1 /* $NetBSD: ip_mroute.c,v 1.143 2016/07/04 04:35:09 knakahara Exp $ */
2
3 /*
4 * Copyright (c) 1992, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Stephen Deering of Stanford University.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
35 */
36
37 /*
38 * Copyright (c) 1989 Stephen Deering
39 *
40 * This code is derived from software contributed to Berkeley by
41 * Stephen Deering of Stanford University.
42 *
43 * Redistribution and use in source and binary forms, with or without
44 * modification, are permitted provided that the following conditions
45 * are met:
46 * 1. Redistributions of source code must retain the above copyright
47 * notice, this list of conditions and the following disclaimer.
48 * 2. Redistributions in binary form must reproduce the above copyright
49 * notice, this list of conditions and the following disclaimer in the
50 * documentation and/or other materials provided with the distribution.
51 * 3. All advertising materials mentioning features or use of this software
52 * must display the following acknowledgement:
53 * This product includes software developed by the University of
54 * California, Berkeley and its contributors.
55 * 4. Neither the name of the University nor the names of its contributors
56 * may be used to endorse or promote products derived from this software
57 * without specific prior written permission.
58 *
59 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
60 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
61 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
62 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
63 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
64 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
65 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
66 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
67 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
68 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
69 * SUCH DAMAGE.
70 *
71 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
72 */
73
74 /*
75 * IP multicast forwarding procedures
76 *
77 * Written by David Waitzman, BBN Labs, August 1988.
78 * Modified by Steve Deering, Stanford, February 1989.
79 * Modified by Mark J. Steiglitz, Stanford, May, 1991
80 * Modified by Van Jacobson, LBL, January 1993
81 * Modified by Ajit Thyagarajan, PARC, August 1993
82 * Modified by Bill Fenner, PARC, April 1994
83 * Modified by Charles M. Hannum, NetBSD, May 1995.
84 * Modified by Ahmed Helmy, SGI, June 1996
85 * Modified by George Edmond Eddy (Rusty), ISI, February 1998
86 * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
87 * Modified by Hitoshi Asaeda, WIDE, August 2000
88 * Modified by Pavlin Radoslavov, ICSI, October 2002
89 *
90 * MROUTING Revision: 1.2
91 * and PIM-SMv2 and PIM-DM support, advanced API support,
92 * bandwidth metering and signaling
93 */
94
95 #include <sys/cdefs.h>
96 __KERNEL_RCSID(0, "$NetBSD: ip_mroute.c,v 1.143 2016/07/04 04:35:09 knakahara Exp $");
97
98 #ifdef _KERNEL_OPT
99 #include "opt_inet.h"
100 #include "opt_ipsec.h"
101 #include "opt_pim.h"
102 #endif
103
104 #ifdef PIM
105 #define _PIM_VT 1
106 #endif
107
108 #include <sys/param.h>
109 #include <sys/systm.h>
110 #include <sys/callout.h>
111 #include <sys/mbuf.h>
112 #include <sys/socket.h>
113 #include <sys/socketvar.h>
114 #include <sys/protosw.h>
115 #include <sys/errno.h>
116 #include <sys/time.h>
117 #include <sys/kernel.h>
118 #include <sys/kmem.h>
119 #include <sys/ioctl.h>
120 #include <sys/syslog.h>
121
122 #include <net/if.h>
123 #include <net/raw_cb.h>
124
125 #include <netinet/in.h>
126 #include <netinet/in_var.h>
127 #include <netinet/in_systm.h>
128 #include <netinet/ip.h>
129 #include <netinet/ip_var.h>
130 #include <netinet/in_pcb.h>
131 #include <netinet/udp.h>
132 #include <netinet/igmp.h>
133 #include <netinet/igmp_var.h>
134 #include <netinet/ip_mroute.h>
135 #ifdef PIM
136 #include <netinet/pim.h>
137 #include <netinet/pim_var.h>
138 #endif
139 #include <netinet/ip_encap.h>
140
141 #ifdef IPSEC
142 #include <netipsec/ipsec.h>
143 #include <netipsec/key.h>
144 #endif
145
146 #define IP_MULTICASTOPTS 0
147 #define M_PULLUP(m, len) \
148 do { \
149 if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \
150 (m) = m_pullup((m), (len)); \
151 } while (/*CONSTCOND*/ 0)
152
153 /*
154 * Globals. All but ip_mrouter and ip_mrtproto could be static,
155 * except for netstat or debugging purposes.
156 */
157 struct socket *ip_mrouter = NULL;
158 int ip_mrtproto = IGMP_DVMRP; /* for netstat only */
159
160 #define NO_RTE_FOUND 0x1
161 #define RTE_FOUND 0x2
162
163 #define MFCHASH(a, g) \
164 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
165 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
166 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
167 u_long mfchash;
168
169 u_char nexpire[MFCTBLSIZ];
170 struct vif viftable[MAXVIFS];
171 struct mrtstat mrtstat;
172 u_int mrtdebug = 0; /* debug level */
173 #define DEBUG_MFC 0x02
174 #define DEBUG_FORWARD 0x04
175 #define DEBUG_EXPIRE 0x08
176 #define DEBUG_XMIT 0x10
177 #define DEBUG_PIM 0x20
178
179 #define VIFI_INVALID ((vifi_t) -1)
180
181 u_int tbfdebug = 0; /* tbf debug level */
182 #ifdef RSVP_ISI
183 u_int rsvpdebug = 0; /* rsvp debug level */
184 #define RSVP_DPRINTF(a) do if (rsvpdebug) printf a; while (/*CONSTCOND*/0)
185 extern struct socket *ip_rsvpd;
186 extern int rsvp_on;
187 #else
188 #define RSVP_DPRINTF(a) do {} while (/*CONSTCOND*/0)
189 #endif /* RSVP_ISI */
190
191 /* vif attachment using sys/netinet/ip_encap.c */
192 static void vif_input(struct mbuf *, int, int);
193 static int vif_encapcheck(struct mbuf *, int, int, void *);
194
195 static const struct encapsw vif_encapsw = {
196 .encapsw4 = {
197 .pr_input = vif_input,
198 .pr_ctlinput = NULL,
199 }
200 };
201
202 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
203 #define UPCALL_EXPIRE 6 /* number of timeouts */
204
205 /*
206 * Define the token bucket filter structures
207 */
208
209 #define TBF_REPROCESS (hz / 100) /* 100x / second */
210
211 static int get_sg_cnt(struct sioc_sg_req *);
212 static int get_vif_cnt(struct sioc_vif_req *);
213 static int ip_mrouter_init(struct socket *, int);
214 static int set_assert(int);
215 static int add_vif(struct vifctl *);
216 static int del_vif(vifi_t *);
217 static void update_mfc_params(struct mfc *, struct mfcctl2 *);
218 static void init_mfc_params(struct mfc *, struct mfcctl2 *);
219 static void expire_mfc(struct mfc *);
220 static int add_mfc(struct sockopt *);
221 #ifdef UPCALL_TIMING
222 static void collate(struct timeval *);
223 #endif
224 static int del_mfc(struct sockopt *);
225 static int set_api_config(struct sockopt *); /* chose API capabilities */
226 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
227 static void expire_upcalls(void *);
228 #ifdef RSVP_ISI
229 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
230 #else
231 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *);
232 #endif
233 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
234 static void encap_send(struct ip *, struct vif *, struct mbuf *);
235 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_int32_t);
236 static void tbf_queue(struct vif *, struct mbuf *);
237 static void tbf_process_q(struct vif *);
238 static void tbf_reprocess_q(void *);
239 static int tbf_dq_sel(struct vif *, struct ip *);
240 static void tbf_send_packet(struct vif *, struct mbuf *);
241 static void tbf_update_tokens(struct vif *);
242 static int priority(struct vif *, struct ip *);
243
244 /*
245 * Bandwidth monitoring
246 */
247 static void free_bw_list(struct bw_meter *);
248 static int add_bw_upcall(struct bw_upcall *);
249 static int del_bw_upcall(struct bw_upcall *);
250 static void bw_meter_receive_packet(struct bw_meter *, int , struct timeval *);
251 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
252 static void bw_upcalls_send(void);
253 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
254 static void unschedule_bw_meter(struct bw_meter *);
255 static void bw_meter_process(void);
256 static void expire_bw_upcalls_send(void *);
257 static void expire_bw_meter_process(void *);
258
259 #ifdef PIM
260 static int pim_register_send(struct ip *, struct vif *,
261 struct mbuf *, struct mfc *);
262 static int pim_register_send_rp(struct ip *, struct vif *,
263 struct mbuf *, struct mfc *);
264 static int pim_register_send_upcall(struct ip *, struct vif *,
265 struct mbuf *, struct mfc *);
266 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
267 #endif
268
269 /*
270 * 'Interfaces' associated with decapsulator (so we can tell
271 * packets that went through it from ones that get reflected
272 * by a broken gateway). These interfaces are never linked into
273 * the system ifnet list & no routes point to them. I.e., packets
274 * can't be sent this way. They only exist as a placeholder for
275 * multicast source verification.
276 */
277 #if 0
278 struct ifnet multicast_decap_if[MAXVIFS];
279 #endif
280
281 #define ENCAP_TTL 64
282 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */
283
284 /* prototype IP hdr for encapsulated packets */
285 struct ip multicast_encap_iphdr = {
286 .ip_hl = sizeof(struct ip) >> 2,
287 .ip_v = IPVERSION,
288 .ip_len = sizeof(struct ip),
289 .ip_ttl = ENCAP_TTL,
290 .ip_p = ENCAP_PROTO,
291 };
292
293 /*
294 * Bandwidth meter variables and constants
295 */
296
297 /*
298 * Pending timeouts are stored in a hash table, the key being the
299 * expiration time. Periodically, the entries are analysed and processed.
300 */
301 #define BW_METER_BUCKETS 1024
302 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
303 struct callout bw_meter_ch;
304 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
305
306 /*
307 * Pending upcalls are stored in a vector which is flushed when
308 * full, or periodically
309 */
310 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX];
311 static u_int bw_upcalls_n; /* # of pending upcalls */
312 struct callout bw_upcalls_ch;
313 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
314
315 #ifdef PIM
316 struct pimstat pimstat;
317
318 /*
319 * Note: the PIM Register encapsulation adds the following in front of a
320 * data packet:
321 *
322 * struct pim_encap_hdr {
323 * struct ip ip;
324 * struct pim_encap_pimhdr pim;
325 * }
326 *
327 */
328
329 struct pim_encap_pimhdr {
330 struct pim pim;
331 uint32_t flags;
332 };
333
334 static struct ip pim_encap_iphdr = {
335 .ip_v = IPVERSION,
336 .ip_hl = sizeof(struct ip) >> 2,
337 .ip_len = sizeof(struct ip),
338 .ip_ttl = ENCAP_TTL,
339 .ip_p = IPPROTO_PIM,
340 };
341
342 static struct pim_encap_pimhdr pim_encap_pimhdr = {
343 {
344 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
345 0, /* reserved */
346 0, /* checksum */
347 },
348 0 /* flags */
349 };
350
351 static struct ifnet multicast_register_if;
352 static vifi_t reg_vif_num = VIFI_INVALID;
353 #endif /* PIM */
354
355
356 /*
357 * Private variables.
358 */
359 static vifi_t numvifs = 0;
360
361 static struct callout expire_upcalls_ch;
362
363 /*
364 * whether or not special PIM assert processing is enabled.
365 */
366 static int pim_assert;
367 /*
368 * Rate limit for assert notification messages, in usec
369 */
370 #define ASSERT_MSG_TIME 3000000
371
372 /*
373 * Kernel multicast routing API capabilities and setup.
374 * If more API capabilities are added to the kernel, they should be
375 * recorded in `mrt_api_support'.
376 */
377 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
378 MRT_MFC_FLAGS_BORDER_VIF |
379 MRT_MFC_RP |
380 MRT_MFC_BW_UPCALL);
381 static u_int32_t mrt_api_config = 0;
382
383 /*
384 * Find a route for a given origin IP address and Multicast group address
385 * Type of service parameter to be added in the future!!!
386 * Statistics are updated by the caller if needed
387 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
388 */
389 static struct mfc *
mfc_find(struct in_addr * o,struct in_addr * g)390 mfc_find(struct in_addr *o, struct in_addr *g)
391 {
392 struct mfc *rt;
393
394 LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
395 if (in_hosteq(rt->mfc_origin, *o) &&
396 in_hosteq(rt->mfc_mcastgrp, *g) &&
397 (rt->mfc_stall == NULL))
398 break;
399 }
400
401 return (rt);
402 }
403
404 /*
405 * Macros to compute elapsed time efficiently
406 * Borrowed from Van Jacobson's scheduling code
407 */
408 #define TV_DELTA(a, b, delta) do { \
409 int xxs; \
410 delta = (a).tv_usec - (b).tv_usec; \
411 xxs = (a).tv_sec - (b).tv_sec; \
412 switch (xxs) { \
413 case 2: \
414 delta += 1000000; \
415 /* fall through */ \
416 case 1: \
417 delta += 1000000; \
418 /* fall through */ \
419 case 0: \
420 break; \
421 default: \
422 delta += (1000000 * xxs); \
423 break; \
424 } \
425 } while (/*CONSTCOND*/ 0)
426
427 #ifdef UPCALL_TIMING
428 u_int32_t upcall_data[51];
429 #endif /* UPCALL_TIMING */
430
431 /*
432 * Handle MRT setsockopt commands to modify the multicast routing tables.
