1 /*-
2 * Copyright (c) 1989 Stephen Deering
3 * Copyright (c) 1992, 1993
4 * The Regents of the University of California. All rights reserved.
5 *
6 * This code is derived from software contributed to Berkeley by
7 * Stephen Deering of Stanford University.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
34 */
35
36 /*
37 * IP multicast forwarding procedures
38 *
39 * Written by David Waitzman, BBN Labs, August 1988.
40 * Modified by Steve Deering, Stanford, February 1989.
41 * Modified by Mark J. Steiglitz, Stanford, May, 1991
42 * Modified by Van Jacobson, LBL, January 1993
43 * Modified by Ajit Thyagarajan, PARC, August 1993
44 * Modified by Bill Fenner, PARC, April 1995
45 * Modified by Ahmed Helmy, SGI, June 1996
46 * Modified by George Edmond Eddy (Rusty), ISI, February 1998
47 * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
48 * Modified by Hitoshi Asaeda, WIDE, August 2000
49 * Modified by Pavlin Radoslavov, ICSI, October 2002
50 *
51 * MROUTING Revision: 3.5
52 * and PIM-SMv2 and PIM-DM support, advanced API support,
53 * bandwidth metering and signaling
54 */
55
56 /*
57 * TODO: Prefix functions with ipmf_.
58 * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
59 * domain attachment (if_afdata) so we can track consumers of that service.
60 * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
61 * move it to socket options.
62 * TODO: Cleanup LSRR removal further.
63 * TODO: Push RSVP stubs into raw_ip.c.
64 * TODO: Use bitstring.h for vif set.
65 * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
66 * TODO: Sync ip6_mroute.c with this file.
67 */
68
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
71
72 #include "opt_inet.h"
73 #include "opt_mrouting.h"
74
75 #define _PIM_VT 1
76
77 #include <sys/param.h>
78 #include <sys/kernel.h>
79 #include <sys/stddef.h>
80 #include <sys/lock.h>
81 #include <sys/ktr.h>
82 #include <sys/malloc.h>
83 #include <sys/mbuf.h>
84 #include <sys/module.h>
85 #include <sys/priv.h>
86 #include <sys/protosw.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/socketvar.h>
90 #include <sys/sockio.h>
91 #include <sys/sx.h>
92 #include <sys/sysctl.h>
93 #include <sys/syslog.h>
94 #include <sys/systm.h>
95 #include <sys/time.h>
96
97 #include <net/if.h>
98 #include <net/netisr.h>
99 #include <net/route.h>
100 #include <net/vnet.h>
101
102 #include <netinet/in.h>
103 #include <netinet/igmp.h>
104 #include <netinet/in_systm.h>
105 #include <netinet/in_var.h>
106 #include <netinet/ip.h>
107 #include <netinet/ip_encap.h>
108 #include <netinet/ip_mroute.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/ip_options.h>
111 #include <netinet/pim.h>
112 #include <netinet/pim_var.h>
113 #include <netinet/udp.h>
114
115 #include <machine/in_cksum.h>
116
117 #ifndef KTR_IPMF
118 #define KTR_IPMF KTR_INET
119 #endif
120
121 #define VIFI_INVALID ((vifi_t) -1)
122 #define M_HASCL(m) ((m)->m_flags & M_EXT)
123
124 static VNET_DEFINE(uint32_t, last_tv_sec); /* last time we processed this */
125 #define V_last_tv_sec VNET(last_tv_sec)
126
127 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
128
129 /*
130 * Locking. We use two locks: one for the virtual interface table and
131 * one for the forwarding table. These locks may be nested in which case
132 * the VIF lock must always be taken first. Note that each lock is used
133 * to cover not only the specific data structure but also related data
134 * structures.
135 */
136
137 static struct mtx mrouter_mtx;
138 #define MROUTER_LOCK() mtx_lock(&mrouter_mtx)
139 #define MROUTER_UNLOCK() mtx_unlock(&mrouter_mtx)
140 #define MROUTER_LOCK_ASSERT() mtx_assert(&mrouter_mtx, MA_OWNED)
141 #define MROUTER_LOCK_INIT() \
142 mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
143 #define MROUTER_LOCK_DESTROY() mtx_destroy(&mrouter_mtx)
144
145 static int ip_mrouter_cnt; /* # of vnets with active mrouters */
146 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
147
148 static VNET_DEFINE(struct mrtstat, mrtstat);
149 #define V_mrtstat VNET(mrtstat)
150 SYSCTL_VNET_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
151 &VNET_NAME(mrtstat), mrtstat,
152 "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
153 "netinet/ip_mroute.h)");
154
155 static VNET_DEFINE(u_long, mfchash);
156 #define V_mfchash VNET(mfchash)
157 #define MFCHASH(a, g) \
158 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
159 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
160 #define MFCHASHSIZE 256
161
162 static u_long mfchashsize; /* Hash size */
163 static VNET_DEFINE(u_char *, nexpire); /* 0..mfchashsize-1 */
164 #define V_nexpire VNET(nexpire)
165 static VNET_DEFINE(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
166 #define V_mfchashtbl VNET(mfchashtbl)
167
168 static struct mtx mfc_mtx;
169 #define MFC_LOCK() mtx_lock(&mfc_mtx)
170 #define MFC_UNLOCK() mtx_unlock(&mfc_mtx)
171 #define MFC_LOCK_ASSERT() mtx_assert(&mfc_mtx, MA_OWNED)
172 #define MFC_LOCK_INIT() \
173 mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
174 #define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx)
175
176 static VNET_DEFINE(vifi_t, numvifs);
177 #define V_numvifs VNET(numvifs)
178 static VNET_DEFINE(struct vif, viftable[MAXVIFS]);
179 #define V_viftable VNET(viftable)
180 SYSCTL_VNET_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
181 &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]",
182 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
183
184 static struct mtx vif_mtx;
185 #define VIF_LOCK() mtx_lock(&vif_mtx)
186 #define VIF_UNLOCK() mtx_unlock(&vif_mtx)
187 #define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED)
188 #define VIF_LOCK_INIT() \
189 mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
190 #define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx)
191
192 static eventhandler_tag if_detach_event_tag = NULL;
193
194 static VNET_DEFINE(struct callout, expire_upcalls_ch);
195 #define V_expire_upcalls_ch VNET(expire_upcalls_ch)
196
197 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
198 #define UPCALL_EXPIRE 6 /* number of timeouts */
199
200 /*
201 * Bandwidth meter variables and constants
202 */
203 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
204 /*
205 * Pending timeouts are stored in a hash table, the key being the
206 * expiration time. Periodically, the entries are analysed and processed.
207 */
208 #define BW_METER_BUCKETS 1024
209 static VNET_DEFINE(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]);
210 #define V_bw_meter_timers VNET(bw_meter_timers)
211 static VNET_DEFINE(struct callout, bw_meter_ch);
212 #define V_bw_meter_ch VNET(bw_meter_ch)
213 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
214
215 /*
216 * Pending upcalls are stored in a vector which is flushed when
217 * full, or periodically
218 */
219 static VNET_DEFINE(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]);
220 #define V_bw_upcalls VNET(bw_upcalls)
221 static VNET_DEFINE(u_int, bw_upcalls_n); /* # of pending upcalls */
222 #define V_bw_upcalls_n VNET(bw_upcalls_n)
223 static VNET_DEFINE(struct callout, bw_upcalls_ch);
224 #define V_bw_upcalls_ch VNET(bw_upcalls_ch)
225
226 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
227
228 static VNET_DEFINE(struct pimstat, pimstat);
229 #define V_pimstat VNET(pimstat)
230
231 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
232 SYSCTL_VNET_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
233 &VNET_NAME(pimstat), pimstat,
234 "PIM Statistics (struct pimstat, netinet/pim_var.h)");
235
236 static u_long pim_squelch_wholepkt = 0;
237 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
238 &pim_squelch_wholepkt, 0,
239 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
240
241 extern struct domain inetdomain;
242 static const struct protosw in_pim_protosw = {
243 .pr_type = SOCK_RAW,
244 .pr_domain = &inetdomain,
245 .pr_protocol = IPPROTO_PIM,
246 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
247 .pr_input = pim_input,
248 .pr_output = (pr_output_t*)rip_output,
249 .pr_ctloutput = rip_ctloutput,
250 .pr_usrreqs = &rip_usrreqs
251 };
252 static const struct encaptab *pim_encap_cookie;
253
254 static int pim_encapcheck(const struct mbuf *, int, int, void *);
255
256 /*
257 * Note: the PIM Register encapsulation adds the following in front of a
258 * data packet:
259 *
260 * struct pim_encap_hdr {
261 * struct ip ip;
262 * struct pim_encap_pimhdr pim;
263 * }
264 *
265 */
266
267 struct pim_encap_pimhdr {
268 struct pim pim;
269 uint32_t flags;
270 };
271 #define PIM_ENCAP_TTL 64
272
273 static struct ip pim_encap_iphdr = {
274 #if BYTE_ORDER == LITTLE_ENDIAN
275 sizeof(struct ip) >> 2,
276 IPVERSION,
277 #else
278 IPVERSION,
279 sizeof(struct ip) >> 2,
280 #endif
281 0, /* tos */
282 sizeof(struct ip), /* total length */
283 0, /* id */
284 0, /* frag offset */
285 PIM_ENCAP_TTL,
286 IPPROTO_PIM,
287 0, /* checksum */
288 };
289
290 static struct pim_encap_pimhdr pim_encap_pimhdr = {
291 {
292 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
293 0, /* reserved */
294 0, /* checksum */
295 },
296 0 /* flags */
297 };
298
299 static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID;
300 #define V_reg_vif_num VNET(reg_vif_num)
301 static VNET_DEFINE(struct ifnet, multicast_register_if);
302 #define V_multicast_register_if VNET(multicast_register_if)
303
304 /*
305 * Private variables.
