1 /* $NetBSD: ip_mroute.c,v 1.90 2005/02/26 22:45:12 perry 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 1.90 2005/02/26 22:45:12 perry 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(const 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 *, struct mbuf *);
216 static int get_version(struct mbuf *);
217 static int set_assert(struct mbuf *);
218 static int get_assert(struct mbuf *);
219 static int add_vif(struct mbuf *);
220 static int del_vif(struct mbuf *);
221 static void update_mfc_params(struct mfc *, struct mfcctl2 *);
222 static void init_mfc_params(struct mfc *, struct mfcctl2 *);
223 static void expire_mfc(struct mfc *);
224 static int add_mfc(struct mbuf *);
225 #ifdef UPCALL_TIMING
226 static void collate(struct timeval *);
227 #endif
228 static int del_mfc(struct mbuf *);
229 static int set_api_config(struct mbuf *); /* chose API capabilities */
230 static int get_api_support(struct mbuf *);
231 static int get_api_config(struct mbuf *);
232 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
233 static void expire_upcalls(void *);
234 #ifdef RSVP_ISI
235 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
236 #else
237 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *);
238 #endif
239 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
240 static void encap_send(struct ip *, struct vif *, struct mbuf *);
241 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_int32_t);
242 static void tbf_queue(struct vif *, struct mbuf *);
243 static void tbf_process_q(struct vif *);
244 static void tbf_reprocess_q(void *);
245 static int tbf_dq_sel(struct vif *, struct ip *);
246 static void tbf_send_packet(struct vif *, struct mbuf *);
247 static void tbf_update_tokens(struct vif *);
248 static int priority(struct vif *, struct ip *);
249
250 /*
251 * Bandwidth monitoring
252 */
253 static void free_bw_list(struct bw_meter *);
254 static int add_bw_upcall(struct mbuf *);
255 static int del_bw_upcall(struct mbuf *);
256 static void bw_meter_receive_packet(struct bw_meter *, int , struct timeval *);
257 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
258 static void bw_upcalls_send(void);
259 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
260 static void unschedule_bw_meter(struct bw_meter *);
261 static void bw_meter_process(void);
262 static void expire_bw_upcalls_send(void *);
263 static void expire_bw_meter_process(void *);
264
265 #ifdef PIM
266 static int pim_register_send(struct ip *, struct vif *,
267 struct mbuf *, struct mfc *);
268 static int pim_register_send_rp(struct ip *, struct vif *,
269 struct mbuf *, struct mfc *);
270 static int pim_register_send_upcall(struct ip *, struct vif *,
271 struct mbuf *, struct mfc *);
272 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
273 #endif
274
275 /*
276 * 'Interfaces' associated with decapsulator (so we can tell
277 * packets that went through it from ones that get reflected
278 * by a broken gateway). These interfaces are never linked into
279 * the system ifnet list & no routes point to them. I.e., packets
280 * can't be sent this way. They only exist as a placeholder for
281 * multicast source verification.
282 */
283 #if 0
284 struct ifnet multicast_decap_if[MAXVIFS];
285 #endif
286
287 #define ENCAP_TTL 64
288 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */
289
290 /* prototype IP hdr for encapsulated packets */
291 struct ip multicast_encap_iphdr = {
292 #if BYTE_ORDER == LITTLE_ENDIAN
293 sizeof(struct ip) >> 2, IPVERSION,
294 #else
295 IPVERSION, sizeof(struct ip) >> 2,
296 #endif
297 0, /* tos */
298 sizeof(struct ip), /* total length */
299 0, /* id */
300 0, /* frag offset */
301 ENCAP_TTL, ENCAP_PROTO,
302 0, /* checksum */
303 };
304
305 /*
306 * Bandwidth meter variables and constants
307 */
308
309 /*
310 * Pending timeouts are stored in a hash table, the key being the
311 * expiration time. Periodically, the entries are analysed and processed.
312 */
313 #define BW_METER_BUCKETS 1024
314 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
315 struct callout bw_meter_ch;
316 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
317
318 /*
319 * Pending upcalls are stored in a vector which is flushed when
320 * full, or periodically
321 */
322 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX];
323 static u_int bw_upcalls_n; /* # of pending upcalls */
324 struct callout bw_upcalls_ch;
325 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
326
327 #ifdef PIM
328 struct pimstat pimstat;
329
330 /*
331 * Note: the PIM Register encapsulation adds the following in front of a
332 * data packet:
333 *
334 * struct pim_encap_hdr {
335 * struct ip ip;
336 * struct pim_encap_pimhdr pim;
337 * }
338 *
339 */
340
341 struct pim_encap_pimhdr {
342 struct pim pim;
343 uint32_t flags;
344 };
345
346 static struct ip pim_encap_iphdr = {
347 #if BYTE_ORDER == LITTLE_ENDIAN
348 sizeof(struct ip) >> 2,
349 IPVERSION,
350 #else
351 IPVERSION,
352 sizeof(struct ip) >> 2,
353 #endif
354 0, /* tos */
355 sizeof(struct ip), /* total length */
356 0, /* id */
357 0, /* frag offset */
358 ENCAP_TTL,
359 IPPROTO_PIM,
360 0, /* checksum */
361 };
362
363 static struct pim_encap_pimhdr pim_encap_pimhdr = {
364 {
365 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
366 0, /* reserved */
367 0, /* checksum */
368 },
369 0 /* flags */
370 };
371
372 static struct ifnet multicast_register_if;
373 static vifi_t reg_vif_num = VIFI_INVALID;
374 #endif /* PIM */
375
376
377 /*
378 * Private variables.
379 */
380 static vifi_t numvifs = 0;
381
382 static struct callout expire_upcalls_ch;
383
384 /*
385 * one-back cache used by vif_encapcheck to locate a tunnel's vif
386 * given a datagram's src ip address.
387 */
388 static struct in_addr last_encap_src;
389 static struct vif *last_encap_vif;
390
391 /*
392 * whether or not special PIM assert processing is enabled.
393 */
394 static int pim_assert;
395 /*
396 * Rate limit for assert notification messages, in usec
397 */
398 #define ASSERT_MSG_TIME 3000000
399
400 /*
401 * Kernel multicast routing API capabilities and setup.
402 * If more API capabilities are added to the kernel, they should be
403 * recorded in `mrt_api_support'.
404 */
405 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
406 MRT_MFC_FLAGS_BORDER_VIF |
407 MRT_MFC_RP |
408 MRT_MFC_BW_UPCALL);
409 static u_int32_t mrt_api_config = 0;
410
411 /*
412 * Find a route for a given origin IP address and Multicast group address
413 * Type of service parameter to be added in the future!!!
414 * Statistics are updated by the caller if needed
415 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
416 */
417 static struct mfc *
418 mfc_find(struct in_addr *o, struct in_addr *g)
419 {
420 struct mfc *rt;
421
422 LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
423 if (in_hosteq(rt->mfc_origin, *o) &&
424 in_hosteq(rt->mfc_mcastgrp, *g) &&
425 (rt->mfc_stall == NULL))
426 break;
427 }
428
429 return (rt);
430 }
431
432 /*
433 * Macros to compute elapsed time efficiently
434 * Borrowed from Van Jacobson's scheduling code
435 */
436 #define TV_DELTA(a, b, delta) do { \
437 int xxs; \
438 delta = (a).tv_usec - (b).tv_usec; \
439 xxs = (a).tv_sec - (b).tv_sec; \
440 switch (xxs) { \
441 case 2: \
442 delta += 1000000; \
443 /* fall through */ \
444 case 1: \
445 delta += 1000000; \
446 /* fall through */ \
447 case 0: \
448 break; \
449 default: \
450 delta += (1000000 * xxs); \
451 break; \
452 } \
453 } while (/*CONSTCOND*/ 0)
454
455 #ifdef UPCALL_TIMING
456 u_int32_t upcall_data[51];
457 #endif /* UPCALL_TIMING */
458
459 /*
460 * Handle MRT setsockopt commands to modify the multicast routing tables.