433 */
434 int
ip_mrouter_set(struct socket * so,struct sockopt * sopt)435 ip_mrouter_set(struct socket *so, struct sockopt *sopt)
436 {
437 int error;
438 int optval;
439 struct vifctl vifc;
440 vifi_t vifi;
441 struct bw_upcall bwuc;
442
443 if (sopt->sopt_name != MRT_INIT && so != ip_mrouter)
444 error = ENOPROTOOPT;
445 else {
446 switch (sopt->sopt_name) {
447 case MRT_INIT:
448 error = sockopt_getint(sopt, &optval);
449 if (error)
450 break;
451
452 error = ip_mrouter_init(so, optval);
453 break;
454 case MRT_DONE:
455 error = ip_mrouter_done();
456 break;
457 case MRT_ADD_VIF:
458 error = sockopt_get(sopt, &vifc, sizeof(vifc));
459 if (error)
460 break;
461 error = add_vif(&vifc);
462 break;
463 case MRT_DEL_VIF:
464 error = sockopt_get(sopt, &vifi, sizeof(vifi));
465 if (error)
466 break;
467 error = del_vif(&vifi);
468 break;
469 case MRT_ADD_MFC:
470 error = add_mfc(sopt);
471 break;
472 case MRT_DEL_MFC:
473 error = del_mfc(sopt);
474 break;
475 case MRT_ASSERT:
476 error = sockopt_getint(sopt, &optval);
477 if (error)
478 break;
479 error = set_assert(optval);
480 break;
481 case MRT_API_CONFIG:
482 error = set_api_config(sopt);
483 break;
484 case MRT_ADD_BW_UPCALL:
485 error = sockopt_get(sopt, &bwuc, sizeof(bwuc));
486 if (error)
487 break;
488 error = add_bw_upcall(&bwuc);
489 break;
490 case MRT_DEL_BW_UPCALL:
491 error = sockopt_get(sopt, &bwuc, sizeof(bwuc));
492 if (error)
493 break;
494 error = del_bw_upcall(&bwuc);
495 break;
496 default:
497 error = ENOPROTOOPT;
498 break;
499 }
500 }
501 return (error);
502 }
503
504 /*
505 * Handle MRT getsockopt commands
506 */
507 int
ip_mrouter_get(struct socket * so,struct sockopt * sopt)508 ip_mrouter_get(struct socket *so, struct sockopt *sopt)
509 {
510 int error;
511
512 if (so != ip_mrouter)
513 error = ENOPROTOOPT;
514 else {
515 switch (sopt->sopt_name) {
516 case MRT_VERSION:
517 error = sockopt_setint(sopt, 0x0305); /* XXX !!!! */
518 break;
519 case MRT_ASSERT:
520 error = sockopt_setint(sopt, pim_assert);
521 break;
522 case MRT_API_SUPPORT:
523 error = sockopt_set(sopt, &mrt_api_support,
524 sizeof(mrt_api_support));
525 break;
526 case MRT_API_CONFIG:
527 error = sockopt_set(sopt, &mrt_api_config,
528 sizeof(mrt_api_config));
529 break;
530 default:
531 error = ENOPROTOOPT;
532 break;
533 }
534 }
535 return (error);
536 }
537
538 /*
539 * Handle ioctl commands to obtain information from the cache
540 */
541 int
mrt_ioctl(struct socket * so,u_long cmd,void * data)542 mrt_ioctl(struct socket *so, u_long cmd, void *data)
543 {
544 int error;
545
546 if (so != ip_mrouter)
547 error = EINVAL;
548 else
549 switch (cmd) {
550 case SIOCGETVIFCNT:
551 error = get_vif_cnt((struct sioc_vif_req *)data);
552 break;
553 case SIOCGETSGCNT:
554 error = get_sg_cnt((struct sioc_sg_req *)data);
555 break;
556 default:
557 error = EINVAL;
558 break;
559 }
560
561 return (error);
562 }
563
564 /*
565 * returns the packet, byte, rpf-failure count for the source group provided
566 */
567 static int
get_sg_cnt(struct sioc_sg_req * req)568 get_sg_cnt(struct sioc_sg_req *req)
569 {
570 int s;
571 struct mfc *rt;
572
573 s = splsoftnet();
574 rt = mfc_find(&req->src, &req->grp);
575 if (rt == NULL) {
576 splx(s);
577 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
578 return (EADDRNOTAVAIL);
579 }
580 req->pktcnt = rt->mfc_pkt_cnt;
581 req->bytecnt = rt->mfc_byte_cnt;
582 req->wrong_if = rt->mfc_wrong_if;
583 splx(s);
584
585 return (0);
586 }
587
588 /*
589 * returns the input and output packet and byte counts on the vif provided
590 */
591 static int
get_vif_cnt(struct sioc_vif_req * req)592 get_vif_cnt(struct sioc_vif_req *req)
593 {
594 vifi_t vifi = req->vifi;
595
596 if (vifi >= numvifs)
597 return (EINVAL);
598
599 req->icount = viftable[vifi].v_pkt_in;
600 req->ocount = viftable[vifi].v_pkt_out;
601 req->ibytes = viftable[vifi].v_bytes_in;
602 req->obytes = viftable[vifi].v_bytes_out;
603
604 return (0);
605 }
606
607 /*
608 * Enable multicast routing
609 */
610 static int
ip_mrouter_init(struct socket * so,int v)611 ip_mrouter_init(struct socket *so, int v)
612 {
613 if (mrtdebug)
614 log(LOG_DEBUG,
615 "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
616 so->so_type, so->so_proto->pr_protocol);
617
618 if (so->so_type != SOCK_RAW ||
619 so->so_proto->pr_protocol != IPPROTO_IGMP)
620 return (EOPNOTSUPP);
621
622 if (v != 1)
623 return (EINVAL);
624
625 if (ip_mrouter != NULL)
626 return (EADDRINUSE);
627
628 ip_mrouter = so;
629
630 mfchashtbl = hashinit(MFCTBLSIZ, HASH_LIST, true, &mfchash);
631 memset((void *)nexpire, 0, sizeof(nexpire));
632
633 pim_assert = 0;
634
635 callout_init(&expire_upcalls_ch, 0);
636 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
637 expire_upcalls, NULL);
638
639 callout_init(&bw_upcalls_ch, 0);
640 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
641 expire_bw_upcalls_send, NULL);
642
643 callout_init(&bw_meter_ch, 0);
644 callout_reset(&bw_meter_ch, BW_METER_PERIOD,
645 expire_bw_meter_process, NULL);
646
647 if (mrtdebug)
648 log(LOG_DEBUG, "ip_mrouter_init\n");
649
650 return (0);
651 }
652
653 /*
654 * Disable multicast routing
655 */
656 int
ip_mrouter_done(void)657 ip_mrouter_done(void)
658 {
659 vifi_t vifi;
660 struct vif *vifp;
661 int i;
662 int s;
663
664 s = splsoftnet();
665
666 /* Clear out all the vifs currently in use. */
667 for (vifi = 0; vifi < numvifs; vifi++) {
668 vifp = &viftable[vifi];
669 if (!in_nullhost(vifp->v_lcl_addr))
670 reset_vif(vifp);
671 }
672
673 numvifs = 0;
674 pim_assert = 0;
675 mrt_api_config = 0;
676
677 callout_stop(&expire_upcalls_ch);
678 callout_stop(&bw_upcalls_ch);
679 callout_stop(&bw_meter_ch);
680
681 /*
682 * Free all multicast forwarding cache entries.
683 */
684 for (i = 0; i < MFCTBLSIZ; i++) {
685 struct mfc *rt, *nrt;
686
687 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
688 nrt = LIST_NEXT(rt, mfc_hash);
689
690 expire_mfc(rt);
691 }
692 }
693
694 memset((void *)nexpire, 0, sizeof(nexpire));
695 hashdone(mfchashtbl, HASH_LIST, mfchash);
696 mfchashtbl = NULL;
697
698 bw_upcalls_n = 0;
699 memset(bw_meter_timers, 0, sizeof(bw_meter_timers));
700
701 /* Reset de-encapsulation cache. */
702
703 ip_mrouter = NULL;
704
705 splx(s);
706
707 if (mrtdebug)
708 log(LOG_DEBUG, "ip_mrouter_done\n");
709
710 return (0);
711 }
712
713 void
ip_mrouter_detach(struct ifnet * ifp)714 ip_mrouter_detach(struct ifnet *ifp)
715 {
716 int vifi, i;
717 struct vif *vifp;
718 struct mfc *rt;
719 struct rtdetq *rte;
720
721 /* XXX not sure about side effect to userland routing daemon */
722 for (vifi = 0; vifi < numvifs; vifi++) {
723 vifp = &viftable[vifi];
724 if (vifp->v_ifp == ifp)
725 reset_vif(vifp);
726 }
727 for (i = 0; i < MFCTBLSIZ; i++) {
728 if (nexpire[i] == 0)
729 continue;
730 LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
731 for (rte = rt->mfc_stall; rte; rte = rte->next) {
732 if (rte->ifp == ifp)
733 rte->ifp = NULL;
734 }
735 }
736 }
737 }
738
739 /*
740 * Set PIM assert processing global
741 */
742 static int
set_assert(int i)743 set_assert(int i)
744 {
745 pim_assert = !!i;
746 return (0);
747 }
748
749 /*
750 * Configure API capabilities
751 */
752 static int
set_api_config(struct sockopt * sopt)753 set_api_config(struct sockopt *sopt)
754 {
755 u_int32_t apival;
756 int i, error;
757
758 /*
759 * We can set the API capabilities only if it is the first operation
760 * after MRT_INIT. I.e.:
761 * - there are no vifs installed
762 * - pim_assert is not enabled
763 * - the MFC table is empty
764 */
765 error = sockopt_get(sopt, &apival, sizeof(apival));
766 if (error)
767 return (error);
768 if (numvifs > 0)
769 return (EPERM);
770 if (pim_assert)
771 return (EPERM);
772 for (i = 0; i < MFCTBLSIZ; i++) {
773 if (LIST_FIRST(&mfchashtbl[i]) != NULL)
774 return (EPERM);
775 }
776
777 mrt_api_config = apival & mrt_api_support;
778 return (0);
779 }
780
781 /*
782 * Add a vif to the vif table
783 */
784 static int
add_vif(struct vifctl * vifcp)785 add_vif(struct vifctl *vifcp)
786 {
787 struct vif *vifp;
788 struct ifaddr *ifa;
789 struct ifnet *ifp;
790 int error, s;
791 struct sockaddr_in sin;
792
793 if (vifcp->vifc_vifi >= MAXVIFS)
794 return (EINVAL);
795 if (in_nullhost(vifcp->vifc_lcl_addr))
796 return (EADDRNOTAVAIL);
797
798 vifp = &viftable[vifcp->vifc_vifi];
799 if (!in_nullhost(vifp->v_lcl_addr))
800 return (EADDRINUSE);
801
802 /* Find the interface with an address in AF_INET family. */
803 #ifdef PIM
804 if (vifcp->vifc_flags & VIFF_REGISTER) {
805 /*
806 * XXX: Because VIFF_REGISTER does not really need a valid
807 * local interface (e.g. it could be 127.0.0.2), we don't
808 * check its address.
809 */
810 ifp = NULL;
811 } else
812 #endif
813 {
814 sockaddr_in_init(&sin, &vifcp->vifc_lcl_addr, 0);
815 ifa = ifa_ifwithaddr(sintosa(&sin));
816 if (ifa == NULL)
817 return (EADDRNOTAVAIL);
818 ifp = ifa->ifa_ifp;
819 }
820
821 if (vifcp->vifc_flags & VIFF_TUNNEL) {
822 if (vifcp->vifc_flags & VIFF_SRCRT) {
823 log(LOG_ERR, "source routed tunnels not supported\n");
824 return (EOPNOTSUPP);
825 }
826
827 /* attach this vif to decapsulator dispatch table */
828 /*
829 * XXX Use addresses in registration so that matching
830 * can be done with radix tree in decapsulator. But,
831 * we need to check inner header for multicast, so
832 * this requires both radix tree lookup and then a
833 * function to check, and this is not supported yet.