306 */
307
308 static u_long X_ip_mcast_src(int);
309 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
310 struct ip_moptions *);
311 static int X_ip_mrouter_done(void);
312 static int X_ip_mrouter_get(struct socket *, struct sockopt *);
313 static int X_ip_mrouter_set(struct socket *, struct sockopt *);
314 static int X_legal_vif_num(int);
315 static int X_mrt_ioctl(u_long, caddr_t, int);
316
317 static int add_bw_upcall(struct bw_upcall *);
318 static int add_mfc(struct mfcctl2 *);
319 static int add_vif(struct vifctl *);
320 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
321 static void bw_meter_process(void);
322 static void bw_meter_receive_packet(struct bw_meter *, int,
323 struct timeval *);
324 static void bw_upcalls_send(void);
325 static int del_bw_upcall(struct bw_upcall *);
326 static int del_mfc(struct mfcctl2 *);
327 static int del_vif(vifi_t);
328 static int del_vif_locked(vifi_t);
329 static void expire_bw_meter_process(void *);
330 static void expire_bw_upcalls_send(void *);
331 static void expire_mfc(struct mfc *);
332 static void expire_upcalls(void *);
333 static void free_bw_list(struct bw_meter *);
334 static int get_sg_cnt(struct sioc_sg_req *);
335 static int get_vif_cnt(struct sioc_vif_req *);
336 static void if_detached_event(void *, struct ifnet *);
337 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
338 static int ip_mrouter_init(struct socket *, int);
339 static __inline struct mfc *
340 mfc_find(struct in_addr *, struct in_addr *);
341 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
342 static struct mbuf *
343 pim_register_prepare(struct ip *, struct mbuf *);
344 static int pim_register_send(struct ip *, struct vif *,
345 struct mbuf *, struct mfc *);
346 static int pim_register_send_rp(struct ip *, struct vif *,
347 struct mbuf *, struct mfc *);
348 static int pim_register_send_upcall(struct ip *, struct vif *,
349 struct mbuf *, struct mfc *);
350 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
351 static void send_packet(struct vif *, struct mbuf *);
352 static int set_api_config(uint32_t *);
353 static int set_assert(int);
354 static int socket_send(struct socket *, struct mbuf *,
355 struct sockaddr_in *);
356 static void unschedule_bw_meter(struct bw_meter *);
357
358 /*
359 * Kernel multicast forwarding API capabilities and setup.
360 * If more API capabilities are added to the kernel, they should be
361 * recorded in `mrt_api_support'.
362 */
363 #define MRT_API_VERSION 0x0305
364
365 static const int mrt_api_version = MRT_API_VERSION;
366 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
367 MRT_MFC_FLAGS_BORDER_VIF |
368 MRT_MFC_RP |
369 MRT_MFC_BW_UPCALL);
370 static VNET_DEFINE(uint32_t, mrt_api_config);
371 #define V_mrt_api_config VNET(mrt_api_config)
372 static VNET_DEFINE(int, pim_assert_enabled);
373 #define V_pim_assert_enabled VNET(pim_assert_enabled)
374 static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */
375
376 /*
377 * Find a route for a given origin IP address and multicast group address.
378 * Statistics must be updated by the caller.
379 */
380 static __inline struct mfc *
381 mfc_find(struct in_addr *o, struct in_addr *g)
382 {
383 struct mfc *rt;
384
385 MFC_LOCK_ASSERT();
386
387 LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
388 if (in_hosteq(rt->mfc_origin, *o) &&
389 in_hosteq(rt->mfc_mcastgrp, *g) &&
390 TAILQ_EMPTY(&rt->mfc_stall))
391 break;
392 }
393
394 return (rt);
395 }
396
397 /*
398 * Handle MRT setsockopt commands to modify the multicast forwarding tables.
399 */
400 static int
401 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
402 {
403 int error, optval;
404 vifi_t vifi;
405 struct vifctl vifc;
406 struct mfcctl2 mfc;
407 struct bw_upcall bw_upcall;
408 uint32_t i;
409
410 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
411 return EPERM;
412
413 error = 0;
414 switch (sopt->sopt_name) {
415 case MRT_INIT:
416 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
417 if (error)
418 break;
419 error = ip_mrouter_init(so, optval);
420 break;
421
422 case MRT_DONE:
423 error = ip_mrouter_done();
424 break;
425
426 case MRT_ADD_VIF:
427 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
428 if (error)
429 break;
430 error = add_vif(&vifc);
431 break;
432
433 case MRT_DEL_VIF:
434 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
435 if (error)
436 break;
437 error = del_vif(vifi);
438 break;
439
440 case MRT_ADD_MFC:
441 case MRT_DEL_MFC:
442 /*
443 * select data size depending on API version.
444 */
445 if (sopt->sopt_name == MRT_ADD_MFC &&
446 V_mrt_api_config & MRT_API_FLAGS_ALL) {
447 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
448 sizeof(struct mfcctl2));
449 } else {
450 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
451 sizeof(struct mfcctl));
452 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
453 sizeof(mfc) - sizeof(struct mfcctl));
454 }
455 if (error)
456 break;
457 if (sopt->sopt_name == MRT_ADD_MFC)
458 error = add_mfc(&mfc);
459 else
460 error = del_mfc(&mfc);
461 break;
462
463 case MRT_ASSERT:
464 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
465 if (error)
466 break;
467 set_assert(optval);
468 break;
469
470 case MRT_API_CONFIG:
471 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
472 if (!error)
473 error = set_api_config(&i);
474 if (!error)
475 error = sooptcopyout(sopt, &i, sizeof i);
476 break;
477
478 case MRT_ADD_BW_UPCALL:
479 case MRT_DEL_BW_UPCALL:
480 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
481 sizeof bw_upcall);
482 if (error)
483 break;
484 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
485 error = add_bw_upcall(&bw_upcall);
486 else
487 error = del_bw_upcall(&bw_upcall);
488 break;
489
490 default:
491 error = EOPNOTSUPP;
492 break;
493 }
494 return error;
495 }
496
497 /*
498 * Handle MRT getsockopt commands
499 */
500 static int
501 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
502 {
503 int error;
504
505 switch (sopt->sopt_name) {
506 case MRT_VERSION:
507 error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
508 break;
509
510 case MRT_ASSERT:
511 error = sooptcopyout(sopt, &V_pim_assert_enabled,
512 sizeof V_pim_assert_enabled);
513 break;
514
515 case MRT_API_SUPPORT:
516 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
517 break;
518
519 case MRT_API_CONFIG:
520 error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
521 break;
522
523 default:
524 error = EOPNOTSUPP;
525 break;
526 }
527 return error;
528 }
529
530 /*
531 * Handle ioctl commands to obtain information from the cache
532 */
533 static int
534 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
535 {
536 int error = 0;
537
538 /*
539 * Currently the only function calling this ioctl routine is rtioctl().
540 * Typically, only root can create the raw socket in order to execute
541 * this ioctl method, however the request might be coming from a prison
542 */
543 error = priv_check(curthread, PRIV_NETINET_MROUTE);
544 if (error)
545 return (error);
546 switch (cmd) {
547 case (SIOCGETVIFCNT):
548 error = get_vif_cnt((struct sioc_vif_req *)data);
549 break;
550
551 case (SIOCGETSGCNT):
552 error = get_sg_cnt((struct sioc_sg_req *)data);
553 break;
554
555 default:
556 error = EINVAL;
557 break;
558 }
559 return error;
560 }
561
562 /*
563 * returns the packet, byte, rpf-failure count for the source group provided
564 */
565 static int
566 get_sg_cnt(struct sioc_sg_req *req)
567 {
568 struct mfc *rt;
569
570 MFC_LOCK();
571 rt = mfc_find(&req->src, &req->grp);
572 if (rt == NULL) {
573 MFC_UNLOCK();
574 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
575 return EADDRNOTAVAIL;
576 }
577 req->pktcnt = rt->mfc_pkt_cnt;
578 req->bytecnt = rt->mfc_byte_cnt;
579 req->wrong_if = rt->mfc_wrong_if;
580 MFC_UNLOCK();
581 return 0;
582 }
583
584 /*
585 * returns the input and output packet and byte counts on the vif provided
586 */
587 static int
588 get_vif_cnt(struct sioc_vif_req *req)
589 {
590 vifi_t vifi = req->vifi;
591
592 VIF_LOCK();
593 if (vifi >= V_numvifs) {
594 VIF_UNLOCK();
595 return EINVAL;
596 }
597
598 req->icount = V_viftable[vifi].v_pkt_in;
599 req->ocount = V_viftable[vifi].v_pkt_out;
600 req->ibytes = V_viftable[vifi].v_bytes_in;
601 req->obytes = V_viftable[vifi].v_bytes_out;
602 VIF_UNLOCK();
603
604 return 0;
605 }
606
607 static void
608 if_detached_event(void *arg __unused, struct ifnet *ifp)
609 {
610 vifi_t vifi;
611 u_long i;
612
613 MROUTER_LOCK();
614
615 if (V_ip_mrouter == NULL) {
616 MROUTER_UNLOCK();
617 return;
618 }
619
620 VIF_LOCK();
621 MFC_LOCK();
622
623 /*
624 * Tear down multicast forwarder state associated with this ifnet.
625 * 1. Walk the vif list, matching vifs against this ifnet.
626 * 2. Walk the multicast forwarding cache (mfc) looking for
627 * inner matches with this vif's index.
628 * 3. Expire any matching multicast forwarding cache entries.
629 * 4. Free vif state. This should disable ALLMULTI on the interface.
630 */
631 for (vifi = 0; vifi < V_numvifs; vifi++) {
632 if (V_viftable[vifi].v_ifp != ifp)
633 continue;
634 for (i = 0; i < mfchashsize; i++) {
635 struct mfc *rt, *nrt;
636 for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
637 nrt = LIST_NEXT(rt, mfc_hash);
638 if (rt->mfc_parent == vifi) {
639 expire_mfc(rt);
640 }
641 }
642 }
643 del_vif_locked(vifi);
644 }
645
646 MFC_UNLOCK();
647 VIF_UNLOCK();
648
649 MROUTER_UNLOCK();
650 }
651
652 /*
653 * Enable multicast forwarding.
654 */
655 static int
656 ip_mrouter_init(struct socket *so, int version)
657 {
658
659 CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
660 so->so_type, so->so_proto->pr_protocol);
661
662 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
663 return EOPNOTSUPP;
664
665 if (version != 1)
666 return ENOPROTOOPT;
667
668 MROUTER_LOCK();
669
670 if (ip_mrouter_unloading) {
671 MROUTER_UNLOCK();
672 return ENOPROTOOPT;
673 }
674
675 if (V_ip_mrouter != NULL) {
676 MROUTER_UNLOCK();
677 return EADDRINUSE;
678 }
679
680 V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
681 HASH_NOWAIT);
682
683 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
684 curvnet);
685 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
686 curvnet);
687 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
688 curvnet);
689
690 V_ip_mrouter = so;
691 ip_mrouter_cnt++;
692
693 MROUTER_UNLOCK();
694
695 CTR1(KTR_IPMF, "%s: done", __func__);
696
697 return 0;
698 }
699
700 /*
701 * Disable multicast forwarding.
702 */
703 static int
704 X_ip_mrouter_done(void)
705 {
706 struct ifnet *ifp;
707 u_long i;
708 vifi_t vifi;
709
710 MROUTER_LOCK();
711
712 if (V_ip_mrouter == NULL) {
713 MROUTER_UNLOCK();
714 return EINVAL;
715 }
716
717 /*
718 * Detach/disable hooks to the reset of the system.
719 */
720 V_ip_mrouter = NULL;
721 ip_mrouter_cnt--;
722 V_mrt_api_config = 0;
723
724 VIF_LOCK();
725
726 /*
727 * For each phyint in use, disable promiscuous reception of all IP
728 * multicasts.
