1 /* $NetBSD: ip_mroute.c,v 1.116 2008/10/01 16:01:51 rmind 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 1.116 2008/10/01 16:01:51 rmind Exp $");
97
98 #include "opt_inet.h"
99 #include "opt_ipsec.h"
100 #include "opt_pim.h"
101
102 #ifdef PIM
103 #define _PIM_VT 1
104 #endif
105
106 #include <sys/param.h>
107 #include <sys/systm.h>
108 #include <sys/callout.h>
109 #include <sys/mbuf.h>
110 #include <sys/socket.h>
111 #include <sys/socketvar.h>
112 #include <sys/protosw.h>
113 #include <sys/errno.h>
114 #include <sys/time.h>
115 #include <sys/kernel.h>
116 #include <sys/ioctl.h>
117 #include <sys/syslog.h>
118
119 #include <net/if.h>
120 #include <net/route.h>
121 #include <net/raw_cb.h>
122
123 #include <netinet/in.h>
124 #include <netinet/in_var.h>
125 #include <netinet/in_systm.h>
126 #include <netinet/ip.h>
127 #include <netinet/ip_var.h>
128 #include <netinet/in_pcb.h>
129 #include <netinet/udp.h>
130 #include <netinet/igmp.h>
131 #include <netinet/igmp_var.h>
132 #include <netinet/ip_mroute.h>
133 #ifdef PIM
134 #include <netinet/pim.h>
135 #include <netinet/pim_var.h>
136 #endif
137 #include <netinet/ip_encap.h>
138
139 #ifdef IPSEC
140 #include <netinet6/ipsec.h>
141 #include <netkey/key.h>
142 #endif
143
144 #ifdef FAST_IPSEC
145 #include <netipsec/ipsec.h>
146 #include <netipsec/key.h>
147 #endif
148
149 #include <machine/stdarg.h>
150
151 #define IP_MULTICASTOPTS 0
152 #define M_PULLUP(m, len) \
153 do { \
154 if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \
155 (m) = m_pullup((m), (len)); \
156 } while (/*CONSTCOND*/ 0)
157
158 /*
159 * Globals. All but ip_mrouter and ip_mrtproto could be static,
160 * except for netstat or debugging purposes.
161 */
162 struct socket *ip_mrouter = NULL;
163 int ip_mrtproto = IGMP_DVMRP; /* for netstat only */
164
165 #define NO_RTE_FOUND 0x1
166 #define RTE_FOUND 0x2
167
168 #define MFCHASH(a, g) \
169 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
170 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
171 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
172 u_long mfchash;
173
174 u_char nexpire[MFCTBLSIZ];
175 struct vif viftable[MAXVIFS];
176 struct mrtstat mrtstat;
177 u_int mrtdebug = 0; /* debug level */
178 #define DEBUG_MFC 0x02
179 #define DEBUG_FORWARD 0x04
180 #define DEBUG_EXPIRE 0x08
181 #define DEBUG_XMIT 0x10
182 #define DEBUG_PIM 0x20
183
184 #define VIFI_INVALID ((vifi_t) -1)
185
186 u_int tbfdebug = 0; /* tbf debug level */
187 #ifdef RSVP_ISI
188 u_int rsvpdebug = 0; /* rsvp debug level */
189 extern struct socket *ip_rsvpd;
190 extern int rsvp_on;
191 #endif /* RSVP_ISI */
192
193 /* vif attachment using sys/netinet/ip_encap.c */
194 static void vif_input(struct mbuf *, ...);
195 static int vif_encapcheck(struct mbuf *, int, int, void *);
196
197 static const struct protosw vif_protosw =
198 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR,
199 vif_input, rip_output, 0, rip_ctloutput,
200 rip_usrreq,
201 0, 0, 0, 0,
202 };
203
204 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
205 #define UPCALL_EXPIRE 6 /* number of timeouts */
206
207 /*
208 * Define the token bucket filter structures
209 */
210
211 #define TBF_REPROCESS (hz / 100) /* 100x / second */
212
213 static int get_sg_cnt(struct sioc_sg_req *);
214 static int get_vif_cnt(struct sioc_vif_req *);
215 static int ip_mrouter_init(struct socket *, int);
216 static int set_assert(int);
217 static int add_vif(struct vifctl *);
218 static int del_vif(vifi_t *);
219 static void update_mfc_params(struct mfc *, struct mfcctl2 *);
220 static void init_mfc_params(struct mfc *, struct mfcctl2 *);
221 static void expire_mfc(struct mfc *);
222 static int add_mfc(struct sockopt *);
223 #ifdef UPCALL_TIMING
224 static void collate(struct timeval *);
225 #endif
226 static int del_mfc(struct sockopt *);
227 static int set_api_config(struct sockopt *); /* chose API capabilities */
228 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
229 static void expire_upcalls(void *);
230 #ifdef RSVP_ISI
231 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
232 #else
233 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *);
234 #endif
235 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
236 static void encap_send(struct ip *, struct vif *, struct mbuf *);
237 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_int32_t);
238 static void tbf_queue(struct vif *, struct mbuf *);
239 static void tbf_process_q(struct vif *);
240 static void tbf_reprocess_q(void *);
241 static int tbf_dq_sel(struct vif *, struct ip *);
242 static void tbf_send_packet(struct vif *, struct mbuf *);
243 static void tbf_update_tokens(struct vif *);
244 static int priority(struct vif *, struct ip *);
245
246 /*
247 * Bandwidth monitoring
248 */
249 static void free_bw_list(struct bw_meter *);
250 static int add_bw_upcall(struct bw_upcall *);
251 static int del_bw_upcall(struct bw_upcall *);
252 static void bw_meter_receive_packet(struct bw_meter *, int , struct timeval *);
253 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
254 static void bw_upcalls_send(void);
255 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
256 static void unschedule_bw_meter(struct bw_meter *);
257 static void bw_meter_process(void);
258 static void expire_bw_upcalls_send(void *);
259 static void expire_bw_meter_process(void *);
260
261 #ifdef PIM
262 static int pim_register_send(struct ip *, struct vif *,
263 struct mbuf *, struct mfc *);
264 static int pim_register_send_rp(struct ip *, struct vif *,
265 struct mbuf *, struct mfc *);
266 static int pim_register_send_upcall(struct ip *, struct vif *,
267 struct mbuf *, struct mfc *);
268 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
269 #endif
270
271 /*
272 * 'Interfaces' associated with decapsulator (so we can tell
273 * packets that went through it from ones that get reflected
274 * by a broken gateway). These interfaces are never linked into
275 * the system ifnet list & no routes point to them. I.e., packets
276 * can't be sent this way. They only exist as a placeholder for
277 * multicast source verification.
278 */
279 #if 0
280 struct ifnet multicast_decap_if[MAXVIFS];
281 #endif
282
283 #define ENCAP_TTL 64
284 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */
285
286 /* prototype IP hdr for encapsulated packets */
287 struct ip multicast_encap_iphdr = {
288 .ip_hl = sizeof(struct ip) >> 2,
289 .ip_v = IPVERSION,
290 .ip_len = sizeof(struct ip),
291 .ip_ttl = ENCAP_TTL,
292 .ip_p = ENCAP_PROTO,
293 };
294
295 /*
296 * Bandwidth meter variables and constants
297 */
298
299 /*
300 * Pending timeouts are stored in a hash table, the key being the
301 * expiration time. Periodically, the entries are analysed and processed.
302 */
303 #define BW_METER_BUCKETS 1024
304 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
305 struct callout bw_meter_ch;
306 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
307
308 /*
309 * Pending upcalls are stored in a vector which is flushed when
310 * full, or periodically
311 */
312 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX];
313 static u_int bw_upcalls_n; /* # of pending upcalls */
314 struct callout bw_upcalls_ch;
315 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
316
317 #ifdef PIM
318 struct pimstat pimstat;
319
320 /*
321 * Note: the PIM Register encapsulation adds the following in front of a
322 * data packet:
323 *
324 * struct pim_encap_hdr {
325 * struct ip ip;
326 * struct pim_encap_pimhdr pim;
327 * }
328 *
329 */
330
331 struct pim_encap_pimhdr {
332 struct pim pim;
333 uint32_t flags;
334 };
335
336 static struct ip pim_encap_iphdr = {
337 .ip_v = IPVERSION,
338 .ip_hl = sizeof(struct ip) >> 2,
339 .ip_len = sizeof(struct ip),
340 .ip_ttl = ENCAP_TTL,
341 .ip_p = IPPROTO_PIM,
342 };
343
344 static struct pim_encap_pimhdr pim_encap_pimhdr = {
345 {
346 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
347 0, /* reserved */
348 0, /* checksum */
349 },
350 0 /* flags */
351 };
352
353 static struct ifnet multicast_register_if;
354 static vifi_t reg_vif_num = VIFI_INVALID;
355 #endif /* PIM */
356
357
358 /*
359 * Private variables.
360 */
361 static vifi_t numvifs = 0;
362
363 static struct callout expire_upcalls_ch;
364
365 /*
366 * whether or not special PIM assert processing is enabled.
367 */
368 static int pim_assert;
369 /*
370 * Rate limit for assert notification messages, in usec
371 */
372 #define ASSERT_MSG_TIME 3000000
373
374 /*
375 * Kernel multicast routing API capabilities and setup.
376 * If more API capabilities are added to the kernel, they should be
377 * recorded in `mrt_api_support'.
378 */
379 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
380 MRT_MFC_FLAGS_BORDER_VIF |
381 MRT_MFC_RP |
382 MRT_MFC_BW_UPCALL);
383 static u_int32_t mrt_api_config = 0;
384
385 /*
386 * Find a route for a given origin IP address and Multicast group address
387 * Type of service parameter to be added in the future!!!
388 * Statistics are updated by the caller if needed
389 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
390 */
391 static struct mfc *
392 mfc_find(struct in_addr *o, struct in_addr *g)
393 {
394 struct mfc *rt;
395
396 LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
397 if (in_hosteq(rt->mfc_origin, *o) &&
398 in_hosteq(rt->mfc_mcastgrp, *g) &&
399 (rt->mfc_stall == NULL))
400 break;
401 }
402
403 return (rt);
404 }
405
406 /*
407 * Macros to compute elapsed time efficiently
408 * Borrowed from Van Jacobson's scheduling code
409 */
410 #define TV_DELTA(a, b, delta) do { \
411 int xxs; \
412 delta = (a).tv_usec - (b).tv_usec; \
413 xxs = (a).tv_sec - (b).tv_sec; \
414 switch (xxs) { \
415 case 2: \
416 delta += 1000000; \
417 /* fall through */ \
418 case 1: \
419 delta += 1000000; \
420 /* fall through */ \
421 case 0: \
422 break; \
423 default: \
424 delta += (1000000 * xxs); \
425 break; \
426 } \
427 } while (/*CONSTCOND*/ 0)
428
429 #ifdef UPCALL_TIMING
430 u_int32_t upcall_data[51];
431 #endif /* UPCALL_TIMING */
432
433 /*
434 * Handle MRT setsockopt commands to modify the multicast routing tables.
