);
1673
1674         return (0);
1675
1676 cleanup:
1677         /* clean up multicast memberships */
1678         if (!sc->sc_naddrs6) {
1679                 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1680                         imm = LIST_FIRST(&im6o->im6o_memberships);
1681                         LIST_REMOVE(imm, i6mm_chain);
1682                         in6_leavegroup(imm);
1683                 }
1684         }
1685         return (error);
1686 }
1687
1688 static int
1689 carp_del_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6)
1690 {
1691         int error = 0;
1692
1693         if (!--sc->sc_naddrs6) {
1694                 struct carp_if *cif = sc->sc_carpdev->if_carp;
1695                 struct ip6_moptions *im6o = &sc->sc_im6o;
1696
1697                 callout_stop(&sc->sc_ad_tmo);
1698                 sc->sc_vhid = -1;
1699                 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1700                         struct in6_multi_mship *imm =
1701                             LIST_FIRST(&im6o->im6o_memberships);
1702
1703                         LIST_REMOVE(imm, i6mm_chain);
1704                         in6_leavegroup(imm);
1705                 }
1706                 im6o->im6o_multicast_ifp = NULL;
1707                 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
1708                 if (TAILQ_EMPTY(&cif->vhif_vrs)) {
1709                         sc->sc_carpdev->if_carp = NULL;
1710                         kfree(cif, M_IFADDR);
1711                 }
1712         }
1713         return (error);
1714 }
1715 #endif /* INET6 */
1716
1717 static int
1718 carp_ioctl(struct ifnet *ifp, u_long cmd, caddr_t addr, struct ucred *cr)
1719 {
1720         struct carp_softc *sc = ifp->if_softc, *vr;
1721         struct carpreq carpr;
1722         struct ifaddr *ifa;
1723         struct ifreq *ifr;
1724         struct ifaliasreq *ifra;
1725         struct ifdrv *ifd;
1726         char devname[IFNAMSIZ];
1727         int error = 0;
1728
1729         ifa = (struct ifaddr *)addr;
1730         ifra = (struct ifaliasreq *)addr;
1731         ifr = (struct ifreq *)addr;
1732         ifd = (struct ifdrv *)addr;
1733
1734         switch (cmd) {
1735         case SIOCSIFADDR:
1736                 switch (ifa->ifa_addr->sa_family) {
1737 #ifdef INET
1738                 case AF_INET:
1739                         ifp->if_flags |= IFF_UP | IFF_RUNNING;
1740                         break;
1741 #endif /* INET */
1742 #ifdef INET6
1743                 case AF_INET6:
1744                         ifp->if_flags |= IFF_UP | IFF_RUNNING;
1745                         error = carp_set_addr6(sc, satosin6(ifa->ifa_addr));
1746                         break;
1747 #endif /* INET6 */
1748                 default:
1749                         error = EAFNOSUPPORT;
1750                         break;
1751                 }
1752                 break;
1753
1754         case SIOCAIFADDR:
1755                 switch (ifa->ifa_addr->sa_family) {
1756 #ifdef INET
1757                 case AF_INET:
1758                         panic("SIOCAIFADDR should never be seen\n");
1759 #endif /* INET */
1760 #ifdef INET6
1761                 case AF_INET6:
1762                         ifp->if_flags |= IFF_UP | IFF_RUNNING;
1763                         error = carp_set_addr6(sc, satosin6(&ifra->ifra_addr));
1764                         break;
1765 #endif /* INET6 */
1766                 default:
1767                         error = EAFNOSUPPORT;
1768                         break;
1769                 }
1770                 break;
1771
1772         case SIOCDIFADDR:
1773                 switch (ifa->ifa_addr->sa_family) {
1774 #ifdef INET
1775                 case AF_INET:
1776                         panic("SIOCDIFADDR should never be seen\n");
1777 #endif /* INET */
1778 #ifdef INET6
1779                 case AF_INET6:
1780                         error = carp_del_addr6(sc, satosin6(&ifra->ifra_addr));
1781                         break;
1782 #endif /* INET6 */
1783                 default:
1784                         error = EAFNOSUPPORT;
1785                         break;
1786                 }
1787                 break;
1788
1789         case SIOCSIFFLAGS:
1790                 if (ifp->if_flags & IFF_UP) {
1791                         if ((ifp->if_flags & IFF_RUNNING) == 0) {
1792                                 ifp->if_flags |= IFF_RUNNING;
1793                                 carp_set_state(sc, INIT);
1794                                 carp_setrun(sc, 0);
1795                         }
1796                 } else if (ifp->if_flags & IFF_RUNNING) {
1797                         carp_stop(sc, 0);
1798                 }
1799                 break;
1800
1801         case SIOCSVH:
1802                 error = priv_check_cred(cr, PRIV_ROOT, NULL_CRED_OKAY);
1803                 if (error)
1804                         break;
1805                 error = copyin(ifr->ifr_data, &carpr, sizeof(carpr));
1806                 if (error)
1807                         break;
1808
1809                 error = 1;
1810                 if ((ifp->if_flags & IFF_RUNNING) &&