834 */
835 error = encap_lock_enter();
836 if (error)
837 return error;
838 vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
839 vif_encapcheck, &vif_encapsw, vifp);
840 encap_lock_exit();
841 if (!vifp->v_encap_cookie)
842 return (EINVAL);
843
844 /* Create a fake encapsulation interface. */
845 ifp = malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK|M_ZERO);
846 snprintf(ifp->if_xname, sizeof(ifp->if_xname),
847 "mdecap%d", vifcp->vifc_vifi);
848
849 /* Prepare cached route entry. */
850 memset(&vifp->v_route, 0, sizeof(vifp->v_route));
851 #ifdef PIM
852 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
853 ifp = &multicast_register_if;
854 if (mrtdebug)
855 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
856 (void *)ifp);
857 if (reg_vif_num == VIFI_INVALID) {
858 memset(ifp, 0, sizeof(*ifp));
859 snprintf(ifp->if_xname, sizeof(ifp->if_xname),
860 "register_vif");
861 ifp->if_flags = IFF_LOOPBACK;
862 memset(&vifp->v_route, 0, sizeof(vifp->v_route));
863 reg_vif_num = vifcp->vifc_vifi;
864 }
865 #endif
866 } else {
867 /* Make sure the interface supports multicast. */
868 if ((ifp->if_flags & IFF_MULTICAST) == 0)
869 return (EOPNOTSUPP);
870
871 /* Enable promiscuous reception of all IP multicasts. */
872 sockaddr_in_init(&sin, &zeroin_addr, 0);
873 error = if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin));
874 if (error)
875 return (error);
876 }
877
878 s = splsoftnet();
879
880 /* Define parameters for the tbf structure. */
881 vifp->tbf_q = NULL;
882 vifp->tbf_t = &vifp->tbf_q;
883 microtime(&vifp->tbf_last_pkt_t);
884 vifp->tbf_n_tok = 0;
885 vifp->tbf_q_len = 0;
886 vifp->tbf_max_q_len = MAXQSIZE;
887
888 vifp->v_flags = vifcp->vifc_flags;
889 vifp->v_threshold = vifcp->vifc_threshold;
890 /* scaling up here allows division by 1024 in critical code */
891 vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
892 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
893 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
894 vifp->v_ifp = ifp;
895 /* Initialize per vif pkt counters. */
896 vifp->v_pkt_in = 0;
897 vifp->v_pkt_out = 0;
898 vifp->v_bytes_in = 0;
899 vifp->v_bytes_out = 0;
900
901 callout_init(&vifp->v_repq_ch, 0);
902
903 #ifdef RSVP_ISI
904 vifp->v_rsvp_on = 0;
905 vifp->v_rsvpd = NULL;
906 #endif /* RSVP_ISI */
907
908 splx(s);
909
910 /* Adjust numvifs up if the vifi is higher than numvifs. */
911 if (numvifs <= vifcp->vifc_vifi)
912 numvifs = vifcp->vifc_vifi + 1;
913
914 if (mrtdebug)
915 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
916 vifcp->vifc_vifi,
917 ntohl(vifcp->vifc_lcl_addr.s_addr),
918 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
919 ntohl(vifcp->vifc_rmt_addr.s_addr),
920 vifcp->vifc_threshold,
921 vifcp->vifc_rate_limit);
922
923 return (0);
924 }
925
926 void
reset_vif(struct vif * vifp)927 reset_vif(struct vif *vifp)
928 {
929 struct mbuf *m, *n;
930 struct ifnet *ifp;
931 struct sockaddr_in sin;
932
933 callout_stop(&vifp->v_repq_ch);
934
935 /* detach this vif from decapsulator dispatch table */
936 encap_lock_enter();
937 encap_detach(vifp->v_encap_cookie);
938 encap_lock_exit();
939 vifp->v_encap_cookie = NULL;
940
941 /*
942 * Free packets queued at the interface
943 */
944 for (m = vifp->tbf_q; m != NULL; m = n) {
945 n = m->m_nextpkt;
946 m_freem(m);
947 }
948
949 if (vifp->v_flags & VIFF_TUNNEL)
950 free(vifp->v_ifp, M_MRTABLE);
951 else if (vifp->v_flags & VIFF_REGISTER) {
952 #ifdef PIM
953 reg_vif_num = VIFI_INVALID;
954 #endif
955 } else {
956 sockaddr_in_init(&sin, &zeroin_addr, 0);
957 ifp = vifp->v_ifp;
958 if_mcast_op(ifp, SIOCDELMULTI, sintosa(&sin));
959 }
960 memset((void *)vifp, 0, sizeof(*vifp));
961 }
962
963 /*
964 * Delete a vif from the vif table
965 */
966 static int
del_vif(vifi_t * vifip)967 del_vif(vifi_t *vifip)
968 {
969 struct vif *vifp;
970 vifi_t vifi;
971 int s;
972
973 if (*vifip >= numvifs)
974 return (EINVAL);
975
976 vifp = &viftable[*vifip];
977 if (in_nullhost(vifp->v_lcl_addr))
978 return (EADDRNOTAVAIL);
979
980 s = splsoftnet();
981
982 reset_vif(vifp);
983
984 /* Adjust numvifs down */
985 for (vifi = numvifs; vifi > 0; vifi--)
986 if (!in_nullhost(viftable[vifi - 1].v_lcl_addr))
987 break;
988 numvifs = vifi;
989
990 splx(s);
991
992 if (mrtdebug)
993 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
994
995 return (0);
996 }
997
998 /*
999 * update an mfc entry without resetting counters and S,G addresses.
1000 */
1001 static void
update_mfc_params(struct mfc * rt,struct mfcctl2 * mfccp)1002 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1003 {
1004 int i;
1005
1006 rt->mfc_parent = mfccp->mfcc_parent;
1007 for (i = 0; i < numvifs; i++) {
1008 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1009 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1010 MRT_MFC_FLAGS_ALL;
1011 }
1012 /* set the RP address */
1013 if (mrt_api_config & MRT_MFC_RP)
1014 rt->mfc_rp = mfccp->mfcc_rp;
1015 else
1016 rt->mfc_rp = zeroin_addr;
1017 }
1018
1019 /*
1020 * fully initialize an mfc entry from the parameter.
1021 */
1022 static void
init_mfc_params(struct mfc * rt,struct mfcctl2 * mfccp)1023 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1024 {
1025 rt->mfc_origin = mfccp->mfcc_origin;
1026 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1027
1028 update_mfc_params(rt, mfccp);
1029
1030 /* initialize pkt counters per src-grp */
1031 rt->mfc_pkt_cnt = 0;
1032 rt->mfc_byte_cnt = 0;
1033 rt->mfc_wrong_if = 0;
1034 timerclear(&rt->mfc_last_assert);
1035 }
1036
1037 static void
expire_mfc(struct mfc * rt)1038 expire_mfc(struct mfc *rt)
1039 {
1040 struct rtdetq *rte, *nrte;
1041
1042 free_bw_list(rt->mfc_bw_meter);
1043
1044 for (rte = rt->mfc_stall; rte != NULL; rte = nrte) {
1045 nrte = rte->next;
1046 m_freem(rte->m);
1047 free(rte, M_MRTABLE);
1048 }
1049
1050 LIST_REMOVE(rt, mfc_hash);
1051 free(rt, M_MRTABLE);
1052 }
1053
1054 /*
1055 * Add an mfc entry
1056 */
1057 static int
add_mfc(struct sockopt * sopt)1058 add_mfc(struct sockopt *sopt)
1059 {
1060 struct mfcctl2 mfcctl2;
1061 struct mfcctl2 *mfccp;
1062 struct mfc *rt;
1063 u_int32_t hash = 0;
1064 struct rtdetq *rte, *nrte;
1065 u_short nstl;
1066 int s;
1067 int error;
1068
1069 /*
1070 * select data size depending on API version.
1071 */
1072 mfccp = &mfcctl2;
1073 memset(&mfcctl2, 0, sizeof(mfcctl2));
1074
1075 if (mrt_api_config & MRT_API_FLAGS_ALL)
1076 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1077 else
1078 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1079
1080 if (error)
1081 return (error);
1082
1083 s = splsoftnet();
1084 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1085
1086 /* If an entry already exists, just update the fields */
1087 if (rt) {
1088 if (mrtdebug & DEBUG_MFC)
1089 log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n",
1090 ntohl(mfccp->mfcc_origin.s_addr),
1091 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1092 mfccp->mfcc_parent);
1093
1094 update_mfc_params(rt, mfccp);
1095
1096 splx(s);
1097 return (0);
1098 }
1099
1100 /*
1101 * Find the entry for which the upcall was made and update
1102 */
1103 nstl = 0;
1104 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1105 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1106 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1107 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1108 rt->mfc_stall != NULL) {
1109 if (nstl++)
1110 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n",
1111 "multiple kernel entries",
1112 ntohl(mfccp->mfcc_origin.s_addr),
1113 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1114 mfccp->mfcc_parent, rt->mfc_stall);
1115
1116 if (mrtdebug & DEBUG_MFC)
1117 log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n",
1118 ntohl(mfccp->mfcc_origin.s_addr),
1119 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1120 mfccp->mfcc_parent, rt->mfc_stall);
1121
1122 rte = rt->mfc_stall;
1123 init_mfc_params(rt, mfccp);
1124 rt->mfc_stall = NULL;
1125
1126 rt->mfc_expire = 0; /* Don't clean this guy up */
1127 nexpire[hash]--;
1128
1129 /* free packets Qed at the end of this entry */
1130 for (; rte != NULL; rte = nrte) {
1131 nrte = rte->next;
1132 if (rte->ifp) {
1133 #ifdef RSVP_ISI
1134 ip_mdq(rte->m, rte->ifp, rt, -1);
1135 #else
1136 ip_mdq(rte->m, rte->ifp, rt);
1137 #endif /* RSVP_ISI */
1138 }
1139 m_freem(rte->m);
1140 #ifdef UPCALL_TIMING
1141 collate(&rte->t);
1142 #endif /* UPCALL_TIMING */
1143 free(rte, M_MRTABLE);
1144 }
1145 }
1146 }
1147
1148 /*
1149 * It is possible that an entry is being inserted without an upcall
1150 */
1151 if (nstl == 0) {
1152 /*
1153 * No mfc; make a new one
1154 */
1155 if (mrtdebug & DEBUG_MFC)
1156 log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n",
1157 ntohl(mfccp->mfcc_origin.s_addr),
1158 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1159 mfccp->mfcc_parent);
1160
1161 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1162 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1163 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1164 init_mfc_params(rt, mfccp);
1165 if (rt->mfc_expire)
1166 nexpire[hash]--;
1167 rt->mfc_expire = 0;
1168 break; /* XXX */
1169 }
1170 }
1171 if (rt == NULL) { /* no upcall, so make a new entry */
1172 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1173 M_NOWAIT);
1174 if (rt == NULL) {
1175 splx(s);
1176 return (ENOBUFS);
1177 }
1178
1179 init_mfc_params(rt, mfccp);
1180 rt->mfc_expire = 0;
1181 rt->mfc_stall = NULL;
1182 rt->mfc_bw_meter = NULL;
1183
1184 /* insert new entry at head of hash chain */
1185 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1186 }
1187 }
1188
1189 splx(s);
1190 return (0);
1191 }
1192
1193 #ifdef UPCALL_TIMING
1194 /*
1195 * collect delay statistics on the upcalls
1196 */
1197 static void
collate(struct timeval * t)1198 collate(struct timeval *t)
1199 {
1200 u_int32_t d;
1201 struct timeval tp;
1202 u_int32_t delta;
1203
1204 microtime(&tp);
1205
1206 if (timercmp(t, &tp, <)) {
1207 TV_DELTA(tp, *t, delta);
1208
1209 d = delta >> 10;
1210 if (d > 50)
1211 d = 50;
1212
1213 ++upcall_data[d];
1214 }
1215 }
1216 #endif /* UPCALL_TIMING */
1217
1218 /*
1219 * Delete an mfc entry
1220 */
1221 static int
del_mfc(struct sockopt * sopt)1222 del_mfc(struct sockopt *sopt)
1223 {
1224 struct mfcctl2 mfcctl2;
1225 struct mfcctl2 *mfccp;
1226 struct mfc *rt;
1227 int s;
1228 int error;
1229
1230 /*
1231 * XXX: for deleting MFC entries the information in entries
1232 * of size "struct mfcctl" is sufficient.
1233 */
1234
1235 mfccp = &mfcctl2;
1236 memset(&mfcctl2, 0, sizeof(mfcctl2));
1237
1238 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1239 if (error) {
1240 /* Try with the size of mfcctl2. */
1241 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1242 if (error)
1243 return (error);
1244 }
1245
1246 if (mrtdebug & DEBUG_MFC)
1247 log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
1248 ntohl(mfccp->mfcc_origin.s_addr),
1249 ntohl(mfccp->mfcc_mcastgrp.s_addr));
1250
1251 s = splsoftnet();
1252
1253 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1254 if (rt == NULL) {
1255 splx(s);
1256 return (EADDRNOTAVAIL);
1257 }
1258
1259 /*
1260 * free the bw_meter entries
1261 */
1262 free_bw_list(rt->mfc_bw_meter);
1263 rt->mfc_bw_meter = NULL;
1264
1265 LIST_REMOVE(rt, mfc_hash);
1266 free(rt, M_MRTABLE);
1267
1268 splx(s);
1269 return (0);
1270 }
1271
1272 static int
socket_send(struct socket * s,struct mbuf * mm,struct sockaddr_in * src)1273 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1274 {
1275 if (s) {
1276 if (sbappendaddr(&s->so_rcv, sintosa(src), mm, NULL) != 0) {
1277 sorwakeup(s);
1278 return (0);
1279 }
1280 }
1281 m_freem(mm);
1282 return (-1);
1283 }
1284
1285 /*
1286 * IP multicast forwarding function. This function assumes that the packet
1287 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1288 * pointed to by "ifp", and the packet is to be relayed to other networks
1289 * that have members of the packet's destination IP multicast group.
1290 *
1291 * The packet is returned unscathed to the caller, unless it is
1292 * erroneous, in which case a non-zero return value tells the caller to
1293 * discard it.
1294 */
1295
1296 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */
1297 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1298
1299 int
1300 #ifdef RSVP_ISI
ip_mforward(struct mbuf * m,struct ifnet * ifp,struct ip_moptions * imo)1301 ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo)
1302 #else
1303 ip_mforward(struct mbuf *m, struct ifnet *ifp)
1304 #endif /* RSVP_ISI */
1305 {
1306 struct ip *ip = mtod(m, struct ip *);
1307 struct mfc *rt;
1308 static int srctun = 0;
1309 struct mbuf *mm;
1310 struct sockaddr_in sin;
1311 int s;
1312 vifi_t vifi;
1313
1314 if (mrtdebug & DEBUG_FORWARD)
1315 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1316 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1317
1318 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1319 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1320 /*
1321 * Packet arrived via a physical interface or
1322 * an encapsulated tunnel or a register_vif.