435 */
436 int
437 ip_mrouter_set(struct socket *so, struct sockopt *sopt)
438 {
439 int error;
440 int optval;
441 struct vifctl vifc;
442 vifi_t vifi;
443 struct bw_upcall bwuc;
444
445 if (sopt->sopt_name != MRT_INIT && so != ip_mrouter)
446 error = ENOPROTOOPT;
447 else {
448 switch (sopt->sopt_name) {
449 case MRT_INIT:
450 error = sockopt_getint(sopt, &optval);
451 if (error)
452 break;
453
454 error = ip_mrouter_init(so, optval);
455 break;
456 case MRT_DONE:
457 error = ip_mrouter_done();
458 break;
459 case MRT_ADD_VIF:
460 error = sockopt_get(sopt, &vifc, sizeof(vifc));
461 if (error)
462 break;
463 error = add_vif(&vifc);
464 break;
465 case MRT_DEL_VIF:
466 error = sockopt_get(sopt, &vifi, sizeof(vifi));
467 if (error)
468 break;
469 error = del_vif(&vifi);
470 break;
471 case MRT_ADD_MFC:
472 error = add_mfc(sopt);
473 break;
474 case MRT_DEL_MFC:
475 error = del_mfc(sopt);
476 break;
477 case MRT_ASSERT:
478 error = sockopt_getint(sopt, &optval);
479 if (error)
480 break;
481 error = set_assert(optval);
482 break;
483 case MRT_API_CONFIG:
484 error = set_api_config(sopt);
485 break;
486 case MRT_ADD_BW_UPCALL:
487 error = sockopt_get(sopt, &bwuc, sizeof(bwuc));
488 if (error)
489 break;
490 error = add_bw_upcall(&bwuc);
491 break;
492 case MRT_DEL_BW_UPCALL:
493 error = sockopt_get(sopt, &bwuc, sizeof(bwuc));
494 if (error)
495 break;
496 error = del_bw_upcall(&bwuc);
497 break;
498 default:
499 error = ENOPROTOOPT;
500 break;
501 }
502 }
503 return (error);
504 }
505
506 /*
507 * Handle MRT getsockopt commands
508 */
509 int
510 ip_mrouter_get(struct socket *so, struct sockopt *sopt)
511 {
512 int error;
513
514 if (so != ip_mrouter)
515 error = ENOPROTOOPT;
516 else {
517 switch (sopt->sopt_name) {
518 case MRT_VERSION:
519 error = sockopt_setint(sopt, 0x0305); /* XXX !!!! */
520 break;
521 case MRT_ASSERT:
522 error = sockopt_setint(sopt, pim_assert);
523 break;
524 case MRT_API_SUPPORT:
525 error = sockopt_set(sopt, &mrt_api_support,
526 sizeof(mrt_api_support));
527 break;
528 case MRT_API_CONFIG:
529 error = sockopt_set(sopt, &mrt_api_config,
530 sizeof(mrt_api_config));
531 break;
532 default:
533 error = ENOPROTOOPT;
534 break;
535 }
536 }
537 return (error);
538 }
539
540 /*
541 * Handle ioctl commands to obtain information from the cache
542 */
543 int
544 mrt_ioctl(struct socket *so, u_long cmd, void *data)
545 {
546 int error;
547
548 if (so != ip_mrouter)
549 error = EINVAL;
550 else
551 switch (cmd) {
552 case SIOCGETVIFCNT:
553 error = get_vif_cnt((struct sioc_vif_req *)data);
554 break;
555 case SIOCGETSGCNT:
556 error = get_sg_cnt((struct sioc_sg_req *)data);
557 break;
558 default:
559 error = EINVAL;
560 break;
561 }
562
563 return (error);
564 }
565
566 /*
567 * returns the packet, byte, rpf-failure count for the source group provided
568 */
569 static int
570 get_sg_cnt(struct sioc_sg_req *req)
571 {
572 int s;
573 struct mfc *rt;
574
575 s = splsoftnet();
576 rt = mfc_find(&req->src, &req->grp);
577 if (rt == NULL) {
578 splx(s);
579 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
580 return (EADDRNOTAVAIL);
581 }
582 req->pktcnt = rt->mfc_pkt_cnt;
583 req->bytecnt = rt->mfc_byte_cnt;
584 req->wrong_if = rt->mfc_wrong_if;
585 splx(s);
586
587 return (0);
588 }
589
590 /*
591 * returns the input and output packet and byte counts on the vif provided
592 */
593 static int
594 get_vif_cnt(struct sioc_vif_req *req)
595 {
596 vifi_t vifi = req->vifi;
597
598 if (vifi >= numvifs)
599 return (EINVAL);
600
601 req->icount = viftable[vifi].v_pkt_in;
602 req->ocount = viftable[vifi].v_pkt_out;
603 req->ibytes = viftable[vifi].v_bytes_in;
604 req->obytes = viftable[vifi].v_bytes_out;
605
606 return (0);
607 }
608
609 /*
610 * Enable multicast routing
611 */
612 static int
613 ip_mrouter_init(struct socket *so, int v)
614 {
615 if (mrtdebug)
616 log(LOG_DEBUG,
617 "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
618 so->so_type, so->so_proto->pr_protocol);
619
620 if (so->so_type != SOCK_RAW ||
621 so->so_proto->pr_protocol != IPPROTO_IGMP)
622 return (EOPNOTSUPP);
623
624 if (v != 1)
625 return (EINVAL);
626
627 if (ip_mrouter != NULL)
628 return (EADDRINUSE);
629
630 ip_mrouter = so;
631
632 mfchashtbl = hashinit(MFCTBLSIZ, HASH_LIST, true, &mfchash);
633 bzero((void *)nexpire, sizeof(nexpire));
634
635 pim_assert = 0;
636
637 callout_init(&expire_upcalls_ch, 0);
638 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
639 expire_upcalls, NULL);
640
641 callout_init(&bw_upcalls_ch, 0);
642 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
643 expire_bw_upcalls_send, NULL);
644
645 callout_init(&bw_meter_ch, 0);
646 callout_reset(&bw_meter_ch, BW_METER_PERIOD,
647 expire_bw_meter_process, NULL);
648
649 if (mrtdebug)
650 log(LOG_DEBUG, "ip_mrouter_init\n");
651
652 return (0);
653 }
654
655 /*
656 * Disable multicast routing
657 */
658 int
659 ip_mrouter_done(void)
660 {
661 vifi_t vifi;
662 struct vif *vifp;
663 int i;
664 int s;
665
666 s = splsoftnet();
667
668 /* Clear out all the vifs currently in use. */
669 for (vifi = 0; vifi < numvifs; vifi++) {
670 vifp = &viftable[vifi];
671 if (!in_nullhost(vifp->v_lcl_addr))
672 reset_vif(vifp);
673 }
674
675 numvifs = 0;
676 pim_assert = 0;
677 mrt_api_config = 0;
678
679 callout_stop(&expire_upcalls_ch);
680 callout_stop(&bw_upcalls_ch);
681 callout_stop(&bw_meter_ch);
682
683 /*
684 * Free all multicast forwarding cache entries.
685 */
686 for (i = 0; i < MFCTBLSIZ; i++) {
687 struct mfc *rt, *nrt;
688
689 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
690 nrt = LIST_NEXT(rt, mfc_hash);
691
692 expire_mfc(rt);
693 }
694 }
695
696 bzero((void *)nexpire, sizeof(nexpire));
697 hashdone(mfchashtbl, HASH_LIST, mfchash);
698 mfchashtbl = NULL;
699
700 bw_upcalls_n = 0;
701 bzero(bw_meter_timers, sizeof(bw_meter_timers));
702
703 /* Reset de-encapsulation cache. */
704
705 ip_mrouter = NULL;
706
707 splx(s);
708
709 if (mrtdebug)
710 log(LOG_DEBUG, "ip_mrouter_done\n");
711
712 return (0);
713 }
714
715 void
716 ip_mrouter_detach(struct ifnet *ifp)
717 {
718 int vifi, i;
719 struct vif *vifp;
720 struct mfc *rt;
721 struct rtdetq *rte;
722
723 /* XXX not sure about side effect to userland routing daemon */
724 for (vifi = 0; vifi < numvifs; vifi++) {
725 vifp = &viftable[vifi];
726 if (vifp->v_ifp == ifp)
727 reset_vif(vifp);
728 }
729 for (i = 0; i < MFCTBLSIZ; i++) {
730 if (nexpire[i] == 0)
731 continue;
732 LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
733 for (rte = rt->mfc_stall; rte; rte = rte->next) {
734 if (rte->ifp == ifp)
735 rte->ifp = NULL;
736 }
737 }
738 }
739 }
740
741 /*
742 * Set PIM assert processing global
743 */
744 static int
745 set_assert(int i)
746 {
747 pim_assert = !!i;
748 return (0);
749 }
750
751 /*
752 * Configure API capabilities
753 */
754 static int
755 set_api_config(struct sockopt *sopt)
756 {
757 u_int32_t apival;
758 int i, error;
759
760 /*
761 * We can set the API capabilities only if it is the first operation
762 * after MRT_INIT. I.e.:
763 * - there are no vifs installed
764 * - pim_assert is not enabled
765 * - the MFC table is empty
766 */
767 error = sockopt_get(sopt, &apival, sizeof(apival));
768 if (error)
769 return (error);
770 if (numvifs > 0)
771 return (EPERM);
772 if (pim_assert)
773 return (EPERM);
774 for (i = 0; i < MFCTBLSIZ; i++) {
775 if (LIST_FIRST(&mfchashtbl[i]) != NULL)
776 return (EPERM);
777 }
778
779 mrt_api_config = apival & mrt_api_support;
780 return (0);
781 }
782
783 /*
784 * Add a vif to the vif table
785 */
786 static int
787 add_vif(struct vifctl *vifcp)
788 {
789 struct vif *vifp;
790 struct ifaddr *ifa;
791 struct ifnet *ifp;
792 struct ifreq ifr;
793 int error, s;
794 struct sockaddr_in sin;
795
796 if (vifcp->vifc_vifi >= MAXVIFS)
797 return (EINVAL);
798 if (in_nullhost(vifcp->vifc_lcl_addr))
799 return (EADDRNOTAVAIL);
800
801 vifp = &viftable[vifcp->vifc_vifi];
802 if (!in_nullhost(vifp->v_lcl_addr))
803 return (EADDRINUSE);
804
805 /* Find the interface with an address in AF_INET family. */
806 #ifdef PIM
807 if (vifcp->vifc_flags & VIFF_REGISTER) {
808 /*
809 * XXX: Because VIFF_REGISTER does not really need a valid
810 * local interface (e.g. it could be 127.0.0.2), we don't
811 * check its address.
812 */
813 ifp = NULL;
814 } else
815 #endif
816 {
817 sockaddr_in_init(&sin, &vifcp->vifc_lcl_addr, 0);
818 ifa = ifa_ifwithaddr(sintosa(&sin));
819 if (ifa == NULL)
820 return (EADDRNOTAVAIL);
821 ifp = ifa->ifa_ifp;
822 }
823
824 if (vifcp->vifc_flags & VIFF_TUNNEL) {
825 if (vifcp->vifc_flags & VIFF_SRCRT) {
826 log(LOG_ERR, "source routed tunnels not supported\n");
827 return (EOPNOTSUPP);
828 }
829
830 /* attach this vif to decapsulator dispatch table */
831 /*
832 * XXX Use addresses in registration so that matching
833 * can be done with radix tree in decapsulator. But,
834 * we need to check inner header for multicast, so
835 * this requires both radix tree lookup and then a
836 * function to check, and this is not supported yet.
837 */
838 vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
839 vif_encapcheck, &vif_protosw, vifp);
840 if (!vifp->v_encap_cookie)
841 return (EINVAL);
842
843 /* Create a fake encapsulation interface. */
844 ifp = (struct ifnet *)malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK);
845 bzero(ifp, sizeof(*ifp));
846 snprintf(ifp->if_xname, sizeof(ifp->if_xname),
847 "mdecap%d", vifcp->vifc_vifi);
848
849 /* Prepare cached route entry. */
850 bzero(&vifp->v_route, 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 bzero(ifp, sizeof(*ifp));
859 snprintf(ifp->if_xname, sizeof(ifp->if_xname),
860 "register_vif");
861 ifp->if_flags = IFF_LOOPBACK;
862 bzero(&vifp->v_route, 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 ifreq_setaddr(SIOCADDMULTI, &ifr, sintosa(&sin));
874 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, &ifr);
875 if (error)
876 return (error);
877 }
878
879 s = splsoftnet();
880
881 /* Define parameters for the tbf structure. */
882 vifp->tbf_q = NULL;
883 vifp->tbf_t = &vifp->tbf_q;
884 microtime(&vifp->tbf_last_pkt_t);
885 vifp->tbf_n_tok = 0;
886 vifp->tbf_q_len = 0;
887 vifp->tbf_max_q_len = MAXQSIZE;
888
889 vifp->v_flags = vifcp->vifc_flags;
890 vifp->v_threshold = vifcp->vifc_threshold;
891 /* scaling up here allows division by 1024 in critical code */
892 vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
893 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
894 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
895 vifp->v_ifp = ifp;
896 /* Initialize per vif pkt counters. */
897 vifp->v_pkt_in = 0;
898 vifp->v_pkt_out = 0;
899 vifp->v_bytes_in = 0;
900 vifp->v_bytes_out = 0;
901
902 callout_init(&vifp->v_repq_ch, 0);
903
904 #ifdef RSVP_ISI
905 vifp->v_rsvp_on = 0;
906 vifp->v_rsvpd = NULL;
907 #endif /* RSVP_ISI */
908
909 splx(s);
910
911 /* Adjust numvifs up if the vifi is higher than numvifs. */
912 if (numvifs <= vifcp->vifc_vifi)
913 numvifs = vifcp->vifc_vifi + 1;
914
915 if (mrtdebug)
916 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
917 vifcp->vifc_vifi,
918 ntohl(vifcp->vifc_lcl_addr.s_addr),
919 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
920 ntohl(vifcp->vifc_rmt_addr.s_addr),
921 vifcp->vifc_threshold,
922 vifcp->vifc_rate_limit);
923
924 return (0);
925 }
926
927 void
928 reset_vif(struct vif *vifp)
929 {
930 struct mbuf *m, *n;
931 struct ifnet *ifp;
932 struct ifreq ifr;
933 struct sockaddr_in sin;
934
935 callout_stop(&vifp->v_repq_ch);
936
937 /* detach this vif from decapsulator dispatch table */
938 encap_detach(vifp->v_encap_cookie);
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 ifreq_setaddr(SIOCDELMULTI, &ifr, sintosa(&sin));
958 ifp = vifp->v_ifp;
959 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, &ifr);
960 }
961 bzero((void *)vifp, sizeof(*vifp));
962 }
963
964 /*
965 * Delete a vif from the vif table
966 */
967 static int
968 del_vif(vifi_t *vifip)
969 {
970 struct vif *vifp;
971 vifi_t vifi;
972 int s;
973
974 if (*vifip >= numvifs)
975 return (EINVAL);
976
977 vifp = &viftable[*vifip];
978 if (in_nullhost(vifp->v_lcl_addr))
979 return (EADDRNOTAVAIL);
980
981 s = splsoftnet();
982
983 reset_vif(vifp);
984
985 /* Adjust numvifs down */
986 for (vifi = numvifs; vifi > 0; vifi--)
987 if (!in_nullhost(viftable[vifi - 1].v_lcl_addr))
988 break;
989 numvifs = vifi;
990
991 splx(s);
992
993 if (mrtdebug)
994 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
995
996 return (0);
997 }
998
999 /*
1000 * update an mfc entry without resetting counters and S,G addresses.