461 */
462 int
463 ip_mrouter_set(struct socket *so, int optname, struct mbuf **m)
464 {
465 int error;
466
467 if (optname != MRT_INIT && so != ip_mrouter)
468 error = ENOPROTOOPT;
469 else
470 switch (optname) {
471 case MRT_INIT:
472 error = ip_mrouter_init(so, *m);
473 break;
474 case MRT_DONE:
475 error = ip_mrouter_done();
476 break;
477 case MRT_ADD_VIF:
478 error = add_vif(*m);
479 break;
480 case MRT_DEL_VIF:
481 error = del_vif(*m);
482 break;
483 case MRT_ADD_MFC:
484 error = add_mfc(*m);
485 break;
486 case MRT_DEL_MFC:
487 error = del_mfc(*m);
488 break;
489 case MRT_ASSERT:
490 error = set_assert(*m);
491 break;
492 case MRT_API_CONFIG:
493 error = set_api_config(*m);
494 break;
495 case MRT_ADD_BW_UPCALL:
496 error = add_bw_upcall(*m);
497 break;
498 case MRT_DEL_BW_UPCALL:
499 error = del_bw_upcall(*m);
500 break;
501 default:
502 error = ENOPROTOOPT;
503 break;
504 }
505
506 if (*m)
507 m_free(*m);
508 return (error);
509 }
510
511 /*
512 * Handle MRT getsockopt commands
513 */
514 int
515 ip_mrouter_get(struct socket *so, int optname, struct mbuf **m)
516 {
517 int error;
518
519 if (so != ip_mrouter)
520 error = ENOPROTOOPT;
521 else {
522 *m = m_get(M_WAIT, MT_SOOPTS);
523 MCLAIM(*m, so->so_mowner);
524
525 switch (optname) {
526 case MRT_VERSION:
527 error = get_version(*m);
528 break;
529 case MRT_ASSERT:
530 error = get_assert(*m);
531 break;
532 case MRT_API_SUPPORT:
533 error = get_api_support(*m);
534 break;
535 case MRT_API_CONFIG:
536 error = get_api_config(*m);
537 break;
538 default:
539 error = ENOPROTOOPT;
540 break;
541 }
542
543 if (error)
544 m_free(*m);
545 }
546
547 return (error);
548 }
549
550 /*
551 * Handle ioctl commands to obtain information from the cache
552 */
553 int
554 mrt_ioctl(struct socket *so, u_long cmd, caddr_t data)
555 {
556 int error;
557
558 if (so != ip_mrouter)
559 error = EINVAL;
560 else
561 switch (cmd) {
562 case SIOCGETVIFCNT:
563 error = get_vif_cnt((struct sioc_vif_req *)data);
564 break;
565 case SIOCGETSGCNT:
566 error = get_sg_cnt((struct sioc_sg_req *)data);
567 break;
568 default:
569 error = EINVAL;
570 break;
571 }
572
573 return (error);
574 }
575
576 /*
577 * returns the packet, byte, rpf-failure count for the source group provided
578 */
579 static int
580 get_sg_cnt(struct sioc_sg_req *req)
581 {
582 int s;
583 struct mfc *rt;
584
585 s = splsoftnet();
586 rt = mfc_find(&req->src, &req->grp);
587 if (rt == NULL) {
588 splx(s);
589 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
590 return (EADDRNOTAVAIL);
591 }
592 req->pktcnt = rt->mfc_pkt_cnt;
593 req->bytecnt = rt->mfc_byte_cnt;
594 req->wrong_if = rt->mfc_wrong_if;
595 splx(s);
596
597 return (0);
598 }
599
600 /*
601 * returns the input and output packet and byte counts on the vif provided
602 */
603 static int
604 get_vif_cnt(struct sioc_vif_req *req)
605 {
606 vifi_t vifi = req->vifi;
607
608 if (vifi >= numvifs)
609 return (EINVAL);
610
611 req->icount = viftable[vifi].v_pkt_in;
612 req->ocount = viftable[vifi].v_pkt_out;
613 req->ibytes = viftable[vifi].v_bytes_in;
614 req->obytes = viftable[vifi].v_bytes_out;
615
616 return (0);
617 }
618
619 /*
620 * Enable multicast routing
621 */
622 static int
623 ip_mrouter_init(struct socket *so, struct mbuf *m)
624 {
625 int *v;
626
627 if (mrtdebug)
628 log(LOG_DEBUG,
629 "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
630 so->so_type, so->so_proto->pr_protocol);
631
632 if (so->so_type != SOCK_RAW ||
633 so->so_proto->pr_protocol != IPPROTO_IGMP)
634 return (EOPNOTSUPP);
635
636 if (m == NULL || m->m_len < sizeof(int))
637 return (EINVAL);
638
639 v = mtod(m, int *);
640 if (*v != 1)
641 return (EINVAL);
642
643 if (ip_mrouter != NULL)
644 return (EADDRINUSE);
645
646 ip_mrouter = so;
647
648 mfchashtbl =
649 hashinit(MFCTBLSIZ, HASH_LIST, M_MRTABLE, M_WAITOK, &mfchash);
650 bzero((caddr_t)nexpire, sizeof(nexpire));
651
652 pim_assert = 0;
653
654 callout_init(&expire_upcalls_ch);
655 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
656 expire_upcalls, NULL);
657
658 callout_init(&bw_upcalls_ch);
659 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
660 expire_bw_upcalls_send, NULL);
661
662 callout_init(&bw_meter_ch);
663 callout_reset(&bw_meter_ch, BW_METER_PERIOD,
664 expire_bw_meter_process, NULL);
665
666 if (mrtdebug)
667 log(LOG_DEBUG, "ip_mrouter_init\n");
668
669 return (0);
670 }
671
672 /*
673 * Disable multicast routing
674 */
675 int
676 ip_mrouter_done(void)
677 {
678 vifi_t vifi;
679 struct vif *vifp;
680 int i;
681 int s;
682
683 s = splsoftnet();
684
685 /* Clear out all the vifs currently in use. */
686 for (vifi = 0; vifi < numvifs; vifi++) {
687 vifp = &viftable[vifi];
688 if (!in_nullhost(vifp->v_lcl_addr))
689 reset_vif(vifp);
690 }
691
692 numvifs = 0;
693 pim_assert = 0;
694 mrt_api_config = 0;
695
696 callout_stop(&expire_upcalls_ch);
697 callout_stop(&bw_upcalls_ch);
698 callout_stop(&bw_meter_ch);
699
700 /*
701 * Free all multicast forwarding cache entries.
702 */
703 for (i = 0; i < MFCTBLSIZ; i++) {
704 struct mfc *rt, *nrt;
705
706 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
707 nrt = LIST_NEXT(rt, mfc_hash);
708
709 expire_mfc(rt);
710 }
711 }
712
713 bzero((caddr_t)nexpire, sizeof(nexpire));
714 free(mfchashtbl, M_MRTABLE);
715 mfchashtbl = NULL;
716
717 bw_upcalls_n = 0;
718 bzero(bw_meter_timers, sizeof(bw_meter_timers));
719
720 /* Reset de-encapsulation cache. */
721
722 ip_mrouter = NULL;
723
724 splx(s);
725
726 if (mrtdebug)
727 log(LOG_DEBUG, "ip_mrouter_done\n");
728
729 return (0);
730 }
731
732 void
733 ip_mrouter_detach(struct ifnet *ifp)
734 {
735 int vifi, i;
736 struct vif *vifp;
737 struct mfc *rt;
738 struct rtdetq *rte;
739
740 /* XXX not sure about side effect to userland routing daemon */
741 for (vifi = 0; vifi < numvifs; vifi++) {
742 vifp = &viftable[vifi];
743 if (vifp->v_ifp == ifp)
744 reset_vif(vifp);
745 }
746 for (i = 0; i < MFCTBLSIZ; i++) {
747 if (nexpire[i] == 0)
748 continue;
749 LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
750 for (rte = rt->mfc_stall; rte; rte = rte->next) {
751 if (rte->ifp == ifp)
752 rte->ifp = NULL;
753 }
754 }
755 }
756 }
757
758 static int
759 get_version(struct mbuf *m)
760 {
761 int *v = mtod(m, int *);
762
763 *v = 0x0305; /* XXX !!!! */
764 m->m_len = sizeof(int);
765 return (0);
766 }
767
768 /*
769 * Set PIM assert processing global
770 */
771 static int
772 set_assert(struct mbuf *m)
773 {
774 int *i;
775
776 if (m == NULL || m->m_len < sizeof(int))
777 return (EINVAL);
778
779 i = mtod(m, int *);
780 pim_assert = !!*i;
781 return (0);
782 }
783
784 /*
785 * Get PIM assert processing global
786 */
787 static int
788 get_assert(struct mbuf *m)
789 {
790 int *i = mtod(m, int *);
791
792 *i = pim_assert;
793 m->m_len = sizeof(int);
794 return (0);
795 }
796
797 /*
798 * Configure API capabilities
799 */
800 static int
801 set_api_config(struct mbuf *m)
802 {
803 int i;
804 u_int32_t *apival;
805
806 if (m == NULL || m->m_len < sizeof(u_int32_t))
807 return (EINVAL);
808
809 apival = mtod(m, u_int32_t *);
810
811 /*
812 * We can set the API capabilities only if it is the first operation
813 * after MRT_INIT. I.e.:
814 * - there are no vifs installed
815 * - pim_assert is not enabled
816 * - the MFC table is empty
817 */
818 if (numvifs > 0) {
819 *apival = 0;
820 return (EPERM);
821 }
822 if (pim_assert) {
823 *apival = 0;
824 return (EPERM);
825 }
826 for (i = 0; i < MFCTBLSIZ; i++) {
827 if (LIST_FIRST(&mfchashtbl[i]) != NULL) {
828 *apival = 0;
829 return (EPERM);
830 }
831 }
832
833 mrt_api_config = *apival & mrt_api_support;
834 *apival = mrt_api_config;
835
836 return (0);
837 }
838
839 /*
840 * Get API capabilities
841 */
842 static int
843 get_api_support(struct mbuf *m)
844 {
845 u_int32_t *apival;
846
847 if (m == NULL || m->m_len < sizeof(u_int32_t))
848 return (EINVAL);
849
850 apival = mtod(m, u_int32_t *);
851
852 *apival = mrt_api_support;
853
854 return (0);
855 }
856
857 /*
858 * Get API configured capabilities
859 */
860 static int
861 get_api_config(struct mbuf *m)
862 {
863 u_int32_t *apival;
864
865 if (m == NULL || m->m_len < sizeof(u_int32_t))
866 return (EINVAL);
867
868 apival = mtod(m, u_int32_t *);
869
870 *apival = mrt_api_config;
871
872 return (0);
873 }
874
875 static struct sockaddr_in sin = { sizeof(sin), AF_INET };
876
877 /*
878 * Add a vif to the vif table
879 */
880 static int
881 add_vif(struct mbuf *m)
882 {
883 struct vifctl *vifcp;
884 struct vif *vifp;
885 struct ifaddr *ifa;
886 struct ifnet *ifp;
887 struct ifreq ifr;
888 int error, s;
889
890 if (m == NULL || m->m_len < sizeof(struct vifctl))
891 return (EINVAL);
892
893 vifcp = mtod(m, struct vifctl *);
894 if (vifcp->vifc_vifi >= MAXVIFS)
895 return (EINVAL);
896 if (in_nullhost(vifcp->vifc_lcl_addr))
897 return (EADDRNOTAVAIL);
898
899 vifp = &viftable[vifcp->vifc_vifi];
900 if (!in_nullhost(vifp->v_lcl_addr))
901 return (EADDRINUSE);
902
903 /* Find the interface with an address in AF_INET family. */
904 #ifdef PIM
905 if (vifcp->vifc_flags & VIFF_REGISTER) {
906 /*
907 * XXX: Because VIFF_REGISTER does not really need a valid
908 * local interface (e.g. it could be 127.0.0.2), we don't
909 * check its address.