1323 */
1324 } else {
1325 /*
1326 * Packet arrived through a source-route tunnel.
1327 * Source-route tunnels are no longer supported.
1328 */
1329 if ((srctun++ % 1000) == 0)
1330 log(LOG_ERR,
1331 "ip_mforward: received source-routed packet from %x\n",
1332 ntohl(ip->ip_src.s_addr));
1333
1334 return (1);
1335 }
1336
1337 /*
1338 * Clear any in-bound checksum flags for this packet.
1339 */
1340 m->m_pkthdr.csum_flags = 0;
1341
1342 #ifdef RSVP_ISI
1343 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1344 if (ip->ip_ttl < MAXTTL)
1345 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1346 if (ip->ip_p == IPPROTO_RSVP) {
1347 struct vif *vifp = viftable + vifi;
1348 RSVP_DPRINTF(("%s: Sending IPPROTO_RSVP from %x to %x"
1349 " on vif %d (%s%s)\n", __func__,
1350 ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
1351 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1352 vifp->v_ifp->if_xname));
1353 }
1354 return (ip_mdq(m, ifp, NULL, vifi));
1355 }
1356 if (ip->ip_p == IPPROTO_RSVP) {
1357 RSVP_DPRINTF(("%s: Warning: IPPROTO_RSVP from %x to %x"
1358 " without vif option\n", __func__,
1359 ntohl(ip->ip_src), ntohl(ip->ip_dst));
1360 }
1361 #endif /* RSVP_ISI */
1362
1363 /*
1364 * Don't forward a packet with time-to-live of zero or one,
1365 * or a packet destined to a local-only group.
1366 */
1367 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1368 return (0);
1369
1370 /*
1371 * Determine forwarding vifs from the forwarding cache table
1372 */
1373 s = splsoftnet();
1374 ++mrtstat.mrts_mfc_lookups;
1375 rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1376
1377 /* Entry exists, so forward if necessary */
1378 if (rt != NULL) {
1379 splx(s);
1380 #ifdef RSVP_ISI
1381 return (ip_mdq(m, ifp, rt, -1));
1382 #else
1383 return (ip_mdq(m, ifp, rt));
1384 #endif /* RSVP_ISI */
1385 } else {
1386 /*
1387 * If we don't have a route for packet's origin,
1388 * Make a copy of the packet & send message to routing daemon
1389 */
1390
1391 struct mbuf *mb0;
1392 struct rtdetq *rte;
1393 u_int32_t hash;
1394 int hlen = ip->ip_hl << 2;
1395 #ifdef UPCALL_TIMING
1396 struct timeval tp;
1397
1398 microtime(&tp);
1399 #endif /* UPCALL_TIMING */
1400
1401 ++mrtstat.mrts_mfc_misses;
1402
1403 mrtstat.mrts_no_route++;
1404 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1405 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1406 ntohl(ip->ip_src.s_addr),
1407 ntohl(ip->ip_dst.s_addr));
1408
1409 /*
1410 * Allocate mbufs early so that we don't do extra work if we are
1411 * just going to fail anyway. Make sure to pullup the header so
1412 * that other people can't step on it.
1413 */
1414 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1415 M_NOWAIT);
1416 if (rte == NULL) {
1417 splx(s);
1418 return (ENOBUFS);
1419 }
1420 mb0 = m_copypacket(m, M_DONTWAIT);
1421 M_PULLUP(mb0, hlen);
1422 if (mb0 == NULL) {
1423 free(rte, M_MRTABLE);
1424 splx(s);
1425 return (ENOBUFS);
1426 }
1427
1428 /* is there an upcall waiting for this flow? */
1429 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1430 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1431 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1432 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1433 rt->mfc_stall != NULL)
1434 break;
1435 }
1436
1437 if (rt == NULL) {
1438 int i;
1439 struct igmpmsg *im;
1440
1441 /*
1442 * Locate the vifi for the incoming interface for
1443 * this packet.
1444 * If none found, drop packet.
1445 */
1446 for (vifi = 0; vifi < numvifs &&
1447 viftable[vifi].v_ifp != ifp; vifi++)
1448 ;
1449 if (vifi >= numvifs) /* vif not found, drop packet */
1450 goto non_fatal;
1451
1452 /* no upcall, so make a new entry */
1453 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1454 M_NOWAIT);
1455 if (rt == NULL)
1456 goto fail;
1457
1458 /*
1459 * Make a copy of the header to send to the user level
1460 * process
1461 */
1462 mm = m_copym(m, 0, hlen, M_DONTWAIT);
1463 M_PULLUP(mm, hlen);
1464 if (mm == NULL)
1465 goto fail1;
1466
1467 /*
1468 * Send message to routing daemon to install
1469 * a route into the kernel table
1470 */
1471
1472 im = mtod(mm, struct igmpmsg *);
1473 im->im_msgtype = IGMPMSG_NOCACHE;
1474 im->im_mbz = 0;
1475 im->im_vif = vifi;
1476
1477 mrtstat.mrts_upcalls++;
1478
1479 sockaddr_in_init(&sin, &ip->ip_src, 0);
1480 if (socket_send(ip_mrouter, mm, &sin) < 0) {
1481 log(LOG_WARNING,
1482 "ip_mforward: ip_mrouter socket queue full\n");
1483 ++mrtstat.mrts_upq_sockfull;
1484 fail1:
1485 free(rt, M_MRTABLE);
1486 fail:
1487 free(rte, M_MRTABLE);
1488 m_freem(mb0);
1489 splx(s);
1490 return (ENOBUFS);
1491 }
1492
1493 /* insert new entry at head of hash chain */
1494 rt->mfc_origin = ip->ip_src;
1495 rt->mfc_mcastgrp = ip->ip_dst;
1496 rt->mfc_pkt_cnt = 0;
1497 rt->mfc_byte_cnt = 0;
1498 rt->mfc_wrong_if = 0;
1499 rt->mfc_expire = UPCALL_EXPIRE;
1500 nexpire[hash]++;
1501 for (i = 0; i < numvifs; i++) {
1502 rt->mfc_ttls[i] = 0;
1503 rt->mfc_flags[i] = 0;
1504 }
1505 rt->mfc_parent = -1;
1506
1507 /* clear the RP address */
1508 rt->mfc_rp = zeroin_addr;
1509
1510 rt->mfc_bw_meter = NULL;
1511
1512 /* link into table */
1513 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1514 /* Add this entry to the end of the queue */
1515 rt->mfc_stall = rte;
1516 } else {
1517 /* determine if q has overflowed */
1518 struct rtdetq **p;
1519 int npkts = 0;
1520
1521 /*
1522 * XXX ouch! we need to append to the list, but we
1523 * only have a pointer to the front, so we have to
1524 * scan the entire list every time.
1525 */
1526 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1527 if (++npkts > MAX_UPQ) {
1528 mrtstat.mrts_upq_ovflw++;
1529 non_fatal:
1530 free(rte, M_MRTABLE);
1531 m_freem(mb0);
1532 splx(s);
1533 return (0);
1534 }
1535
1536 /* Add this entry to the end of the queue */
1537 *p = rte;
1538 }
1539
1540 rte->next = NULL;
1541 rte->m = mb0;
1542 rte->ifp = ifp;
1543 #ifdef UPCALL_TIMING
1544 rte->t = tp;
1545 #endif /* UPCALL_TIMING */
1546
1547 splx(s);
1548
1549 return (0);
1550 }
1551 }
1552
1553
1554 /*ARGSUSED*/
1555 static void
1556 expire_upcalls(void *v)
1557 {
1558 int i;
1559 int s;
1560
1561 s = splsoftnet();
1562
1563 for (i = 0; i < MFCTBLSIZ; i++) {
1564 struct mfc *rt, *nrt;
1565
1566 if (nexpire[i] == 0)
1567 continue;
1568
1569 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1570 nrt = LIST_NEXT(rt, mfc_hash);
1571
1572 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1573 continue;
1574 nexpire[i]--;
1575
1576 /*
1577 * free the bw_meter entries
1578 */
1579 while (rt->mfc_bw_meter != NULL) {
1580 struct bw_meter *x = rt->mfc_bw_meter;
1581
1582 rt->mfc_bw_meter = x->bm_mfc_next;
1583 kmem_free(x, sizeof(*x));
1584 }
1585
1586 ++mrtstat.mrts_cache_cleanups;
1587 if (mrtdebug & DEBUG_EXPIRE)
1588 log(LOG_DEBUG,
1589 "expire_upcalls: expiring (%x %x)\n",
1590 ntohl(rt->mfc_origin.s_addr),
1591 ntohl(rt->mfc_mcastgrp.s_addr));
1592
1593 expire_mfc(rt);
1594 }
1595 }
1596
1597 splx(s);
1598 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1599 expire_upcalls, NULL);
1600 }
1601
1602 /*
1603 * Packet forwarding routine once entry in the cache is made
1604 */
1605 static int
1606 #ifdef RSVP_ISI
1607 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1608 #else
1609 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt)
1610 #endif /* RSVP_ISI */
1611 {
1612 struct ip *ip = mtod(m, struct ip *);
1613 vifi_t vifi;
1614 struct vif *vifp;
1615 struct sockaddr_in sin;
1616 int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1617
1618 /*
1619 * Macro to send packet on vif. Since RSVP packets don't get counted on
1620 * input, they shouldn't get counted on output, so statistics keeping is
1621 * separate.
1622 */
1623 #define MC_SEND(ip, vifp, m) do { \
1624 if ((vifp)->v_flags & VIFF_TUNNEL) \
1625 encap_send((ip), (vifp), (m)); \
1626 else \
1627 phyint_send((ip), (vifp), (m)); \
1628 } while (/*CONSTCOND*/ 0)
1629
1630 #ifdef RSVP_ISI
1631 /*
1632 * If xmt_vif is not -1, send on only the requested vif.
1633 *
1634 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1635 */
1636 if (xmt_vif < numvifs) {
1637 #ifdef PIM
1638 if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1639 pim_register_send(ip, viftable + xmt_vif, m, rt);
1640 else
1641 #endif
1642 MC_SEND(ip, viftable + xmt_vif, m);
1643 return (1);
1644 }
1645 #endif /* RSVP_ISI */
1646
1647 /*
1648 * Don't forward if it didn't arrive from the parent vif for its origin.
1649 */
1650 vifi = rt->mfc_parent;
1651 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1652 /* came in the wrong interface */
1653 if (mrtdebug & DEBUG_FORWARD)
1654 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1655 ifp, vifi,
1656 vifi >= numvifs ? 0 : viftable[vifi].v_ifp);
1657 ++mrtstat.mrts_wrong_if;
1658 ++rt->mfc_wrong_if;
1659 /*
1660 * If we are doing PIM assert processing, send a message
1661 * to the routing daemon.
1662 *
1663 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1664 * can complete the SPT switch, regardless of the type
1665 * of the iif (broadcast media, GRE tunnel, etc).
1666 */
1667 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1668 struct timeval now;
1669 u_int32_t delta;
1670
1671 #ifdef PIM
1672 if (ifp == &multicast_register_if)
1673 pimstat.pims_rcv_registers_wrongiif++;
1674 #endif
1675
1676 /* Get vifi for the incoming packet */
1677 for (vifi = 0;
1678 vifi < numvifs && viftable[vifi].v_ifp != ifp;
1679 vifi++)
1680 ;
1681 if (vifi >= numvifs) {
1682 /* The iif is not found: ignore the packet. */
1683 return (0);
1684 }
1685
1686 if (rt->mfc_flags[vifi] &
1687 MRT_MFC_FLAGS_DISABLE_WRONGVIF) {
1688 /* WRONGVIF disabled: ignore the packet */
1689 return (0);
1690 }
1691
1692 microtime(&now);
1693
1694 TV_DELTA(rt->mfc_last_assert, now, delta);
1695
1696 if (delta > ASSERT_MSG_TIME) {
1697 struct igmpmsg *im;
1698 int hlen = ip->ip_hl << 2;
1699 struct mbuf *mm =
1700 m_copym(m, 0, hlen, M_DONTWAIT);
1701
1702 M_PULLUP(mm, hlen);
1703 if (mm == NULL)
1704 return (ENOBUFS);
1705
1706 rt->mfc_last_assert = now;
1707
1708 im = mtod(mm, struct igmpmsg *);
1709 im->im_msgtype = IGMPMSG_WRONGVIF;
1710 im->im_mbz = 0;
1711 im->im_vif = vifi;
1712
1713 mrtstat.mrts_upcalls++;
1714
1715 sockaddr_in_init(&sin, &im->im_src, 0);
1716 if (socket_send(ip_mrouter, mm, &sin) < 0) {
1717 log(LOG_WARNING,
1718 "ip_mforward: ip_mrouter socket queue full\n");
1719 ++mrtstat.mrts_upq_sockfull;
1720 return (ENOBUFS);
1721 }
1722 }
1723 }
1724 return (0);
1725 }
1726
1727 /* If I sourced this packet, it counts as output, else it was input. */
1728 if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1729 viftable[vifi].v_pkt_out++;
1730 viftable[vifi].v_bytes_out += plen;
1731 } else {
1732 viftable[vifi].v_pkt_in++;
1733 viftable[vifi].v_bytes_in += plen;
1734 }
1735 rt->mfc_pkt_cnt++;
1736 rt->mfc_byte_cnt += plen;
1737
1738 /*
1739 * For each vif, decide if a copy of the packet should be forwarded.