1001 */
1002 static void
1003 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1004 {
1005 int i;
1006
1007 rt->mfc_parent = mfccp->mfcc_parent;
1008 for (i = 0; i < numvifs; i++) {
1009 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1010 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1011 MRT_MFC_FLAGS_ALL;
1012 }
1013 /* set the RP address */
1014 if (mrt_api_config & MRT_MFC_RP)
1015 rt->mfc_rp = mfccp->mfcc_rp;
1016 else
1017 rt->mfc_rp = zeroin_addr;
1018 }
1019
1020 /*
1021 * fully initialize an mfc entry from the parameter.
1022 */
1023 static void
1024 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1025 {
1026 rt->mfc_origin = mfccp->mfcc_origin;
1027 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1028
1029 update_mfc_params(rt, mfccp);
1030
1031 /* initialize pkt counters per src-grp */
1032 rt->mfc_pkt_cnt = 0;
1033 rt->mfc_byte_cnt = 0;
1034 rt->mfc_wrong_if = 0;
1035 timerclear(&rt->mfc_last_assert);
1036 }
1037
1038 static void
1039 expire_mfc(struct mfc *rt)
1040 {
1041 struct rtdetq *rte, *nrte;
1042
1043 free_bw_list(rt->mfc_bw_meter);
1044
1045 for (rte = rt->mfc_stall; rte != NULL; rte = nrte) {
1046 nrte = rte->next;
1047 m_freem(rte->m);
1048 free(rte, M_MRTABLE);
1049 }
1050
1051 LIST_REMOVE(rt, mfc_hash);
1052 free(rt, M_MRTABLE);
1053 }
1054
1055 /*
1056 * Add an mfc entry
1057 */
1058 static int
1059 add_mfc(struct sockopt *sopt)
1060 {
1061 struct mfcctl2 mfcctl2;
1062 struct mfcctl2 *mfccp;
1063 struct mfc *rt;
1064 u_int32_t hash = 0;
1065 struct rtdetq *rte, *nrte;
1066 u_short nstl;
1067 int s;
1068 int mfcctl_size = sizeof(struct mfcctl);
1069 int error;
1070
1071 if (mrt_api_config & MRT_API_FLAGS_ALL)
1072 mfcctl_size = sizeof(struct mfcctl2);
1073
1074 /*
1075 * select data size depending on API version.
1076 */
1077 mfccp = &mfcctl2;
1078 memset(&mfcctl2, 0, sizeof(mfcctl2));
1079
1080 if (mrt_api_config & MRT_API_FLAGS_ALL)
1081 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1082 else
1083 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1084
1085 if (error)
1086 return (error);
1087
1088 s = splsoftnet();
1089 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1090
1091 /* If an entry already exists, just update the fields */
1092 if (rt) {
1093 if (mrtdebug & DEBUG_MFC)
1094 log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n",
1095 ntohl(mfccp->mfcc_origin.s_addr),
1096 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1097 mfccp->mfcc_parent);
1098
1099 update_mfc_params(rt, mfccp);
1100
1101 splx(s);
1102 return (0);
1103 }
1104
1105 /*
1106 * Find the entry for which the upcall was made and update
1107 */
1108 nstl = 0;
1109 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1110 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1111 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1112 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1113 rt->mfc_stall != NULL) {
1114 if (nstl++)
1115 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n",
1116 "multiple kernel entries",
1117 ntohl(mfccp->mfcc_origin.s_addr),
1118 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1119 mfccp->mfcc_parent, rt->mfc_stall);
1120
1121 if (mrtdebug & DEBUG_MFC)
1122 log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n",
1123 ntohl(mfccp->mfcc_origin.s_addr),
1124 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1125 mfccp->mfcc_parent, rt->mfc_stall);
1126
1127 rte = rt->mfc_stall;
1128 init_mfc_params(rt, mfccp);
1129 rt->mfc_stall = NULL;
1130
1131 rt->mfc_expire = 0; /* Don't clean this guy up */
1132 nexpire[hash]--;
1133
1134 /* free packets Qed at the end of this entry */
1135 for (; rte != NULL; rte = nrte) {
1136 nrte = rte->next;
1137 if (rte->ifp) {
1138 #ifdef RSVP_ISI
1139 ip_mdq(rte->m, rte->ifp, rt, -1);
1140 #else
1141 ip_mdq(rte->m, rte->ifp, rt);
1142 #endif /* RSVP_ISI */
1143 }
1144 m_freem(rte->m);
1145 #ifdef UPCALL_TIMING
1146 collate(&rte->t);
1147 #endif /* UPCALL_TIMING */
1148 free(rte, M_MRTABLE);
1149 }
1150 }
1151 }
1152
1153 /*
1154 * It is possible that an entry is being inserted without an upcall
1155 */
1156 if (nstl == 0) {
1157 /*
1158 * No mfc; make a new one
1159 */
1160 if (mrtdebug & DEBUG_MFC)
1161 log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n",
1162 ntohl(mfccp->mfcc_origin.s_addr),
1163 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1164 mfccp->mfcc_parent);
1165
1166 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1167 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1168 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1169 init_mfc_params(rt, mfccp);
1170 if (rt->mfc_expire)
1171 nexpire[hash]--;
1172 rt->mfc_expire = 0;
1173 break; /* XXX */
1174 }
1175 }
1176 if (rt == NULL) { /* no upcall, so make a new entry */
1177 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1178 M_NOWAIT);
1179 if (rt == NULL) {
1180 splx(s);
1181 return (ENOBUFS);
1182 }
1183
1184 init_mfc_params(rt, mfccp);
1185 rt->mfc_expire = 0;
1186 rt->mfc_stall = NULL;
1187 rt->mfc_bw_meter = NULL;
1188
1189 /* insert new entry at head of hash chain */
1190 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1191 }
1192 }
1193
1194 splx(s);
1195 return (0);
1196 }
1197
1198 #ifdef UPCALL_TIMING
1199 /*
1200 * collect delay statistics on the upcalls
1201 */
1202 static void
1203 collate(struct timeval *t)
1204 {
1205 u_int32_t d;
1206 struct timeval tp;
1207 u_int32_t delta;
1208
1209 microtime(&tp);
1210
1211 if (timercmp(t, &tp, <)) {
1212 TV_DELTA(tp, *t, delta);
1213
1214 d = delta >> 10;
1215 if (d > 50)
1216 d = 50;
1217
1218 ++upcall_data[d];
1219 }
1220 }
1221 #endif /* UPCALL_TIMING */
1222
1223 /*
1224 * Delete an mfc entry
1225 */
1226 static int
1227 del_mfc(struct sockopt *sopt)
1228 {
1229 struct mfcctl2 mfcctl2;
1230 struct mfcctl2 *mfccp;
1231 struct mfc *rt;
1232 int s;
1233 int error;
1234
1235 /*
1236 * XXX: for deleting MFC entries the information in entries
1237 * of size "struct mfcctl" is sufficient.
1238 */
1239
1240 mfccp = &mfcctl2;
1241 memset(&mfcctl2, 0, sizeof(mfcctl2));
1242
1243 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1244 if (error) {
1245 /* Try with the size of mfcctl2. */
1246 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1247 if (error)
1248 return (error);
1249 }
1250
1251 if (mrtdebug & DEBUG_MFC)
1252 log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
1253 ntohl(mfccp->mfcc_origin.s_addr),
1254 ntohl(mfccp->mfcc_mcastgrp.s_addr));
1255
1256 s = splsoftnet();
1257
1258 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1259 if (rt == NULL) {
1260 splx(s);
1261 return (EADDRNOTAVAIL);
1262 }
1263
1264 /*
1265 * free the bw_meter entries
1266 */
1267 free_bw_list(rt->mfc_bw_meter);
1268 rt->mfc_bw_meter = NULL;
1269
1270 LIST_REMOVE(rt, mfc_hash);
1271 free(rt, M_MRTABLE);
1272
1273 splx(s);
1274 return (0);
1275 }
1276
1277 static int
1278 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1279 {
1280 if (s) {
1281 if (sbappendaddr(&s->so_rcv, sintosa(src), mm,
1282 (struct mbuf *)NULL) != 0) {
1283 sorwakeup(s);
1284 return (0);
1285 }
1286 }
1287 m_freem(mm);
1288 return (-1);
1289 }
1290
1291 /*
1292 * IP multicast forwarding function. This function assumes that the packet
1293 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1294 * pointed to by "ifp", and the packet is to be relayed to other networks
1295 * that have members of the packet's destination IP multicast group.
1296 *
1297 * The packet is returned unscathed to the caller, unless it is
1298 * erroneous, in which case a non-zero return value tells the caller to
1299 * discard it.
1300 */
1301
1302 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */
1303 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1304
1305 int
1306 #ifdef RSVP_ISI
1307 ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo)
1308 #else
1309 ip_mforward(struct mbuf *m, struct ifnet *ifp)
1310 #endif /* RSVP_ISI */
1311 {
1312 struct ip *ip = mtod(m, struct ip *);
1313 struct mfc *rt;
1314 static int srctun = 0;
1315 struct mbuf *mm;
1316 struct sockaddr_in sin;
1317 int s;
1318 vifi_t vifi;
1319
1320 if (mrtdebug & DEBUG_FORWARD)
1321 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1322 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1323
1324 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1325 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1326 /*
1327 * Packet arrived via a physical interface or
1328 * an encapsulated tunnel or a register_vif.
1329 */
1330 } else {
1331 /*
1332 * Packet arrived through a source-route tunnel.
1333 * Source-route tunnels are no longer supported.
1334 */
1335 if ((srctun++ % 1000) == 0)
1336 log(LOG_ERR,
1337 "ip_mforward: received source-routed packet from %x\n",
1338 ntohl(ip->ip_src.s_addr));
1339
1340 return (1);
1341 }
1342
1343 /*
1344 * Clear any in-bound checksum flags for this packet.
1345 */
1346 m->m_pkthdr.csum_flags = 0;
1347
1348 #ifdef RSVP_ISI
1349 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1350 if (ip->ip_ttl < MAXTTL)
1351 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1352 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1353 struct vif *vifp = viftable + vifi;
1354 printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n",
1355 ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
1356 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1357 vifp->v_ifp->if_xname);
1358 }
1359 return (ip_mdq(m, ifp, (struct mfc *)NULL, vifi));
1360 }
1361 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1362 printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n",
1363 ntohl(ip->ip_src), ntohl(ip->ip_dst));
1364 }
1365 #endif /* RSVP_ISI */
1366
1367 /*
1368 * Don't forward a packet with time-to-live of zero or one,
1369 * or a packet destined to a local-only group.
1370 */
1371 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1372 return (0);
1373
1374 /*
1375 * Determine forwarding vifs from the forwarding cache table
1376 */
1377 s = splsoftnet();
1378 ++mrtstat.mrts_mfc_lookups;
1379 rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1380
1381 /* Entry exists, so forward if necessary */
1382 if (rt != NULL) {
1383 splx(s);
1384 #ifdef RSVP_ISI
1385 return (ip_mdq(m, ifp, rt, -1));
1386 #else
1387 return (ip_mdq(m, ifp, rt));
1388 #endif /* RSVP_ISI */
1389 } else {
1390 /*
1391 * If we don't have a route for packet's origin,
1392 * Make a copy of the packet & send message to routing daemon
1393 */
1394
1395 struct mbuf *mb0;
1396 struct rtdetq *rte;
1397 u_int32_t hash;
1398 int hlen = ip->ip_hl << 2;
1399 #ifdef UPCALL_TIMING
1400 struct timeval tp;
1401
1402 microtime(&tp);
1403 #endif /* UPCALL_TIMING */
1404
1405 ++mrtstat.mrts_mfc_misses;
1406
1407 mrtstat.mrts_no_route++;
1408 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1409 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1410 ntohl(ip->ip_src.s_addr),
1411 ntohl(ip->ip_dst.s_addr));
1412
1413 /*
1414 * Allocate mbufs early so that we don't do extra work if we are
1415 * just going to fail anyway. Make sure to pullup the header so
1416 * that other people can't step on it.