910 */
911 ifp = NULL;
912 } else
913 #endif
914 {
915 sin.sin_addr = vifcp->vifc_lcl_addr;
916 ifa = ifa_ifwithaddr(sintosa(&sin));
917 if (ifa == NULL)
918 return (EADDRNOTAVAIL);
919 ifp = ifa->ifa_ifp;
920 }
921
922 if (vifcp->vifc_flags & VIFF_TUNNEL) {
923 if (vifcp->vifc_flags & VIFF_SRCRT) {
924 log(LOG_ERR, "source routed tunnels not supported\n");
925 return (EOPNOTSUPP);
926 }
927
928 /* attach this vif to decapsulator dispatch table */
929 vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
930 vif_encapcheck, &vif_protosw, vifp);
931 if (!vifp->v_encap_cookie)
932 return (EINVAL);
933
934 /* Create a fake encapsulation interface. */
935 ifp = (struct ifnet *)malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK);
936 bzero(ifp, sizeof(*ifp));
937 snprintf(ifp->if_xname, sizeof(ifp->if_xname),
938 "mdecap%d", vifcp->vifc_vifi);
939
940 /* Prepare cached route entry. */
941 bzero(&vifp->v_route, sizeof(vifp->v_route));
942 #ifdef PIM
943 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
944 ifp = &multicast_register_if;
945 if (mrtdebug)
946 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
947 (void *)ifp);
948 if (reg_vif_num == VIFI_INVALID) {
949 bzero(ifp, sizeof(*ifp));
950 snprintf(ifp->if_xname, sizeof(ifp->if_xname),
951 "register_vif");
952 ifp->if_flags = IFF_LOOPBACK;
953 bzero(&vifp->v_route, sizeof(vifp->v_route));
954 reg_vif_num = vifcp->vifc_vifi;
955 }
956 #endif
957 } else {
958 /* Make sure the interface supports multicast. */
959 if ((ifp->if_flags & IFF_MULTICAST) == 0)
960 return (EOPNOTSUPP);
961
962 /* Enable promiscuous reception of all IP multicasts. */
963 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
964 satosin(&ifr.ifr_addr)->sin_family = AF_INET;
965 satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
966 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
967 if (error)
968 return (error);
969 }
970
971 s = splsoftnet();
972
973 /* Define parameters for the tbf structure. */
974 vifp->tbf_q = NULL;
975 vifp->tbf_t = &vifp->tbf_q;
976 microtime(&vifp->tbf_last_pkt_t);
977 vifp->tbf_n_tok = 0;
978 vifp->tbf_q_len = 0;
979 vifp->tbf_max_q_len = MAXQSIZE;
980
981 vifp->v_flags = vifcp->vifc_flags;
982 vifp->v_threshold = vifcp->vifc_threshold;
983 /* scaling up here allows division by 1024 in critical code */
984 vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
985 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
986 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
987 vifp->v_ifp = ifp;
988 /* Initialize per vif pkt counters. */
989 vifp->v_pkt_in = 0;
990 vifp->v_pkt_out = 0;
991 vifp->v_bytes_in = 0;
992 vifp->v_bytes_out = 0;
993
994 callout_init(&vifp->v_repq_ch);
995
996 #ifdef RSVP_ISI
997 vifp->v_rsvp_on = 0;
998 vifp->v_rsvpd = NULL;
999 #endif /* RSVP_ISI */
1000
1001 splx(s);
1002
1003 /* Adjust numvifs up if the vifi is higher than numvifs. */
1004 if (numvifs <= vifcp->vifc_vifi)
1005 numvifs = vifcp->vifc_vifi + 1;
1006
1007 if (mrtdebug)
1008 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
1009 vifcp->vifc_vifi,
1010 ntohl(vifcp->vifc_lcl_addr.s_addr),
1011 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
1012 ntohl(vifcp->vifc_rmt_addr.s_addr),
1013 vifcp->vifc_threshold,
1014 vifcp->vifc_rate_limit);
1015
1016 return (0);
1017 }
1018
1019 void
1020 reset_vif(struct vif *vifp)
1021 {
1022 struct mbuf *m, *n;
1023 struct ifnet *ifp;
1024 struct ifreq ifr;
1025
1026 callout_stop(&vifp->v_repq_ch);
1027
1028 /* detach this vif from decapsulator dispatch table */
1029 encap_detach(vifp->v_encap_cookie);
1030 vifp->v_encap_cookie = NULL;
1031
1032 /*
1033 * Free packets queued at the interface
1034 */
1035 for (m = vifp->tbf_q; m != NULL; m = n) {
1036 n = m->m_nextpkt;
1037 m_freem(m);
1038 }
1039
1040 if (vifp->v_flags & VIFF_TUNNEL) {
1041 free(vifp->v_ifp, M_MRTABLE);
1042 if (vifp == last_encap_vif) {
1043 last_encap_vif = NULL;
1044 last_encap_src = zeroin_addr;
1045 }
1046 } else if (vifp->v_flags & VIFF_REGISTER) {
1047 #ifdef PIM
1048 reg_vif_num = VIFI_INVALID;
1049 #endif
1050 } else {
1051 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
1052 satosin(&ifr.ifr_addr)->sin_family = AF_INET;
1053 satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
1054 ifp = vifp->v_ifp;
1055 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
1056 }
1057 bzero((caddr_t)vifp, sizeof(*vifp));
1058 }
1059
1060 /*
1061 * Delete a vif from the vif table
1062 */
1063 static int
1064 del_vif(struct mbuf *m)
1065 {
1066 vifi_t *vifip;
1067 struct vif *vifp;
1068 vifi_t vifi;
1069 int s;
1070
1071 if (m == NULL || m->m_len < sizeof(vifi_t))
1072 return (EINVAL);
1073
1074 vifip = mtod(m, vifi_t *);
1075 if (*vifip >= numvifs)
1076 return (EINVAL);
1077
1078 vifp = &viftable[*vifip];
1079 if (in_nullhost(vifp->v_lcl_addr))
1080 return (EADDRNOTAVAIL);
1081
1082 s = splsoftnet();
1083
1084 reset_vif(vifp);
1085
1086 /* Adjust numvifs down */
1087 for (vifi = numvifs; vifi > 0; vifi--)
1088 if (!in_nullhost(viftable[vifi - 1].v_lcl_addr))
1089 break;
1090 numvifs = vifi;
1091
1092 splx(s);
1093
1094 if (mrtdebug)
1095 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
1096
1097 return (0);
1098 }
1099
1100 /*
1101 * update an mfc entry without resetting counters and S,G addresses.
1102 */
1103 static void
1104 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1105 {
1106 int i;
1107
1108 rt->mfc_parent = mfccp->mfcc_parent;
1109 for (i = 0; i < numvifs; i++) {
1110 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1111 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1112 MRT_MFC_FLAGS_ALL;
1113 }
1114 /* set the RP address */
1115 if (mrt_api_config & MRT_MFC_RP)
1116 rt->mfc_rp = mfccp->mfcc_rp;
1117 else
1118 rt->mfc_rp = zeroin_addr;
1119 }
1120
1121 /*
1122 * fully initialize an mfc entry from the parameter.
1123 */
1124 static void
1125 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1126 {
1127 rt->mfc_origin = mfccp->mfcc_origin;
1128 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1129
1130 update_mfc_params(rt, mfccp);
1131
1132 /* initialize pkt counters per src-grp */
1133 rt->mfc_pkt_cnt = 0;
1134 rt->mfc_byte_cnt = 0;
1135 rt->mfc_wrong_if = 0;
1136 timerclear(&rt->mfc_last_assert);
1137 }
1138
1139 static void
1140 expire_mfc(struct mfc *rt)
1141 {
1142 struct rtdetq *rte, *nrte;
1143
1144 free_bw_list(rt->mfc_bw_meter);
1145
1146 for (rte = rt->mfc_stall; rte != NULL; rte = nrte) {
1147 nrte = rte->next;
1148 m_freem(rte->m);
1149 free(rte, M_MRTABLE);
1150 }
1151
1152 LIST_REMOVE(rt, mfc_hash);
1153 free(rt, M_MRTABLE);
1154 }
1155
1156 /*
1157 * Add an mfc entry
1158 */
1159 static int
1160 add_mfc(struct mbuf *m)
1161 {
1162 struct mfcctl2 mfcctl2;
1163 struct mfcctl2 *mfccp;
1164 struct mfc *rt;
1165 u_int32_t hash = 0;
1166 struct rtdetq *rte, *nrte;
1167 u_short nstl;
1168 int s;
1169 int mfcctl_size = sizeof(struct mfcctl);
1170
1171 if (mrt_api_config & MRT_API_FLAGS_ALL)
1172 mfcctl_size = sizeof(struct mfcctl2);
1173
1174 if (m == NULL || m->m_len < mfcctl_size)
1175 return (EINVAL);
1176
1177 /*
1178 * select data size depending on API version.