1811                     sc->sc_state != INIT && carpr.carpr_state != sc->sc_state) {
1812                         switch (carpr.carpr_state) {
1813                         case BACKUP:
1814                                 callout_stop(&sc->sc_ad_tmo);
1815                                 carp_set_state(sc, BACKUP);
1816                                 carp_setrun(sc, 0);
1817                                 carp_setroute(sc, RTM_DELETE);
1818                                 break;
1819
1820                         case MASTER:
1821                                 carp_master_down(sc);
1822                                 break;
1823
1824                         default:
1825                                 break;
1826                         }
1827                 }
1828                 if (carpr.carpr_vhid > 0) {
1829                         if (carpr.carpr_vhid > 255) {
1830                                 error = EINVAL;
1831                                 break;
1832                         }
1833                         if (sc->sc_carpdev) {
1834                                 struct carp_if *cif = sc->sc_carpdev->if_carp;
1835
1836                                 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
1837                                         if (vr != sc &&
1838                                             vr->sc_vhid == carpr.carpr_vhid)
1839                                                 return EEXIST;
1840                                 }
1841                         }
1842                         sc->sc_vhid = carpr.carpr_vhid;
1843                         IF_LLADDR(ifp)[0] = 0;
1844                         IF_LLADDR(ifp)[1] = 0;
1845                         IF_LLADDR(ifp)[2] = 0x5e;
1846                         IF_LLADDR(ifp)[3] = 0;
1847                         IF_LLADDR(ifp)[4] = 1;
1848                         IF_LLADDR(ifp)[5] = sc->sc_vhid;
1849                         error--;
1850                 }
1851                 if (carpr.carpr_advbase > 0 || carpr.carpr_advskew > 0) {
1852                         if (carpr.carpr_advskew >= 255) {
1853                                 error = EINVAL;
1854                                 break;
1855                         }
1856                         if (carpr.carpr_advbase > 255) {
1857                                 error = EINVAL;
1858                                 break;
1859                         }
1860                         sc->sc_advbase = carpr.carpr_advbase;
1861                         sc->sc_advskew = carpr.carpr_advskew;
1862                         error--;
1863                 }
1864                 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key));
1865                 if (error > 0) {
1866                         error = EINVAL;
1867                 } else {
1868                         error = 0;
1869                         carp_setrun(sc, 0);
1870                 }
1871                 break;
1872
1873         case SIOCGVH:
1874                 bzero(&carpr, sizeof(carpr));
1875                 carpr.carpr_state = sc->sc_state;
1876                 carpr.carpr_vhid = sc->sc_vhid;
1877                 carpr.carpr_advbase = sc->sc_advbase;
1878                 carpr.carpr_advskew = sc->sc_advskew;
1879                 error = priv_check_cred(cr, PRIV_ROOT, NULL_CRED_OKAY);
1880                 if (error == 0) {
1881                         bcopy(sc->sc_key, carpr.carpr_key,
1882                               sizeof(carpr.carpr_key));
1883                 }
1884
1885                 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr));
1886                 break;
1887
1888         case SIOCGDRVSPEC:
1889                 switch (ifd->ifd_cmd) {
1890                 case CARPGDEVNAME:
1891                         if (ifd->ifd_len != sizeof(devname))
1892                                 error = EINVAL;
1893                         break;
1894
1895                 case CARPGVHADDR:
1896                         break;
1897
1898                 default:
1899                         error = EINVAL;
1900                         break;
1901                 }
1902                 if (error)
1903                         break;
1904
1905                 switch (ifd->ifd_cmd) {
1906                 case CARPGVHADDR:
1907                         error = carp_get_vhaddr(sc, ifd);
1908                         break;
1909
1910                 case CARPGDEVNAME:
1911                         bzero(devname, sizeof(devname));
1912                         if (sc->sc_carpdev != NULL) {
1913                                 strlcpy(devname, sc->sc_carpdev->if_xname,
1914                                         sizeof(devname));
1915                         }
1916                         error = copyout(devname, ifd->ifd_data,
1917                                         sizeof(devname));
1918                         break;
1919                 }
1920                 break;
1921
1922         default:
1923                 error = EINVAL;
1924                 break;
1925         }
1926         carp_hmac_prepare(sc);
1927         return error;
1928 }
1929
1930 /*
1931  * XXX: this is looutput. We should eventually use it from there.