1417 */
1418 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1419 M_NOWAIT);
1420 if (rte == NULL) {
1421 splx(s);
1422 return (ENOBUFS);
1423 }
1424 mb0 = m_copypacket(m, M_DONTWAIT);
1425 M_PULLUP(mb0, hlen);
1426 if (mb0 == NULL) {
1427 free(rte, M_MRTABLE);
1428 splx(s);
1429 return (ENOBUFS);
1430 }
1431
1432 /* is there an upcall waiting for this flow? */
1433 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1434 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1435 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1436 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1437 rt->mfc_stall != NULL)
1438 break;
1439 }
1440
1441 if (rt == NULL) {
1442 int i;
1443 struct igmpmsg *im;
1444
1445 /*
1446 * Locate the vifi for the incoming interface for
1447 * this packet.
1448 * If none found, drop packet.
1449 */
1450 for (vifi = 0; vifi < numvifs &&
1451 viftable[vifi].v_ifp != ifp; vifi++)
1452 ;
1453 if (vifi >= numvifs) /* vif not found, drop packet */
1454 goto non_fatal;
1455
1456 /* no upcall, so make a new entry */
1457 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1458 M_NOWAIT);
1459 if (rt == NULL)
1460 goto fail;
1461
1462 /*
1463 * Make a copy of the header to send to the user level
1464 * process
1465 */
1466 mm = m_copym(m, 0, hlen, M_DONTWAIT);
1467 M_PULLUP(mm, hlen);
1468 if (mm == NULL)
1469 goto fail1;
1470
1471 /*
1472 * Send message to routing daemon to install
1473 * a route into the kernel table
1474 */
1475
1476 im = mtod(mm, struct igmpmsg *);
1477 im->im_msgtype = IGMPMSG_NOCACHE;
1478 im->im_mbz = 0;
1479 im->im_vif = vifi;
1480
1481 mrtstat.mrts_upcalls++;
1482
1483 sockaddr_in_init(&sin, &ip->ip_src, 0);
1484 if (socket_send(ip_mrouter, mm, &sin) < 0) {
1485 log(LOG_WARNING,
1486 "ip_mforward: ip_mrouter socket queue full\n");
1487 ++mrtstat.mrts_upq_sockfull;
1488 fail1:
1489 free(rt, M_MRTABLE);
1490 fail:
1491 free(rte, M_MRTABLE);
1492 m_freem(mb0);
1493 splx(s);
1494 return (ENOBUFS);
1495 }
1496
1497 /* insert new entry at head of hash chain */
1498 rt->mfc_origin = ip->ip_src;
1499 rt->mfc_mcastgrp = ip->ip_dst;
1500 rt->mfc_pkt_cnt = 0;
1501 rt->mfc_byte_cnt = 0;
1502 rt->mfc_wrong_if = 0;
1503 rt->mfc_expire = UPCALL_EXPIRE;
1504 nexpire[hash]++;
1505 for (i = 0; i < numvifs; i++) {
1506 rt->mfc_ttls[i] = 0;
1507 rt->mfc_flags[i] = 0;
1508 }
1509 rt->mfc_parent = -1;
1510
1511 /* clear the RP address */
1512 rt->mfc_rp = zeroin_addr;
1513
1514 rt->mfc_bw_meter = NULL;
1515
1516 /* link into table */
1517 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1518 /* Add this entry to the end of the queue */
1519 rt->mfc_stall = rte;
1520 } else {
1521 /* determine if q has overflowed */
1522 struct rtdetq **p;
1523 int npkts = 0;
1524
1525 /*
1526 * XXX ouch! we need to append to the list, but we
1527 * only have a pointer to the front, so we have to
1528 * scan the entire list every time.
1529 */
1530 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1531 if (++npkts > MAX_UPQ) {
1532 mrtstat.mrts_upq_ovflw++;
1533 non_fatal:
1534 free(rte, M_MRTABLE);
1535 m_freem(mb0);
1536 splx(s);
1537 return (0);
1538 }
1539
1540 /* Add this entry to the end of the queue */
1541 *p = rte;
1542 }
1543
1544 rte->next = NULL;
1545 rte->m = mb0;
1546 rte->ifp = ifp;
1547 #ifdef UPCALL_TIMING
1548 rte->t = tp;
1549 #endif /* UPCALL_TIMING */
1550
1551 splx(s);
1552
1553 return (0);
1554 }
1555 }
1556
1557
1558 /*ARGSUSED*/
1559 static void
1560 expire_upcalls(void *v)
1561 {
1562 int i;
1563 int s;
1564
1565 s = splsoftnet();
1566
1567 for (i = 0; i < MFCTBLSIZ; i++) {
1568 struct mfc *rt, *nrt;
1569
1570 if (nexpire[i] == 0)
1571 continue;
1572
1573 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1574 nrt = LIST_NEXT(rt, mfc_hash);
1575
1576 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1577 continue;
1578 nexpire[i]--;
1579
1580 /*
1581 * free the bw_meter entries
1582 */
1583 while (rt->mfc_bw_meter != NULL) {
1584 struct bw_meter *x = rt->mfc_bw_meter;
1585
1586 rt->mfc_bw_meter = x->bm_mfc_next;
1587 free(x, M_BWMETER);
1588 }
1589
1590 ++mrtstat.mrts_cache_cleanups;
1591 if (mrtdebug & DEBUG_EXPIRE)
1592 log(LOG_DEBUG,
1593 "expire_upcalls: expiring (%x %x)\n",
1594 ntohl(rt->mfc_origin.s_addr),
1595 ntohl(rt->mfc_mcastgrp.s_addr));
1596
1597 expire_mfc(rt);
1598 }
1599 }
1600
1601 splx(s);
1602 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1603 expire_upcalls, NULL);
1604 }
1605
1606 /*
1607 * Packet forwarding routine once entry in the cache is made
1608 */
1609 static int
1610 #ifdef RSVP_ISI
1611 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1612 #else
1613 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt)
1614 #endif /* RSVP_ISI */
1615 {
1616 struct ip *ip = mtod(m, struct ip *);
1617 vifi_t vifi;
1618 struct vif *vifp;
1619 struct sockaddr_in sin;
1620 int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1621
1622 /*
1623 * Macro to send packet on vif. Since RSVP packets don't get counted on
1624 * input, they shouldn't get counted on output, so statistics keeping is
1625 * separate.
1626 */
1627 #define MC_SEND(ip, vifp, m) do { \
1628 if ((vifp)->v_flags & VIFF_TUNNEL) \
1629 encap_send((ip), (vifp), (m)); \
1630 else \
1631 phyint_send((ip), (vifp), (m)); \
1632 } while (/*CONSTCOND*/ 0)
1633
1634 #ifdef RSVP_ISI
1635 /*
1636 * If xmt_vif is not -1, send on only the requested vif.
1637 *
1638 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1639 */
1640 if (xmt_vif < numvifs) {
1641 #ifdef PIM
1642 if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1643 pim_register_send(ip, viftable + xmt_vif, m, rt);
1644 else
1645 #endif
1646 MC_SEND(ip, viftable + xmt_vif, m);
1647 return (1);
1648 }
1649 #endif /* RSVP_ISI */
1650
1651 /*
1652 * Don't forward if it didn't arrive from the parent vif for its origin.
1653 */
1654 vifi = rt->mfc_parent;
1655 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1656 /* came in the wrong interface */
1657 if (mrtdebug & DEBUG_FORWARD)
1658 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1659 ifp, vifi,
1660 vifi >= numvifs ? 0 : viftable[vifi].v_ifp);
1661 ++mrtstat.mrts_wrong_if;
1662 ++rt->mfc_wrong_if;
1663 /*
1664 * If we are doing PIM assert processing, send a message
1665 * to the routing daemon.
1666 *
1667 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1668 * can complete the SPT switch, regardless of the type
1669 * of the iif (broadcast media, GRE tunnel, etc).
1670 */
1671 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1672 struct timeval now;
1673 u_int32_t delta;
1674
1675 #ifdef PIM
1676 if (ifp == &multicast_register_if)
1677 pimstat.pims_rcv_registers_wrongiif++;
1678 #endif
1679
1680 /* Get vifi for the incoming packet */
1681 for (vifi = 0;
1682 vifi < numvifs && viftable[vifi].v_ifp != ifp;
1683 vifi++)
1684 ;
1685 if (vifi >= numvifs) {
1686 /* The iif is not found: ignore the packet. */
1687 return (0);
1688 }
1689
1690 if (rt->mfc_flags[vifi] &
1691 MRT_MFC_FLAGS_DISABLE_WRONGVIF) {
1692 /* WRONGVIF disabled: ignore the packet */
1693 return (0);
1694 }
1695
1696 microtime(&now);
1697
1698 TV_DELTA(rt->mfc_last_assert, now, delta);
1699
1700 if (delta > ASSERT_MSG_TIME) {
1701 struct igmpmsg *im;
1702 int hlen = ip->ip_hl << 2;
1703 struct mbuf *mm =
1704 m_copym(m, 0, hlen, M_DONTWAIT);
1705
1706 M_PULLUP(mm, hlen);
1707 if (mm == NULL)
1708 return (ENOBUFS);
1709
1710 rt->mfc_last_assert = now;
1711
1712 im = mtod(mm, struct igmpmsg *);
1713 im->im_msgtype = IGMPMSG_WRONGVIF;
1714 im->im_mbz = 0;
1715 im->im_vif = vifi;
1716
1717 mrtstat.mrts_upcalls++;
1718
1719 sockaddr_in_init(&sin, &im->im_src, 0);
1720 if (socket_send(ip_mrouter, mm, &sin) < 0) {
1721 log(LOG_WARNING,
1722 "ip_mforward: ip_mrouter socket queue full\n");
1723 ++mrtstat.mrts_upq_sockfull;
1724 return (ENOBUFS);
1725 }
1726 }
1727 }
1728 return (0);
1729 }
1730
1731 /* If I sourced this packet, it counts as output, else it was input. */
1732 if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1733 viftable[vifi].v_pkt_out++;
1734 viftable[vifi].v_bytes_out += plen;
1735 } else {
1736 viftable[vifi].v_pkt_in++;
1737 viftable[vifi].v_bytes_in += plen;
1738 }
1739 rt->mfc_pkt_cnt++;
1740 rt->mfc_byte_cnt += plen;
1741
1742 /*
1743 * For each vif, decide if a copy of the packet should be forwarded.
1744 * Forward if:
1745 * - the ttl exceeds the vif's threshold
1746 * - there are group members downstream on interface
1747 */
1748 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1749 if ((rt->mfc_ttls[vifi] > 0) &&
1750 (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1751 vifp->v_pkt_out++;
1752 vifp->v_bytes_out += plen;
1753 #ifdef PIM
1754 if (vifp->v_flags & VIFF_REGISTER)
1755 pim_register_send(ip, vifp, m, rt);
1756 else
1757 #endif
1758 MC_SEND(ip, vifp, m);
1759 }
1760
1761 /*
1762 * Perform upcall-related bw measuring.
1763 */
1764 if (rt->mfc_bw_meter != NULL) {
1765 struct bw_meter *x;
1766 struct timeval now;
1767
1768 microtime(&now);
1769 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1770 bw_meter_receive_packet(x, plen, &now);
1771 }
1772
1773 return (0);
1774 }
1775
1776 #ifdef RSVP_ISI
1777 /*
1778 * check if a vif number is legal/ok. This is used by ip_output.
1779 */
1780 int
1781 legal_vif_num(int vif)
1782 {
1783 if (vif >= 0 && vif < numvifs)
1784 return (1);
1785 else
1786 return (0);
1787 }
1788 #endif /* RSVP_ISI */
1789
1790 static void
1791 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1792 {
1793 struct mbuf *mb_copy;
1794 int hlen = ip->ip_hl << 2;
1795
1796 /*
1797 * Make a new reference to the packet; make sure that
1798 * the IP header is actually copied, not just referenced,
1799 * so that ip_output() only scribbles on the copy.
1800 */
1801 mb_copy = m_copypacket(m, M_DONTWAIT);
1802 M_PULLUP(mb_copy, hlen);
1803 if (mb_copy == NULL)
1804 return;
1805
1806 if (vifp->v_rate_limit <= 0)
1807 tbf_send_packet(vifp, mb_copy);
1808 else
1809 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *),
1810 ntohs(ip->ip_len));
1811 }
1812
1813 static void
1814 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1815 {
1816 struct mbuf *mb_copy;
1817 struct ip *ip_copy;
1818 int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr);
1819
1820 /* Take care of delayed checksums */
1821 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1822 in_delayed_cksum(m);
1823 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1824 }
1825
1826 /*
1827 * copy the old packet & pullup it's IP header into the
1828 * new mbuf so we can modify it. Try to fill the new
1829 * mbuf since if we don't the ethernet driver will.
1830 */
1831 MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1832 if (mb_copy == NULL)
1833 return;
1834 mb_copy->m_data += max_linkhdr;
1835 mb_copy->m_pkthdr.len = len;
1836 mb_copy->m_len = sizeof(multicast_encap_iphdr);
1837
1838 if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) {
1839 m_freem(mb_copy);
1840 return;
1841 }
1842 i = MHLEN - max_linkhdr;
1843 if (i > len)
1844 i = len;
1845 mb_copy = m_pullup(mb_copy, i);
1846 if (mb_copy == NULL)
1847 return;
1848
1849 /*
1850 * fill in the encapsulating IP header.