1179 */
1180 if (mrt_api_config & MRT_API_FLAGS_ALL) {
1181 struct mfcctl2 *mp2 = mtod(m, struct mfcctl2 *);
1182 bcopy(mp2, (caddr_t)&mfcctl2, sizeof(*mp2));
1183 } else {
1184 struct mfcctl *mp = mtod(m, struct mfcctl *);
1185 bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp));
1186 bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl),
1187 sizeof(mfcctl2) - sizeof(struct mfcctl));
1188 }
1189 mfccp = &mfcctl2;
1190
1191 s = splsoftnet();
1192 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1193
1194 /* If an entry already exists, just update the fields */
1195 if (rt) {
1196 if (mrtdebug & DEBUG_MFC)
1197 log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n",
1198 ntohl(mfccp->mfcc_origin.s_addr),
1199 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1200 mfccp->mfcc_parent);
1201
1202 update_mfc_params(rt, mfccp);
1203
1204 splx(s);
1205 return (0);
1206 }
1207
1208 /*
1209 * Find the entry for which the upcall was made and update
1210 */
1211 nstl = 0;
1212 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1213 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1214 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1215 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1216 rt->mfc_stall != NULL) {
1217 if (nstl++)
1218 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n",
1219 "multiple kernel entries",
1220 ntohl(mfccp->mfcc_origin.s_addr),
1221 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1222 mfccp->mfcc_parent, rt->mfc_stall);
1223
1224 if (mrtdebug & DEBUG_MFC)
1225 log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n",
1226 ntohl(mfccp->mfcc_origin.s_addr),
1227 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1228 mfccp->mfcc_parent, rt->mfc_stall);
1229
1230 rte = rt->mfc_stall;
1231 init_mfc_params(rt, mfccp);
1232 rt->mfc_stall = NULL;
1233
1234 rt->mfc_expire = 0; /* Don't clean this guy up */
1235 nexpire[hash]--;
1236
1237 /* free packets Qed at the end of this entry */
1238 for (; rte != NULL; rte = nrte) {
1239 nrte = rte->next;
1240 if (rte->ifp) {
1241 #ifdef RSVP_ISI
1242 ip_mdq(rte->m, rte->ifp, rt, -1);
1243 #else
1244 ip_mdq(rte->m, rte->ifp, rt);
1245 #endif /* RSVP_ISI */
1246 }
1247 m_freem(rte->m);
1248 #ifdef UPCALL_TIMING
1249 collate(&rte->t);
1250 #endif /* UPCALL_TIMING */
1251 free(rte, M_MRTABLE);
1252 }
1253 }
1254 }
1255
1256 /*
1257 * It is possible that an entry is being inserted without an upcall
1258 */
1259 if (nstl == 0) {
1260 /*
1261 * No mfc; make a new one
1262 */
1263 if (mrtdebug & DEBUG_MFC)
1264 log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n",
1265 ntohl(mfccp->mfcc_origin.s_addr),
1266 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1267 mfccp->mfcc_parent);
1268
1269 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1270 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1271 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1272 init_mfc_params(rt, mfccp);
1273 if (rt->mfc_expire)
1274 nexpire[hash]--;
1275 rt->mfc_expire = 0;
1276 break; /* XXX */
1277 }
1278 }
1279 if (rt == NULL) { /* no upcall, so make a new entry */
1280 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1281 M_NOWAIT);
1282 if (rt == NULL) {
1283 splx(s);
1284 return (ENOBUFS);
1285 }
1286
1287 init_mfc_params(rt, mfccp);
1288 rt->mfc_expire = 0;
1289 rt->mfc_stall = NULL;
1290 rt->mfc_bw_meter = NULL;
1291
1292 /* insert new entry at head of hash chain */
1293 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1294 }
1295 }
1296
1297 splx(s);
1298 return (0);
1299 }
1300
1301 #ifdef UPCALL_TIMING
1302 /*
1303 * collect delay statistics on the upcalls
1304 */
1305 static void
1306 collate(struct timeval *t)
1307 {
1308 u_int32_t d;
1309 struct timeval tp;
1310 u_int32_t delta;
1311
1312 microtime(&tp);
1313
1314 if (timercmp(t, &tp, <)) {
1315 TV_DELTA(tp, *t, delta);
1316
1317 d = delta >> 10;
1318 if (d > 50)
1319 d = 50;
1320
1321 ++upcall_data[d];
1322 }
1323 }
1324 #endif /* UPCALL_TIMING */
1325
1326 /*
1327 * Delete an mfc entry
1328 */
1329 static int
1330 del_mfc(struct mbuf *m)
1331 {
1332 struct mfcctl2 mfcctl2;
1333 struct mfcctl2 *mfccp;
1334 struct mfc *rt;
1335 int s;
1336 int mfcctl_size = sizeof(struct mfcctl);
1337 struct mfcctl *mp = mtod(m, struct mfcctl *);
1338
1339 /*
1340 * XXX: for deleting MFC entries the information in entries
1341 * of size "struct mfcctl" is sufficient.
1342 */
1343
1344 if (m == NULL || m->m_len < mfcctl_size)
1345 return (EINVAL);
1346
1347 bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp));
1348 bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl),
1349 sizeof(mfcctl2) - sizeof(struct mfcctl));
1350
1351 mfccp = &mfcctl2;
1352
1353 if (mrtdebug & DEBUG_MFC)
1354 log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
1355 ntohl(mfccp->mfcc_origin.s_addr),
1356 ntohl(mfccp->mfcc_mcastgrp.s_addr));
1357
1358 s = splsoftnet();
1359
1360 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1361 if (rt == NULL) {
1362 splx(s);
1363 return (EADDRNOTAVAIL);
1364 }
1365
1366 /*
1367 * free the bw_meter entries
1368 */
1369 free_bw_list(rt->mfc_bw_meter);
1370 rt->mfc_bw_meter = NULL;
1371
1372 LIST_REMOVE(rt, mfc_hash);
1373 free(rt, M_MRTABLE);
1374
1375 splx(s);
1376 return (0);
1377 }
1378
1379 static int
1380 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1381 {
1382 if (s) {
1383 if (sbappendaddr(&s->so_rcv, sintosa(src), mm,
1384 (struct mbuf *)NULL) != 0) {
1385 sorwakeup(s);
1386 return (0);
1387 }
1388 }
1389 m_freem(mm);
1390 return (-1);
1391 }
1392
1393 /*
1394 * IP multicast forwarding function. This function assumes that the packet
1395 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1396 * pointed to by "ifp", and the packet is to be relayed to other networks
1397 * that have members of the packet's destination IP multicast group.
1398 *
1399 * The packet is returned unscathed to the caller, unless it is
1400 * erroneous, in which case a non-zero return value tells the caller to
1401 * discard it.
1402 */
1403
1404 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */
1405 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1406
1407 int
1408 #ifdef RSVP_ISI
1409 ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo)
1410 #else
1411 ip_mforward(struct mbuf *m, struct ifnet *ifp)
1412 #endif /* RSVP_ISI */
1413 {
1414 struct ip *ip = mtod(m, struct ip *);
1415 struct mfc *rt;
1416 static int srctun = 0;
1417 struct mbuf *mm;
1418 int s;
1419 vifi_t vifi;
1420
1421 if (mrtdebug & DEBUG_FORWARD)
1422 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1423 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1424
1425 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1426 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1427 /*
1428 * Packet arrived via a physical interface or
1429 * an encapsulated tunnel or a register_vif.
1430 */
1431 } else {
1432 /*
1433 * Packet arrived through a source-route tunnel.
1434 * Source-route tunnels are no longer supported.
1435 */
1436 if ((srctun++ % 1000) == 0)
1437 log(LOG_ERR,
1438 "ip_mforward: received source-routed packet from %x\n",
1439 ntohl(ip->ip_src.s_addr));
1440
1441 return (1);
1442 }
1443
1444 #ifdef RSVP_ISI
1445 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1446 if (ip->ip_ttl < 255)
1447 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1448 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1449 struct vif *vifp = viftable + vifi;
1450 printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n",
1451 ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
1452 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1453 vifp->v_ifp->if_xname);
1454 }
1455 return (ip_mdq(m, ifp, (struct mfc *)NULL, vifi));
1456 }
1457 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1458 printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n",
1459 ntohl(ip->ip_src), ntohl(ip->ip_dst));
1460 }
1461 #endif /* RSVP_ISI */
1462
1463 /*
1464 * Don't forward a packet with time-to-live of zero or one,
1465 * or a packet destined to a local-only group.