729 */
730 for (vifi = 0; vifi < V_numvifs; vifi++) {
731 if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
732 !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
733 ifp = V_viftable[vifi].v_ifp;
734 if_allmulti(ifp, 0);
735 }
736 }
737 bzero((caddr_t)V_viftable, sizeof(V_viftable));
738 V_numvifs = 0;
739 V_pim_assert_enabled = 0;
740
741 VIF_UNLOCK();
742
743 callout_stop(&V_expire_upcalls_ch);
744 callout_stop(&V_bw_upcalls_ch);
745 callout_stop(&V_bw_meter_ch);
746
747 MFC_LOCK();
748
749 /*
750 * Free all multicast forwarding cache entries.
751 * Do not use hashdestroy(), as we must perform other cleanup.
752 */
753 for (i = 0; i < mfchashsize; i++) {
754 struct mfc *rt, *nrt;
755 for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
756 nrt = LIST_NEXT(rt, mfc_hash);
757 expire_mfc(rt);
758 }
759 }
760 free(V_mfchashtbl, M_MRTABLE);
761 V_mfchashtbl = NULL;
762
763 bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
764
765 V_bw_upcalls_n = 0;
766 bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
767
768 MFC_UNLOCK();
769
770 V_reg_vif_num = VIFI_INVALID;
771
772 MROUTER_UNLOCK();
773
774 CTR1(KTR_IPMF, "%s: done", __func__);
775
776 return 0;
777 }
778
779 /*
780 * Set PIM assert processing global
781 */
782 static int
783 set_assert(int i)
784 {
785 if ((i != 1) && (i != 0))
786 return EINVAL;
787
788 V_pim_assert_enabled = i;
789
790 return 0;
791 }
792
793 /*
794 * Configure API capabilities
795 */
796 int
797 set_api_config(uint32_t *apival)
798 {
799 u_long i;
800
801 /*
802 * We can set the API capabilities only if it is the first operation
803 * after MRT_INIT. I.e.:
804 * - there are no vifs installed
805 * - pim_assert is not enabled
806 * - the MFC table is empty
807 */
808 if (V_numvifs > 0) {
809 *apival = 0;
810 return EPERM;
811 }
812 if (V_pim_assert_enabled) {
813 *apival = 0;
814 return EPERM;
815 }
816
817 MFC_LOCK();
818
819 for (i = 0; i < mfchashsize; i++) {
820 if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
821 MFC_UNLOCK();
822 *apival = 0;
823 return EPERM;
824 }
825 }
826
827 MFC_UNLOCK();
828
829 V_mrt_api_config = *apival & mrt_api_support;
830 *apival = V_mrt_api_config;
831
832 return 0;
833 }
834
835 /*
836 * Add a vif to the vif table
837 */
838 static int
839 add_vif(struct vifctl *vifcp)
840 {
841 struct vif *vifp = V_viftable + vifcp->vifc_vifi;
842 struct sockaddr_in sin = {sizeof sin, AF_INET};
843 struct ifaddr *ifa;
844 struct ifnet *ifp;
845 int error;
846
847 VIF_LOCK();
848 if (vifcp->vifc_vifi >= MAXVIFS) {
849 VIF_UNLOCK();
850 return EINVAL;
851 }
852 /* rate limiting is no longer supported by this code */
853 if (vifcp->vifc_rate_limit != 0) {
854 log(LOG_ERR, "rate limiting is no longer supported\n");
855 VIF_UNLOCK();
856 return EINVAL;
857 }
858 if (!in_nullhost(vifp->v_lcl_addr)) {
859 VIF_UNLOCK();
860 return EADDRINUSE;
861 }
862 if (in_nullhost(vifcp->vifc_lcl_addr)) {
863 VIF_UNLOCK();
864 return EADDRNOTAVAIL;
865 }
866
867 /* Find the interface with an address in AF_INET family */
868 if (vifcp->vifc_flags & VIFF_REGISTER) {
869 /*
870 * XXX: Because VIFF_REGISTER does not really need a valid
871 * local interface (e.g. it could be 127.0.0.2), we don't
872 * check its address.
873 */
874 ifp = NULL;
875 } else {
876 sin.sin_addr = vifcp->vifc_lcl_addr;
877 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
878 if (ifa == NULL) {
879 VIF_UNLOCK();
880 return EADDRNOTAVAIL;
881 }
882 ifp = ifa->ifa_ifp;
883 ifa_free(ifa);
884 }
885
886 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
887 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
888 VIF_UNLOCK();
889 return EOPNOTSUPP;
890 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
891 ifp = &V_multicast_register_if;
892 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
893 if (V_reg_vif_num == VIFI_INVALID) {
894 if_initname(&V_multicast_register_if, "register_vif", 0);
895 V_multicast_register_if.if_flags = IFF_LOOPBACK;
896 V_reg_vif_num = vifcp->vifc_vifi;
897 }
898 } else { /* Make sure the interface supports multicast */
899 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
900 VIF_UNLOCK();
901 return EOPNOTSUPP;
902 }
903
904 /* Enable promiscuous reception of all IP multicasts from the if */
905 error = if_allmulti(ifp, 1);
906 if (error) {
907 VIF_UNLOCK();
908 return error;
909 }
910 }
911
912 vifp->v_flags = vifcp->vifc_flags;
913 vifp->v_threshold = vifcp->vifc_threshold;
914 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
915 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
916 vifp->v_ifp = ifp;
917 /* initialize per vif pkt counters */
918 vifp->v_pkt_in = 0;
919 vifp->v_pkt_out = 0;
920 vifp->v_bytes_in = 0;
921 vifp->v_bytes_out = 0;
922
923 /* Adjust numvifs up if the vifi is higher than numvifs */
924 if (V_numvifs <= vifcp->vifc_vifi)
925 V_numvifs = vifcp->vifc_vifi + 1;
926
927 VIF_UNLOCK();
928
929 CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__,
930 (int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_addr),
931 (int)vifcp->vifc_threshold);
932
933 return 0;
934 }
935
936 /*
937 * Delete a vif from the vif table
938 */
939 static int
940 del_vif_locked(vifi_t vifi)
941 {
942 struct vif *vifp;
943
944 VIF_LOCK_ASSERT();
945
946 if (vifi >= V_numvifs) {
947 return EINVAL;
948 }
949 vifp = &V_viftable[vifi];
950 if (in_nullhost(vifp->v_lcl_addr)) {
951 return EADDRNOTAVAIL;
952 }
953
954 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
955 if_allmulti(vifp->v_ifp, 0);
956
957 if (vifp->v_flags & VIFF_REGISTER)
958 V_reg_vif_num = VIFI_INVALID;
959
960 bzero((caddr_t)vifp, sizeof (*vifp));
961
962 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
963
964 /* Adjust numvifs down */
965 for (vifi = V_numvifs; vifi > 0; vifi--)
966 if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
967 break;
968 V_numvifs = vifi;
969
970 return 0;
971 }
972
973 static int
974 del_vif(vifi_t vifi)
975 {
976 int cc;
977
978 VIF_LOCK();
979 cc = del_vif_locked(vifi);
980 VIF_UNLOCK();
981
982 return cc;
983 }
984
985 /*
986 * update an mfc entry without resetting counters and S,G addresses.
987 */
988 static void
989 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
990 {
991 int i;
992
993 rt->mfc_parent = mfccp->mfcc_parent;
994 for (i = 0; i < V_numvifs; i++) {
995 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
996 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
997 MRT_MFC_FLAGS_ALL;
998 }
999 /* set the RP address */
1000 if (V_mrt_api_config & MRT_MFC_RP)
1001 rt->mfc_rp = mfccp->mfcc_rp;
1002 else
1003 rt->mfc_rp.s_addr = INADDR_ANY;
1004 }
1005
1006 /*
1007 * fully initialize an mfc entry from the parameter.
1008 */
1009 static void
1010 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1011 {
1012 rt->mfc_origin = mfccp->mfcc_origin;
1013 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1014
1015 update_mfc_params(rt, mfccp);
1016
1017 /* initialize pkt counters per src-grp */
1018 rt->mfc_pkt_cnt = 0;
1019 rt->mfc_byte_cnt = 0;
1020 rt->mfc_wrong_if = 0;
1021 timevalclear(&rt->mfc_last_assert);
1022 }
1023
1024 static void
1025 expire_mfc(struct mfc *rt)
1026 {
1027 struct rtdetq *rte, *nrte;
1028
1029 MFC_LOCK_ASSERT();
1030
1031 free_bw_list(rt->mfc_bw_meter);
1032
1033 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1034 m_freem(rte->m);
1035 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1036 free(rte, M_MRTABLE);
1037 }
1038
1039 LIST_REMOVE(rt, mfc_hash);
1040 free(rt, M_MRTABLE);
1041 }
1042
1043 /*
1044 * Add an mfc entry
1045 */
1046 static int
1047 add_mfc(struct mfcctl2 *mfccp)
1048 {
1049 struct mfc *rt;
1050 struct rtdetq *rte, *nrte;
1051 u_long hash = 0;
1052 u_short nstl;
1053
1054 VIF_LOCK();
1055 MFC_LOCK();
1056
1057 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1058
1059 /* If an entry already exists, just update the fields */
1060 if (rt) {
1061 CTR4(KTR_IPMF, "%s: update mfc orig %s group %lx parent %x",
1062 __func__, inet_ntoa(mfccp->mfcc_origin),
1063 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1064 mfccp->mfcc_parent);
1065 update_mfc_params(rt, mfccp);
1066 MFC_UNLOCK();
1067 VIF_UNLOCK();
1068 return (0);
1069 }
1070
1071 /*
1072 * Find the entry for which the upcall was made and update
1073 */
1074 nstl = 0;
1075 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1076 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1077 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1078 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1079 !TAILQ_EMPTY(&rt->mfc_stall)) {
1080 CTR5(KTR_IPMF,
1081 "%s: add mfc orig %s group %lx parent %x qh %p",
1082 __func__, inet_ntoa(mfccp->mfcc_origin),
1083 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1084 mfccp->mfcc_parent,
1085 TAILQ_FIRST(&rt->mfc_stall));
1086 if (nstl++)
1087 CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1088
1089 init_mfc_params(rt, mfccp);
1090 rt->mfc_expire = 0; /* Don't clean this guy up */
1091 V_nexpire[hash]--;
1092
1093 /* Free queued packets, but attempt to forward them first. */
1094 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1095 if (rte->ifp != NULL)
1096 ip_mdq(rte->m, rte->ifp, rt, -1);
1097 m_freem(rte->m);
1098 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1099 rt->mfc_nstall--;
1100 free(rte, M_MRTABLE);
1101 }
1102 }
1103 }
1104
1105 /*
1106 * It is possible that an entry is being inserted without an upcall
1107 */
1108 if (nstl == 0) {
1109 CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1110 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1111 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1112 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1113 init_mfc_params(rt, mfccp);
1114 if (rt->mfc_expire)
1115 V_nexpire[hash]--;
1116 rt->mfc_expire = 0;
1117 break; /* XXX */
1118 }
1119 }
1120
1121 if (rt == NULL) { /* no upcall, so make a new entry */
1122 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1123 if (rt == NULL) {
1124 MFC_UNLOCK();
1125 VIF_UNLOCK();
1126 return (ENOBUFS);
1127 }
1128
1129 init_mfc_params(rt, mfccp);
1130 TAILQ_INIT(&rt->mfc_stall);
1131 rt->mfc_nstall = 0;
1132
1133 rt->mfc_expire = 0;
1134 rt->mfc_bw_meter = NULL;
1135
1136 /* insert new entry at head of hash chain */
1137 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1138 }
1139 }
1140
1141 MFC_UNLOCK();
1142 VIF_UNLOCK();
1143
1144 return (0);
1145 }
1146
1147 /*
1148 * Delete an mfc entry
1149 */
1150 static int
1151 del_mfc(struct mfcctl2 *mfccp)
1152 {
1153 struct in_addr origin;
1154 struct in_addr mcastgrp;
1155 struct mfc *rt;
1156
1157 origin = mfccp->mfcc_origin;
1158 mcastgrp = mfccp->mfcc_mcastgrp;
1159
1160 CTR3(KTR_IPMF, "%s: delete mfc orig %s group %lx", __func__,
1161 inet_ntoa(origin), (u_long)ntohl(mcastgrp.s_addr));
1162
1163 MFC_LOCK();
1164
1165 rt = mfc_find(&origin, &mcastgrp);
1166 if (rt == NULL) {
1167 MFC_UNLOCK();
1168 return EADDRNOTAVAIL;
1169 }
1170
1171 /*
1172 * free the bw_meter entries
1173 */
1174 free_bw_list(rt->mfc_bw_meter);
1175 rt->mfc_bw_meter = NULL;
1176
1177 LIST_REMOVE(rt, mfc_hash);
1178 free(rt, M_MRTABLE);
1179
1180 MFC_UNLOCK();
1181
1182 return (0);
1183 }
1184
1185 /*
1186 * Send a message to the routing daemon on the multicast routing socket.