1851 */
1852 ip_copy = mtod(mb_copy, struct ip *);
1853 *ip_copy = multicast_encap_iphdr;
1854 if (len < IP_MINFRAGSIZE)
1855 ip_copy->ip_id = 0;
1856 else
1857 ip_copy->ip_id = ip_newid(NULL);
1858 ip_copy->ip_len = htons(len);
1859 ip_copy->ip_src = vifp->v_lcl_addr;
1860 ip_copy->ip_dst = vifp->v_rmt_addr;
1861
1862 /*
1863 * turn the encapsulated IP header back into a valid one.
1864 */
1865 ip = (struct ip *)((char *)ip_copy + sizeof(multicast_encap_iphdr));
1866 --ip->ip_ttl;
1867 ip->ip_sum = 0;
1868 mb_copy->m_data += sizeof(multicast_encap_iphdr);
1869 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1870 mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1871
1872 if (vifp->v_rate_limit <= 0)
1873 tbf_send_packet(vifp, mb_copy);
1874 else
1875 tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len));
1876 }
1877
1878 /*
1879 * De-encapsulate a packet and feed it back through ip input.
1880 */
1881 static void
1882 vif_input(struct mbuf *m, ...)
1883 {
1884 int off, proto;
1885 va_list ap;
1886 struct vif *vifp;
1887 int s;
1888 struct ifqueue *ifq;
1889
1890 va_start(ap, m);
1891 off = va_arg(ap, int);
1892 proto = va_arg(ap, int);
1893 va_end(ap);
1894
1895 vifp = (struct vif *)encap_getarg(m);
1896 if (!vifp || proto != ENCAP_PROTO) {
1897 m_freem(m);
1898 mrtstat.mrts_bad_tunnel++;
1899 return;
1900 }
1901
1902 m_adj(m, off);
1903 m->m_pkthdr.rcvif = vifp->v_ifp;
1904 ifq = &ipintrq;
1905 s = splnet();
1906 if (IF_QFULL(ifq)) {
1907 IF_DROP(ifq);
1908 m_freem(m);
1909 } else {
1910 IF_ENQUEUE(ifq, m);
1911 /*
1912 * normally we would need a "schednetisr(NETISR_IP)"
1913 * here but we were called by ip_input and it is going
1914 * to loop back & try to dequeue the packet we just
1915 * queued as soon as we return so we avoid the
1916 * unnecessary software interrrupt.
1917 */
1918 }
1919 splx(s);
1920 }
1921
1922 /*
1923 * Check if the packet should be received on the vif denoted by arg.
1924 * (The encap selection code will call this once per vif since each is
1925 * registered separately.)
1926 */
1927 static int
1928 vif_encapcheck(struct mbuf *m, int off, int proto, void *arg)
1929 {
1930 struct vif *vifp;
1931 struct ip ip;
1932
1933 #ifdef DIAGNOSTIC
1934 if (!arg || proto != IPPROTO_IPV4)
1935 panic("unexpected arg in vif_encapcheck");
1936 #endif
1937
1938 /*
1939 * Accept the packet only if the inner heaader is multicast
1940 * and the outer header matches a tunnel-mode vif. Order
1941 * checks in the hope that common non-matching packets will be
1942 * rejected quickly. Assume that unicast IPv4 traffic in a
1943 * parallel tunnel (e.g. gif(4)) is unlikely.
1944 */
1945
1946 /* Obtain the outer IP header and the vif pointer. */
1947 m_copydata((struct mbuf *)m, 0, sizeof(ip), (void *)&ip);
1948 vifp = (struct vif *)arg;
1949
1950 /*
1951 * The outer source must match the vif's remote peer address.
1952 * For a multicast router with several tunnels, this is the
1953 * only check that will fail on packets in other tunnels,
1954 * assuming the local address is the same.
1955 */
1956 if (!in_hosteq(vifp->v_rmt_addr, ip.ip_src))
1957 return 0;
1958
1959 /* The outer destination must match the vif's local address. */
1960 if (!in_hosteq(vifp->v_lcl_addr, ip.ip_dst))
1961 return 0;
1962
1963 /* The vif must be of tunnel type. */
1964 if ((vifp->v_flags & VIFF_TUNNEL) == 0)
1965 return 0;
1966
1967 /* Check that the inner destination is multicast. */
1968 m_copydata((struct mbuf *)m, off, sizeof(ip), (void *)&ip);
1969 if (!IN_MULTICAST(ip.ip_dst.s_addr))
1970 return 0;
1971
1972 /*
1973 * We have checked that both the outer src and dst addresses
1974 * match the vif, and that the inner destination is multicast
1975 * (224/5). By claiming more than 64, we intend to
1976 * preferentially take packets that also match a parallel
1977 * gif(4).
1978 */
1979 return 32 + 32 + 5;
1980 }
1981
1982 /*
1983 * Token bucket filter module
1984 */
1985 static void
1986 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len)
1987 {
1988
1989 if (len > MAX_BKT_SIZE) {
1990 /* drop if packet is too large */
1991 mrtstat.mrts_pkt2large++;
1992 m_freem(m);
1993 return;
1994 }
1995
1996 tbf_update_tokens(vifp);
1997
1998 /*
1999 * If there are enough tokens, and the queue is empty, send this packet
2000 * out immediately. Otherwise, try to insert it on this vif's queue.
2001 */
2002 if (vifp->tbf_q_len == 0) {
2003 if (len <= vifp->tbf_n_tok) {
2004 vifp->tbf_n_tok -= len;
2005 tbf_send_packet(vifp, m);
2006 } else {
2007 /* queue packet and timeout till later */
2008 tbf_queue(vifp, m);
2009 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2010 tbf_reprocess_q, vifp);
2011 }
2012 } else {
2013 if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
2014 !tbf_dq_sel(vifp, ip)) {
2015 /* queue full, and couldn't make room */
2016 mrtstat.mrts_q_overflow++;
2017 m_freem(m);
2018 } else {
2019 /* queue length low enough, or made room */
2020 tbf_queue(vifp, m);
2021 tbf_process_q(vifp);
2022 }
2023 }
2024 }
2025
2026 /*
2027 * adds a packet to the queue at the interface
2028 */
2029 static void
2030 tbf_queue(struct vif *vifp, struct mbuf *m)
2031 {
2032 int s = splsoftnet();
2033
2034 /* insert at tail */
2035 *vifp->tbf_t = m;
2036 vifp->tbf_t = &m->m_nextpkt;
2037 vifp->tbf_q_len++;
2038
2039 splx(s);
2040 }
2041
2042
2043 /*
2044 * processes the queue at the interface
2045 */
2046 static void
2047 tbf_process_q(struct vif *vifp)
2048 {
2049 struct mbuf *m;
2050 int len;
2051 int s = splsoftnet();
2052
2053 /*
2054 * Loop through the queue at the interface and send as many packets
2055 * as possible.
2056 */
2057 for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) {
2058 len = ntohs(mtod(m, struct ip *)->ip_len);
2059
2060 /* determine if the packet can be sent */
2061 if (len <= vifp->tbf_n_tok) {
2062 /* if so,
2063 * reduce no of tokens, dequeue the packet,
2064 * send the packet.
2065 */
2066 if ((vifp->tbf_q = m->m_nextpkt) == NULL)
2067 vifp->tbf_t = &vifp->tbf_q;
2068 --vifp->tbf_q_len;
2069
2070 m->m_nextpkt = NULL;
2071 vifp->tbf_n_tok -= len;
2072 tbf_send_packet(vifp, m);
2073 } else
2074 break;
2075 }
2076 splx(s);
2077 }
2078
2079 static void
2080 tbf_reprocess_q(void *arg)
2081 {
2082 struct vif *vifp = arg;
2083
2084 if (ip_mrouter == NULL)
2085 return;
2086
2087 tbf_update_tokens(vifp);
2088 tbf_process_q(vifp);
2089
2090 if (vifp->tbf_q_len != 0)
2091 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2092 tbf_reprocess_q, vifp);
2093 }
2094
2095 /* function that will selectively discard a member of the queue
2096 * based on the precedence value and the priority
2097 */
2098 static int
2099 tbf_dq_sel(struct vif *vifp, struct ip *ip)
2100 {
2101 u_int p;
2102 struct mbuf **mp, *m;
2103 int s = splsoftnet();
2104
2105 p = priority(vifp, ip);
2106
2107 for (mp = &vifp->tbf_q, m = *mp;
2108 m != NULL;
2109 mp = &m->m_nextpkt, m = *mp) {
2110 if (p > priority(vifp, mtod(m, struct ip *))) {
2111 if ((*mp = m->m_nextpkt) == NULL)
2112 vifp->tbf_t = mp;
2113 --vifp->tbf_q_len;
2114
2115 m_freem(m);
2116 mrtstat.mrts_drop_sel++;
2117 splx(s);
2118 return (1);
2119 }
2120 }
2121 splx(s);
2122 return (0);
2123 }
2124
2125 static void
2126 tbf_send_packet(struct vif *vifp, struct mbuf *m)
2127 {
2128 int error;
2129 int s = splsoftnet();
2130
2131 if (vifp->v_flags & VIFF_TUNNEL) {
2132 /* If tunnel options */
2133 ip_output(m, (struct mbuf *)NULL, &vifp->v_route,
2134 IP_FORWARDING, (struct ip_moptions *)NULL,
2135 (struct socket *)NULL);
2136 } else {
2137 /* if physical interface option, extract the options and then send */
2138 struct ip_moptions imo;
2139
2140 imo.imo_multicast_ifp = vifp->v_ifp;
2141 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2142 imo.imo_multicast_loop = 1;
2143 #ifdef RSVP_ISI
2144 imo.imo_multicast_vif = -1;
2145 #endif
2146
2147 error = ip_output(m, NULL, NULL, IP_FORWARDING|IP_MULTICASTOPTS,
2148 &imo, NULL);
2149
2150 if (mrtdebug & DEBUG_XMIT)
2151 log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
2152 (long)(vifp - viftable), error);
2153 }
2154 splx(s);
2155 }
2156
2157 /* determine the current time and then
2158 * the elapsed time (between the last time and time now)
2159 * in milliseconds & update the no. of tokens in the bucket
2160 */
2161 static void
2162 tbf_update_tokens(struct vif *vifp)
2163 {
2164 struct timeval tp;
2165 u_int32_t tm;
2166 int s = splsoftnet();
2167
2168 microtime(&tp);
2169
2170 TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
2171
2172 /*
2173 * This formula is actually
2174 * "time in seconds" * "bytes/second".
2175 *
2176 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2177 *
2178 * The (1000/1024) was introduced in add_vif to optimize
2179 * this divide into a shift.
2180 */
2181 vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
2182 vifp->tbf_last_pkt_t = tp;
2183
2184 if (vifp->tbf_n_tok > MAX_BKT_SIZE)
2185 vifp->tbf_n_tok = MAX_BKT_SIZE;
2186
2187 splx(s);
2188 }
2189
2190 static int
2191 priority(struct vif *vifp, struct ip *ip)
2192 {
2193 int prio = 50; /* the lowest priority -- default case */
2194
2195 /* temporary hack; may add general packet classifier some day */
2196
2197 /*
2198 * The UDP port space is divided up into four priority ranges:
2199 * [0, 16384) : unclassified - lowest priority
2200 * [16384, 32768) : audio - highest priority
2201 * [32768, 49152) : whiteboard - medium priority
2202 * [49152, 65536) : video - low priority
2203 */
2204 if (ip->ip_p == IPPROTO_UDP) {
2205 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2206
2207 switch (ntohs(udp->uh_dport) & 0xc000) {
2208 case 0x4000:
2209 prio = 70;
2210 break;
2211 case 0x8000:
2212 prio = 60;
2213 break;
2214 case 0xc000:
2215 prio = 55;
2216 break;
2217 }
2218
2219 if (tbfdebug > 1)
2220 log(LOG_DEBUG, "port %x prio %d\n",
2221 ntohs(udp->uh_dport), prio);
2222 }
2223
2224 return (prio);
2225 }
2226
2227 /*
2228 * End of token bucket filter modifications
2229 */
2230 #ifdef RSVP_ISI
2231 int
2232 ip_rsvp_vif_init(struct socket *so, struct mbuf *m)
2233 {
2234 int vifi, s;
2235
2236 if (rsvpdebug)
2237 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2238 so->so_type, so->so_proto->pr_protocol);
2239
2240 if (so->so_type != SOCK_RAW ||
2241 so->so_proto->pr_protocol != IPPROTO_RSVP)
2242 return (EOPNOTSUPP);
2243
2244 /* Check mbuf. */
2245 if (m == NULL || m->m_len != sizeof(int)) {
2246 return (EINVAL);
2247 }
2248 vifi = *(mtod(m, int *));
2249
2250 if (rsvpdebug)
2251 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",
2252 vifi, rsvp_on);
2253
2254 s = splsoftnet();
2255
2256 /* Check vif. */
2257 if (!legal_vif_num(vifi)) {
2258 splx(s);
2259 return (EADDRNOTAVAIL);
2260 }
2261
2262 /* Check if socket is available. */
2263 if (viftable[vifi].v_rsvpd != NULL) {
2264 splx(s);
2265 return (EADDRINUSE);
2266 }
2267
2268 viftable[vifi].v_rsvpd = so;
2269 /*
2270 * This may seem silly, but we need to be sure we don't over-increment
2271 * the RSVP counter, in case something slips up.