1466 */
1467 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1468 return (0);
1469
1470 /*
1471 * Determine forwarding vifs from the forwarding cache table
1472 */
1473 s = splsoftnet();
1474 ++mrtstat.mrts_mfc_lookups;
1475 rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1476
1477 /* Entry exists, so forward if necessary */
1478 if (rt != NULL) {
1479 splx(s);
1480 #ifdef RSVP_ISI
1481 return (ip_mdq(m, ifp, rt, -1));
1482 #else
1483 return (ip_mdq(m, ifp, rt));
1484 #endif /* RSVP_ISI */
1485 } else {
1486 /*
1487 * If we don't have a route for packet's origin,
1488 * Make a copy of the packet & send message to routing daemon
1489 */
1490
1491 struct mbuf *mb0;
1492 struct rtdetq *rte;
1493 u_int32_t hash;
1494 int hlen = ip->ip_hl << 2;
1495 #ifdef UPCALL_TIMING
1496 struct timeval tp;
1497
1498 microtime(&tp);
1499 #endif /* UPCALL_TIMING */
1500
1501 ++mrtstat.mrts_mfc_misses;
1502
1503 mrtstat.mrts_no_route++;
1504 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1505 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1506 ntohl(ip->ip_src.s_addr),
1507 ntohl(ip->ip_dst.s_addr));
1508
1509 /*
1510 * Allocate mbufs early so that we don't do extra work if we are
1511 * just going to fail anyway. Make sure to pullup the header so
1512 * that other people can't step on it.
1513 */
1514 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1515 M_NOWAIT);
1516 if (rte == NULL) {
1517 splx(s);
1518 return (ENOBUFS);
1519 }
1520 mb0 = m_copy(m, 0, M_COPYALL);
1521 M_PULLUP(mb0, hlen);
1522 if (mb0 == NULL) {
1523 free(rte, M_MRTABLE);
1524 splx(s);
1525 return (ENOBUFS);
1526 }
1527
1528 /* is there an upcall waiting for this flow? */
1529 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1530 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1531 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1532 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1533 rt->mfc_stall != NULL)
1534 break;
1535 }
1536
1537 if (rt == NULL) {
1538 int i;
1539 struct igmpmsg *im;
1540
1541 /*
1542 * Locate the vifi for the incoming interface for
1543 * this packet.
1544 * If none found, drop packet.
1545 */
1546 for (vifi = 0; vifi < numvifs &&
1547 viftable[vifi].v_ifp != ifp; vifi++)
1548 ;
1549 if (vifi >= numvifs) /* vif not found, drop packet */
1550 goto non_fatal;
1551
1552 /* no upcall, so make a new entry */
1553 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1554 M_NOWAIT);
1555 if (rt == NULL)
1556 goto fail;
1557
1558 /*
1559 * Make a copy of the header to send to the user level
1560 * process
1561 */
1562 mm = m_copy(m, 0, hlen);
1563 M_PULLUP(mm, hlen);
1564 if (mm == NULL)
1565 goto fail1;
1566
1567 /*
1568 * Send message to routing daemon to install
1569 * a route into the kernel table
1570 */
1571
1572 im = mtod(mm, struct igmpmsg *);
1573 im->im_msgtype = IGMPMSG_NOCACHE;
1574 im->im_mbz = 0;
1575 im->im_vif = vifi;
1576
1577 mrtstat.mrts_upcalls++;
1578
1579 sin.sin_addr = ip->ip_src;
1580 if (socket_send(ip_mrouter, mm, &sin) < 0) {
1581 log(LOG_WARNING,
1582 "ip_mforward: ip_mrouter socket queue full\n");
1583 ++mrtstat.mrts_upq_sockfull;
1584 fail1:
1585 free(rt, M_MRTABLE);
1586 fail:
1587 free(rte, M_MRTABLE);
1588 m_freem(mb0);
1589 splx(s);
1590 return (ENOBUFS);
1591 }
1592
1593 /* insert new entry at head of hash chain */
1594 rt->mfc_origin = ip->ip_src;
1595 rt->mfc_mcastgrp = ip->ip_dst;
1596 rt->mfc_pkt_cnt = 0;
1597 rt->mfc_byte_cnt = 0;
1598 rt->mfc_wrong_if = 0;
1599 rt->mfc_expire = UPCALL_EXPIRE;
1600 nexpire[hash]++;
1601 for (i = 0; i < numvifs; i++) {
1602 rt->mfc_ttls[i] = 0;
1603 rt->mfc_flags[i] = 0;
1604 }
1605 rt->mfc_parent = -1;
1606
1607 /* clear the RP address */
1608 rt->mfc_rp = zeroin_addr;
1609
1610 rt->mfc_bw_meter = NULL;
1611
1612 /* link into table */
1613 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1614 /* Add this entry to the end of the queue */
1615 rt->mfc_stall = rte;
1616 } else {
1617 /* determine if q has overflowed */
1618 struct rtdetq **p;
1619 int npkts = 0;
1620
1621 /*
1622 * XXX ouch! we need to append to the list, but we
1623 * only have a pointer to the front, so we have to
1624 * scan the entire list every time.
1625 */
1626 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1627 if (++npkts > MAX_UPQ) {
1628 mrtstat.mrts_upq_ovflw++;
1629 non_fatal:
1630 free(rte, M_MRTABLE);
1631 m_freem(mb0);
1632 splx(s);
1633 return (0);
1634 }
1635
1636 /* Add this entry to the end of the queue */
1637 *p = rte;
1638 }
1639
1640 rte->next = NULL;
1641 rte->m = mb0;
1642 rte->ifp = ifp;
1643 #ifdef UPCALL_TIMING
1644 rte->t = tp;
1645 #endif /* UPCALL_TIMING */
1646
1647 splx(s);
1648
1649 return (0);
1650 }
1651 }
1652
1653
1654 /*ARGSUSED*/
1655 static void
1656 expire_upcalls(void *v)
1657 {
1658 int i;
1659 int s;
1660
1661 s = splsoftnet();
1662
1663 for (i = 0; i < MFCTBLSIZ; i++) {
1664 struct mfc *rt, *nrt;
1665
1666 if (nexpire[i] == 0)
1667 continue;
1668
1669 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1670 nrt = LIST_NEXT(rt, mfc_hash);
1671
1672 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1673 continue;
1674 nexpire[i]--;
1675
1676 /*
1677 * free the bw_meter entries
1678 */
1679 while (rt->mfc_bw_meter != NULL) {
1680 struct bw_meter *x = rt->mfc_bw_meter;
1681
1682 rt->mfc_bw_meter = x->bm_mfc_next;
1683 free(x, M_BWMETER);
1684 }
1685
1686 ++mrtstat.mrts_cache_cleanups;
1687 if (mrtdebug & DEBUG_EXPIRE)
1688 log(LOG_DEBUG,
1689 "expire_upcalls: expiring (%x %x)\n",
1690 ntohl(rt->mfc_origin.s_addr),
1691 ntohl(rt->mfc_mcastgrp.s_addr));
1692
1693 expire_mfc(rt);
1694 }
1695 }
1696
1697 splx(s);
1698 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1699 expire_upcalls, NULL);
1700 }
1701
1702 /*
1703 * Packet forwarding routine once entry in the cache is made
1704 */
1705 static int
1706 #ifdef RSVP_ISI
1707 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1708 #else
1709 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt)
1710 #endif /* RSVP_ISI */
1711 {
1712 struct ip *ip = mtod(m, struct ip *);
1713 vifi_t vifi;
1714 struct vif *vifp;
1715 int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1716
1717 /*
1718 * Macro to send packet on vif. Since RSVP packets don't get counted on
1719 * input, they shouldn't get counted on output, so statistics keeping is
1720 * separate.
1721 */
1722 #define MC_SEND(ip, vifp, m) do { \
1723 if ((vifp)->v_flags & VIFF_TUNNEL) \
1724 encap_send((ip), (vifp), (m)); \
1725 else \
1726 phyint_send((ip), (vifp), (m)); \
1727 } while (/*CONSTCOND*/ 0)
1728
1729 #ifdef RSVP_ISI
1730 /*
1731 * If xmt_vif is not -1, send on only the requested vif.
1732 *
1733 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1734 */
1735 if (xmt_vif < numvifs) {
1736 #ifdef PIM
1737 if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1738 pim_register_send(ip, viftable + xmt_vif, m, rt);
1739 else
1740 #endif
1741 MC_SEND(ip, viftable + xmt_vif, m);
1742 return (1);
1743 }
1744 #endif /* RSVP_ISI */
1745
1746 /*
1747 * Don't forward if it didn't arrive from the parent vif for its origin.
1748 */
1749 vifi = rt->mfc_parent;
1750 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1751 /* came in the wrong interface */
1752 if (mrtdebug & DEBUG_FORWARD)
1753 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1754 ifp, vifi,
1755 vifi >= numvifs ? 0 : viftable[vifi].v_ifp);
1756 ++mrtstat.mrts_wrong_if;
1757 ++rt->mfc_wrong_if;
1758 /*
1759 * If we are doing PIM assert processing, send a message
1760 * to the routing daemon.
1761 *
1762 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1763 * can complete the SPT switch, regardless of the type
1764 * of the iif (broadcast media, GRE tunnel, etc).