1932  */
1933 static int
1934 carp_looutput(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
1935     struct rtentry *rt)
1936 {
1937         uint32_t af;
1938
1939         M_ASSERTPKTHDR(m); /* check if we have the packet header */
1940
1941         if (rt && rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
1942                 m_freem(m);
1943                 return (rt->rt_flags & RTF_BLACKHOLE ? 0 :
1944                         rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
1945         }
1946
1947         ifp->if_opackets++;
1948         ifp->if_obytes += m->m_pkthdr.len;
1949
1950         /* BPF writes need to be handled specially. */
1951         if (dst->sa_family == AF_UNSPEC) {
1952                 bcopy(dst->sa_data, &af, sizeof(af));
1953                 dst->sa_family = af;
1954         }
1955
1956 #if 1   /* XXX */
1957         switch (dst->sa_family) {
1958         case AF_INET:
1959         case AF_INET6:
1960         case AF_IPX:
1961         case AF_APPLETALK:
1962                 break;
1963
1964         default:
1965                 m_freem(m);
1966                 return (EAFNOSUPPORT);
1967         }
1968 #endif
1969         return (if_simloop(ifp, m, dst->sa_family, 0));
1970 }
1971
1972 /*
1973  * Start output on carp interface. This function should never be called.
1974  */
1975 static void
1976 carp_start(struct ifnet *ifp)
1977 {
1978 #ifdef DEBUG
1979         kprintf("%s: start called\n", ifp->if_xname);
1980 #endif
1981 }
1982
1983 int
1984 carp_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1985     struct rtentry *rt)
1986 {
1987         struct m_tag *mtag;
1988         struct carp_softc *sc;
1989         struct ifnet *carp_ifp;
1990         struct ether_header *eh;
1991
1992         if (!sa)
1993                 return (0);
1994
1995         switch (sa->sa_family) {
1996 #ifdef INET
1997         case AF_INET:
1998                 break;
1999 #endif /* INET */
2000 #ifdef INET6
2001         case AF_INET6:
2002                 break;
2003 #endif /* INET6 */
2004         default:
2005                 return (0);
2006         }
2007
2008         mtag = m_tag_find(m, PACKET_TAG_CARP, NULL);
2009         if (mtag == NULL)
2010                 return (0);
2011
2012         bcopy(mtag + 1, &carp_ifp, sizeof(struct ifnet *));
2013         sc = carp_ifp->if_softc;
2014
2015         /* Set the source MAC address to Virtual Router MAC Address */
2016         switch (ifp->if_type) {
2017         case IFT_ETHER:
2018         case IFT_L2VLAN:
2019                 eh = mtod(m, struct ether_header *);
2020                 eh->ether_shost[0] = 0;
2021                 eh->ether_shost[1] = 0;
2022                 eh->ether_shost[2] = 0x5e;
2023                 eh->ether_shost[3] = 0;
2024                 eh->ether_shost[4] = 1;
2025                 eh->ether_shost[5] = sc->sc_vhid;
2026                 break;
2027
2028         default:
2029                 if_printf(ifp, "carp is not supported for this "
2030                           "interface type\n");
2031                 return (EOPNOTSUPP);
2032         }
2033         return (0);
2034 }
2035
2036 static void
2037 carp_set_state(struct carp_softc *sc, int state)
2038 {
2039         struct ifnet *cifp = &sc->sc_if;
2040
2041         if (sc->sc_state == state)
2042                 return;
2043         sc->sc_state = state;
2044
2045         switch (sc->sc_state) {
2046         case BACKUP:
2047                 cifp->if_link_state = LINK_STATE_DOWN;
2048                 break;
2049
2050         case MASTER:
2051                 cifp->if_link_state = LINK_STATE_UP;
2052                 break;
2053
2054         default:
2055                 cifp->if_link_state = LINK_STATE_UNKNOWN;
2056                 break;
2057         }
2058         rt_ifmsg(cifp);
2059 }
2060
2061 void
2062 carp_carpdev_state(void *v)
2063 {
2064         struct carp_if *cif = v;
2065         struct carp_softc *sc;
2066