2272 */
2273 if (!viftable[vifi].v_rsvp_on) {
2274 viftable[vifi].v_rsvp_on = 1;
2275 rsvp_on++;
2276 }
2277
2278 splx(s);
2279 return (0);
2280 }
2281
2282 int
2283 ip_rsvp_vif_done(struct socket *so, struct mbuf *m)
2284 {
2285 int vifi, s;
2286
2287 if (rsvpdebug)
2288 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2289 so->so_type, so->so_proto->pr_protocol);
2290
2291 if (so->so_type != SOCK_RAW ||
2292 so->so_proto->pr_protocol != IPPROTO_RSVP)
2293 return (EOPNOTSUPP);
2294
2295 /* Check mbuf. */
2296 if (m == NULL || m->m_len != sizeof(int)) {
2297 return (EINVAL);
2298 }
2299 vifi = *(mtod(m, int *));
2300
2301 s = splsoftnet();
2302
2303 /* Check vif. */
2304 if (!legal_vif_num(vifi)) {
2305 splx(s);
2306 return (EADDRNOTAVAIL);
2307 }
2308
2309 if (rsvpdebug)
2310 printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
2311 viftable[vifi].v_rsvpd, so);
2312
2313 viftable[vifi].v_rsvpd = NULL;
2314 /*
2315 * This may seem silly, but we need to be sure we don't over-decrement
2316 * the RSVP counter, in case something slips up.
2317 */
2318 if (viftable[vifi].v_rsvp_on) {
2319 viftable[vifi].v_rsvp_on = 0;
2320 rsvp_on--;
2321 }
2322
2323 splx(s);
2324 return (0);
2325 }
2326
2327 void
2328 ip_rsvp_force_done(struct socket *so)
2329 {
2330 int vifi, s;
2331
2332 /* Don't bother if it is not the right type of socket. */
2333 if (so->so_type != SOCK_RAW ||
2334 so->so_proto->pr_protocol != IPPROTO_RSVP)
2335 return;
2336
2337 s = splsoftnet();
2338
2339 /*
2340 * The socket may be attached to more than one vif...this
2341 * is perfectly legal.
2342 */
2343 for (vifi = 0; vifi < numvifs; vifi++) {
2344 if (viftable[vifi].v_rsvpd == so) {
2345 viftable[vifi].v_rsvpd = NULL;
2346 /*
2347 * This may seem silly, but we need to be sure we don't
2348 * over-decrement the RSVP counter, in case something
2349 * slips up.
2350 */
2351 if (viftable[vifi].v_rsvp_on) {
2352 viftable[vifi].v_rsvp_on = 0;
2353 rsvp_on--;
2354 }
2355 }
2356 }
2357
2358 splx(s);
2359 return;
2360 }
2361
2362 void
2363 rsvp_input(struct mbuf *m, struct ifnet *ifp)
2364 {
2365 int vifi, s;
2366 struct ip *ip = mtod(m, struct ip *);
2367 struct sockaddr_in rsvp_src;
2368
2369 if (rsvpdebug)
2370 printf("rsvp_input: rsvp_on %d\n", rsvp_on);
2371
2372 /*
2373 * Can still get packets with rsvp_on = 0 if there is a local member
2374 * of the group to which the RSVP packet is addressed. But in this
2375 * case we want to throw the packet away.
2376 */
2377 if (!rsvp_on) {
2378 m_freem(m);
2379 return;
2380 }
2381
2382 /*
2383 * If the old-style non-vif-associated socket is set, then use
2384 * it and ignore the new ones.
2385 */
2386 if (ip_rsvpd != NULL) {
2387 if (rsvpdebug)
2388 printf("rsvp_input: "
2389 "Sending packet up old-style socket\n");
2390 rip_input(m); /*XXX*/
2391 return;
2392 }
2393
2394 s = splsoftnet();
2395
2396 if (rsvpdebug)
2397 printf("rsvp_input: check vifs\n");
2398
2399 /* Find which vif the packet arrived on. */
2400 for (vifi = 0; vifi < numvifs; vifi++) {
2401 if (viftable[vifi].v_ifp == ifp)
2402 break;
2403 }
2404
2405 if (vifi == numvifs) {
2406 /* Can't find vif packet arrived on. Drop packet. */
2407 if (rsvpdebug)
2408 printf("rsvp_input: "
2409 "Can't find vif for packet...dropping it.\n");
2410 m_freem(m);
2411 splx(s);
2412 return;
2413 }
2414
2415 if (rsvpdebug)
2416 printf("rsvp_input: check socket\n");
2417
2418 if (viftable[vifi].v_rsvpd == NULL) {
2419 /*
2420 * drop packet, since there is no specific socket for this
2421 * interface
2422 */
2423 if (rsvpdebug)
2424 printf("rsvp_input: No socket defined for vif %d\n",
2425 vifi);
2426 m_freem(m);
2427 splx(s);
2428 return;
2429 }
2430
2431 sockaddr_in_init(&rsvp_src, &ip->ip_src, 0);
2432
2433 if (rsvpdebug && m)
2434 printf("rsvp_input: m->m_len = %d, sbspace() = %d\n",
2435 m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv));
2436
2437 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2438 if (rsvpdebug)
2439 printf("rsvp_input: Failed to append to socket\n");
2440 else
2441 if (rsvpdebug)
2442 printf("rsvp_input: send packet up\n");
2443
2444 splx(s);
2445 }
2446 #endif /* RSVP_ISI */
2447
2448 /*
2449 * Code for bandwidth monitors
2450 */
2451
2452 /*
2453 * Define common interface for timeval-related methods
2454 */
2455 #define BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp)
2456 #define BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp))
2457 #define BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp))
2458
2459 static uint32_t
2460 compute_bw_meter_flags(struct bw_upcall *req)
2461 {
2462 uint32_t flags = 0;
2463
2464 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2465 flags |= BW_METER_UNIT_PACKETS;
2466 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2467 flags |= BW_METER_UNIT_BYTES;
2468 if (req->bu_flags & BW_UPCALL_GEQ)
2469 flags |= BW_METER_GEQ;
2470 if (req->bu_flags & BW_UPCALL_LEQ)
2471 flags |= BW_METER_LEQ;
2472
2473 return flags;
2474 }
2475
2476 /*
2477 * Add a bw_meter entry
2478 */
2479 static int
2480 add_bw_upcall(struct bw_upcall *req)
2481 {
2482 int s;
2483 struct mfc *mfc;
2484 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2485 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2486 struct timeval now;
2487 struct bw_meter *x;
2488 uint32_t flags;
2489
2490 if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2491 return EOPNOTSUPP;
2492
2493 /* Test if the flags are valid */
2494 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2495 return EINVAL;
2496 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2497 return EINVAL;
2498 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2499 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2500 return EINVAL;
2501
2502 /* Test if the threshold time interval is valid */
2503 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2504 return EINVAL;
2505
2506 flags = compute_bw_meter_flags(req);
2507
2508 /*
2509 * Find if we have already same bw_meter entry
2510 */
2511 s = splsoftnet();
2512 mfc = mfc_find(&req->bu_src, &req->bu_dst);
2513 if (mfc == NULL) {
2514 splx(s);
2515 return EADDRNOTAVAIL;
2516 }
2517 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2518 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2519 &req->bu_threshold.b_time, ==)) &&
2520 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2521 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2522 (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
2523 splx(s);
2524 return 0; /* XXX Already installed */
2525 }
2526 }
2527
2528 /* Allocate the new bw_meter entry */
2529 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
2530 if (x == NULL) {
2531 splx(s);
2532 return ENOBUFS;
2533 }
2534
2535 /* Set the new bw_meter entry */
2536 x->bm_threshold.b_time = req->bu_threshold.b_time;
2537 microtime(&now);
2538 x->bm_start_time = now;
2539 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2540 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2541 x->bm_measured.b_packets = 0;
2542 x->bm_measured.b_bytes = 0;
2543 x->bm_flags = flags;
2544 x->bm_time_next = NULL;
2545 x->bm_time_hash = BW_METER_BUCKETS;
2546
2547 /* Add the new bw_meter entry to the front of entries for this MFC */
2548 x->bm_mfc = mfc;
2549 x->bm_mfc_next = mfc->mfc_bw_meter;
2550 mfc->mfc_bw_meter = x;
2551 schedule_bw_meter(x, &now);
2552 splx(s);
2553
2554 return 0;
2555 }
2556
2557 static void
2558 free_bw_list(struct bw_meter *list)
2559 {
2560 while (list != NULL) {
2561 struct bw_meter *x = list;
2562
2563 list = list->bm_mfc_next;
2564 unschedule_bw_meter(x);
2565 free(x, M_BWMETER);
2566 }
2567 }
2568
2569 /*
2570 * Delete one or multiple bw_meter entries
2571 */
2572 static int
2573 del_bw_upcall(struct bw_upcall *req)
2574 {
2575 int s;
2576 struct mfc *mfc;
2577 struct bw_meter *x;
2578
2579 if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2580 return EOPNOTSUPP;
2581
2582 s = splsoftnet();
2583 /* Find the corresponding MFC entry */
2584 mfc = mfc_find(&req->bu_src, &req->bu_dst);
2585 if (mfc == NULL) {
2586 splx(s);
2587 return EADDRNOTAVAIL;
2588 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2589 /*
2590 * Delete all bw_meter entries for this mfc
2591 */
2592 struct bw_meter *list;
2593
2594 list = mfc->mfc_bw_meter;
2595 mfc->mfc_bw_meter = NULL;
2596 free_bw_list(list);
2597 splx(s);
2598 return 0;
2599 } else { /* Delete a single bw_meter entry */
2600 struct bw_meter *prev;
2601 uint32_t flags = 0;
2602
2603 flags = compute_bw_meter_flags(req);
2604
2605 /* Find the bw_meter entry to delete */
2606 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2607 prev = x, x = x->bm_mfc_next) {
2608 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2609 &req->bu_threshold.b_time, ==)) &&
2610 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2611 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2612 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2613 break;
2614 }
2615 if (x != NULL) { /* Delete entry from the list for this MFC */
2616 if (prev != NULL)
2617 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
2618 else
2619 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2620
2621 unschedule_bw_meter(x);
2622 splx(s);
2623 /* Free the bw_meter entry */
2624 free(x, M_BWMETER);
2625 return 0;
2626 } else {
2627 splx(s);
2628 return EINVAL;
2629 }
2630 }
2631 /* NOTREACHED */
2632 }
2633
2634 /*
2635 * Perform bandwidth measurement processing that may result in an upcall
2636 */
2637 static void
2638 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2639 {
2640 struct timeval delta;
2641
2642 delta = *nowp;
2643 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2644
2645 if (x->bm_flags & BW_METER_GEQ) {
2646 /*
2647 * Processing for ">=" type of bw_meter entry
2648 */
2649 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2650 /* Reset the bw_meter entry */
2651 x->bm_start_time = *nowp;
2652 x->bm_measured.b_packets = 0;
2653 x->bm_measured.b_bytes = 0;
2654 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2655 }
2656
2657 /* Record that a packet is received */
2658 x->bm_measured.b_packets++;
2659 x->bm_measured.b_bytes += plen;
2660
2661 /*
2662 * Test if we should deliver an upcall
2663 */
2664 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2665 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2666 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2667 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2668 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2669 /* Prepare an upcall for delivery */
2670 bw_meter_prepare_upcall(x, nowp);
2671 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2672 }
2673 }
2674 } else if (x->bm_flags & BW_METER_LEQ) {
2675 /*
2676 * Processing for "<=" type of bw_meter entry
2677 */
2678 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2679 /*
2680 * We are behind time with the multicast forwarding table
2681 * scanning for "<=" type of bw_meter entries, so test now
2682 * if we should deliver an upcall.
2683 */
2684 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2685 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2686 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2687 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2688 /* Prepare an upcall for delivery */
2689 bw_meter_prepare_upcall(x, nowp);
2690 }
2691 /* Reschedule the bw_meter entry */
2692 unschedule_bw_meter(x);
2693 schedule_bw_meter(x, nowp);
2694 }
2695
2696 /* Record that a packet is received */
2697 x->bm_measured.b_packets++;
2698 x->bm_measured.b_bytes += plen;
2699
2700 /*
2701 * Test if we should restart the measuring interval
2702 */
2703 if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2704 x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2705 (x->bm_flags & BW_METER_UNIT_BYTES &&
2706 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2707 /* Don't restart the measuring interval */
2708 } else {
2709 /* Do restart the measuring interval */
2710 /*
2711 * XXX: note that we don't unschedule and schedule, because this
2712 * might be too much overhead per packet. Instead, when we process
2713 * all entries for a given timer hash bin, we check whether it is
2714 * really a timeout. If not, we reschedule at that time.