1187 */
1188 static int
1189 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1190 {
1191 if (s) {
1192 SOCKBUF_LOCK(&s->so_rcv);
1193 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1194 NULL) != 0) {
1195 sorwakeup_locked(s);
1196 return 0;
1197 }
1198 SOCKBUF_UNLOCK(&s->so_rcv);
1199 }
1200 m_freem(mm);
1201 return -1;
1202 }
1203
1204 /*
1205 * IP multicast forwarding function. This function assumes that the packet
1206 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1207 * pointed to by "ifp", and the packet is to be relayed to other networks
1208 * that have members of the packet's destination IP multicast group.
1209 *
1210 * The packet is returned unscathed to the caller, unless it is
1211 * erroneous, in which case a non-zero return value tells the caller to
1212 * discard it.
1213 */
1214
1215 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1216
1217 static int
1218 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1219 struct ip_moptions *imo)
1220 {
1221 struct mfc *rt;
1222 int error;
1223 vifi_t vifi;
1224
1225 CTR3(KTR_IPMF, "ip_mforward: delete mfc orig %s group %lx ifp %p",
1226 inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1227
1228 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1229 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1230 /*
1231 * Packet arrived via a physical interface or
1232 * an encapsulated tunnel or a register_vif.
1233 */
1234 } else {
1235 /*
1236 * Packet arrived through a source-route tunnel.
1237 * Source-route tunnels are no longer supported.
1238 */
1239 return (1);
1240 }
1241
1242 VIF_LOCK();
1243 MFC_LOCK();
1244 if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1245 if (ip->ip_ttl < MAXTTL)
1246 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1247 error = ip_mdq(m, ifp, NULL, vifi);
1248 MFC_UNLOCK();
1249 VIF_UNLOCK();
1250 return error;
1251 }
1252
1253 /*
1254 * Don't forward a packet with time-to-live of zero or one,
1255 * or a packet destined to a local-only group.
1256 */
1257 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1258 MFC_UNLOCK();
1259 VIF_UNLOCK();
1260 return 0;
1261 }
1262
1263 /*
1264 * Determine forwarding vifs from the forwarding cache table
1265 */
1266 MRTSTAT_INC(mrts_mfc_lookups);
1267 rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1268
1269 /* Entry exists, so forward if necessary */
1270 if (rt != NULL) {
1271 error = ip_mdq(m, ifp, rt, -1);
1272 MFC_UNLOCK();
1273 VIF_UNLOCK();
1274 return error;
1275 } else {
1276 /*
1277 * If we don't have a route for packet's origin,
1278 * Make a copy of the packet & send message to routing daemon
1279 */
1280
1281 struct mbuf *mb0;
1282 struct rtdetq *rte;
1283 u_long hash;
1284 int hlen = ip->ip_hl << 2;
1285
1286 MRTSTAT_INC(mrts_mfc_misses);
1287 MRTSTAT_INC(mrts_no_route);
1288 CTR2(KTR_IPMF, "ip_mforward: no mfc for (%s,%lx)",
1289 inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr));
1290
1291 /*
1292 * Allocate mbufs early so that we don't do extra work if we are
1293 * just going to fail anyway. Make sure to pullup the header so
1294 * that other people can't step on it.
1295 */
1296 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1297 M_NOWAIT|M_ZERO);
1298 if (rte == NULL) {
1299 MFC_UNLOCK();
1300 VIF_UNLOCK();
1301 return ENOBUFS;
1302 }
1303
1304 mb0 = m_copypacket(m, M_DONTWAIT);
1305 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1306 mb0 = m_pullup(mb0, hlen);
1307 if (mb0 == NULL) {
1308 free(rte, M_MRTABLE);
1309 MFC_UNLOCK();
1310 VIF_UNLOCK();
1311 return ENOBUFS;
1312 }
1313
1314 /* is there an upcall waiting for this flow ? */
1315 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1316 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1317 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1318 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1319 !TAILQ_EMPTY(&rt->mfc_stall))
1320 break;
1321 }
1322
1323 if (rt == NULL) {
1324 int i;
1325 struct igmpmsg *im;
1326 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1327 struct mbuf *mm;
1328
1329 /*
1330 * Locate the vifi for the incoming interface for this packet.
1331 * If none found, drop packet.
1332 */
1333 for (vifi = 0; vifi < V_numvifs &&
1334 V_viftable[vifi].v_ifp != ifp; vifi++)
1335 ;
1336 if (vifi >= V_numvifs) /* vif not found, drop packet */
1337 goto non_fatal;
1338
1339 /* no upcall, so make a new entry */
1340 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1341 if (rt == NULL)
1342 goto fail;
1343
1344 /* Make a copy of the header to send to the user level process */
1345 mm = m_copy(mb0, 0, hlen);
1346 if (mm == NULL)
1347 goto fail1;
1348
1349 /*
1350 * Send message to routing daemon to install
1351 * a route into the kernel table
1352 */
1353
1354 im = mtod(mm, struct igmpmsg *);
1355 im->im_msgtype = IGMPMSG_NOCACHE;
1356 im->im_mbz = 0;
1357 im->im_vif = vifi;
1358
1359 MRTSTAT_INC(mrts_upcalls);
1360
1361 k_igmpsrc.sin_addr = ip->ip_src;
1362 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1363 CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1364 MRTSTAT_INC(mrts_upq_sockfull);
1365 fail1:
1366 free(rt, M_MRTABLE);
1367 fail:
1368 free(rte, M_MRTABLE);
1369 m_freem(mb0);
1370 MFC_UNLOCK();
1371 VIF_UNLOCK();
1372 return ENOBUFS;
1373 }
1374
1375 /* insert new entry at head of hash chain */
1376 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1377 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1378 rt->mfc_expire = UPCALL_EXPIRE;
1379 V_nexpire[hash]++;
1380 for (i = 0; i < V_numvifs; i++) {
1381 rt->mfc_ttls[i] = 0;
1382 rt->mfc_flags[i] = 0;
1383 }
1384 rt->mfc_parent = -1;
1385
1386 /* clear the RP address */
1387 rt->mfc_rp.s_addr = INADDR_ANY;
1388 rt->mfc_bw_meter = NULL;
1389
1390 /* initialize pkt counters per src-grp */
1391 rt->mfc_pkt_cnt = 0;
1392 rt->mfc_byte_cnt = 0;
1393 rt->mfc_wrong_if = 0;
1394 timevalclear(&rt->mfc_last_assert);
1395
1396 TAILQ_INIT(&rt->mfc_stall);
1397 rt->mfc_nstall = 0;
1398
1399 /* link into table */
1400 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1401 TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1402 rt->mfc_nstall++;
1403
1404 } else {
1405 /* determine if queue has overflowed */
1406 if (rt->mfc_nstall > MAX_UPQ) {
1407 MRTSTAT_INC(mrts_upq_ovflw);
1408 non_fatal:
1409 free(rte, M_MRTABLE);
1410 m_freem(mb0);
1411 MFC_UNLOCK();
1412 VIF_UNLOCK();
1413 return (0);
1414 }
1415 TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1416 rt->mfc_nstall++;
1417 }
1418
1419 rte->m = mb0;
1420 rte->ifp = ifp;
1421
1422 MFC_UNLOCK();
1423 VIF_UNLOCK();
1424
1425 return 0;
1426 }
1427 }
1428
1429 /*
1430 * Clean up the cache entry if upcall is not serviced
1431 */
1432 static void
1433 expire_upcalls(void *arg)
1434 {
1435 u_long i;
1436
1437 CURVNET_SET((struct vnet *) arg);
1438
1439 MFC_LOCK();
1440
1441 for (i = 0; i < mfchashsize; i++) {
1442 struct mfc *rt, *nrt;
1443
1444 if (V_nexpire[i] == 0)
1445 continue;
1446
1447 for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
1448 nrt = LIST_NEXT(rt, mfc_hash);
1449
1450 if (TAILQ_EMPTY(&rt->mfc_stall))
1451 continue;
1452
1453 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1454 continue;
1455
1456 /*
1457 * free the bw_meter entries
1458 */
1459 while (rt->mfc_bw_meter != NULL) {
1460 struct bw_meter *x = rt->mfc_bw_meter;
1461
1462 rt->mfc_bw_meter = x->bm_mfc_next;
1463 free(x, M_BWMETER);
1464 }
1465
1466 MRTSTAT_INC(mrts_cache_cleanups);
1467 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1468 (u_long)ntohl(rt->mfc_origin.s_addr),
1469 (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1470
1471 expire_mfc(rt);
1472 }
1473 }
1474
1475 MFC_UNLOCK();
1476
1477 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1478 curvnet);
1479
1480 CURVNET_RESTORE();
1481 }
1482
1483 /*
1484 * Packet forwarding routine once entry in the cache is made
1485 */
1486 static int
1487 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1488 {
1489 struct ip *ip = mtod(m, struct ip *);
1490 vifi_t vifi;
1491 int plen = ip->ip_len;
1492
1493 VIF_LOCK_ASSERT();
1494
1495 /*
1496 * If xmt_vif is not -1, send on only the requested vif.