1740 * Forward if:
1741 * - the ttl exceeds the vif's threshold
1742 * - there are group members downstream on interface
1743 */
1744 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1745 if ((rt->mfc_ttls[vifi] > 0) &&
1746 (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1747 vifp->v_pkt_out++;
1748 vifp->v_bytes_out += plen;
1749 #ifdef PIM
1750 if (vifp->v_flags & VIFF_REGISTER)
1751 pim_register_send(ip, vifp, m, rt);
1752 else
1753 #endif
1754 MC_SEND(ip, vifp, m);
1755 }
1756
1757 /*
1758 * Perform upcall-related bw measuring.
1759 */
1760 if (rt->mfc_bw_meter != NULL) {
1761 struct bw_meter *x;
1762 struct timeval now;
1763
1764 microtime(&now);
1765 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1766 bw_meter_receive_packet(x, plen, &now);
1767 }
1768
1769 return (0);
1770 }
1771
1772 #ifdef RSVP_ISI
1773 /*
1774 * check if a vif number is legal/ok. This is used by ip_output.
1775 */
1776 int
1777 legal_vif_num(int vif)
1778 {
1779 if (vif >= 0 && vif < numvifs)
1780 return (1);
1781 else
1782 return (0);
1783 }
1784 #endif /* RSVP_ISI */
1785
1786 static void
1787 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1788 {
1789 struct mbuf *mb_copy;
1790 int hlen = ip->ip_hl << 2;
1791
1792 /*
1793 * Make a new reference to the packet; make sure that
1794 * the IP header is actually copied, not just referenced,
1795 * so that ip_output() only scribbles on the copy.
1796 */
1797 mb_copy = m_copypacket(m, M_DONTWAIT);
1798 M_PULLUP(mb_copy, hlen);
1799 if (mb_copy == NULL)
1800 return;
1801
1802 if (vifp->v_rate_limit <= 0)
1803 tbf_send_packet(vifp, mb_copy);
1804 else
1805 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *),
1806 ntohs(ip->ip_len));
1807 }
1808
1809 static void
1810 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1811 {
1812 struct mbuf *mb_copy;
1813 struct ip *ip_copy;
1814 int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr);
1815
1816 /* Take care of delayed checksums */
1817 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1818 in_delayed_cksum(m);
1819 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1820 }
1821
1822 /*
1823 * copy the old packet & pullup its IP header into the
1824 * new mbuf so we can modify it. Try to fill the new
1825 * mbuf since if we don't the ethernet driver will.
1826 */
1827 MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1828 if (mb_copy == NULL)
1829 return;
1830 mb_copy->m_data += max_linkhdr;
1831 mb_copy->m_pkthdr.len = len;
1832 mb_copy->m_len = sizeof(multicast_encap_iphdr);
1833
1834 if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) {
1835 m_freem(mb_copy);
1836 return;
1837 }
1838 i = MHLEN - max_linkhdr;
1839 if (i > len)
1840 i = len;
1841 mb_copy = m_pullup(mb_copy, i);
1842 if (mb_copy == NULL)
1843 return;
1844
1845 /*
1846 * fill in the encapsulating IP header.
1847 */
1848 ip_copy = mtod(mb_copy, struct ip *);
1849 *ip_copy = multicast_encap_iphdr;
1850 if (len < IP_MINFRAGSIZE)
1851 ip_copy->ip_id = 0;
1852 else
1853 ip_copy->ip_id = ip_newid(NULL);
1854 ip_copy->ip_len = htons(len);
1855 ip_copy->ip_src = vifp->v_lcl_addr;
1856 ip_copy->ip_dst = vifp->v_rmt_addr;
1857
1858 /*
1859 * turn the encapsulated IP header back into a valid one.
1860 */
1861 ip = (struct ip *)((char *)ip_copy + sizeof(multicast_encap_iphdr));
1862 --ip->ip_ttl;
1863 ip->ip_sum = 0;
1864 mb_copy->m_data += sizeof(multicast_encap_iphdr);
1865 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1866 mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1867
1868 if (vifp->v_rate_limit <= 0)
1869 tbf_send_packet(vifp, mb_copy);
1870 else
1871 tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len));
1872 }
1873
1874 /*
1875 * De-encapsulate a packet and feed it back through ip input.
1876 */
1877 static void
1878 vif_input(struct mbuf *m, int off, int proto)
1879 {
1880 struct vif *vifp;
1881
1882 vifp = (struct vif *)encap_getarg(m);
1883 if (!vifp || proto != ENCAP_PROTO) {
1884 m_freem(m);
1885 mrtstat.mrts_bad_tunnel++;
1886 return;
1887 }
1888
1889 m_adj(m, off);
1890 m_set_rcvif(m, vifp->v_ifp);
1891
1892 if (__predict_false(!pktq_enqueue(ip_pktq, m, 0))) {
1893 m_freem(m);
1894 }
1895 }
1896
1897 /*
1898 * Check if the packet should be received on the vif denoted by arg.
1899 * (The encap selection code will call this once per vif since each is
1900 * registered separately.)
1901 */
1902 static int
1903 vif_encapcheck(struct mbuf *m, int off, int proto, void *arg)
1904 {
1905 struct vif *vifp;
1906 struct ip ip;
1907
1908 #ifdef DIAGNOSTIC
1909 if (!arg || proto != IPPROTO_IPV4)
1910 panic("unexpected arg in vif_encapcheck");
1911 #endif
1912
1913 /*
1914 * Accept the packet only if the inner heaader is multicast
1915 * and the outer header matches a tunnel-mode vif. Order
1916 * checks in the hope that common non-matching packets will be
1917 * rejected quickly. Assume that unicast IPv4 traffic in a
1918 * parallel tunnel (e.g. gif(4)) is unlikely.
1919 */
1920
1921 /* Obtain the outer IP header and the vif pointer. */
1922 m_copydata((struct mbuf *)m, 0, sizeof(ip), (void *)&ip);
1923 vifp = (struct vif *)arg;
1924
1925 /*
1926 * The outer source must match the vif's remote peer address.
1927 * For a multicast router with several tunnels, this is the
1928 * only check that will fail on packets in other tunnels,
1929 * assuming the local address is the same.
1930 */
1931 if (!in_hosteq(vifp->v_rmt_addr, ip.ip_src))
1932 return 0;
1933
1934 /* The outer destination must match the vif's local address. */
1935 if (!in_hosteq(vifp->v_lcl_addr, ip.ip_dst))
1936 return 0;
1937
1938 /* The vif must be of tunnel type. */
1939 if ((vifp->v_flags & VIFF_TUNNEL) == 0)
1940 return 0;
1941
1942 /* Check that the inner destination is multicast. */
1943 m_copydata((struct mbuf *)m, off, sizeof(ip), (void *)&ip);
1944 if (!IN_MULTICAST(ip.ip_dst.s_addr))
1945 return 0;
1946
1947 /*
1948 * We have checked that both the outer src and dst addresses
1949 * match the vif, and that the inner destination is multicast
1950 * (224/5). By claiming more than 64, we intend to
1951 * preferentially take packets that also match a parallel
1952 * gif(4).
1953 */
1954 return 32 + 32 + 5;
1955 }
1956
1957 /*
1958 * Token bucket filter module
1959 */
1960 static void
1961 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len)
1962 {
1963
1964 if (len > MAX_BKT_SIZE) {
1965 /* drop if packet is too large */
1966 mrtstat.mrts_pkt2large++;
1967 m_freem(m);
1968 return;
1969 }
1970
1971 tbf_update_tokens(vifp);
1972
1973 /*
1974 * If there are enough tokens, and the queue is empty, send this packet
1975 * out immediately. Otherwise, try to insert it on this vif's queue.
1976 */
1977 if (vifp->tbf_q_len == 0) {
1978 if (len <= vifp->tbf_n_tok) {
1979 vifp->tbf_n_tok -= len;
1980 tbf_send_packet(vifp, m);
1981 } else {
1982 /* queue packet and timeout till later */
1983 tbf_queue(vifp, m);
1984 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
1985 tbf_reprocess_q, vifp);
1986 }
1987 } else {
1988 if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
1989 !tbf_dq_sel(vifp, ip)) {
1990 /* queue full, and couldn't make room */
1991 mrtstat.mrts_q_overflow++;
1992 m_freem(m);
1993 } else {
1994 /* queue length low enough, or made room */
1995 tbf_queue(vifp, m);
1996 tbf_process_q(vifp);
1997 }
1998 }
1999 }
2000
2001 /*
2002 * adds a packet to the queue at the interface
2003 */
2004 static void
2005 tbf_queue(struct vif *vifp, struct mbuf *m)
2006 {
2007 int s = splsoftnet();
2008
2009 /* insert at tail */
2010 *vifp->tbf_t = m;
2011 vifp->tbf_t = &m->m_nextpkt;
2012 vifp->tbf_q_len++;
2013
2014 splx(s);
2015 }
2016
2017
2018 /*
2019 * processes the queue at the interface
2020 */
2021 static void
2022 tbf_process_q(struct vif *vifp)
2023 {
2024 struct mbuf *m;
2025 int len;
2026 int s = splsoftnet();
2027
2028 /*
2029 * Loop through the queue at the interface and send as many packets
2030 * as possible.
2031 */
2032 for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) {
2033 len = ntohs(mtod(m, struct ip *)->ip_len);
2034
2035 /* determine if the packet can be sent */
2036 if (len <= vifp->tbf_n_tok) {
2037 /* if so,
2038 * reduce no of tokens, dequeue the packet,
2039 * send the packet.
2040 */
2041 if ((vifp->tbf_q = m->m_nextpkt) == NULL)
2042 vifp->tbf_t = &vifp->tbf_q;
2043 --vifp->tbf_q_len;
2044
2045 m->m_nextpkt = NULL;
2046 vifp->tbf_n_tok -= len;
2047 tbf_send_packet(vifp, m);
2048 } else
2049 break;
2050 }
2051 splx(s);
2052 }
2053
2054 static void
2055 tbf_reprocess_q(void *arg)
2056 {
2057 struct vif *vifp = arg;
2058
2059 if (ip_mrouter == NULL)
2060 return;
2061
2062 tbf_update_tokens(vifp);
2063 tbf_process_q(vifp);
2064
2065 if (vifp->tbf_q_len != 0)
2066 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2067 tbf_reprocess_q, vifp);
2068 }
2069
2070 /* function that will selectively discard a member of the queue
2071 * based on the precedence value and the priority
2072 */
2073 static int
2074 tbf_dq_sel(struct vif *vifp, struct ip *ip)
2075 {
2076 u_int p;
2077 struct mbuf **mp, *m;
2078 int s = splsoftnet();
2079
2080 p = priority(vifp, ip);
2081
2082 for (mp = &vifp->tbf_q, m = *mp;
2083 m != NULL;
2084 mp = &m->m_nextpkt, m = *mp) {
2085 if (p > priority(vifp, mtod(m, struct ip *))) {
2086 if ((*mp = m->m_nextpkt) == NULL)
2087 vifp->tbf_t = mp;
2088 --vifp->tbf_q_len;
2089
2090 m_freem(m);
2091 mrtstat.mrts_drop_sel++;
2092 splx(s);
2093 return (1);
2094 }
2095 }
2096 splx(s);
2097 return (0);
2098 }
2099
2100 static void
2101 tbf_send_packet(struct vif *vifp, struct mbuf *m)
2102 {
2103 int error;
2104 int s = splsoftnet();
2105
2106 if (vifp->v_flags & VIFF_TUNNEL) {
2107 /* If tunnel options */
2108 ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL);
2109 } else {
2110 /* if physical interface option, extract the options and then send */
2111 struct ip_moptions imo;
2112
2113 imo.imo_multicast_if_index = if_get_index(vifp->v_ifp);
2114 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2115 imo.imo_multicast_loop = 1;
2116 #ifdef RSVP_ISI
2117 imo.imo_multicast_vif = -1;
2118 #endif
2119
2120 error = ip_output(m, NULL, NULL, IP_FORWARDING|IP_MULTICASTOPTS,
2121 &imo, NULL);
2122
2123 if (mrtdebug & DEBUG_XMIT)
2124 log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
2125 (long)(vifp - viftable), error);
2126 }
2127 splx(s);
2128 }
2129
2130 /* determine the current time and then
2131 * the elapsed time (between the last time and time now)
2132 * in milliseconds & update the no. of tokens in the bucket
2133 */
2134 static void
2135 tbf_update_tokens(struct vif *vifp)
2136 {
2137 struct timeval tp;
2138 u_int32_t tm;
2139 int s = splsoftnet();
2140
2141 microtime(&tp);
2142
2143 TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
2144
2145 /*
2146 * This formula is actually
2147 * "time in seconds" * "bytes/second".
2148 *
2149 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2150 *
2151 * The (1000/1024) was introduced in add_vif to optimize
2152 * this divide into a shift.