2715 */
2716 x->bm_start_time = *nowp;
2717 x->bm_measured.b_packets = 0;
2718 x->bm_measured.b_bytes = 0;
2719 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2720 }
2721 }
2722 }
2723
2724 /*
2725 * Prepare a bandwidth-related upcall
2726 */
2727 static void
2728 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2729 {
2730 struct timeval delta;
2731 struct bw_upcall *u;
2732
2733 /*
2734 * Compute the measured time interval
2735 */
2736 delta = *nowp;
2737 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2738
2739 /*
2740 * If there are too many pending upcalls, deliver them now
2741 */
2742 if (bw_upcalls_n >= BW_UPCALLS_MAX)
2743 bw_upcalls_send();
2744
2745 /*
2746 * Set the bw_upcall entry
2747 */
2748 u = &bw_upcalls[bw_upcalls_n++];
2749 u->bu_src = x->bm_mfc->mfc_origin;
2750 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2751 u->bu_threshold.b_time = x->bm_threshold.b_time;
2752 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2753 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2754 u->bu_measured.b_time = delta;
2755 u->bu_measured.b_packets = x->bm_measured.b_packets;
2756 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2757 u->bu_flags = 0;
2758 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2759 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2760 if (x->bm_flags & BW_METER_UNIT_BYTES)
2761 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2762 if (x->bm_flags & BW_METER_GEQ)
2763 u->bu_flags |= BW_UPCALL_GEQ;
2764 if (x->bm_flags & BW_METER_LEQ)
2765 u->bu_flags |= BW_UPCALL_LEQ;
2766 }
2767
2768 /*
2769 * Send the pending bandwidth-related upcalls
2770 */
2771 static void
2772 bw_upcalls_send(void)
2773 {
2774 struct mbuf *m;
2775 int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2776 struct sockaddr_in k_igmpsrc = {
2777 .sin_len = sizeof(k_igmpsrc),
2778 .sin_family = AF_INET,
2779 };
2780 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2781 0, /* unused2 */
2782 IGMPMSG_BW_UPCALL,/* im_msgtype */
2783 0, /* im_mbz */
2784 0, /* im_vif */
2785 0, /* unused3 */
2786 { 0 }, /* im_src */
2787 { 0 } }; /* im_dst */
2788
2789 if (bw_upcalls_n == 0)
2790 return; /* No pending upcalls */
2791
2792 bw_upcalls_n = 0;
2793
2794 /*
2795 * Allocate a new mbuf, initialize it with the header and
2796 * the payload for the pending calls.
2797 */
2798 MGETHDR(m, M_DONTWAIT, MT_HEADER);
2799 if (m == NULL) {
2800 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2801 return;
2802 }
2803
2804 m->m_len = m->m_pkthdr.len = 0;
2805 m_copyback(m, 0, sizeof(struct igmpmsg), (void *)&igmpmsg);
2806 m_copyback(m, sizeof(struct igmpmsg), len, (void *)&bw_upcalls[0]);
2807
2808 /*
2809 * Send the upcalls
2810 * XXX do we need to set the address in k_igmpsrc ?
2811 */
2812 mrtstat.mrts_upcalls++;
2813 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2814 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2815 ++mrtstat.mrts_upq_sockfull;
2816 }
2817 }
2818
2819 /*
2820 * Compute the timeout hash value for the bw_meter entries
2821 */
2822 #define BW_METER_TIMEHASH(bw_meter, hash) \
2823 do { \
2824 struct timeval next_timeval = (bw_meter)->bm_start_time; \
2825 \
2826 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2827 (hash) = next_timeval.tv_sec; \
2828 if (next_timeval.tv_usec) \
2829 (hash)++; /* XXX: make sure we don't timeout early */ \
2830 (hash) %= BW_METER_BUCKETS; \
2831 } while (/*CONSTCOND*/ 0)
2832
2833 /*
2834 * Schedule a timer to process periodically bw_meter entry of type "<="
2835 * by linking the entry in the proper hash bucket.
2836 */
2837 static void
2838 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2839 {
2840 int time_hash;
2841
2842 if (!(x->bm_flags & BW_METER_LEQ))
2843 return; /* XXX: we schedule timers only for "<=" entries */
2844
2845 /*
2846 * Reset the bw_meter entry
2847 */
2848 x->bm_start_time = *nowp;
2849 x->bm_measured.b_packets = 0;
2850 x->bm_measured.b_bytes = 0;
2851 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2852
2853 /*
2854 * Compute the timeout hash value and insert the entry
2855 */
2856 BW_METER_TIMEHASH(x, time_hash);
2857 x->bm_time_next = bw_meter_timers[time_hash];
2858 bw_meter_timers[time_hash] = x;
2859 x->bm_time_hash = time_hash;
2860 }
2861
2862 /*
2863 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2864 * by removing the entry from the proper hash bucket.
2865 */
2866 static void
2867 unschedule_bw_meter(struct bw_meter *x)
2868 {
2869 int time_hash;
2870 struct bw_meter *prev, *tmp;
2871
2872 if (!(x->bm_flags & BW_METER_LEQ))
2873 return; /* XXX: we schedule timers only for "<=" entries */
2874
2875 /*
2876 * Compute the timeout hash value and delete the entry
2877 */
2878 time_hash = x->bm_time_hash;
2879 if (time_hash >= BW_METER_BUCKETS)
2880 return; /* Entry was not scheduled */
2881
2882 for (prev = NULL, tmp = bw_meter_timers[time_hash];
2883 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2884 if (tmp == x)
2885 break;
2886
2887 if (tmp == NULL)
2888 panic("unschedule_bw_meter: bw_meter entry not found");
2889
2890 if (prev != NULL)
2891 prev->bm_time_next = x->bm_time_next;
2892 else
2893 bw_meter_timers[time_hash] = x->bm_time_next;
2894
2895 x->bm_time_next = NULL;
2896 x->bm_time_hash = BW_METER_BUCKETS;
2897 }
2898
2899 /*
2900 * Process all "<=" type of bw_meter that should be processed now,
2901 * and for each entry prepare an upcall if necessary. Each processed
2902 * entry is rescheduled again for the (periodic) processing.
2903 *
2904 * This is run periodically (once per second normally). On each round,
2905 * all the potentially matching entries are in the hash slot that we are
2906 * looking at.
2907 */
2908 static void
2909 bw_meter_process(void)
2910 {
2911 int s;
2912 static uint32_t last_tv_sec; /* last time we processed this */
2913
2914 uint32_t loops;
2915 int i;
2916 struct timeval now, process_endtime;
2917
2918 microtime(&now);
2919 if (last_tv_sec == now.tv_sec)
2920 return; /* nothing to do */
2921
2922 loops = now.tv_sec - last_tv_sec;
2923 last_tv_sec = now.tv_sec;
2924 if (loops > BW_METER_BUCKETS)
2925 loops = BW_METER_BUCKETS;
2926
2927 s = splsoftnet();
2928 /*
2929 * Process all bins of bw_meter entries from the one after the last
2930 * processed to the current one. On entry, i points to the last bucket
2931 * visited, so we need to increment i at the beginning of the loop.
2932 */
2933 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2934 struct bw_meter *x, *tmp_list;
2935
2936 if (++i >= BW_METER_BUCKETS)
2937 i = 0;
2938
2939 /* Disconnect the list of bw_meter entries from the bin */
2940 tmp_list = bw_meter_timers[i];
2941 bw_meter_timers[i] = NULL;
2942
2943 /* Process the list of bw_meter entries */
2944 while (tmp_list != NULL) {
2945 x = tmp_list;
2946 tmp_list = tmp_list->bm_time_next;
2947
2948 /* Test if the time interval is over */
2949 process_endtime = x->bm_start_time;
2950 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2951 if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2952 /* Not yet: reschedule, but don't reset */
2953 int time_hash;
2954
2955 BW_METER_TIMEHASH(x, time_hash);
2956 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2957 /*
2958 * XXX: somehow the bin processing is a bit ahead of time.
2959 * Put the entry in the next bin.
2960 */
2961 if (++time_hash >= BW_METER_BUCKETS)
2962 time_hash = 0;
2963 }
2964 x->bm_time_next = bw_meter_timers[time_hash];
2965 bw_meter_timers[time_hash] = x;
2966 x->bm_time_hash = time_hash;
2967
2968 continue;
2969 }
2970
2971 /*
2972 * Test if we should deliver an upcall
2973 */
2974 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2975 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2976 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2977 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2978 /* Prepare an upcall for delivery */
2979 bw_meter_prepare_upcall(x, &now);
2980 }
2981
2982 /*
2983 * Reschedule for next processing
2984 */
2985 schedule_bw_meter(x, &now);
2986 }
2987 }
2988
2989 /* Send all upcalls that are pending delivery */
2990 bw_upcalls_send();
2991
2992 splx(s);
2993 }
2994
2995 /*
2996 * A periodic function for sending all upcalls that are pending delivery
2997 */
2998 static void
2999 expire_bw_upcalls_send(void *unused)
3000 {
3001 int s;
3002
3003 s = splsoftnet();
3004 bw_upcalls_send();
3005 splx(s);
3006
3007 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
3008 expire_bw_upcalls_send, NULL);
3009 }
3010
3011 /*
3012 * A periodic function for periodic scanning of the multicast forwarding
3013 * table for processing all "<=" bw_meter entries.
3014 */
3015 static void
3016 expire_bw_meter_process(void *unused)
3017 {
3018 if (mrt_api_config & MRT_MFC_BW_UPCALL)
3019 bw_meter_process();
3020
3021 callout_reset(&bw_meter_ch, BW_METER_PERIOD,
3022 expire_bw_meter_process, NULL);
3023 }
3024
3025 /*
3026 * End of bandwidth monitoring code
3027 */
3028
3029 #ifdef PIM
3030 /*
3031 * Send the packet up to the user daemon, or eventually do kernel encapsulation
3032 */
3033 static int
3034 pim_register_send(struct ip *ip, struct vif *vifp,
3035 struct mbuf *m, struct mfc *rt)
3036 {
3037 struct mbuf *mb_copy, *mm;
3038
3039 if (mrtdebug & DEBUG_PIM)
3040 log(LOG_DEBUG, "pim_register_send: ");
3041
3042 mb_copy = pim_register_prepare(ip, m);
3043 if (mb_copy == NULL)
3044 return ENOBUFS;
3045
3046 /*
3047 * Send all the fragments. Note that the mbuf for each fragment
3048 * is freed by the sending machinery.
3049 */
3050 for (mm = mb_copy; mm; mm = mb_copy) {
3051 mb_copy = mm->m_nextpkt;
3052 mm->m_nextpkt = NULL;
3053 mm = m_pullup(mm, sizeof(struct ip));
3054 if (mm != NULL) {
3055 ip = mtod(mm, struct ip *);
3056 if ((mrt_api_config & MRT_MFC_RP) &&
3057 !in_nullhost(rt->mfc_rp)) {
3058 pim_register_send_rp(ip, vifp, mm, rt);
3059 } else {
3060 pim_register_send_upcall(ip, vifp, mm, rt);
3061 }
3062 }
3063 }
3064
3065 return 0;
3066 }
3067
3068 /*
3069 * Return a copy of the data packet that is ready for PIM Register
3070 * encapsulation.
3071 * XXX: Note that in the returned copy the IP header is a valid one.
3072 */
3073 static struct mbuf *
3074 pim_register_prepare(struct ip *ip, struct mbuf *m)
3075 {
3076 struct mbuf *mb_copy = NULL;
3077 int mtu;
3078
3079 /* Take care of delayed checksums */
3080 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3081 in_delayed_cksum(m);
3082 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3083 }
3084
3085 /*
3086 * Copy the old packet & pullup its IP header into the
3087 * new mbuf so we can modify it.
3088 */
3089 mb_copy = m_copypacket(m, M_DONTWAIT);
3090 if (mb_copy == NULL)
3091 return NULL;
3092 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3093 if (mb_copy == NULL)
3094 return NULL;
3095
3096 /* take care of the TTL */
3097 ip = mtod(mb_copy, struct ip *);
3098 --ip->ip_ttl;
3099
3100 /* Compute the MTU after the PIM Register encapsulation */
3101 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3102
3103 if (ntohs(ip->ip_len) <= mtu) {
3104 /* Turn the IP header into a valid one */
3105 ip->ip_sum = 0;
3106 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3107 } else {
3108 /* Fragment the packet */
3109 if (ip_fragment(mb_copy, NULL, mtu) != 0) {
3110 /* XXX: mb_copy was freed by ip_fragment() */
3111 return NULL;
3112 }
3113 }
3114 return mb_copy;
3115 }
3116
3117 /*
3118 * Send an upcall with the data packet to the user-level process.