2067         TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list)
2068                 carp_sc_state(sc);
2069 }
2070
2071 static void
2072 carp_sc_state(struct carp_softc *sc)
2073 {
2074         if (!(sc->sc_carpdev->if_flags & IFF_UP)) {
2075                 callout_stop(&sc->sc_ad_tmo);
2076                 callout_stop(&sc->sc_md_tmo);
2077                 callout_stop(&sc->sc_md6_tmo);
2078                 carp_set_state(sc, INIT);
2079                 carp_setrun(sc, 0);
2080                 if (!sc->sc_suppress) {
2081                         carp_suppress_preempt++;
2082                         if (carp_suppress_preempt == 1)
2083                                 carp_send_ad_all();
2084                 }
2085                 sc->sc_suppress = 1;
2086         } else {
2087                 carp_set_state(sc, INIT);
2088                 carp_setrun(sc, 0);
2089                 if (sc->sc_suppress)
2090                         carp_suppress_preempt--;
2091                 sc->sc_suppress = 0;
2092         }
2093 }
2094
2095 static void
2096 carp_stop(struct carp_softc *sc, int detach)
2097 {
2098         sc->sc_if.if_flags &= ~IFF_RUNNING;
2099
2100         callout_stop(&sc->sc_ad_tmo);
2101         callout_stop(&sc->sc_md_tmo);
2102         callout_stop(&sc->sc_md6_tmo);
2103
2104         if (!detach && sc->sc_state == MASTER)
2105                 carp_send_ad(sc);
2106
2107         if (sc->sc_suppress)
2108                 carp_suppress_preempt--;
2109         sc->sc_suppress = 0;
2110
2111         if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS)
2112                 carp_suppress_preempt--;
2113         sc->sc_sendad_errors = 0;
2114         sc->sc_sendad_success = 0;
2115
2116         carp_set_state(sc, INIT);
2117         carp_setrun(sc, 0);
2118 }
2119
2120 static void
2121 carp_reset(struct carp_softc *sc, int detach)
2122 {
2123         struct ifnet *cifp = &sc->sc_if;
2124
2125         carp_stop(sc, detach);
2126         if (!sc->sc_dead && (cifp->if_flags & IFF_UP))
2127                 cifp->if_flags |= IFF_RUNNING;
2128 }
2129
2130 static int
2131 carp_activate_vhaddr(struct carp_softc *sc, struct carp_vhaddr *vha,
2132                      struct ifnet *ifp, const struct in_ifaddr *ia_if, int own)
2133 {
2134         struct ip_moptions *imo = &sc->sc_imo;
2135         struct carp_if *cif;
2136         struct carp_softc *vr, *after = NULL;
2137         int onlist, error;
2138 #ifdef INVARIANTS
2139         int assert_onlist;
2140 #endif
2141
2142         KKASSERT(vha->vha_ia != NULL);
2143
2144         KASSERT(ia_if != NULL, ("NULL backing address\n"));
2145         KASSERT(vha->vha_iaback == NULL, ("%p is already activated\n", vha));
2146         KASSERT((vha->vha_flags & CARP_VHAF_OWNER) == 0,
2147                 ("inactive vhaddr %p is the address owner\n", vha));
2148
2149         KASSERT(sc->sc_carpdev == NULL || sc->sc_carpdev == ifp,
2150                 ("%s is already on %s\n", sc->sc_if.if_xname,
2151                  sc->sc_carpdev->if_xname));
2152
2153         KASSERT(imo->imo_multicast_ifp == NULL ||
2154                 imo->imo_multicast_ifp == ifp,
2155                 ("%s didn't leave mcast group on %s\n",
2156                  sc->sc_if.if_xname, imo->imo_multicast_ifp->if_xname));
2157
2158         if (imo->imo_num_memberships == 0) {
2159                 struct in_addr addr;
2160
2161                 addr.s_addr = htonl(INADDR_CARP_GROUP);
2162                 if ((imo->imo_membership[0] = in_addmulti(&addr, ifp)) == NULL)
2163                         return ENOBUFS;
2164                 imo->imo_num_memberships++;
2165                 imo->imo_multicast_ifp = ifp;
2166                 imo->imo_multicast_ttl = CARP_DFLTTL;
2167                 imo->imo_multicast_loop = 0;
2168         }
2169
2170         if (!ifp->if_carp) {
2171                 KASSERT(sc->sc_carpdev == NULL,
2172                         ("%s is already on %s\n", sc->sc_if.if_xname,
2173                          sc->sc_carpdev->if_xname));