1497 *
1498 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1499 */
1500 if (xmt_vif < V_numvifs) {
1501 if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1502 pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1503 else
1504 phyint_send(ip, V_viftable + xmt_vif, m);
1505 return 1;
1506 }
1507
1508 /*
1509 * Don't forward if it didn't arrive from the parent vif for its origin.
1510 */
1511 vifi = rt->mfc_parent;
1512 if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1513 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1514 __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1515 MRTSTAT_INC(mrts_wrong_if);
1516 ++rt->mfc_wrong_if;
1517 /*
1518 * If we are doing PIM assert processing, send a message
1519 * to the routing daemon.
1520 *
1521 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1522 * can complete the SPT switch, regardless of the type
1523 * of the iif (broadcast media, GRE tunnel, etc).
1524 */
1525 if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1526 V_viftable[vifi].v_ifp) {
1527
1528 if (ifp == &V_multicast_register_if)
1529 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1530
1531 /* Get vifi for the incoming packet */
1532 for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1533 vifi++)
1534 ;
1535 if (vifi >= V_numvifs)
1536 return 0; /* The iif is not found: ignore the packet. */
1537
1538 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1539 return 0; /* WRONGVIF disabled: ignore the packet */
1540
1541 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1542 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1543 struct igmpmsg *im;
1544 int hlen = ip->ip_hl << 2;
1545 struct mbuf *mm = m_copy(m, 0, hlen);
1546
1547 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1548 mm = m_pullup(mm, hlen);
1549 if (mm == NULL)
1550 return ENOBUFS;
1551
1552 im = mtod(mm, struct igmpmsg *);
1553 im->im_msgtype = IGMPMSG_WRONGVIF;
1554 im->im_mbz = 0;
1555 im->im_vif = vifi;
1556
1557 MRTSTAT_INC(mrts_upcalls);
1558
1559 k_igmpsrc.sin_addr = im->im_src;
1560 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1561 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1562 MRTSTAT_INC(mrts_upq_sockfull);
1563 return ENOBUFS;
1564 }
1565 }
1566 }
1567 return 0;
1568 }
1569
1570
1571 /* If I sourced this packet, it counts as output, else it was input. */
1572 if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1573 V_viftable[vifi].v_pkt_out++;
1574 V_viftable[vifi].v_bytes_out += plen;
1575 } else {
1576 V_viftable[vifi].v_pkt_in++;
1577 V_viftable[vifi].v_bytes_in += plen;
1578 }
1579 rt->mfc_pkt_cnt++;
1580 rt->mfc_byte_cnt += plen;
1581
1582 /*
1583 * For each vif, decide if a copy of the packet should be forwarded.
1584 * Forward if:
1585 * - the ttl exceeds the vif's threshold
1586 * - there are group members downstream on interface
1587 */
1588 for (vifi = 0; vifi < V_numvifs; vifi++)
1589 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1590 V_viftable[vifi].v_pkt_out++;
1591 V_viftable[vifi].v_bytes_out += plen;
1592 if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1593 pim_register_send(ip, V_viftable + vifi, m, rt);
1594 else
1595 phyint_send(ip, V_viftable + vifi, m);
1596 }
1597
1598 /*
1599 * Perform upcall-related bw measuring.
1600 */
1601 if (rt->mfc_bw_meter != NULL) {
1602 struct bw_meter *x;
1603 struct timeval now;
1604
1605 microtime(&now);
1606 MFC_LOCK_ASSERT();
1607 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1608 bw_meter_receive_packet(x, plen, &now);
1609 }
1610
1611 return 0;
1612 }
1613
1614 /*
1615 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1616 */
1617 static int
1618 X_legal_vif_num(int vif)
1619 {
1620 int ret;
1621
1622 ret = 0;
1623 if (vif < 0)
1624 return (ret);
1625
1626 VIF_LOCK();
1627 if (vif < V_numvifs)
1628 ret = 1;
1629 VIF_UNLOCK();
1630
1631 return (ret);
1632 }
1633
1634 /*
1635 * Return the local address used by this vif
1636 */
1637 static u_long
1638 X_ip_mcast_src(int vifi)
1639 {
1640 in_addr_t addr;
1641
1642 addr = INADDR_ANY;
1643 if (vifi < 0)
1644 return (addr);
1645
1646 VIF_LOCK();
1647 if (vifi < V_numvifs)
1648 addr = V_viftable[vifi].v_lcl_addr.s_addr;
1649 VIF_UNLOCK();
1650
1651 return (addr);
1652 }
1653
1654 static void
1655 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1656 {
1657 struct mbuf *mb_copy;
1658 int hlen = ip->ip_hl << 2;
1659
1660 VIF_LOCK_ASSERT();
1661
1662 /*
1663 * Make a new reference to the packet; make sure that
1664 * the IP header is actually copied, not just referenced,
1665 * so that ip_output() only scribbles on the copy.
1666 */
1667 mb_copy = m_copypacket(m, M_DONTWAIT);
1668 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1669 mb_copy = m_pullup(mb_copy, hlen);
1670 if (mb_copy == NULL)
1671 return;
1672
1673 send_packet(vifp, mb_copy);
1674 }
1675
1676 static void
1677 send_packet(struct vif *vifp, struct mbuf *m)
1678 {
1679 struct ip_moptions imo;
1680 struct in_multi *imm[2];
1681 int error;
1682
1683 VIF_LOCK_ASSERT();
1684
1685 imo.imo_multicast_ifp = vifp->v_ifp;
1686 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1687 imo.imo_multicast_loop = 1;
1688 imo.imo_multicast_vif = -1;
1689 imo.imo_num_memberships = 0;
1690 imo.imo_max_memberships = 2;
1691 imo.imo_membership = &imm[0];
1692
1693 /*
1694 * Re-entrancy should not be a problem here, because
1695 * the packets that we send out and are looped back at us
1696 * should get rejected because they appear to come from
1697 * the loopback interface, thus preventing looping.
1698 */
1699 error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1700 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1701 (ptrdiff_t)(vifp - V_viftable), error);
1702 }
1703
1704 /*
1705 * Stubs for old RSVP socket shim implementation.
1706 */
1707
1708 static int
1709 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1710 {
1711
1712 return (EOPNOTSUPP);
1713 }
1714
1715 static void
1716 X_ip_rsvp_force_done(struct socket *so __unused)
1717 {
1718
1719 }
1720
1721 static void
1722 X_rsvp_input(struct mbuf *m, int off __unused)
1723 {
1724
1725 if (!V_rsvp_on)
1726 m_freem(m);
1727 }
1728
1729 /*
1730 * Code for bandwidth monitors
1731 */
1732
1733 /*
1734 * Define common interface for timeval-related methods
1735 */
1736 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1737 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1738 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1739
1740 static uint32_t
1741 compute_bw_meter_flags(struct bw_upcall *req)
1742 {
1743 uint32_t flags = 0;
1744
1745 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1746 flags |= BW_METER_UNIT_PACKETS;
1747 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1748 flags |= BW_METER_UNIT_BYTES;
1749 if (req->bu_flags & BW_UPCALL_GEQ)
1750 flags |= BW_METER_GEQ;
1751 if (req->bu_flags & BW_UPCALL_LEQ)
1752 flags |= BW_METER_LEQ;
1753
1754 return flags;
1755 }
1756
1757 /*
1758 * Add a bw_meter entry
1759 */
1760 static int
1761 add_bw_upcall(struct bw_upcall *req)
1762 {
1763 struct mfc *mfc;
1764 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1765 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1766 struct timeval now;
1767 struct bw_meter *x;
1768 uint32_t flags;
1769
1770 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1771 return EOPNOTSUPP;
1772
1773 /* Test if the flags are valid */
1774 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1775 return EINVAL;
1776 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1777 return EINVAL;
1778 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1779 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1780 return EINVAL;
1781
1782 /* Test if the threshold time interval is valid */
1783 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1784 return EINVAL;
1785
1786 flags = compute_bw_meter_flags(req);
1787
1788 /*
1789 * Find if we have already same bw_meter entry
1790 */
1791 MFC_LOCK();
1792 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1793 if (mfc == NULL) {
1794 MFC_UNLOCK();
1795 return EADDRNOTAVAIL;
1796 }
1797 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1798 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1799 &req->bu_threshold.b_time, ==)) &&
1800 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1801 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1802 (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
1803 MFC_UNLOCK();
1804 return 0; /* XXX Already installed */
1805 }
1806 }
1807
1808 /* Allocate the new bw_meter entry */
1809 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1810 if (x == NULL) {
1811 MFC_UNLOCK();
1812 return ENOBUFS;
1813 }
1814
1815 /* Set the new bw_meter entry */
1816 x->bm_threshold.b_time = req->bu_threshold.b_time;
1817 microtime(&now);
1818 x->bm_start_time = now;
1819 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1820 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1821 x->bm_measured.b_packets = 0;
1822 x->bm_measured.b_bytes = 0;
1823 x->bm_flags = flags;
1824 x->bm_time_next = NULL;
1825 x->bm_time_hash = BW_METER_BUCKETS;
1826
1827 /* Add the new bw_meter entry to the front of entries for this MFC */
1828 x->bm_mfc = mfc;
1829 x->bm_mfc_next = mfc->mfc_bw_meter;
1830 mfc->mfc_bw_meter = x;
1831 schedule_bw_meter(x, &now);
1832 MFC_UNLOCK();
1833
1834 return 0;
1835 }
1836
1837 static void
1838 free_bw_list(struct bw_meter *list)
1839 {
1840 while (list != NULL) {
1841 struct bw_meter *x = list;
1842
1843 list = list->bm_mfc_next;
1844 unschedule_bw_meter(x);
1845 free(x, M_BWMETER);
1846 }
1847 }
1848
1849 /*
1850 * Delete one or multiple bw_meter entries
1851 */
1852 static int
1853 del_bw_upcall(struct bw_upcall *req)
1854 {
1855 struct mfc *mfc;
1856 struct bw_meter *x;
1857
1858 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1859 return EOPNOTSUPP;
1860
1861 MFC_LOCK();
1862
1863 /* Find the corresponding MFC entry */
1864 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1865 if (mfc == NULL) {
1866 MFC_UNLOCK();
1867 return EADDRNOTAVAIL;
1868 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1869 /*
1870 * Delete all bw_meter entries for this mfc
1871 */
1872 struct bw_meter *list;
1873
1874 list = mfc->mfc_bw_meter;
1875 mfc->mfc_bw_meter = NULL;
1876 free_bw_list(list);
1877 MFC_UNLOCK();
1878 return 0;
1879 } else { /* Delete a single bw_meter entry */
1880 struct bw_meter *prev;
1881 uint32_t flags = 0;
1882
1883 flags = compute_bw_meter_flags(req);
1884
1885 /* Find the bw_meter entry to delete */
1886 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1887 prev = x, x = x->bm_mfc_next) {
1888 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1889 &req->bu_threshold.b_time, ==)) &&
1890 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1891 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1892 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
1893 break;
1894 }
1895 if (x != NULL) { /* Delete entry from the list for this MFC */
1896 if (prev != NULL)
1897 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
1898 else
1899 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1900
1901 unschedule_bw_meter(x);
1902 MFC_UNLOCK();
1903 /* Free the bw_meter entry */
1904 free(x, M_BWMETER);
1905 return 0;
1906 } else {
1907 MFC_UNLOCK();
1908 return EINVAL;
1909 }
1910 }
1911 /* NOTREACHED */
1912 }
1913
1914 /*
1915 * Perform bandwidth measurement processing that may result in an upcall
1916 */
1917 static void
1918 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1919 {
1920 struct timeval delta;
1921
1922 MFC_LOCK_ASSERT();
1923
1924 delta = *nowp;
1925 BW_TIMEVALDECR(&delta, &x->bm_start_time);
1926
1927 if (x->bm_flags & BW_METER_GEQ) {
1928 /*
1929 * Processing for ">=" type of bw_meter entry
1930 */
1931 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1932 /* Reset the bw_meter entry */
1933 x->bm_start_time = *nowp;
1934 x->bm_measured.b_packets = 0;
1935 x->bm_measured.b_bytes = 0;
1936 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1937 }
1938
1939 /* Record that a packet is received */
1940 x->bm_measured.b_packets++;
1941 x->bm_measured.b_bytes += plen;
1942
1943 /*
1944 * Test if we should deliver an upcall
1945 */
1946 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1947 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1948 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1949 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1950 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1951 /* Prepare an upcall for delivery */
1952 bw_meter_prepare_upcall(x, nowp);
1953 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1954 }
1955 }
1956 } else if (x->bm_flags & BW_METER_LEQ) {
1957 /*
1958 * Processing for "<=" type of bw_meter entry
1959 */
1960 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1961 /*
1962 * We are behind time with the multicast forwarding table
1963 * scanning for "<=" type of bw_meter entries, so test now
1964 * if we should deliver an upcall.