2153 */
2154 vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
2155 vifp->tbf_last_pkt_t = tp;
2156
2157 if (vifp->tbf_n_tok > MAX_BKT_SIZE)
2158 vifp->tbf_n_tok = MAX_BKT_SIZE;
2159
2160 splx(s);
2161 }
2162
2163 static int
2164 priority(struct vif *vifp, struct ip *ip)
2165 {
2166 int prio = 50; /* the lowest priority -- default case */
2167
2168 /* temporary hack; may add general packet classifier some day */
2169
2170 /*
2171 * The UDP port space is divided up into four priority ranges:
2172 * [0, 16384) : unclassified - lowest priority
2173 * [16384, 32768) : audio - highest priority
2174 * [32768, 49152) : whiteboard - medium priority
2175 * [49152, 65536) : video - low priority
2176 */
2177 if (ip->ip_p == IPPROTO_UDP) {
2178 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2179
2180 switch (ntohs(udp->uh_dport) & 0xc000) {
2181 case 0x4000:
2182 prio = 70;
2183 break;
2184 case 0x8000:
2185 prio = 60;
2186 break;
2187 case 0xc000:
2188 prio = 55;
2189 break;
2190 }
2191
2192 if (tbfdebug > 1)
2193 log(LOG_DEBUG, "port %x prio %d\n",
2194 ntohs(udp->uh_dport), prio);
2195 }
2196
2197 return (prio);
2198 }
2199
2200 /*
2201 * End of token bucket filter modifications
2202 */
2203 #ifdef RSVP_ISI
2204 int
2205 ip_rsvp_vif_init(struct socket *so, struct mbuf *m)
2206 {
2207 int vifi, s;
2208
2209 RSVP_DPRINTF(("%s: so_type = %d, pr_protocol = %d\n", __func__
2210 so->so_type, so->so_proto->pr_protocol));
2211
2212 if (so->so_type != SOCK_RAW ||
2213 so->so_proto->pr_protocol != IPPROTO_RSVP)
2214 return (EOPNOTSUPP);
2215
2216 /* Check mbuf. */
2217 if (m == NULL || m->m_len != sizeof(int)) {
2218 return (EINVAL);
2219 }
2220 vifi = *(mtod(m, int *));
2221
2222 RSVP_DPRINTF(("%s: vif = %d rsvp_on = %d\n", __func__, vifi, rsvp_on));
2223
2224 s = splsoftnet();
2225
2226 /* Check vif. */
2227 if (!legal_vif_num(vifi)) {
2228 splx(s);
2229 return (EADDRNOTAVAIL);
2230 }
2231
2232 /* Check if socket is available. */
2233 if (viftable[vifi].v_rsvpd != NULL) {
2234 splx(s);
2235 return (EADDRINUSE);
2236 }
2237
2238 viftable[vifi].v_rsvpd = so;
2239 /*
2240 * This may seem silly, but we need to be sure we don't over-increment
2241 * the RSVP counter, in case something slips up.
2242 */
2243 if (!viftable[vifi].v_rsvp_on) {
2244 viftable[vifi].v_rsvp_on = 1;
2245 rsvp_on++;
2246 }
2247
2248 splx(s);
2249 return (0);
2250 }
2251
2252 int
2253 ip_rsvp_vif_done(struct socket *so, struct mbuf *m)
2254 {
2255 int vifi, s;
2256
2257 RSVP_DPRINTF(("%s: so_type = %d, pr_protocol = %d\n", __func__,
2258 so->so_type, so->so_proto->pr_protocol));
2259
2260 if (so->so_type != SOCK_RAW ||
2261 so->so_proto->pr_protocol != IPPROTO_RSVP)
2262 return (EOPNOTSUPP);
2263
2264 /* Check mbuf. */
2265 if (m == NULL || m->m_len != sizeof(int)) {
2266 return (EINVAL);
2267 }
2268 vifi = *(mtod(m, int *));
2269
2270 s = splsoftnet();
2271
2272 /* Check vif. */
2273 if (!legal_vif_num(vifi)) {
2274 splx(s);
2275 return (EADDRNOTAVAIL);
2276 }
2277
2278 RSVP_DPRINTF(("%s: v_rsvpd = %x so = %x\n", __func__,
2279 viftable[vifi].v_rsvpd, so));
2280
2281 viftable[vifi].v_rsvpd = NULL;
2282 /*
2283 * This may seem silly, but we need to be sure we don't over-decrement
2284 * the RSVP counter, in case something slips up.
2285 */
2286 if (viftable[vifi].v_rsvp_on) {
2287 viftable[vifi].v_rsvp_on = 0;
2288 rsvp_on--;
2289 }
2290
2291 splx(s);
2292 return (0);
2293 }
2294
2295 void
2296 ip_rsvp_force_done(struct socket *so)
2297 {
2298 int vifi, s;
2299
2300 /* Don't bother if it is not the right type of socket. */
2301 if (so->so_type != SOCK_RAW ||
2302 so->so_proto->pr_protocol != IPPROTO_RSVP)
2303 return;
2304
2305 s = splsoftnet();
2306
2307 /*
2308 * The socket may be attached to more than one vif...this
2309 * is perfectly legal.
2310 */
2311 for (vifi = 0; vifi < numvifs; vifi++) {
2312 if (viftable[vifi].v_rsvpd == so) {
2313 viftable[vifi].v_rsvpd = NULL;
2314 /*
2315 * This may seem silly, but we need to be sure we don't
2316 * over-decrement the RSVP counter, in case something
2317 * slips up.
2318 */
2319 if (viftable[vifi].v_rsvp_on) {
2320 viftable[vifi].v_rsvp_on = 0;
2321 rsvp_on--;
2322 }
2323 }
2324 }
2325
2326 splx(s);
2327 return;
2328 }
2329
2330 void
2331 rsvp_input(struct mbuf *m, struct ifnet *ifp)
2332 {
2333 int vifi, s;
2334 struct ip *ip = mtod(m, struct ip *);
2335 struct sockaddr_in rsvp_src;
2336
2337 RSVP_DPRINTF(("%s: rsvp_on %d\n", __func__, rsvp_on));
2338
2339 /*
2340 * Can still get packets with rsvp_on = 0 if there is a local member
2341 * of the group to which the RSVP packet is addressed. But in this
2342 * case we want to throw the packet away.
2343 */
2344 if (!rsvp_on) {
2345 m_freem(m);
2346 return;
2347 }
2348
2349 /*
2350 * If the old-style non-vif-associated socket is set, then use
2351 * it and ignore the new ones.
2352 */
2353 if (ip_rsvpd != NULL) {
2354 RSVP_DPRINTF(("%s: Sending packet up old-style socket\n",
2355 __func__));
2356 rip_input(m); /*XXX*/
2357 return;
2358 }
2359
2360 s = splsoftnet();
2361
2362 RSVP_DPRINTF(("%s: check vifs\n", __func__));
2363
2364 /* Find which vif the packet arrived on. */
2365 for (vifi = 0; vifi < numvifs; vifi++) {
2366 if (viftable[vifi].v_ifp == ifp)
2367 break;
2368 }
2369
2370 if (vifi == numvifs) {
2371 /* Can't find vif packet arrived on. Drop packet. */
2372 RSVP_DPRINTF("%s: Can't find vif for packet...dropping it.\n",
2373 __func__));
2374 m_freem(m);
2375 splx(s);
2376 return;
2377 }
2378
2379 RSVP_DPRINTF(("%s: check socket\n", __func__));
2380
2381 if (viftable[vifi].v_rsvpd == NULL) {
2382 /*
2383 * drop packet, since there is no specific socket for this
2384 * interface
2385 */
2386 RSVP_DPRINTF(("%s: No socket defined for vif %d\n", __func__,
2387 vifi));
2388 m_freem(m);
2389 splx(s);
2390 return;
2391 }
2392
2393 sockaddr_in_init(&rsvp_src, &ip->ip_src, 0);
2394
2395 if (m)
2396 RSVP_DPRINTF(("%s: m->m_len = %d, sbspace() = %d\n", __func__,
2397 m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv)));
2398
2399 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2400 RSVP_DPRINTF(("%s: Failed to append to socket\n", __func__));
2401 else
2402 RSVP_DPRINTF(("%s: send packet up\n", __func__));
2403
2404 splx(s);
2405 }
2406 #endif /* RSVP_ISI */
2407
2408 /*
2409 * Code for bandwidth monitors
2410 */
2411
2412 /*
2413 * Define common interface for timeval-related methods
2414 */
2415 #define BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp)
2416 #define BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp))
2417 #define BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp))
2418
2419 static uint32_t
2420 compute_bw_meter_flags(struct bw_upcall *req)
2421 {
2422 uint32_t flags = 0;
2423
2424 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2425 flags |= BW_METER_UNIT_PACKETS;
2426 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2427 flags |= BW_METER_UNIT_BYTES;
2428 if (req->bu_flags & BW_UPCALL_GEQ)
2429 flags |= BW_METER_GEQ;
2430 if (req->bu_flags & BW_UPCALL_LEQ)
2431 flags |= BW_METER_LEQ;
2432
2433 return flags;
2434 }
2435
2436 /*
2437 * Add a bw_meter entry
2438 */
2439 static int
2440 add_bw_upcall(struct bw_upcall *req)
2441 {
2442 int s;
2443 struct mfc *mfc;
2444 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2445 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2446 struct timeval now;
2447 struct bw_meter *x;
2448 uint32_t flags;
2449
2450 if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2451 return EOPNOTSUPP;
2452
2453 /* Test if the flags are valid */
2454 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2455 return EINVAL;
2456 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2457 return EINVAL;
2458 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2459 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2460 return EINVAL;
2461
2462 /* Test if the threshold time interval is valid */
2463 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2464 return EINVAL;
2465
2466 flags = compute_bw_meter_flags(req);
2467
2468 /*
2469 * Find if we have already same bw_meter entry
2470 */
2471 s = splsoftnet();
2472 mfc = mfc_find(&req->bu_src, &req->bu_dst);
2473 if (mfc == NULL) {
2474 splx(s);
2475 return EADDRNOTAVAIL;
2476 }
2477 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2478 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2479 &req->bu_threshold.b_time, ==)) &&
2480 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2481 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2482 (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
2483 splx(s);
2484 return 0; /* XXX Already installed */
2485 }
2486 }
2487
2488 /* Allocate the new bw_meter entry */
2489 x = kmem_intr_alloc(sizeof(*x), KM_NOSLEEP);
2490 if (x == NULL) {
2491 splx(s);
2492 return ENOBUFS;
2493 }
2494
2495 /* Set the new bw_meter entry */
2496 x->bm_threshold.b_time = req->bu_threshold.b_time;
2497 microtime(&now);
2498 x->bm_start_time = now;
2499 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2500 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2501 x->bm_measured.b_packets = 0;
2502 x->bm_measured.b_bytes = 0;
2503 x->bm_flags = flags;
2504 x->bm_time_next = NULL;
2505 x->bm_time_hash = BW_METER_BUCKETS;
2506
2507 /* Add the new bw_meter entry to the front of entries for this MFC */
2508 x->bm_mfc = mfc;
2509 x->bm_mfc_next = mfc->mfc_bw_meter;
2510 mfc->mfc_bw_meter = x;
2511 schedule_bw_meter(x, &now);
2512 splx(s);
2513
2514 return 0;
2515 }
2516
2517 static void
2518 free_bw_list(struct bw_meter *list)
2519 {
2520 while (list != NULL) {
2521 struct bw_meter *x = list;
2522
2523 list = list->bm_mfc_next;
2524 unschedule_bw_meter(x);
2525 kmem_free(x, sizeof(*x));
2526 }
2527 }
2528
2529 /*
2530 * Delete one or multiple bw_meter entries
2531 */
2532 static int
2533 del_bw_upcall(struct bw_upcall *req)
2534 {
2535 int s;
2536 struct mfc *mfc;
2537 struct bw_meter *x;
2538
2539 if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2540 return EOPNOTSUPP;
2541
2542 s = splsoftnet();
2543 /* Find the corresponding MFC entry */
2544 mfc = mfc_find(&req->bu_src, &req->bu_dst);
2545 if (mfc == NULL) {
2546 splx(s);
2547 return EADDRNOTAVAIL;
2548 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2549 /*
2550 * Delete all bw_meter entries for this mfc
2551 */
2552 struct bw_meter *list;
2553
2554 list = mfc->mfc_bw_meter;
2555 mfc->mfc_bw_meter = NULL;
2556 free_bw_list(list);
2557 splx(s);
2558 return 0;
2559 } else { /* Delete a single bw_meter entry */
2560 struct bw_meter *prev;
2561 uint32_t flags = 0;
2562
2563 flags = compute_bw_meter_flags(req);
2564
2565 /* Find the bw_meter entry to delete */
2566 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2567 prev = x, x = x->bm_mfc_next) {
2568 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2569 &req->bu_threshold.b_time, ==)) &&
2570 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2571 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2572 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2573 break;
2574 }
2575 if (x != NULL) { /* Delete entry from the list for this MFC */
2576 if (prev != NULL)
2577 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
2578 else
2579 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2580
2581 unschedule_bw_meter(x);
2582 splx(s);
2583 /* Free the bw_meter entry */
2584 kmem_free(x, sizeof(*x));
2585 return 0;
2586 } else {
2587 splx(s);
2588 return EINVAL;
2589 }
2590 }
2591 /* NOTREACHED */
2592 }
2593
2594 /*
2595 * Perform bandwidth measurement processing that may result in an upcall
2596 */
2597 static void
2598 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2599 {
2600 struct timeval delta;
2601
2602 delta = *nowp;
2603 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2604
2605 if (x->bm_flags & BW_METER_GEQ) {
2606 /*
2607 * Processing for ">=" type of bw_meter entry
2608 */
2609 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2610 /* Reset the bw_meter entry */
2611 x->bm_start_time = *nowp;
2612 x->bm_measured.b_packets = 0;
2613 x->bm_measured.b_bytes = 0;
2614 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2615 }
2616
2617 /* Record that a packet is received */
2618 x->bm_measured.b_packets++;
2619 x->bm_measured.b_bytes += plen;
2620
2621 /*
2622 * Test if we should deliver an upcall
2623 */
2624 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2625 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2626 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2627 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2628 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2629 /* Prepare an upcall for delivery */
2630 bw_meter_prepare_upcall(x, nowp);
2631 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2632 }
2633 }
2634 } else if (x->bm_flags & BW_METER_LEQ) {
2635 /*
2636 * Processing for "<=" type of bw_meter entry
2637 */
2638 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2639 /*
2640 * We are behind time with the multicast forwarding table
2641 * scanning for "<=" type of bw_meter entries, so test now
2642 * if we should deliver an upcall.