2174
2175                 cif = kmalloc(sizeof(*cif), M_CARP, M_WAITOK | M_ZERO);
2176
2177                 error = ifpromisc(ifp, 1);
2178                 if (error) {
2179                         kfree(cif, M_CARP);
2180                         goto cleanup;
2181                 }
2182
2183                 TAILQ_INIT(&cif->vhif_vrs);
2184                 ifp->if_carp = cif;
2185         } else {
2186                 cif = ifp->if_carp;
2187                 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
2188                         if (vr != sc && vr->sc_vhid == sc->sc_vhid) {
2189                                 error = EINVAL;
2190                                 goto cleanup;
2191                         }
2192                 }
2193         }
2194
2195 #ifdef INVARIANTS
2196         if (sc->sc_carpdev != NULL)
2197                 assert_onlist = 1;
2198         else
2199                 assert_onlist = 0;
2200 #endif
2201         sc->sc_ia = ia_if;
2202         sc->sc_carpdev = ifp;
2203
2204         cif = ifp->if_carp;
2205         onlist = 0;
2206         TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
2207                 if (vr == sc)
2208                         onlist = 1;
2209                 if (vr->sc_vhid < sc->sc_vhid)
2210                         after = vr;
2211         }
2212
2213 #ifdef INVARIANTS
2214         if (assert_onlist) {
2215                 KASSERT(onlist, ("%s is not on %s carp list\n",
2216                         sc->sc_if.if_xname, ifp->if_xname));
2217         } else {
2218                 KASSERT(!onlist, ("%s is already on %s carp list\n",
2219                         sc->sc_if.if_xname, ifp->if_xname));
2220         }
2221 #endif
2222
2223         if (!onlist) {
2224                 /* We're trying to keep things in order */
2225                 if (after == NULL)
2226                         TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list);
2227                 else
2228                         TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list);
2229         }
2230
2231         vha->vha_iaback = ia_if;
2232         sc->sc_naddrs++;
2233
2234         if (own) {
2235                 vha->vha_flags |= CARP_VHAF_OWNER;
2236
2237                 /* XXX save user configured advskew? */
2238                 sc->sc_advskew = 0;
2239         }
2240
2241         carp_hmac_prepare(sc);
2242         carp_set_state(sc, INIT);
2243         carp_setrun(sc, 0);
2244         return 0;
2245 cleanup:
2246         carp_multicast_cleanup(sc);
2247         return error;
2248 }
2249
2250 static void
2251 carp_deactivate_vhaddr(struct carp_softc *sc, struct carp_vhaddr *vha)
2252 {
2253         KKASSERT(vha->vha_ia != NULL);
2254
2255         carp_hmac_prepare(sc);
2256
2257         if (vha->vha_iaback == NULL) {
2258                 KASSERT((vha->vha_flags & CARP_VHAF_OWNER) == 0,
2259                         ("inactive vhaddr %p is the address owner\n", vha));
2260                 return;
2261         }
2262
2263         vha->vha_flags &= ~CARP_VHAF_OWNER;
2264
2265         KKASSERT(sc->sc_naddrs > 0);
2266         vha->vha_iaback = NULL;
2267         sc->sc_naddrs--;
2268         if (!sc->sc_naddrs) {
2269                 if (sc->sc_naddrs6) {
2270                         carp_multicast_cleanup(sc);
2271                         sc->sc_ia = NULL;
2272                 } else {
2273                         carp_detach(sc, 0);
2274                 }
2275         }
2276 }
2277
2278 static void
2279 carp_link_addrs(struct carp_softc *sc, struct ifnet *ifp, struct ifaddr *ifa_if)
2280 {
2281         struct carp_vhaddr *vha;
2282         struct in_ifaddr *ia_if;
2283
2284         KKASSERT(ifa_if->ifa_addr->sa_family == AF_INET);
2285         ia_if = ifatoia(ifa_if);
2286
2287         if ((ia_if->ia_flags & IFA_ROUTE) == 0)
2288                 return;
2289
2290         /*
2291          * Test each inactive vhaddr against the newly added address.