1765 */
1766 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1767 struct timeval now;
1768 u_int32_t delta;
1769
1770 #ifdef PIM
1771 if (ifp == &multicast_register_if)
1772 pimstat.pims_rcv_registers_wrongiif++;
1773 #endif
1774
1775 /* Get vifi for the incoming packet */
1776 for (vifi = 0;
1777 vifi < numvifs && viftable[vifi].v_ifp != ifp;
1778 vifi++)
1779 ;
1780 if (vifi >= numvifs) {
1781 /* The iif is not found: ignore the packet. */
1782 return (0);
1783 }
1784
1785 if (rt->mfc_flags[vifi] &
1786 MRT_MFC_FLAGS_DISABLE_WRONGVIF) {
1787 /* WRONGVIF disabled: ignore the packet */
1788 return (0);
1789 }
1790
1791 microtime(&now);
1792
1793 TV_DELTA(rt->mfc_last_assert, now, delta);
1794
1795 if (delta > ASSERT_MSG_TIME) {
1796 struct igmpmsg *im;
1797 int hlen = ip->ip_hl << 2;
1798 struct mbuf *mm = m_copy(m, 0, hlen);
1799
1800 M_PULLUP(mm, hlen);
1801 if (mm == NULL)
1802 return (ENOBUFS);
1803
1804 rt->mfc_last_assert = now;
1805
1806 im = mtod(mm, struct igmpmsg *);
1807 im->im_msgtype = IGMPMSG_WRONGVIF;
1808 im->im_mbz = 0;
1809 im->im_vif = vifi;
1810
1811 mrtstat.mrts_upcalls++;
1812
1813 sin.sin_addr = im->im_src;
1814 if (socket_send(ip_mrouter, mm, &sin) < 0) {
1815 log(LOG_WARNING,
1816 "ip_mforward: ip_mrouter socket queue full\n");
1817 ++mrtstat.mrts_upq_sockfull;
1818 return (ENOBUFS);
1819 }
1820 }
1821 }
1822 return (0);
1823 }
1824
1825 /* If I sourced this packet, it counts as output, else it was input. */
1826 if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1827 viftable[vifi].v_pkt_out++;
1828 viftable[vifi].v_bytes_out += plen;
1829 } else {
1830 viftable[vifi].v_pkt_in++;
1831 viftable[vifi].v_bytes_in += plen;
1832 }
1833 rt->mfc_pkt_cnt++;
1834 rt->mfc_byte_cnt += plen;
1835
1836 /*
1837 * For each vif, decide if a copy of the packet should be forwarded.
1838 * Forward if:
1839 * - the ttl exceeds the vif's threshold
1840 * - there are group members downstream on interface
1841 */
1842 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1843 if ((rt->mfc_ttls[vifi] > 0) &&
1844 (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1845 vifp->v_pkt_out++;
1846 vifp->v_bytes_out += plen;
1847 #ifdef PIM
1848 if (vifp->v_flags & VIFF_REGISTER)
1849 pim_register_send(ip, vifp, m, rt);
1850 else
1851 #endif
1852 MC_SEND(ip, vifp, m);
1853 }
1854
1855 /*
1856 * Perform upcall-related bw measuring.
1857 */
1858 if (rt->mfc_bw_meter != NULL) {
1859 struct bw_meter *x;
1860 struct timeval now;
1861
1862 microtime(&now);
1863 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1864 bw_meter_receive_packet(x, plen, &now);
1865 }
1866
1867 return (0);
1868 }
1869
1870 #ifdef RSVP_ISI
1871 /*
1872 * check if a vif number is legal/ok. This is used by ip_output.
1873 */
1874 int
1875 legal_vif_num(int vif)
1876 {
1877 if (vif >= 0 && vif < numvifs)
1878 return (1);
1879 else
1880 return (0);
1881 }
1882 #endif /* RSVP_ISI */
1883
1884 static void
1885 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1886 {
1887 struct mbuf *mb_copy;
1888 int hlen = ip->ip_hl << 2;
1889
1890 /*
1891 * Make a new reference to the packet; make sure that
1892 * the IP header is actually copied, not just referenced,
1893 * so that ip_output() only scribbles on the copy.
1894 */
1895 mb_copy = m_copy(m, 0, M_COPYALL);
1896 M_PULLUP(mb_copy, hlen);
1897 if (mb_copy == NULL)
1898 return;
1899
1900 if (vifp->v_rate_limit <= 0)
1901 tbf_send_packet(vifp, mb_copy);
1902 else
1903 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *),
1904 ntohs(ip->ip_len));
1905 }
1906
1907 static void
1908 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1909 {
1910 struct mbuf *mb_copy;
1911 struct ip *ip_copy;
1912 int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr);
1913
1914 /* Take care of delayed checksums */
1915 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1916 in_delayed_cksum(m);
1917 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1918 }
1919
1920 /*
1921 * copy the old packet & pullup it's IP header into the
1922 * new mbuf so we can modify it. Try to fill the new
1923 * mbuf since if we don't the ethernet driver will.
1924 */
1925 MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1926 if (mb_copy == NULL)
1927 return;
1928 mb_copy->m_data += max_linkhdr;
1929 mb_copy->m_pkthdr.len = len;
1930 mb_copy->m_len = sizeof(multicast_encap_iphdr);
1931
1932 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1933 m_freem(mb_copy);
1934 return;
1935 }
1936 i = MHLEN - max_linkhdr;
1937 if (i > len)
1938 i = len;
1939 mb_copy = m_pullup(mb_copy, i);
1940 if (mb_copy == NULL)
1941 return;
1942
1943 /*
1944 * fill in the encapsulating IP header.
1945 */
1946 ip_copy = mtod(mb_copy, struct ip *);
1947 *ip_copy = multicast_encap_iphdr;
1948 ip_copy->ip_id = ip_newid();
1949 ip_copy->ip_len = htons(len);
1950 ip_copy->ip_src = vifp->v_lcl_addr;
1951 ip_copy->ip_dst = vifp->v_rmt_addr;
1952
1953 /*
1954 * turn the encapsulated IP header back into a valid one.
1955 */
1956 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1957 --ip->ip_ttl;
1958 ip->ip_sum = 0;
1959 mb_copy->m_data += sizeof(multicast_encap_iphdr);
1960 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1961 mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1962
1963 if (vifp->v_rate_limit <= 0)
1964 tbf_send_packet(vifp, mb_copy);
1965 else
1966 tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len));
1967 }
1968
1969 /*
1970 * De-encapsulate a packet and feed it back through ip input.
1971 */
1972 static void
1973 vif_input(struct mbuf *m, ...)
1974 {
1975 int off, proto;
1976 va_list ap;
1977 struct vif *vifp;
1978 int s;
1979 struct ifqueue *ifq;
1980
1981 va_start(ap, m);
1982 off = va_arg(ap, int);
1983 proto = va_arg(ap, int);
1984 va_end(ap);
1985
1986 vifp = (struct vif *)encap_getarg(m);
1987 if (!vifp || proto != AF_INET) {
1988 m_freem(m);
1989 mrtstat.mrts_bad_tunnel++;
1990 return;
1991 }
1992
1993 m_adj(m, off);
1994 m->m_pkthdr.rcvif = vifp->v_ifp;
1995 ifq = &ipintrq;
1996 s = splnet();
1997 if (IF_QFULL(ifq)) {
1998 IF_DROP(ifq);
1999 m_freem(m);
2000 } else {
2001 IF_ENQUEUE(ifq, m);
2002 /*
2003 * normally we would need a "schednetisr(NETISR_IP)"
2004 * here but we were called by ip_input and it is going
2005 * to loop back & try to dequeue the packet we just
2006 * queued as soon as we return so we avoid the
2007 * unnecessary software interrrupt.
2008 */
2009 }
2010 splx(s);
2011 }
2012
2013 /*
2014 * Check if the packet should be grabbed by us.
2015 */
2016 static int
2017 vif_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2018 {
2019 struct vif *vifp;
2020 struct ip ip;
2021
2022 #ifdef DIAGNOSTIC
2023 if (!arg || proto != IPPROTO_IPV4)
2024 panic("unexpected arg in vif_encapcheck");
2025 #endif
2026
2027 /*
2028 * do not grab the packet if it's not to a multicast destination or if
2029 * we don't have an encapsulating tunnel with the source.
2030 * Note: This code assumes that the remote site IP address
2031 * uniquely identifies the tunnel (i.e., that this site has
2032 * at most one tunnel with the remote site).
2033 */
2034
2035 /* LINTED const cast */
2036 m_copydata((struct mbuf *)m, off, sizeof(ip), (caddr_t)&ip);
2037 if (!IN_MULTICAST(ip.ip_dst.s_addr))
2038 return 0;
2039
2040 /* LINTED const cast */
2041 m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip);
2042 if (!in_hosteq(ip.ip_src, last_encap_src)) {
2043 vifp = (struct vif *)arg;
2044 if (vifp->v_flags & VIFF_TUNNEL &&
2045 in_hosteq(vifp->v_rmt_addr, ip.ip_src))
2046 ;
2047 else
2048 return 0;
2049 last_encap_vif = vifp;
2050 last_encap_src = ip.ip_src;
2051 } else
2052 vifp = last_encap_vif;
2053
2054 /* 32bit match, since we have checked ip_src only */
2055 return 32;
2056 }
2057
2058 /*
2059 * Token bucket filter module
2060 */
2061 static void
2062 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len)
2063 {
2064
2065 if (len > MAX_BKT_SIZE) {
2066 /* drop if packet is too large */
2067 mrtstat.mrts_pkt2large++;
2068 m_freem(m);
2069 return;
2070 }
2071
2072 tbf_update_tokens(vifp);
2073
2074 /*
2075 * If there are enough tokens, and the queue is empty, send this packet
2076 * out immediately. Otherwise, try to insert it on this vif's queue.