2643 */
2644 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2645 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2646 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2647 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2648 /* Prepare an upcall for delivery */
2649 bw_meter_prepare_upcall(x, nowp);
2650 }
2651 /* Reschedule the bw_meter entry */
2652 unschedule_bw_meter(x);
2653 schedule_bw_meter(x, nowp);
2654 }
2655
2656 /* Record that a packet is received */
2657 x->bm_measured.b_packets++;
2658 x->bm_measured.b_bytes += plen;
2659
2660 /*
2661 * Test if we should restart the measuring interval
2662 */
2663 if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2664 x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2665 (x->bm_flags & BW_METER_UNIT_BYTES &&
2666 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2667 /* Don't restart the measuring interval */
2668 } else {
2669 /* Do restart the measuring interval */
2670 /*
2671 * XXX: note that we don't unschedule and schedule, because this
2672 * might be too much overhead per packet. Instead, when we process
2673 * all entries for a given timer hash bin, we check whether it is
2674 * really a timeout. If not, we reschedule at that time.
2675 */
2676 x->bm_start_time = *nowp;
2677 x->bm_measured.b_packets = 0;
2678 x->bm_measured.b_bytes = 0;
2679 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2680 }
2681 }
2682 }
2683
2684 /*
2685 * Prepare a bandwidth-related upcall
2686 */
2687 static void
2688 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2689 {
2690 struct timeval delta;
2691 struct bw_upcall *u;
2692
2693 /*
2694 * Compute the measured time interval
2695 */
2696 delta = *nowp;
2697 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2698
2699 /*
2700 * If there are too many pending upcalls, deliver them now
2701 */
2702 if (bw_upcalls_n >= BW_UPCALLS_MAX)
2703 bw_upcalls_send();
2704
2705 /*
2706 * Set the bw_upcall entry
2707 */
2708 u = &bw_upcalls[bw_upcalls_n++];
2709 u->bu_src = x->bm_mfc->mfc_origin;
2710 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2711 u->bu_threshold.b_time = x->bm_threshold.b_time;
2712 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2713 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2714 u->bu_measured.b_time = delta;
2715 u->bu_measured.b_packets = x->bm_measured.b_packets;
2716 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2717 u->bu_flags = 0;
2718 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2719 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2720 if (x->bm_flags & BW_METER_UNIT_BYTES)
2721 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2722 if (x->bm_flags & BW_METER_GEQ)
2723 u->bu_flags |= BW_UPCALL_GEQ;
2724 if (x->bm_flags & BW_METER_LEQ)
2725 u->bu_flags |= BW_UPCALL_LEQ;
2726 }
2727
2728 /*
2729 * Send the pending bandwidth-related upcalls
2730 */
2731 static void
2732 bw_upcalls_send(void)
2733 {
2734 struct mbuf *m;
2735 int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2736 struct sockaddr_in k_igmpsrc = {
2737 .sin_len = sizeof(k_igmpsrc),
2738 .sin_family = AF_INET,
2739 };
2740 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2741 0, /* unused2 */
2742 IGMPMSG_BW_UPCALL,/* im_msgtype */
2743 0, /* im_mbz */
2744 0, /* im_vif */
2745 0, /* unused3 */
2746 { 0 }, /* im_src */
2747 { 0 } }; /* im_dst */
2748
2749 if (bw_upcalls_n == 0)
2750 return; /* No pending upcalls */
2751
2752 bw_upcalls_n = 0;
2753
2754 /*
2755 * Allocate a new mbuf, initialize it with the header and
2756 * the payload for the pending calls.
2757 */
2758 MGETHDR(m, M_DONTWAIT, MT_HEADER);
2759 if (m == NULL) {
2760 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2761 return;
2762 }
2763
2764 m->m_len = m->m_pkthdr.len = 0;
2765 m_copyback(m, 0, sizeof(struct igmpmsg), (void *)&igmpmsg);
2766 m_copyback(m, sizeof(struct igmpmsg), len, (void *)&bw_upcalls[0]);
2767
2768 /*
2769 * Send the upcalls
2770 * XXX do we need to set the address in k_igmpsrc ?
2771 */
2772 mrtstat.mrts_upcalls++;
2773 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2774 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2775 ++mrtstat.mrts_upq_sockfull;
2776 }
2777 }
2778
2779 /*
2780 * Compute the timeout hash value for the bw_meter entries
2781 */
2782 #define BW_METER_TIMEHASH(bw_meter, hash) \
2783 do { \
2784 struct timeval next_timeval = (bw_meter)->bm_start_time; \
2785 \
2786 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2787 (hash) = next_timeval.tv_sec; \
2788 if (next_timeval.tv_usec) \
2789 (hash)++; /* XXX: make sure we don't timeout early */ \
2790 (hash) %= BW_METER_BUCKETS; \
2791 } while (/*CONSTCOND*/ 0)
2792
2793 /*
2794 * Schedule a timer to process periodically bw_meter entry of type "<="
2795 * by linking the entry in the proper hash bucket.
2796 */
2797 static void
2798 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2799 {
2800 int time_hash;
2801
2802 if (!(x->bm_flags & BW_METER_LEQ))
2803 return; /* XXX: we schedule timers only for "<=" entries */
2804
2805 /*
2806 * Reset the bw_meter entry
2807 */
2808 x->bm_start_time = *nowp;
2809 x->bm_measured.b_packets = 0;
2810 x->bm_measured.b_bytes = 0;
2811 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2812
2813 /*
2814 * Compute the timeout hash value and insert the entry
2815 */
2816 BW_METER_TIMEHASH(x, time_hash);
2817 x->bm_time_next = bw_meter_timers[time_hash];
2818 bw_meter_timers[time_hash] = x;
2819 x->bm_time_hash = time_hash;
2820 }
2821
2822 /*
2823 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2824 * by removing the entry from the proper hash bucket.
2825 */
2826 static void
2827 unschedule_bw_meter(struct bw_meter *x)
2828 {
2829 int time_hash;
2830 struct bw_meter *prev, *tmp;
2831
2832 if (!(x->bm_flags & BW_METER_LEQ))
2833 return; /* XXX: we schedule timers only for "<=" entries */
2834
2835 /*
2836 * Compute the timeout hash value and delete the entry
2837 */
2838 time_hash = x->bm_time_hash;
2839 if (time_hash >= BW_METER_BUCKETS)
2840 return; /* Entry was not scheduled */
2841
2842 for (prev = NULL, tmp = bw_meter_timers[time_hash];
2843 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2844 if (tmp == x)
2845 break;
2846
2847 if (tmp == NULL)
2848 panic("unschedule_bw_meter: bw_meter entry not found");
2849
2850 if (prev != NULL)
2851 prev->bm_time_next = x->bm_time_next;
2852 else
2853 bw_meter_timers[time_hash] = x->bm_time_next;
2854
2855 x->bm_time_next = NULL;
2856 x->bm_time_hash = BW_METER_BUCKETS;
2857 }
2858
2859 /*
2860 * Process all "<=" type of bw_meter that should be processed now,
2861 * and for each entry prepare an upcall if necessary. Each processed
2862 * entry is rescheduled again for the (periodic) processing.
2863 *
2864 * This is run periodically (once per second normally). On each round,
2865 * all the potentially matching entries are in the hash slot that we are
2866 * looking at.
2867 */
2868 static void
2869 bw_meter_process(void)
2870 {
2871 int s;
2872 static uint32_t last_tv_sec; /* last time we processed this */
2873
2874 uint32_t loops;
2875 int i;
2876 struct timeval now, process_endtime;
2877
2878 microtime(&now);
2879 if (last_tv_sec == now.tv_sec)
2880 return; /* nothing to do */
2881
2882 loops = now.tv_sec - last_tv_sec;
2883 last_tv_sec = now.tv_sec;
2884 if (loops > BW_METER_BUCKETS)
2885 loops = BW_METER_BUCKETS;
2886
2887 s = splsoftnet();
2888 /*
2889 * Process all bins of bw_meter entries from the one after the last
2890 * processed to the current one. On entry, i points to the last bucket
2891 * visited, so we need to increment i at the beginning of the loop.
2892 */
2893 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2894 struct bw_meter *x, *tmp_list;
2895
2896 if (++i >= BW_METER_BUCKETS)
2897 i = 0;
2898
2899 /* Disconnect the list of bw_meter entries from the bin */
2900 tmp_list = bw_meter_timers[i];
2901 bw_meter_timers[i] = NULL;
2902
2903 /* Process the list of bw_meter entries */
2904 while (tmp_list != NULL) {
2905 x = tmp_list;
2906 tmp_list = tmp_list->bm_time_next;
2907
2908 /* Test if the time interval is over */
2909 process_endtime = x->bm_start_time;
2910 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2911 if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2912 /* Not yet: reschedule, but don't reset */
2913 int time_hash;
2914
2915 BW_METER_TIMEHASH(x, time_hash);
2916 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2917 /*
2918 * XXX: somehow the bin processing is a bit ahead of time.
2919 * Put the entry in the next bin.
2920 */
2921 if (++time_hash >= BW_METER_BUCKETS)
2922 time_hash = 0;
2923 }
2924 x->bm_time_next = bw_meter_timers[time_hash];
2925 bw_meter_timers[time_hash] = x;
2926 x->bm_time_hash = time_hash;
2927
2928 continue;
2929 }
2930
2931 /*
2932 * Test if we should deliver an upcall
2933 */
2934 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2935 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2936 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2937 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2938 /* Prepare an upcall for delivery */
2939 bw_meter_prepare_upcall(x, &now);
2940 }
2941
2942 /*
2943 * Reschedule for next processing
2944 */
2945 schedule_bw_meter(x, &now);
2946 }
2947 }
2948
2949 /* Send all upcalls that are pending delivery */
2950 bw_upcalls_send();
2951
2952 splx(s);
2953 }
2954
2955 /*
2956 * A periodic function for sending all upcalls that are pending delivery
2957 */
2958 static void
2959 expire_bw_upcalls_send(void *unused)
2960 {
2961 int s;
2962
2963 s = splsoftnet();
2964 bw_upcalls_send();
2965 splx(s);
2966
2967 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
2968 expire_bw_upcalls_send, NULL);
2969 }
2970
2971 /*
2972 * A periodic function for periodic scanning of the multicast forwarding
2973 * table for processing all "<=" bw_meter entries.
2974 */
2975 static void
2976 expire_bw_meter_process(void *unused)
2977 {
2978 if (mrt_api_config & MRT_MFC_BW_UPCALL)
2979 bw_meter_process();
2980
2981 callout_reset(&bw_meter_ch, BW_METER_PERIOD,
2982 expire_bw_meter_process, NULL);
2983 }
2984
2985 /*
2986 * End of bandwidth monitoring code
2987 */
2988
2989 #ifdef PIM
2990 /*
2991 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2992 */
2993 static int
2994 pim_register_send(struct ip *ip, struct vif *vifp,
2995 struct mbuf *m, struct mfc *rt)
2996 {
2997 struct mbuf *mb_copy, *mm;
2998
2999 if (mrtdebug & DEBUG_PIM)
3000 log(LOG_DEBUG, "pim_register_send: \n");
3001
3002 mb_copy = pim_register_prepare(ip, m);
3003 if (mb_copy == NULL)
3004 return ENOBUFS;
3005
3006 /*
3007 * Send all the fragments. Note that the mbuf for each fragment
3008 * is freed by the sending machinery.
3009 */
3010 for (mm = mb_copy; mm; mm = mb_copy) {
3011 mb_copy = mm->m_nextpkt;
3012 mm->m_nextpkt = NULL;
3013 mm = m_pullup(mm, sizeof(struct ip));
3014 if (mm != NULL) {
3015 ip = mtod(mm, struct ip *);
3016 if ((mrt_api_config & MRT_MFC_RP) &&
3017 !in_nullhost(rt->mfc_rp)) {
3018 pim_register_send_rp(ip, vifp, mm, rt);
3019 } else {
3020 pim_register_send_upcall(ip, vifp, mm, rt);
3021 }
3022 }
3023 }
3024
3025 return 0;
3026 }
3027
3028 /*
3029 * Return a copy of the data packet that is ready for PIM Register
3030 * encapsulation.
3031 * XXX: Note that in the returned copy the IP header is a valid one.
3032 */
3033 static struct mbuf *
3034 pim_register_prepare(struct ip *ip, struct mbuf *m)
3035 {
3036 struct mbuf *mb_copy = NULL;
3037 int mtu;
3038
3039 /* Take care of delayed checksums */
3040 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3041 in_delayed_cksum(m);
3042 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3043 }
3044
3045 /*
3046 * Copy the old packet & pullup its IP header into the
3047 * new mbuf so we can modify it.
3048 */
3049 mb_copy = m_copypacket(m, M_DONTWAIT);
3050 if (mb_copy == NULL)
3051 return NULL;
3052 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3053 if (mb_copy == NULL)
3054 return NULL;
3055
3056 /* take care of the TTL */
3057 ip = mtod(mb_copy, struct ip *);
3058 --ip->ip_ttl;
3059
3060 /* Compute the MTU after the PIM Register encapsulation */
3061 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3062
3063 if (ntohs(ip->ip_len) <= mtu) {
3064 /* Turn the IP header into a valid one */
3065 ip->ip_sum = 0;
3066 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3067 } else {
3068 /* Fragment the packet */
3069 if (ip_fragment(mb_copy, NULL, mtu) != 0) {
3070 /* XXX: mb_copy was freed by ip_fragment() */
3071 return NULL;
3072 }
3073 }
3074 return mb_copy;
3075 }
3076
3077 /*
3078 * Send an upcall with the data packet to the user-level process.