3119 */
3120 static int
3121 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3122 struct mbuf *mb_copy, struct mfc *rt)
3123 {
3124 struct mbuf *mb_first;
3125 int len = ntohs(ip->ip_len);
3126 struct igmpmsg *im;
3127 struct sockaddr_in k_igmpsrc = {
3128 .sin_len = sizeof(k_igmpsrc),
3129 .sin_family = AF_INET,
3130 };
3131
3132 /*
3133 * Add a new mbuf with an upcall header
3134 */
3135 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3136 if (mb_first == NULL) {
3137 m_freem(mb_copy);
3138 return ENOBUFS;
3139 }
3140 mb_first->m_data += max_linkhdr;
3141 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3142 mb_first->m_len = sizeof(struct igmpmsg);
3143 mb_first->m_next = mb_copy;
3144
3145 /* Send message to routing daemon */
3146 im = mtod(mb_first, struct igmpmsg *);
3147 im->im_msgtype = IGMPMSG_WHOLEPKT;
3148 im->im_mbz = 0;
3149 im->im_vif = vifp - viftable;
3150 im->im_src = ip->ip_src;
3151 im->im_dst = ip->ip_dst;
3152
3153 k_igmpsrc.sin_addr = ip->ip_src;
3154
3155 mrtstat.mrts_upcalls++;
3156
3157 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3158 if (mrtdebug & DEBUG_PIM)
3159 log(LOG_WARNING,
3160 "mcast: pim_register_send_upcall: ip_mrouter socket queue full");
3161 ++mrtstat.mrts_upq_sockfull;
3162 return ENOBUFS;
3163 }
3164
3165 /* Keep statistics */
3166 pimstat.pims_snd_registers_msgs++;
3167 pimstat.pims_snd_registers_bytes += len;
3168
3169 return 0;
3170 }
3171
3172 /*
3173 * Encapsulate the data packet in PIM Register message and send it to the RP.
3174 */
3175 static int
3176 pim_register_send_rp(struct ip *ip, struct vif *vifp,
3177 struct mbuf *mb_copy, struct mfc *rt)
3178 {
3179 struct mbuf *mb_first;
3180 struct ip *ip_outer;
3181 struct pim_encap_pimhdr *pimhdr;
3182 int len = ntohs(ip->ip_len);
3183 vifi_t vifi = rt->mfc_parent;
3184
3185 if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
3186 m_freem(mb_copy);
3187 return EADDRNOTAVAIL; /* The iif vif is invalid */
3188 }
3189
3190 /*
3191 * Add a new mbuf with the encapsulating header
3192 */
3193 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3194 if (mb_first == NULL) {
3195 m_freem(mb_copy);
3196 return ENOBUFS;
3197 }
3198 mb_first->m_data += max_linkhdr;
3199 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3200 mb_first->m_next = mb_copy;
3201
3202 mb_first->m_pkthdr.len = len + mb_first->m_len;
3203
3204 /*
3205 * Fill in the encapsulating IP and PIM header
3206 */
3207 ip_outer = mtod(mb_first, struct ip *);
3208 *ip_outer = pim_encap_iphdr;
3209 if (mb_first->m_pkthdr.len < IP_MINFRAGSIZE)
3210 ip_outer->ip_id = 0;
3211 else
3212 ip_outer->ip_id = ip_newid(NULL);
3213 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
3214 sizeof(pim_encap_pimhdr));
3215 ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3216 ip_outer->ip_dst = rt->mfc_rp;
3217 /*
3218 * Copy the inner header TOS to the outer header, and take care of the
3219 * IP_DF bit.
3220 */
3221 ip_outer->ip_tos = ip->ip_tos;
3222 if (ntohs(ip->ip_off) & IP_DF)
3223 ip_outer->ip_off |= htons(IP_DF);
3224 pimhdr = (struct pim_encap_pimhdr *)((char *)ip_outer
3225 + sizeof(pim_encap_iphdr));
3226 *pimhdr = pim_encap_pimhdr;
3227 /* If the iif crosses a border, set the Border-bit */
3228 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3229 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3230
3231 mb_first->m_data += sizeof(pim_encap_iphdr);
3232 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3233 mb_first->m_data -= sizeof(pim_encap_iphdr);
3234
3235 if (vifp->v_rate_limit == 0)
3236 tbf_send_packet(vifp, mb_first);
3237 else
3238 tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len));
3239
3240 /* Keep statistics */
3241 pimstat.pims_snd_registers_msgs++;
3242 pimstat.pims_snd_registers_bytes += len;
3243
3244 return 0;
3245 }
3246
3247 /*
3248 * PIM-SMv2 and PIM-DM messages processing.
3249 * Receives and verifies the PIM control messages, and passes them
3250 * up to the listening socket, using rip_input().
3251 * The only message with special processing is the PIM_REGISTER message
3252 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3253 * is passed to if_simloop().
3254 */
3255 void
3256 pim_input(struct mbuf *m, ...)
3257 {
3258 struct ip *ip = mtod(m, struct ip *);
3259 struct pim *pim;
3260 int minlen;
3261 int datalen;
3262 int ip_tos;
3263 int proto;
3264 int iphlen;
3265 va_list ap;
3266
3267 va_start(ap, m);
3268 iphlen = va_arg(ap, int);
3269 proto = va_arg(ap, int);
3270 va_end(ap);
3271
3272 datalen = ntohs(ip->ip_len) - iphlen;
3273
3274 /* Keep statistics */
3275 pimstat.pims_rcv_total_msgs++;
3276 pimstat.pims_rcv_total_bytes += datalen;
3277
3278 /*
3279 * Validate lengths
3280 */
3281 if (datalen < PIM_MINLEN) {
3282 pimstat.pims_rcv_tooshort++;
3283 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3284 datalen, (u_long)ip->ip_src.s_addr);
3285 m_freem(m);
3286 return;
3287 }
3288
3289 /*
3290 * If the packet is at least as big as a REGISTER, go agead
3291 * and grab the PIM REGISTER header size, to avoid another
3292 * possible m_pullup() later.
3293 *
3294 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
3295 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3296 */
3297 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3298 /*
3299 * Get the IP and PIM headers in contiguous memory, and
3300 * possibly the PIM REGISTER header.
3301 */
3302 if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3303 (m = m_pullup(m, minlen)) == NULL) {
3304 log(LOG_ERR, "pim_input: m_pullup failure\n");
3305 return;
3306 }
3307 /* m_pullup() may have given us a new mbuf so reset ip. */
3308 ip = mtod(m, struct ip *);
3309 ip_tos = ip->ip_tos;
3310
3311 /* adjust mbuf to point to the PIM header */
3312 m->m_data += iphlen;
3313 m->m_len -= iphlen;
3314 pim = mtod(m, struct pim *);
3315
3316 /*
3317 * Validate checksum. If PIM REGISTER, exclude the data packet.
3318 *
3319 * XXX: some older PIMv2 implementations don't make this distinction,
3320 * so for compatibility reason perform the checksum over part of the
3321 * message, and if error, then over the whole message.
3322 */
3323 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3324 /* do nothing, checksum okay */
3325 } else if (in_cksum(m, datalen)) {
3326 pimstat.pims_rcv_badsum++;
3327 if (mrtdebug & DEBUG_PIM)
3328 log(LOG_DEBUG, "pim_input: invalid checksum");
3329 m_freem(m);
3330 return;
3331 }
3332
3333 /* PIM version check */
3334 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3335 pimstat.pims_rcv_badversion++;
3336 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3337 PIM_VT_V(pim->pim_vt), PIM_VERSION);
3338 m_freem(m);
3339 return;
3340 }
3341
3342 /* restore mbuf back to the outer IP */
3343 m->m_data -= iphlen;
3344 m->m_len += iphlen;
3345
3346 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3347 /*
3348 * Since this is a REGISTER, we'll make a copy of the register
3349 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3350 * routing daemon.
3351 */
3352 int s;
3353 struct sockaddr_in dst = {
3354 .sin_len = sizeof(dst),
3355 .sin_family = AF_INET,
3356 };
3357 struct mbuf *mcp;
3358 struct ip *encap_ip;
3359 u_int32_t *reghdr;
3360 struct ifnet *vifp;
3361
3362 s = splsoftnet();
3363 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3364 splx(s);
3365 if (mrtdebug & DEBUG_PIM)
3366 log(LOG_DEBUG,
3367 "pim_input: register vif not set: %d\n", reg_vif_num);
3368 m_freem(m);
3369 return;
3370 }
3371 /* XXX need refcnt? */
3372 vifp = viftable[reg_vif_num].v_ifp;
3373 splx(s);
3374
3375 /*
3376 * Validate length
3377 */
3378 if (datalen < PIM_REG_MINLEN) {
3379 pimstat.pims_rcv_tooshort++;
3380 pimstat.pims_rcv_badregisters++;
3381 log(LOG_ERR,
3382 "pim_input: register packet size too small %d from %lx\n",
3383 datalen, (u_long)ip->ip_src.s_addr);
3384 m_freem(m);
3385 return;
3386 }
3387
3388 reghdr = (u_int32_t *)(pim + 1);
3389 encap_ip = (struct ip *)(reghdr + 1);
3390
3391 if (mrtdebug & DEBUG_PIM) {
3392 log(LOG_DEBUG,
3393 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3394 (u_long)ntohl(encap_ip->ip_src.s_addr),
3395 (u_long)ntohl(encap_ip->ip_dst.s_addr),
3396 ntohs(encap_ip->ip_len));
3397 }
3398
3399 /* verify the version number of the inner packet */
3400 if (encap_ip->ip_v != IPVERSION) {
3401 pimstat.pims_rcv_badregisters++;
3402 if (mrtdebug & DEBUG_PIM) {
3403 log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3404 "of the inner packet\n", encap_ip->ip_v);
3405 }
3406 m_freem(m);
3407 return;
3408 }
3409
3410 /* verify the inner packet is destined to a mcast group */
3411 if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) {
3412 pimstat.pims_rcv_badregisters++;
3413 if (mrtdebug & DEBUG_PIM)
3414 log(LOG_DEBUG,
3415 "pim_input: inner packet of register is not "
3416 "multicast %lx\n",
3417 (u_long)ntohl(encap_ip->ip_dst.s_addr));
3418 m_freem(m);
3419 return;
3420 }
3421
3422 /* If a NULL_REGISTER, pass it to the daemon */
3423 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3424 goto pim_input_to_daemon;
3425
3426 /*
3427 * Copy the TOS from the outer IP header to the inner IP header.
3428 */
3429 if (encap_ip->ip_tos != ip_tos) {
3430 /* Outer TOS -> inner TOS */
3431 encap_ip->ip_tos = ip_tos;
3432 /* Recompute the inner header checksum. Sigh... */
3433
3434 /* adjust mbuf to point to the inner IP header */
3435 m->m_data += (iphlen + PIM_MINLEN);
3436 m->m_len -= (iphlen + PIM_MINLEN);
3437
3438 encap_ip->ip_sum = 0;
3439 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3440
3441 /* restore mbuf to point back to the outer IP header */
3442 m->m_data -= (iphlen + PIM_MINLEN);
3443 m->m_len += (iphlen + PIM_MINLEN);
3444 }
3445
3446 /*
3447 * Decapsulate the inner IP packet and loopback to forward it
3448 * as a normal multicast packet. Also, make a copy of the
3449 * outer_iphdr + pimhdr + reghdr + encap_iphdr
3450 * to pass to the daemon later, so it can take the appropriate
3451 * actions (e.g., send back PIM_REGISTER_STOP).
3452 * XXX: here m->m_data points to the outer IP header.
3453 */
3454 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_DONTWAIT);
3455 if (mcp == NULL) {
3456 log(LOG_ERR,
3457 "pim_input: pim register: could not copy register head\n");
3458 m_freem(m);
3459 return;
3460 }
3461
3462 /* Keep statistics */
3463 /* XXX: registers_bytes include only the encap. mcast pkt */
3464 pimstat.pims_rcv_registers_msgs++;
3465 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3466
3467 /*
3468 * forward the inner ip packet; point m_data at the inner ip.
3469 */
3470 m_adj(m, iphlen + PIM_MINLEN);
3471
3472 if (mrtdebug & DEBUG_PIM) {
3473 log(LOG_DEBUG,
3474 "pim_input: forwarding decapsulated register: "
3475 "src %lx, dst %lx, vif %d\n",
3476 (u_long)ntohl(encap_ip->ip_src.s_addr),
3477 (u_long)ntohl(encap_ip->ip_dst.s_addr),
3478 reg_vif_num);
3479 }
3480 /* NB: vifp was collected above; can it change on us? */
3481 looutput(vifp, m, (struct sockaddr *)&dst, (struct rtentry *)NULL);
3482
3483 /* prepare the register head to send to the mrouting daemon */
3484 m = mcp;
3485 }
3486
3487 pim_input_to_daemon:
3488 /*
3489 * Pass the PIM message up to the daemon; if it is a Register message,
3490 * pass the 'head' only up to the daemon. This includes the
3491 * outer IP header, PIM header, PIM-Register header and the
3492 * inner IP header.
3493 * XXX: the outer IP header pkt size of a Register is not adjust to
3494 * reflect the fact that the inner multicast data is truncated.
3495 */
3496 rip_input(m, iphlen, proto);
3497
3498 return;
3499 }
3500 #endif /* PIM */
Cache object: 2eda539c9b2fe588321e3556fa506701
|