2077 */
2078 if (vifp->tbf_q_len == 0) {
2079 if (len <= vifp->tbf_n_tok) {
2080 vifp->tbf_n_tok -= len;
2081 tbf_send_packet(vifp, m);
2082 } else {
2083 /* queue packet and timeout till later */
2084 tbf_queue(vifp, m);
2085 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2086 tbf_reprocess_q, vifp);
2087 }
2088 } else {
2089 if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
2090 !tbf_dq_sel(vifp, ip)) {
2091 /* queue full, and couldn't make room */
2092 mrtstat.mrts_q_overflow++;
2093 m_freem(m);
2094 } else {
2095 /* queue length low enough, or made room */
2096 tbf_queue(vifp, m);
2097 tbf_process_q(vifp);
2098 }
2099 }
2100 }
2101
2102 /*
2103 * adds a packet to the queue at the interface
2104 */
2105 static void
2106 tbf_queue(struct vif *vifp, struct mbuf *m)
2107 {
2108 int s = splsoftnet();
2109
2110 /* insert at tail */
2111 *vifp->tbf_t = m;
2112 vifp->tbf_t = &m->m_nextpkt;
2113 vifp->tbf_q_len++;
2114
2115 splx(s);
2116 }
2117
2118
2119 /*
2120 * processes the queue at the interface
2121 */
2122 static void
2123 tbf_process_q(struct vif *vifp)
2124 {
2125 struct mbuf *m;
2126 int len;
2127 int s = splsoftnet();
2128
2129 /*
2130 * Loop through the queue at the interface and send as many packets
2131 * as possible.
2132 */
2133 for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) {
2134 len = ntohs(mtod(m, struct ip *)->ip_len);
2135
2136 /* determine if the packet can be sent */
2137 if (len <= vifp->tbf_n_tok) {
2138 /* if so,
2139 * reduce no of tokens, dequeue the packet,
2140 * send the packet.
2141 */
2142 if ((vifp->tbf_q = m->m_nextpkt) == NULL)
2143 vifp->tbf_t = &vifp->tbf_q;
2144 --vifp->tbf_q_len;
2145
2146 m->m_nextpkt = NULL;
2147 vifp->tbf_n_tok -= len;
2148 tbf_send_packet(vifp, m);
2149 } else
2150 break;
2151 }
2152 splx(s);
2153 }
2154
2155 static void
2156 tbf_reprocess_q(void *arg)
2157 {
2158 struct vif *vifp = arg;
2159
2160 if (ip_mrouter == NULL)
2161 return;
2162
2163 tbf_update_tokens(vifp);
2164 tbf_process_q(vifp);
2165
2166 if (vifp->tbf_q_len != 0)
2167 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2168 tbf_reprocess_q, vifp);
2169 }
2170
2171 /* function that will selectively discard a member of the queue
2172 * based on the precedence value and the priority
2173 */
2174 static int
2175 tbf_dq_sel(struct vif *vifp, struct ip *ip)
2176 {
2177 u_int p;
2178 struct mbuf **mp, *m;
2179 int s = splsoftnet();
2180
2181 p = priority(vifp, ip);
2182
2183 for (mp = &vifp->tbf_q, m = *mp;
2184 m != NULL;
2185 mp = &m->m_nextpkt, m = *mp) {
2186 if (p > priority(vifp, mtod(m, struct ip *))) {
2187 if ((*mp = m->m_nextpkt) == NULL)
2188 vifp->tbf_t = mp;
2189 --vifp->tbf_q_len;
2190
2191 m_freem(m);
2192 mrtstat.mrts_drop_sel++;
2193 splx(s);
2194 return (1);
2195 }
2196 }
2197 splx(s);
2198 return (0);
2199 }
2200
2201 static void
2202 tbf_send_packet(struct vif *vifp, struct mbuf *m)
2203 {
2204 int error;
2205 int s = splsoftnet();
2206
2207 if (vifp->v_flags & VIFF_TUNNEL) {
2208 /* If tunnel options */
2209 ip_output(m, (struct mbuf *)NULL, &vifp->v_route,
2210 IP_FORWARDING, (struct ip_moptions *)NULL,
2211 (struct socket *)NULL);
2212 } else {
2213 /* if physical interface option, extract the options and then send */
2214 struct ip_moptions imo;
2215
2216 imo.imo_multicast_ifp = vifp->v_ifp;
2217 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2218 imo.imo_multicast_loop = 1;
2219 #ifdef RSVP_ISI
2220 imo.imo_multicast_vif = -1;
2221 #endif
2222
2223 error = ip_output(m, (struct mbuf *)NULL, (struct route *)NULL,
2224 IP_FORWARDING|IP_MULTICASTOPTS, &imo,
2225 (struct socket *)NULL);
2226
2227 if (mrtdebug & DEBUG_XMIT)
2228 log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
2229 (long)(vifp - viftable), error);
2230 }
2231 splx(s);
2232 }
2233
2234 /* determine the current time and then
2235 * the elapsed time (between the last time and time now)
2236 * in milliseconds & update the no. of tokens in the bucket
2237 */
2238 static void
2239 tbf_update_tokens(struct vif *vifp)
2240 {
2241 struct timeval tp;
2242 u_int32_t tm;
2243 int s = splsoftnet();
2244
2245 microtime(&tp);
2246
2247 TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
2248
2249 /*
2250 * This formula is actually
2251 * "time in seconds" * "bytes/second".
2252 *
2253 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2254 *
2255 * The (1000/1024) was introduced in add_vif to optimize
2256 * this divide into a shift.
2257 */
2258 vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
2259 vifp->tbf_last_pkt_t = tp;
2260
2261 if (vifp->tbf_n_tok > MAX_BKT_SIZE)
2262 vifp->tbf_n_tok = MAX_BKT_SIZE;
2263
2264 splx(s);
2265 }
2266
2267 static int
2268 priority(struct vif *vifp, struct ip *ip)
2269 {
2270 int prio = 50; /* the lowest priority -- default case */
2271
2272 /* temporary hack; may add general packet classifier some day */
2273
2274 /*
2275 * The UDP port space is divided up into four priority ranges:
2276 * [0, 16384) : unclassified - lowest priority
2277 * [16384, 32768) : audio - highest priority
2278 * [32768, 49152) : whiteboard - medium priority
2279 * [49152, 65536) : video - low priority
2280 */
2281 if (ip->ip_p == IPPROTO_UDP) {
2282 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2283
2284 switch (ntohs(udp->uh_dport) & 0xc000) {
2285 case 0x4000:
2286 prio = 70;
2287 break;
2288 case 0x8000:
2289 prio = 60;
2290 break;
2291 case 0xc000:
2292 prio = 55;
2293 break;
2294 }
2295
2296 if (tbfdebug > 1)
2297 log(LOG_DEBUG, "port %x prio %d\n",
2298 ntohs(udp->uh_dport), prio);
2299 }
2300
2301 return (prio);
2302 }
2303
2304 /*
2305 * End of token bucket filter modifications
2306 */
2307 #ifdef RSVP_ISI
2308 int
2309 ip_rsvp_vif_init(struct socket *so, struct mbuf *m)
2310 {
2311 int vifi, s;
2312
2313 if (rsvpdebug)
2314 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2315 so->so_type, so->so_proto->pr_protocol);
2316
2317 if (so->so_type != SOCK_RAW ||
2318 so->so_proto->pr_protocol != IPPROTO_RSVP)
2319 return (EOPNOTSUPP);
2320
2321 /* Check mbuf. */
2322 if (m == NULL || m->m_len != sizeof(int)) {
2323 return (EINVAL);
2324 }
2325 vifi = *(mtod(m, int *));
2326
2327 if (rsvpdebug)
2328 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",
2329 vifi, rsvp_on);
2330
2331 s = splsoftnet();
2332
2333 /* Check vif. */
2334 if (!legal_vif_num(vifi)) {
2335 splx(s);
2336 return (EADDRNOTAVAIL);
2337 }
2338
2339 /* Check if socket is available. */
2340 if (viftable[vifi].v_rsvpd != NULL) {
2341 splx(s);
2342 return (EADDRINUSE);
2343 }
2344
2345 viftable[vifi].v_rsvpd = so;
2346 /*
2347 * This may seem silly, but we need to be sure we don't over-increment
2348 * the RSVP counter, in case something slips up.
2349 */
2350 if (!viftable[vifi].v_rsvp_on) {
2351 viftable[vifi].v_rsvp_on = 1;
2352 rsvp_on++;
2353 }
2354
2355 splx(s);
2356 return (0);
2357 }
2358
2359 int
2360 ip_rsvp_vif_done(struct socket *so, struct mbuf *m)
2361 {
2362 int vifi, s;
2363
2364 if (rsvpdebug)
2365 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2366 so->so_type, so->so_proto->pr_protocol);
2367
2368 if (so->so_type != SOCK_RAW ||
2369 so->so_proto->pr_protocol != IPPROTO_RSVP)
2370 return (EOPNOTSUPP);
2371
2372 /* Check mbuf. */
2373 if (m == NULL || m->m_len != sizeof(int)) {
2374 return (EINVAL);
2375 }
2376 vifi = *(mtod(m, int *));
2377
2378 s = splsoftnet();
2379
2380 /* Check vif. */
2381 if (!legal_vif_num(vifi)) {
2382 splx(s);
2383 return (EADDRNOTAVAIL);
2384 }
2385
2386 if (rsvpdebug)
2387 printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
2388 viftable[vifi].v_rsvpd, so);
2389
2390 viftable[vifi].v_rsvpd = NULL;
2391 /*
2392 * This may seem silly, but we need to be sure we don't over-decrement
2393 * the RSVP counter, in case something slips up.