3079 */
3080 static int
3081 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3082 struct mbuf *mb_copy, struct mfc *rt)
3083 {
3084 struct mbuf *mb_first;
3085 int len = ntohs(ip->ip_len);
3086 struct igmpmsg *im;
3087 struct sockaddr_in k_igmpsrc = {
3088 .sin_len = sizeof(k_igmpsrc),
3089 .sin_family = AF_INET,
3090 };
3091
3092 /*
3093 * Add a new mbuf with an upcall header
3094 */
3095 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3096 if (mb_first == NULL) {
3097 m_freem(mb_copy);
3098 return ENOBUFS;
3099 }
3100 mb_first->m_data += max_linkhdr;
3101 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3102 mb_first->m_len = sizeof(struct igmpmsg);
3103 mb_first->m_next = mb_copy;
3104
3105 /* Send message to routing daemon */
3106 im = mtod(mb_first, struct igmpmsg *);
3107 im->im_msgtype = IGMPMSG_WHOLEPKT;
3108 im->im_mbz = 0;
3109 im->im_vif = vifp - viftable;
3110 im->im_src = ip->ip_src;
3111 im->im_dst = ip->ip_dst;
3112
3113 k_igmpsrc.sin_addr = ip->ip_src;
3114
3115 mrtstat.mrts_upcalls++;
3116
3117 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3118 if (mrtdebug & DEBUG_PIM)
3119 log(LOG_WARNING,
3120 "mcast: pim_register_send_upcall: ip_mrouter socket queue full\n");
3121 ++mrtstat.mrts_upq_sockfull;
3122 return ENOBUFS;
3123 }
3124
3125 /* Keep statistics */
3126 pimstat.pims_snd_registers_msgs++;
3127 pimstat.pims_snd_registers_bytes += len;
3128
3129 return 0;
3130 }
3131
3132 /*
3133 * Encapsulate the data packet in PIM Register message and send it to the RP.
3134 */
3135 static int
3136 pim_register_send_rp(struct ip *ip, struct vif *vifp,
3137 struct mbuf *mb_copy, struct mfc *rt)
3138 {
3139 struct mbuf *mb_first;
3140 struct ip *ip_outer;
3141 struct pim_encap_pimhdr *pimhdr;
3142 int len = ntohs(ip->ip_len);
3143 vifi_t vifi = rt->mfc_parent;
3144
3145 if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
3146 m_freem(mb_copy);
3147 return EADDRNOTAVAIL; /* The iif vif is invalid */
3148 }
3149
3150 /*
3151 * Add a new mbuf with the encapsulating header
3152 */
3153 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3154 if (mb_first == NULL) {
3155 m_freem(mb_copy);
3156 return ENOBUFS;
3157 }
3158 mb_first->m_data += max_linkhdr;
3159 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3160 mb_first->m_next = mb_copy;
3161
3162 mb_first->m_pkthdr.len = len + mb_first->m_len;
3163
3164 /*
3165 * Fill in the encapsulating IP and PIM header
3166 */
3167 ip_outer = mtod(mb_first, struct ip *);
3168 *ip_outer = pim_encap_iphdr;
3169 if (mb_first->m_pkthdr.len < IP_MINFRAGSIZE)
3170 ip_outer->ip_id = 0;
3171 else
3172 ip_outer->ip_id = ip_newid(NULL);
3173 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
3174 sizeof(pim_encap_pimhdr));
3175 ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3176 ip_outer->ip_dst = rt->mfc_rp;
3177 /*
3178 * Copy the inner header TOS to the outer header, and take care of the
3179 * IP_DF bit.
3180 */
3181 ip_outer->ip_tos = ip->ip_tos;
3182 if (ntohs(ip->ip_off) & IP_DF)
3183 ip_outer->ip_off |= htons(IP_DF);
3184 pimhdr = (struct pim_encap_pimhdr *)((char *)ip_outer
3185 + sizeof(pim_encap_iphdr));
3186 *pimhdr = pim_encap_pimhdr;
3187 /* If the iif crosses a border, set the Border-bit */
3188 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3189 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3190
3191 mb_first->m_data += sizeof(pim_encap_iphdr);
3192 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3193 mb_first->m_data -= sizeof(pim_encap_iphdr);
3194
3195 if (vifp->v_rate_limit == 0)
3196 tbf_send_packet(vifp, mb_first);
3197 else
3198 tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len));
3199
3200 /* Keep statistics */
3201 pimstat.pims_snd_registers_msgs++;
3202 pimstat.pims_snd_registers_bytes += len;
3203
3204 return 0;
3205 }
3206
3207 /*
3208 * PIM-SMv2 and PIM-DM messages processing.
3209 * Receives and verifies the PIM control messages, and passes them
3210 * up to the listening socket, using rip_input().
3211 * The only message with special processing is the PIM_REGISTER message
3212 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3213 * is passed to if_simloop().
3214 */
3215 void
3216 pim_input(struct mbuf *m, ...)
3217 {
3218 struct ip *ip = mtod(m, struct ip *);
3219 struct pim *pim;
3220 int minlen;
3221 int datalen;
3222 int ip_tos;
3223 int proto;
3224 int iphlen;
3225 va_list ap;
3226
3227 va_start(ap, m);
3228 iphlen = va_arg(ap, int);
3229 proto = va_arg(ap, int);
3230 va_end(ap);
3231
3232 datalen = ntohs(ip->ip_len) - iphlen;
3233
3234 /* Keep statistics */
3235 pimstat.pims_rcv_total_msgs++;
3236 pimstat.pims_rcv_total_bytes += datalen;
3237
3238 /*
3239 * Validate lengths
3240 */
3241 if (datalen < PIM_MINLEN) {
3242 pimstat.pims_rcv_tooshort++;
3243 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3244 datalen, (u_long)ip->ip_src.s_addr);
3245 m_freem(m);
3246 return;
3247 }
3248
3249 /*
3250 * If the packet is at least as big as a REGISTER, go agead
3251 * and grab the PIM REGISTER header size, to avoid another
3252 * possible m_pullup() later.
3253 *
3254 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
3255 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3256 */
3257 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3258 /*
3259 * Get the IP and PIM headers in contiguous memory, and
3260 * possibly the PIM REGISTER header.
3261 */
3262 if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3263 (m = m_pullup(m, minlen)) == NULL) {
3264 log(LOG_ERR, "pim_input: m_pullup failure\n");
3265 return;
3266 }
3267 /* m_pullup() may have given us a new mbuf so reset ip. */
3268 ip = mtod(m, struct ip *);
3269 ip_tos = ip->ip_tos;
3270
3271 /* adjust mbuf to point to the PIM header */
3272 m->m_data += iphlen;
3273 m->m_len -= iphlen;
3274 pim = mtod(m, struct pim *);
3275
3276 /*
3277 * Validate checksum. If PIM REGISTER, exclude the data packet.
3278 *
3279 * XXX: some older PIMv2 implementations don't make this distinction,
3280 * so for compatibility reason perform the checksum over part of the
3281 * message, and if error, then over the whole message.
3282 */
3283 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3284 /* do nothing, checksum okay */
3285 } else if (in_cksum(m, datalen)) {
3286 pimstat.pims_rcv_badsum++;
3287 if (mrtdebug & DEBUG_PIM)
3288 log(LOG_DEBUG, "pim_input: invalid checksum\n");
3289 m_freem(m);
3290 return;
3291 }
3292
3293 /* PIM version check */
3294 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3295 pimstat.pims_rcv_badversion++;
3296 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3297 PIM_VT_V(pim->pim_vt), PIM_VERSION);
3298 m_freem(m);
3299 return;
3300 }
3301
3302 /* restore mbuf back to the outer IP */
3303 m->m_data -= iphlen;
3304 m->m_len += iphlen;
3305
3306 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3307 /*
3308 * Since this is a REGISTER, we'll make a copy of the register
3309 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3310 * routing daemon.
3311 */
3312 int s;
3313 struct sockaddr_in dst = {
3314 .sin_len = sizeof(dst),
3315 .sin_family = AF_INET,
3316 };
3317 struct mbuf *mcp;
3318 struct ip *encap_ip;
3319 u_int32_t *reghdr;
3320 struct ifnet *vifp;
3321
3322 s = splsoftnet();
3323 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3324 splx(s);
3325 if (mrtdebug & DEBUG_PIM)
3326 log(LOG_DEBUG,
3327 "pim_input: register vif not set: %d\n", reg_vif_num);
3328 m_freem(m);
3329 return;
3330 }
3331 /* XXX need refcnt? */
3332 vifp = viftable[reg_vif_num].v_ifp;
3333 splx(s);
3334
3335 /*
3336 * Validate length
3337 */
3338 if (datalen < PIM_REG_MINLEN) {
3339 pimstat.pims_rcv_tooshort++;
3340 pimstat.pims_rcv_badregisters++;
3341 log(LOG_ERR,
3342 "pim_input: register packet size too small %d from %lx\n",
3343 datalen, (u_long)ip->ip_src.s_addr);
3344 m_freem(m);
3345 return;
3346 }
3347
3348 reghdr = (u_int32_t *)(pim + 1);
3349 encap_ip = (struct ip *)(reghdr + 1);
3350
3351 if (mrtdebug & DEBUG_PIM) {
3352 log(LOG_DEBUG,
3353 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3354 (u_long)ntohl(encap_ip->ip_src.s_addr),
3355 (u_long)ntohl(encap_ip->ip_dst.s_addr),
3356 ntohs(encap_ip->ip_len));
3357 }
3358
3359 /* verify the version number of the inner packet */
3360 if (encap_ip->ip_v != IPVERSION) {
3361 pimstat.pims_rcv_badregisters++;
3362 if (mrtdebug & DEBUG_PIM) {
3363 log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3364 "of the inner packet\n", encap_ip->ip_v);
3365 }
3366 m_freem(m);
3367 return;
3368 }
3369
3370 /* verify the inner packet is destined to a mcast group */
3371 if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) {
3372 pimstat.pims_rcv_badregisters++;
3373 if (mrtdebug & DEBUG_PIM)
3374 log(LOG_DEBUG,
3375 "pim_input: inner packet of register is not "
3376 "multicast %lx\n",
3377 (u_long)ntohl(encap_ip->ip_dst.s_addr));
3378 m_freem(m);
3379 return;
3380 }
3381
3382 /* If a NULL_REGISTER, pass it to the daemon */
3383 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3384 goto pim_input_to_daemon;
3385
3386 /*
3387 * Copy the TOS from the outer IP header to the inner IP header.
3388 */
3389 if (encap_ip->ip_tos != ip_tos) {
3390 /* Outer TOS -> inner TOS */
3391 encap_ip->ip_tos = ip_tos;
3392 /* Recompute the inner header checksum. Sigh... */
3393
3394 /* adjust mbuf to point to the inner IP header */
3395 m->m_data += (iphlen + PIM_MINLEN);
3396 m->m_len -= (iphlen + PIM_MINLEN);
3397
3398 encap_ip->ip_sum = 0;
3399 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3400
3401 /* restore mbuf to point back to the outer IP header */
3402 m->m_data -= (iphlen + PIM_MINLEN);
3403 m->m_len += (iphlen + PIM_MINLEN);
3404 }
3405
3406 /*
3407 * Decapsulate the inner IP packet and loopback to forward it
3408 * as a normal multicast packet. Also, make a copy of the
3409 * outer_iphdr + pimhdr + reghdr + encap_iphdr
3410 * to pass to the daemon later, so it can take the appropriate
3411 * actions (e.g., send back PIM_REGISTER_STOP).
3412 * XXX: here m->m_data points to the outer IP header.
3413 */
3414 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_DONTWAIT);
3415 if (mcp == NULL) {
3416 log(LOG_ERR,
3417 "pim_input: pim register: could not copy register head\n");
3418 m_freem(m);
3419 return;
3420 }
3421
3422 /* Keep statistics */
3423 /* XXX: registers_bytes include only the encap. mcast pkt */
3424 pimstat.pims_rcv_registers_msgs++;
3425 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3426
3427 /*
3428 * forward the inner ip packet; point m_data at the inner ip.
3429 */
3430 m_adj(m, iphlen + PIM_MINLEN);
3431
3432 if (mrtdebug & DEBUG_PIM) {
3433 log(LOG_DEBUG,
3434 "pim_input: forwarding decapsulated register: "
3435 "src %lx, dst %lx, vif %d\n",
3436 (u_long)ntohl(encap_ip->ip_src.s_addr),
3437 (u_long)ntohl(encap_ip->ip_dst.s_addr),
3438 reg_vif_num);
3439 }
3440 /* NB: vifp was collected above; can it change on us? */
3441 looutput(vifp, m, (struct sockaddr *)&dst, NULL);
3442
3443 /* prepare the register head to send to the mrouting daemon */
3444 m = mcp;
3445 }
3446
3447 pim_input_to_daemon:
3448 /*
3449 * Pass the PIM message up to the daemon; if it is a Register message,
3450 * pass the 'head' only up to the daemon. This includes the
3451 * outer IP header, PIM header, PIM-Register header and the
3452 * inner IP header.
3453 * XXX: the outer IP header pkt size of a Register is not adjust to
3454 * reflect the fact that the inner multicast data is truncated.
3455 */
3456 rip_input(m, iphlen, proto);
3457
3458 return;
3459 }
3460 #endif /* PIM */
3461