1965 */
1966 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1967 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1968 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1969 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1970 /* Prepare an upcall for delivery */
1971 bw_meter_prepare_upcall(x, nowp);
1972 }
1973 /* Reschedule the bw_meter entry */
1974 unschedule_bw_meter(x);
1975 schedule_bw_meter(x, nowp);
1976 }
1977
1978 /* Record that a packet is received */
1979 x->bm_measured.b_packets++;
1980 x->bm_measured.b_bytes += plen;
1981
1982 /*
1983 * Test if we should restart the measuring interval
1984 */
1985 if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1986 x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1987 (x->bm_flags & BW_METER_UNIT_BYTES &&
1988 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1989 /* Don't restart the measuring interval */
1990 } else {
1991 /* Do restart the measuring interval */
1992 /*
1993 * XXX: note that we don't unschedule and schedule, because this
1994 * might be too much overhead per packet. Instead, when we process
1995 * all entries for a given timer hash bin, we check whether it is
1996 * really a timeout. If not, we reschedule at that time.
1997 */
1998 x->bm_start_time = *nowp;
1999 x->bm_measured.b_packets = 0;
2000 x->bm_measured.b_bytes = 0;
2001 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2002 }
2003 }
2004 }
2005
2006 /*
2007 * Prepare a bandwidth-related upcall
2008 */
2009 static void
2010 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2011 {
2012 struct timeval delta;
2013 struct bw_upcall *u;
2014
2015 MFC_LOCK_ASSERT();
2016
2017 /*
2018 * Compute the measured time interval
2019 */
2020 delta = *nowp;
2021 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2022
2023 /*
2024 * If there are too many pending upcalls, deliver them now
2025 */
2026 if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2027 bw_upcalls_send();
2028
2029 /*
2030 * Set the bw_upcall entry
2031 */
2032 u = &V_bw_upcalls[V_bw_upcalls_n++];
2033 u->bu_src = x->bm_mfc->mfc_origin;
2034 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2035 u->bu_threshold.b_time = x->bm_threshold.b_time;
2036 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2037 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2038 u->bu_measured.b_time = delta;
2039 u->bu_measured.b_packets = x->bm_measured.b_packets;
2040 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2041 u->bu_flags = 0;
2042 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2043 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2044 if (x->bm_flags & BW_METER_UNIT_BYTES)
2045 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2046 if (x->bm_flags & BW_METER_GEQ)
2047 u->bu_flags |= BW_UPCALL_GEQ;
2048 if (x->bm_flags & BW_METER_LEQ)
2049 u->bu_flags |= BW_UPCALL_LEQ;
2050 }
2051
2052 /*
2053 * Send the pending bandwidth-related upcalls
2054 */
2055 static void
2056 bw_upcalls_send(void)
2057 {
2058 struct mbuf *m;
2059 int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2060 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2061 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2062 0, /* unused2 */
2063 IGMPMSG_BW_UPCALL,/* im_msgtype */
2064 0, /* im_mbz */
2065 0, /* im_vif */
2066 0, /* unused3 */
2067 { 0 }, /* im_src */
2068 { 0 } }; /* im_dst */
2069
2070 MFC_LOCK_ASSERT();
2071
2072 if (V_bw_upcalls_n == 0)
2073 return; /* No pending upcalls */
2074
2075 V_bw_upcalls_n = 0;
2076
2077 /*
2078 * Allocate a new mbuf, initialize it with the header and
2079 * the payload for the pending calls.
2080 */
2081 MGETHDR(m, M_DONTWAIT, MT_DATA);
2082 if (m == NULL) {
2083 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2084 return;
2085 }
2086
2087 m->m_len = m->m_pkthdr.len = 0;
2088 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2089 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2090
2091 /*
2092 * Send the upcalls
2093 * XXX do we need to set the address in k_igmpsrc ?
2094 */
2095 MRTSTAT_INC(mrts_upcalls);
2096 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2097 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2098 MRTSTAT_INC(mrts_upq_sockfull);
2099 }
2100 }
2101
2102 /*
2103 * Compute the timeout hash value for the bw_meter entries
2104 */
2105 #define BW_METER_TIMEHASH(bw_meter, hash) \
2106 do { \
2107 struct timeval next_timeval = (bw_meter)->bm_start_time; \
2108 \
2109 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2110 (hash) = next_timeval.tv_sec; \
2111 if (next_timeval.tv_usec) \
2112 (hash)++; /* XXX: make sure we don't timeout early */ \
2113 (hash) %= BW_METER_BUCKETS; \
2114 } while (0)
2115
2116 /*
2117 * Schedule a timer to process periodically bw_meter entry of type "<="
2118 * by linking the entry in the proper hash bucket.
2119 */
2120 static void
2121 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2122 {
2123 int time_hash;
2124
2125 MFC_LOCK_ASSERT();
2126
2127 if (!(x->bm_flags & BW_METER_LEQ))
2128 return; /* XXX: we schedule timers only for "<=" entries */
2129
2130 /*
2131 * Reset the bw_meter entry
2132 */
2133 x->bm_start_time = *nowp;
2134 x->bm_measured.b_packets = 0;
2135 x->bm_measured.b_bytes = 0;
2136 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2137
2138 /*
2139 * Compute the timeout hash value and insert the entry
2140 */
2141 BW_METER_TIMEHASH(x, time_hash);
2142 x->bm_time_next = V_bw_meter_timers[time_hash];
2143 V_bw_meter_timers[time_hash] = x;
2144 x->bm_time_hash = time_hash;
2145 }
2146
2147 /*
2148 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2149 * by removing the entry from the proper hash bucket.
2150 */
2151 static void
2152 unschedule_bw_meter(struct bw_meter *x)
2153 {
2154 int time_hash;
2155 struct bw_meter *prev, *tmp;
2156
2157 MFC_LOCK_ASSERT();
2158
2159 if (!(x->bm_flags & BW_METER_LEQ))
2160 return; /* XXX: we schedule timers only for "<=" entries */
2161
2162 /*
2163 * Compute the timeout hash value and delete the entry
2164 */
2165 time_hash = x->bm_time_hash;
2166 if (time_hash >= BW_METER_BUCKETS)
2167 return; /* Entry was not scheduled */
2168
2169 for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2170 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2171 if (tmp == x)
2172 break;
2173
2174 if (tmp == NULL)
2175 panic("unschedule_bw_meter: bw_meter entry not found");
2176
2177 if (prev != NULL)
2178 prev->bm_time_next = x->bm_time_next;
2179 else
2180 V_bw_meter_timers[time_hash] = x->bm_time_next;
2181
2182 x->bm_time_next = NULL;
2183 x->bm_time_hash = BW_METER_BUCKETS;
2184 }
2185
2186
2187 /*
2188 * Process all "<=" type of bw_meter that should be processed now,
2189 * and for each entry prepare an upcall if necessary. Each processed
2190 * entry is rescheduled again for the (periodic) processing.
2191 *
2192 * This is run periodically (once per second normally). On each round,
2193 * all the potentially matching entries are in the hash slot that we are
2194 * looking at.
2195 */
2196 static void
2197 bw_meter_process()
2198 {
2199 uint32_t loops;
2200 int i;
2201 struct timeval now, process_endtime;
2202
2203 microtime(&now);
2204 if (V_last_tv_sec == now.tv_sec)
2205 return; /* nothing to do */
2206
2207 loops = now.tv_sec - V_last_tv_sec;
2208 V_last_tv_sec = now.tv_sec;
2209 if (loops > BW_METER_BUCKETS)
2210 loops = BW_METER_BUCKETS;
2211
2212 MFC_LOCK();
2213 /*
2214 * Process all bins of bw_meter entries from the one after the last
2215 * processed to the current one. On entry, i points to the last bucket
2216 * visited, so we need to increment i at the beginning of the loop.
2217 */
2218 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2219 struct bw_meter *x, *tmp_list;
2220
2221 if (++i >= BW_METER_BUCKETS)
2222 i = 0;
2223
2224 /* Disconnect the list of bw_meter entries from the bin */
2225 tmp_list = V_bw_meter_timers[i];
2226 V_bw_meter_timers[i] = NULL;
2227
2228 /* Process the list of bw_meter entries */
2229 while (tmp_list != NULL) {
2230 x = tmp_list;
2231 tmp_list = tmp_list->bm_time_next;
2232
2233 /* Test if the time interval is over */
2234 process_endtime = x->bm_start_time;
2235 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2236 if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2237 /* Not yet: reschedule, but don't reset */
2238 int time_hash;
2239
2240 BW_METER_TIMEHASH(x, time_hash);
2241 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2242 /*
2243 * XXX: somehow the bin processing is a bit ahead of time.