2292          * If the newly added address could be the backing address,
2293          * then activate the matching vhaddr.
2294          */
2295         TAILQ_FOREACH(vha, &sc->sc_vha_list, vha_link) {
2296                 const struct in_ifaddr *ia;
2297                 u_long iaddr;
2298                 int own;
2299
2300                 if (vha->vha_iaback != NULL)
2301                         continue;
2302
2303                 ia = vha->vha_ia;
2304                 iaddr = ntohl(ia->ia_addr.sin_addr.s_addr);
2305
2306                 if ((iaddr & ia_if->ia_subnetmask) != ia_if->ia_subnet)
2307                         continue;
2308
2309                 own = 0;
2310                 if (ia->ia_addr.sin_addr.s_addr ==
2311                     ia_if->ia_addr.sin_addr.s_addr)
2312                         own = 1;
2313
2314                 carp_activate_vhaddr(sc, vha, ifp, ia_if, own);
2315         }
2316 }
2317
2318 static void
2319 carp_unlink_addrs(struct carp_softc *sc, struct ifnet *ifp,
2320                   struct ifaddr *ifa_if)
2321 {
2322         struct carp_vhaddr *vha;
2323         struct in_ifaddr *ia_if;
2324
2325         KKASSERT(ifa_if->ifa_addr->sa_family == AF_INET);
2326         ia_if = ifatoia(ifa_if);
2327
2328         /*
2329          * Ad src address is deleted; set it to NULL.
2330          * Following loop will try pick up a new ad src address
2331          * if one of the vhaddr could retain its backing address.
2332          */
2333         if (sc->sc_ia == ia_if)
2334                 sc->sc_ia = NULL;
2335
2336         /*
2337          * Test each active vhaddr against the deleted address.
2338          * If the deleted address is vhaddr address's backing
2339          * address, then deactivate the vhaddr.
2340          */
2341         TAILQ_FOREACH(vha, &sc->sc_vha_list, vha_link) {
2342                 if (vha->vha_iaback == NULL)
2343                         continue;
2344
2345                 if (vha->vha_iaback == ia_if)
2346                         carp_deactivate_vhaddr(sc, vha);
2347                 else if (sc->sc_ia == NULL)
2348                         sc->sc_ia = vha->vha_iaback;
2349         }
2350 }
2351
2352 static void
2353 carp_update_addrs(struct carp_softc *sc)
2354 {
2355         struct carp_vhaddr *vha;
2356
2357         KKASSERT(sc->sc_carpdev == NULL);
2358
2359         TAILQ_FOREACH(vha, &sc->sc_vha_list, vha_link)
2360                 carp_config_vhaddr(sc, vha);
2361 }
2362
2363 static void
2364 carp_ifaddr(void *arg __unused, struct ifnet *ifp,
2365             enum ifaddr_event event, struct ifaddr *ifa)
2366 {
2367         struct carp_softc *sc;
2368
2369         if (ifa->ifa_addr->sa_family != AF_INET)
2370                 return;
2371
2372         if (ifp->if_type == IFT_CARP) {
2373                 /*
2374                  * Address is changed on carp(4) interface
2375                  */
2376                 switch (event) {
2377                 case IFADDR_EVENT_ADD:
2378                         carp_add_addr(ifp->if_softc, ifa);
2379                         break;
2380
2381                 case IFADDR_EVENT_CHANGE:
2382                         carp_config_addr(ifp->if_softc, ifa);
2383                         break;
2384
2385                 case IFADDR_EVENT_DELETE:
2386                         carp_del_addr(ifp->if_softc, ifa);
2387                         break;
2388                 }
2389                 return;
2390         }
2391
2392         /*
2393          * Address is changed on non-carp(4) interface
2394          */
2395         if ((ifp->if_flags & IFF_MULTICAST) == 0)
2396                 return;
2397
2398         crit_enter();
2399         LIST_FOREACH(sc, &carpif_list, sc_next) {
2400                 if (sc->sc_carpdev != NULL && sc->sc_carpdev != ifp) {
2401                         /* Not the parent iface; skip */
2402                         continue;
2403                 }
2404
2405                 switch (event) {
2406                 case IFADDR_EVENT_ADD:
2407                         carp_link_addrs(sc, ifp, ifa);
2408                         break;
2409
2410                 case IFADDR_EVENT_DELETE:
2411                         if (sc->sc_carpdev != NULL) {
2412                                 carp_unlink_addrs(sc, ifp, ifa);
2413                                 if (sc->sc_carpdev == NULL)
2414                                         carp_update_addrs(sc);
2415                         } else {
2416                                 /*
2417                                  * The carp(4) interface didn't have a
2418                                  * parent iface, so it is not possible
2419                                  * that it will contain any address to
2420                                  * be unlinked.
2421                                  */
2422                         }
2423                         break;
2424
2425                 case IFADDR_EVENT_CHANGE:
2426                         if (sc->sc_carpdev == NULL) {
2427                                 /*
2428                                  * The carp(4) interface didn't have a
2429                                  * parent iface, so it is not possible
2430                                  * that it will contain any address to
2431                                  * be updated.
2432                                  */
2433                                 carp_link_addrs(sc, ifp, ifa);
2434                         } else {
2435                                 /*
2436                                  * First try breaking tie with the old
2437                                  * address.  Then see whether we could
2438                                  * link certain vhaddr to the new address.
2439                                  * If that fails, i.e. carpdev is NULL,
2440                                  * we try a global update.
2441                                  *
2442                                  * NOTE: The above order is critical.
2443                                  */
2444                                 carp_unlink_addrs(sc, ifp, ifa);
2445                                 carp_link_addrs(sc, ifp, ifa);
2446                                 if (sc->sc_carpdev == NULL)
2447                                         carp_update_addrs(sc);
2448                         }
2449                         break;
2450                 }
2451         }
2452         crit_exit();
2453 }
2454
2455 static int
2456 carp_modevent(module_t mod, int type, void *data)
2457 {
2458         switch (type) {
2459         case MOD_LOAD:
2460                 LIST_INIT(&carpif_list);
2461                 carp_ifdetach_event =
2462                 EVENTHANDLER_REGISTER(ifnet_detach_event, carp_ifdetach, NULL,
2463                                       EVENTHANDLER_PRI_ANY);
2464                 carp_ifaddr_event =
2465                 EVENTHANDLER_REGISTER(ifaddr_event, carp_ifaddr, NULL,
2466                                       EVENTHANDLER_PRI_ANY);
2467                 if_clone_attach(&carp_cloner);
2468                 break;
2469
2470         case MOD_UNLOAD:
2471                 EVENTHANDLER_DEREGISTER(ifnet_detach_event,
2472                                         carp_ifdetach_event);
2473                 EVENTHANDLER_DEREGISTER(ifaddr_event,
2474                                         carp_ifaddr_event);
2475                 if_clone_detach(&carp_cloner);
2476                 break;
2477
2478         default:
2479                 return (EINVAL);
2480         }
2481         return (0);
2482 }
2483
2484 static moduledata_t carp_mod = {
2485         "carp",
2486         carp_modevent,
2487         0
2488 };
2489 DECLARE_MODULE(carp, carp_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);