2244 * Put the entry in the next bin.
2245 */
2246 if (++time_hash >= BW_METER_BUCKETS)
2247 time_hash = 0;
2248 }
2249 x->bm_time_next = V_bw_meter_timers[time_hash];
2250 V_bw_meter_timers[time_hash] = x;
2251 x->bm_time_hash = time_hash;
2252
2253 continue;
2254 }
2255
2256 /*
2257 * Test if we should deliver an upcall
2258 */
2259 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2260 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2261 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2262 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2263 /* Prepare an upcall for delivery */
2264 bw_meter_prepare_upcall(x, &now);
2265 }
2266
2267 /*
2268 * Reschedule for next processing
2269 */
2270 schedule_bw_meter(x, &now);
2271 }
2272 }
2273
2274 /* Send all upcalls that are pending delivery */
2275 bw_upcalls_send();
2276
2277 MFC_UNLOCK();
2278 }
2279
2280 /*
2281 * A periodic function for sending all upcalls that are pending delivery
2282 */
2283 static void
2284 expire_bw_upcalls_send(void *arg)
2285 {
2286 CURVNET_SET((struct vnet *) arg);
2287
2288 MFC_LOCK();
2289 bw_upcalls_send();
2290 MFC_UNLOCK();
2291
2292 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2293 curvnet);
2294 CURVNET_RESTORE();
2295 }
2296
2297 /*
2298 * A periodic function for periodic scanning of the multicast forwarding
2299 * table for processing all "<=" bw_meter entries.
2300 */
2301 static void
2302 expire_bw_meter_process(void *arg)
2303 {
2304 CURVNET_SET((struct vnet *) arg);
2305
2306 if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2307 bw_meter_process();
2308
2309 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2310 curvnet);
2311 CURVNET_RESTORE();
2312 }
2313
2314 /*
2315 * End of bandwidth monitoring code
2316 */
2317
2318 /*
2319 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2320 *
2321 */
2322 static int
2323 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2324 struct mfc *rt)
2325 {
2326 struct mbuf *mb_copy, *mm;
2327
2328 /*
2329 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2330 * rendezvous point was unspecified, and we were told not to.
2331 */
2332 if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2333 in_nullhost(rt->mfc_rp))
2334 return 0;
2335
2336 mb_copy = pim_register_prepare(ip, m);
2337 if (mb_copy == NULL)
2338 return ENOBUFS;
2339
2340 /*
2341 * Send all the fragments. Note that the mbuf for each fragment
2342 * is freed by the sending machinery.
2343 */
2344 for (mm = mb_copy; mm; mm = mb_copy) {
2345 mb_copy = mm->m_nextpkt;
2346 mm->m_nextpkt = 0;
2347 mm = m_pullup(mm, sizeof(struct ip));
2348 if (mm != NULL) {
2349 ip = mtod(mm, struct ip *);
2350 if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2351 pim_register_send_rp(ip, vifp, mm, rt);
2352 } else {
2353 pim_register_send_upcall(ip, vifp, mm, rt);
2354 }
2355 }
2356 }
2357
2358 return 0;
2359 }
2360
2361 /*
2362 * Return a copy of the data packet that is ready for PIM Register
2363 * encapsulation.
2364 * XXX: Note that in the returned copy the IP header is a valid one.
2365 */
2366 static struct mbuf *
2367 pim_register_prepare(struct ip *ip, struct mbuf *m)
2368 {
2369 struct mbuf *mb_copy = NULL;
2370 int mtu;
2371
2372 /* Take care of delayed checksums */
2373 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2374 in_delayed_cksum(m);
2375 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2376 }
2377
2378 /*
2379 * Copy the old packet & pullup its IP header into the
2380 * new mbuf so we can modify it.
2381 */
2382 mb_copy = m_copypacket(m, M_DONTWAIT);
2383 if (mb_copy == NULL)
2384 return NULL;
2385 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2386 if (mb_copy == NULL)
2387 return NULL;
2388
2389 /* take care of the TTL */
2390 ip = mtod(mb_copy, struct ip *);
2391 --ip->ip_ttl;
2392
2393 /* Compute the MTU after the PIM Register encapsulation */
2394 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2395
2396 if (ip->ip_len <= mtu) {
2397 /* Turn the IP header into a valid one */
2398 ip->ip_len = htons(ip->ip_len);
2399 ip->ip_off = htons(ip->ip_off);
2400 ip->ip_sum = 0;
2401 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2402 } else {
2403 /* Fragment the packet */
2404 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) {
2405 m_freem(mb_copy);
2406 return NULL;
2407 }
2408 }
2409 return mb_copy;
2410 }
2411
2412 /*
2413 * Send an upcall with the data packet to the user-level process.
2414 */
2415 static int
2416 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2417 struct mbuf *mb_copy, struct mfc *rt)
2418 {
2419 struct mbuf *mb_first;
2420 int len = ntohs(ip->ip_len);
2421 struct igmpmsg *im;
2422 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2423
2424 VIF_LOCK_ASSERT();
2425
2426 /*
2427 * Add a new mbuf with an upcall header
2428 */
2429 MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2430 if (mb_first == NULL) {
2431 m_freem(mb_copy);
2432 return ENOBUFS;
2433 }
2434 mb_first->m_data += max_linkhdr;
2435 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2436 mb_first->m_len = sizeof(struct igmpmsg);
2437 mb_first->m_next = mb_copy;
2438
2439 /* Send message to routing daemon */
2440 im = mtod(mb_first, struct igmpmsg *);
2441 im->im_msgtype = IGMPMSG_WHOLEPKT;
2442 im->im_mbz = 0;
2443 im->im_vif = vifp - V_viftable;
2444 im->im_src = ip->ip_src;
2445 im->im_dst = ip->ip_dst;
2446
2447 k_igmpsrc.sin_addr = ip->ip_src;
2448
2449 MRTSTAT_INC(mrts_upcalls);
2450
2451 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2452 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2453 MRTSTAT_INC(mrts_upq_sockfull);
2454 return ENOBUFS;
2455 }
2456
2457 /* Keep statistics */
2458 PIMSTAT_INC(pims_snd_registers_msgs);
2459 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2460
2461 return 0;
2462 }
2463
2464 /*
2465 * Encapsulate the data packet in PIM Register message and send it to the RP.
2466 */
2467 static int
2468 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2469 struct mfc *rt)
2470 {
2471 struct mbuf *mb_first;
2472 struct ip *ip_outer;
2473 struct pim_encap_pimhdr *pimhdr;
2474 int len = ntohs(ip->ip_len);
2475 vifi_t vifi = rt->mfc_parent;
2476
2477 VIF_LOCK_ASSERT();
2478
2479 if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2480 m_freem(mb_copy);
2481 return EADDRNOTAVAIL; /* The iif vif is invalid */
2482 }
2483
2484 /*
2485 * Add a new mbuf with the encapsulating header
2486 */
2487 MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2488 if (mb_first == NULL) {
2489 m_freem(mb_copy);
2490 return ENOBUFS;
2491 }
2492 mb_first->m_data += max_linkhdr;
2493 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2494 mb_first->m_next = mb_copy;
2495
2496 mb_first->m_pkthdr.len = len + mb_first->m_len;
2497
2498 /*
2499 * Fill in the encapsulating IP and PIM header
2500 */
2501 ip_outer = mtod(mb_first, struct ip *);
2502 *ip_outer = pim_encap_iphdr;
2503 ip_outer->ip_id = ip_newid();
2504 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2505 ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2506 ip_outer->ip_dst = rt->mfc_rp;
2507 /*
2508 * Copy the inner header TOS to the outer header, and take care of the
2509 * IP_DF bit.
2510 */
2511 ip_outer->ip_tos = ip->ip_tos;
2512 if (ntohs(ip->ip_off) & IP_DF)
2513 ip_outer->ip_off |= IP_DF;
2514 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2515 + sizeof(pim_encap_iphdr));
2516 *pimhdr = pim_encap_pimhdr;
2517 /* If the iif crosses a border, set the Border-bit */
2518 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2519 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2520
2521 mb_first->m_data += sizeof(pim_encap_iphdr);
2522 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2523 mb_first->m_data -= sizeof(pim_encap_iphdr);
2524
2525 send_packet(vifp, mb_first);
2526
2527 /* Keep statistics */
2528 PIMSTAT_INC(pims_snd_registers_msgs);
2529 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2530
2531 return 0;
2532 }
2533
2534 /*
2535 * pim_encapcheck() is called by the encap4_input() path at runtime to
2536 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2537 * into the kernel.
2538 */
2539 static int
2540 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2541 {
2542
2543 #ifdef DIAGNOSTIC
2544 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2545 #endif
2546 if (proto != IPPROTO_PIM)
2547 return 0; /* not for us; reject the datagram. */
2548
2549 return 64; /* claim the datagram. */
2550 }
2551
2552 /*
2553 * PIM-SMv2 and PIM-DM messages processing.
2554 * Receives and verifies the PIM control messages, and passes them
2555 * up to the listening socket, using rip_input().
2556 * The only message with special processing is the PIM_REGISTER message
2557 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2558 * is passed to if_simloop().
2559 */
2560 void
2561 pim_input(struct mbuf *m, int off)
2562 {
2563 struct ip *ip = mtod(m, struct ip *);
2564 struct pim *pim;
2565 int minlen;
2566 int datalen = ip->ip_len;
2567 int ip_tos;
2568 int iphlen = off;
2569
2570 /* Keep statistics */
2571 PIMSTAT_INC(pims_rcv_total_msgs);
2572 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2573
2574 /*
2575 * Validate lengths
2576 */
2577 if (datalen < PIM_MINLEN) {
2578 PIMSTAT_INC(pims_rcv_tooshort);
2579 CTR3(KTR_IPMF, "%s: short packet (%d) from %s",
2580 __func__, datalen, inet_ntoa(ip->ip_src));
2581 m_freem(m);
2582 return;
2583 }
2584
2585 /*
2586 * If the packet is at least as big as a REGISTER, go agead
2587 * and grab the PIM REGISTER header size, to avoid another
2588 * possible m_pullup() later.
2589 *
2590 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2591 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2592 */
2593 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2594 /*
2595 * Get the IP and PIM headers in contiguous memory, and
2596 * possibly the PIM REGISTER header.
2597 */
2598 if ((m->m_flags & M_EXT || m->m_len < minlen) &&
2599 (m = m_pullup(m, minlen)) == 0) {
2600 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2601 return;
2602 }
2603
2604 /* m_pullup() may have given us a new mbuf so reset ip. */
2605 ip = mtod(m, struct ip *);
2606 ip_tos = ip->ip_tos;
2607
2608 /* adjust mbuf to point to the PIM header */
2609 m->m_data += iphlen;
2610 m->m_len -= iphlen;
2611 pim = mtod(m, struct pim *);
2612
2613 /*
2614 * Validate checksum. If PIM REGISTER, exclude the data packet.
2615 *
2616 * XXX: some older PIMv2 implementations don't make this distinction,
2617 * so for compatibility reason perform the checksum over part of the
2618 * message, and if error, then over the whole message.
2619 */
2620 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2621 /* do nothing, checksum okay */
2622 } else if (in_cksum(m, datalen)) {
2623 PIMSTAT_INC(pims_rcv_badsum);
2624 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2625 m_freem(m);
2626 return;
2627 }
2628
2629 /* PIM version check */
2630 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2631 PIMSTAT_INC(pims_rcv_badversion);
2632 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2633 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2634 m_freem(m);
2635 return;
2636 }
2637
2638 /* restore mbuf back to the outer IP */
2639 m->m_data -= iphlen;
2640 m->m_len += iphlen;
2641
2642 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2643 /*
2644 * Since this is a REGISTER, we'll make a copy of the register
2645 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2646 * routing daemon.
2647 */
2648 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2649 struct mbuf *mcp;
2650 struct ip *encap_ip;
2651 u_int32_t *reghdr;
2652 struct ifnet *vifp;
2653
2654 VIF_LOCK();
2655 if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2656 VIF_UNLOCK();
2657 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2658 (int)V_reg_vif_num);
2659 m_freem(m);
2660 return;
2661 }
2662 /* XXX need refcnt? */
2663 vifp = V_viftable[V_reg_vif_num].v_ifp;
2664 VIF_UNLOCK();
2665
2666 /*
2667 * Validate length
2668 */
2669 if (datalen < PIM_REG_MINLEN) {
2670 PIMSTAT_INC(pims_rcv_tooshort);
2671 PIMSTAT_INC(pims_rcv_badregisters);
2672 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2673 m_freem(m);
2674 return;
2675 }
2676
2677 reghdr = (u_int32_t *)(pim + 1);
2678 encap_ip = (struct ip *)(reghdr + 1);
2679
2680 CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d",
2681 __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len));
2682
2683 /* verify the version number of the inner packet */
2684 if (encap_ip->ip_v != IPVERSION) {
2685 PIMSTAT_INC(pims_rcv_badregisters);
2686 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2687 m_freem(m);
2688 return;
2689 }
2690
2691 /* verify the inner packet is destined to a mcast group */
2692 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2693 PIMSTAT_INC(pims_rcv_badregisters);
2694 CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__,
2695 inet_ntoa(encap_ip->ip_dst));
2696 m_freem(m);
2697 return;
2698 }
2699
2700 /* If a NULL_REGISTER, pass it to the daemon */
2701 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2702 goto pim_input_to_daemon;
2703
2704 /*
2705 * Copy the TOS from the outer IP header to the inner IP header.
2706 */
2707 if (encap_ip->ip_tos != ip_tos) {
2708 /* Outer TOS -> inner TOS */
2709 encap_ip->ip_tos = ip_tos;
2710 /* Recompute the inner header checksum. Sigh... */
2711
2712 /* adjust mbuf to point to the inner IP header */
2713 m->m_data += (iphlen + PIM_MINLEN);
2714 m->m_len -= (iphlen + PIM_MINLEN);
2715
2716 encap_ip->ip_sum = 0;
2717 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2718
2719 /* restore mbuf to point back to the outer IP header */
2720 m->m_data -= (iphlen + PIM_MINLEN);
2721 m->m_len += (iphlen + PIM_MINLEN);
2722 }
2723
2724 /*
2725 * Decapsulate the inner IP packet and loopback to forward it
2726 * as a normal multicast packet. Also, make a copy of the
2727 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2728 * to pass to the daemon later, so it can take the appropriate
2729 * actions (e.g., send back PIM_REGISTER_STOP).
2730 * XXX: here m->m_data points to the outer IP header.
2731 */
2732 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
2733 if (mcp == NULL) {
2734 CTR1(KTR_IPMF, "%s: m_copy() failed", __func__);
2735 m_freem(m);
2736 return;
2737 }
2738
2739 /* Keep statistics */
2740 /* XXX: registers_bytes include only the encap. mcast pkt */
2741 PIMSTAT_INC(pims_rcv_registers_msgs);
2742 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2743
2744 /*
2745 * forward the inner ip packet; point m_data at the inner ip.
2746 */
2747 m_adj(m, iphlen + PIM_MINLEN);
2748
2749 CTR4(KTR_IPMF,
2750 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2751 __func__,
2752 (u_long)ntohl(encap_ip->ip_src.s_addr),
2753 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2754 (int)V_reg_vif_num);
2755
2756 /* NB: vifp was collected above; can it change on us? */
2757 if_simloop(vifp, m, dst.sin_family, 0);
2758
2759 /* prepare the register head to send to the mrouting daemon */
2760 m = mcp;
2761 }
2762
2763 pim_input_to_daemon:
2764 /*
2765 * Pass the PIM message up to the daemon; if it is a Register message,
2766 * pass the 'head' only up to the daemon. This includes the
2767 * outer IP header, PIM header, PIM-Register header and the
2768 * inner IP header.
2769 * XXX: the outer IP header pkt size of a Register is not adjust to
2770 * reflect the fact that the inner multicast data is truncated.
2771 */
2772 rip_input(m, iphlen);
2773
2774 return;
2775 }
2776
2777 static int
2778 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2779 {
2780 struct mfc *rt;
2781 int error, i;
2782
2783 if (req->newptr)
2784 return (EPERM);
2785 if (V_mfchashtbl == NULL) /* XXX unlocked */
2786 return (0);
2787 error = sysctl_wire_old_buffer(req, 0);
2788 if (error)
2789 return (error);
2790
2791 MFC_LOCK();
2792 for (i = 0; i < mfchashsize; i++) {
2793 LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2794 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2795 if (error)
2796 goto out_locked;
2797 }
2798 }
2799 out_locked:
2800 MFC_UNLOCK();
2801 return (error);
2802 }
2803
2804 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2805 sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2806 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2807
2808 static void
2809 vnet_mroute_init(const void *unused __unused)
2810 {
2811
2812 MALLOC(V_nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2813 bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2814 callout_init(&V_expire_upcalls_ch, CALLOUT_MPSAFE);
2815 callout_init(&V_bw_upcalls_ch, CALLOUT_MPSAFE);
2816 callout_init(&V_bw_meter_ch, CALLOUT_MPSAFE);
2817 }
2818
2819 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mroute_init,
2820 NULL);
2821
2822 static void
2823 vnet_mroute_uninit(const void *unused __unused)
2824 {
2825
2826 FREE(V_nexpire, M_MRTABLE);
2827 V_nexpire = NULL;
2828 }
2829
2830 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE,
2831 vnet_mroute_uninit, NULL);
2832
2833 static int
2834 ip_mroute_modevent(module_t mod, int type, void *unused)
2835 {
2836
2837 switch (type) {
2838 case MOD_LOAD:
2839 MROUTER_LOCK_INIT();
2840
2841 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2842 if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2843 if (if_detach_event_tag == NULL) {
2844 printf("ip_mroute: unable to register "
2845 "ifnet_departure_event handler\n");
2846 MROUTER_LOCK_DESTROY();
2847 return (EINVAL);
2848 }
2849
2850 MFC_LOCK_INIT();
2851 VIF_LOCK_INIT();
2852
2853 mfchashsize = MFCHASHSIZE;
2854 if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2855 !powerof2(mfchashsize)) {
2856 printf("WARNING: %s not a power of 2; using default\n",
2857 "net.inet.ip.mfchashsize");
2858 mfchashsize = MFCHASHSIZE;
2859 }
2860
2861 pim_squelch_wholepkt = 0;
2862 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2863 &pim_squelch_wholepkt);
2864
2865 pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2866 pim_encapcheck, &in_pim_protosw, NULL);
2867 if (pim_encap_cookie == NULL) {
2868 printf("ip_mroute: unable to attach pim encap\n");
2869 VIF_LOCK_DESTROY();
2870 MFC_LOCK_DESTROY();
2871 MROUTER_LOCK_DESTROY();
2872 return (EINVAL);
2873 }
2874
2875 ip_mcast_src = X_ip_mcast_src;
2876 ip_mforward = X_ip_mforward;
2877 ip_mrouter_done = X_ip_mrouter_done;
2878 ip_mrouter_get = X_ip_mrouter_get;
2879 ip_mrouter_set = X_ip_mrouter_set;
2880
2881 ip_rsvp_force_done = X_ip_rsvp_force_done;
2882 ip_rsvp_vif = X_ip_rsvp_vif;
2883
2884 legal_vif_num = X_legal_vif_num;
2885 mrt_ioctl = X_mrt_ioctl;
2886 rsvp_input_p = X_rsvp_input;
2887 break;
2888
2889 case MOD_UNLOAD:
2890 /*
2891 * Typically module unload happens after the user-level
2892 * process has shutdown the kernel services (the check
2893 * below insures someone can't just yank the module out
2894 * from under a running process). But if the module is
2895 * just loaded and then unloaded w/o starting up a user
2896 * process we still need to cleanup.
2897 */
2898 MROUTER_LOCK();
2899 if (ip_mrouter_cnt != 0) {
2900 MROUTER_UNLOCK();
2901 return (EINVAL);
2902 }
2903 ip_mrouter_unloading = 1;
2904 MROUTER_UNLOCK();
2905
2906 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2907
2908 if (pim_encap_cookie) {
2909 encap_detach(pim_encap_cookie);
2910 pim_encap_cookie = NULL;
2911 }
2912
2913 ip_mcast_src = NULL;
2914 ip_mforward = NULL;
2915 ip_mrouter_done = NULL;
2916 ip_mrouter_get = NULL;
2917 ip_mrouter_set = NULL;
2918
2919 ip_rsvp_force_done = NULL;
2920 ip_rsvp_vif = NULL;
2921
2922 legal_vif_num = NULL;
2923 mrt_ioctl = NULL;
2924 rsvp_input_p = NULL;
2925
2926 VIF_LOCK_DESTROY();
2927 MFC_LOCK_DESTROY();
2928 MROUTER_LOCK_DESTROY();
2929 break;
2930
2931 default:
2932 return EOPNOTSUPP;
2933 }
2934 return 0;
2935 }
2936
2937 static moduledata_t ip_mroutemod = {
2938 "ip_mroute",
2939 ip_mroute_modevent,
2940 0
2941 };
2942
2943 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_MIDDLE);
Cache object: ede4b3c6f1c15e0ef321930588dbedbe
|