2394 */
2395 if (viftable[vifi].v_rsvp_on) {
2396 viftable[vifi].v_rsvp_on = 0;
2397 rsvp_on--;
2398 }
2399
2400 splx(s);
2401 return (0);
2402 }
2403
2404 void
2405 ip_rsvp_force_done(struct socket *so)
2406 {
2407 int vifi, s;
2408
2409 /* Don't bother if it is not the right type of socket. */
2410 if (so->so_type != SOCK_RAW ||
2411 so->so_proto->pr_protocol != IPPROTO_RSVP)
2412 return;
2413
2414 s = splsoftnet();
2415
2416 /*
2417 * The socket may be attached to more than one vif...this
2418 * is perfectly legal.
2419 */
2420 for (vifi = 0; vifi < numvifs; vifi++) {
2421 if (viftable[vifi].v_rsvpd == so) {
2422 viftable[vifi].v_rsvpd = NULL;
2423 /*
2424 * This may seem silly, but we need to be sure we don't
2425 * over-decrement the RSVP counter, in case something
2426 * slips up.
2427 */
2428 if (viftable[vifi].v_rsvp_on) {
2429 viftable[vifi].v_rsvp_on = 0;
2430 rsvp_on--;
2431 }
2432 }
2433 }
2434
2435 splx(s);
2436 return;
2437 }
2438
2439 void
2440 rsvp_input(struct mbuf *m, struct ifnet *ifp)
2441 {
2442 int vifi, s;
2443 struct ip *ip = mtod(m, struct ip *);
2444 static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET };
2445
2446 if (rsvpdebug)
2447 printf("rsvp_input: rsvp_on %d\n", rsvp_on);
2448
2449 /*
2450 * Can still get packets with rsvp_on = 0 if there is a local member
2451 * of the group to which the RSVP packet is addressed. But in this
2452 * case we want to throw the packet away.
2453 */
2454 if (!rsvp_on) {
2455 m_freem(m);
2456 return;
2457 }
2458
2459 /*
2460 * If the old-style non-vif-associated socket is set, then use
2461 * it and ignore the new ones.
2462 */
2463 if (ip_rsvpd != NULL) {
2464 if (rsvpdebug)
2465 printf("rsvp_input: "
2466 "Sending packet up old-style socket\n");
2467 rip_input(m); /*XXX*/
2468 return;
2469 }
2470
2471 s = splsoftnet();
2472
2473 if (rsvpdebug)
2474 printf("rsvp_input: check vifs\n");
2475
2476 /* Find which vif the packet arrived on. */
2477 for (vifi = 0; vifi < numvifs; vifi++) {
2478 if (viftable[vifi].v_ifp == ifp)
2479 break;
2480 }
2481
2482 if (vifi == numvifs) {
2483 /* Can't find vif packet arrived on. Drop packet. */
2484 if (rsvpdebug)
2485 printf("rsvp_input: "
2486 "Can't find vif for packet...dropping it.\n");
2487 m_freem(m);
2488 splx(s);
2489 return;
2490 }
2491
2492 if (rsvpdebug)
2493 printf("rsvp_input: check socket\n");
2494
2495 if (viftable[vifi].v_rsvpd == NULL) {
2496 /*
2497 * drop packet, since there is no specific socket for this
2498 * interface
2499 */
2500 if (rsvpdebug)
2501 printf("rsvp_input: No socket defined for vif %d\n",
2502 vifi);
2503 m_freem(m);
2504 splx(s);
2505 return;
2506 }
2507
2508 rsvp_src.sin_addr = ip->ip_src;
2509
2510 if (rsvpdebug && m)
2511 printf("rsvp_input: m->m_len = %d, sbspace() = %d\n",
2512 m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv));
2513
2514 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2515 if (rsvpdebug)
2516 printf("rsvp_input: Failed to append to socket\n");
2517 else
2518 if (rsvpdebug)
2519 printf("rsvp_input: send packet up\n");
2520
2521 splx(s);
2522 }
2523 #endif /* RSVP_ISI */
2524
2525 /*
2526 * Code for bandwidth monitors
2527 */
2528
2529 /*
2530 * Define common interface for timeval-related methods
2531 */
2532 #define BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp)
2533 #define BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp))
2534 #define BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp))
2535
2536 static uint32_t
2537 compute_bw_meter_flags(struct bw_upcall *req)
2538 {
2539 uint32_t flags = 0;
2540
2541 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2542 flags |= BW_METER_UNIT_PACKETS;
2543 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2544 flags |= BW_METER_UNIT_BYTES;
2545 if (req->bu_flags & BW_UPCALL_GEQ)
2546 flags |= BW_METER_GEQ;
2547 if (req->bu_flags & BW_UPCALL_LEQ)
2548 flags |= BW_METER_LEQ;
2549
2550 return flags;
2551 }
2552
2553 /*
2554 * Add a bw_meter entry
2555 */
2556 static int
2557 add_bw_upcall(struct mbuf *m)
2558 {
2559 int s;
2560 struct mfc *mfc;
2561 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2562 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2563 struct timeval now;
2564 struct bw_meter *x;
2565 uint32_t flags;
2566 struct bw_upcall *req;
2567
2568 if (m == NULL || m->m_len < sizeof(struct bw_upcall))
2569 return EINVAL;
2570
2571 req = mtod(m, struct bw_upcall *);
2572
2573 if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2574 return EOPNOTSUPP;
2575
2576 /* Test if the flags are valid */
2577 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2578 return EINVAL;
2579 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2580 return EINVAL;
2581 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2582 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2583 return EINVAL;
2584
2585 /* Test if the threshold time interval is valid */
2586 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2587 return EINVAL;
2588
2589 flags = compute_bw_meter_flags(req);
2590
2591 /*
2592 * Find if we have already same bw_meter entry
2593 */
2594 s = splsoftnet();
2595 mfc = mfc_find(&req->bu_src, &req->bu_dst);
2596 if (mfc == NULL) {
2597 splx(s);
2598 return EADDRNOTAVAIL;
2599 }
2600 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2601 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2602 &req->bu_threshold.b_time, ==)) &&
2603 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2604 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2605 (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
2606 splx(s);
2607 return 0; /* XXX Already installed */
2608 }
2609 }
2610
2611 /* Allocate the new bw_meter entry */
2612 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
2613 if (x == NULL) {
2614 splx(s);
2615 return ENOBUFS;
2616 }
2617
2618 /* Set the new bw_meter entry */
2619 x->bm_threshold.b_time = req->bu_threshold.b_time;
2620 microtime(&now);
2621 x->bm_start_time = now;
2622 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2623 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2624 x->bm_measured.b_packets = 0;
2625 x->bm_measured.b_bytes = 0;
2626 x->bm_flags = flags;
2627 x->bm_time_next = NULL;
2628 x->bm_time_hash = BW_METER_BUCKETS;
2629
2630 /* Add the new bw_meter entry to the front of entries for this MFC */
2631 x->bm_mfc = mfc;
2632 x->bm_mfc_next = mfc->mfc_bw_meter;
2633 mfc->mfc_bw_meter = x;
2634 schedule_bw_meter(x, &now);
2635 splx(s);
2636
2637 return 0;
2638 }
2639
2640 static void
2641 free_bw_list(struct bw_meter *list)
2642 {
2643 while (list != NULL) {
2644 struct bw_meter *x = list;
2645
2646 list = list->bm_mfc_next;
2647 unschedule_bw_meter(x);
2648 free(x, M_BWMETER);
2649 }
2650 }
2651
2652 /*
2653 * Delete one or multiple bw_meter entries
2654 */
2655 static int
2656 del_bw_upcall(struct mbuf *m)
2657 {
2658 int s;
2659 struct mfc *mfc;
2660 struct bw_meter *x;
2661 struct bw_upcall *req;
2662
2663 if (m == NULL || m->m_len < sizeof(struct bw_upcall))
2664 return EINVAL;
2665
2666 req = mtod(m, struct bw_upcall *);
2667
2668 if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2669 return EOPNOTSUPP;
2670
2671 s = splsoftnet();
2672 /* Find the corresponding MFC entry */
2673 mfc = mfc_find(&req->bu_src, &req->bu_dst);
2674 if (mfc == NULL) {
2675 splx(s);
2676 return EADDRNOTAVAIL;
2677 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2678 /*
2679 * Delete all bw_meter entries for this mfc
2680 */
2681 struct bw_meter *list;
2682
2683 list = mfc->mfc_bw_meter;
2684 mfc->mfc_bw_meter = NULL;
2685 free_bw_list(list);
2686 splx(s);
2687 return 0;
2688 } else { /* Delete a single bw_meter entry */
2689 struct bw_meter *prev;
2690 uint32_t flags = 0;
2691
2692 flags = compute_bw_meter_flags(req);
2693
2694 /* Find the bw_meter entry to delete */
2695 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2696 prev = x, x = x->bm_mfc_next) {
2697 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2698 &req->bu_threshold.b_time, ==)) &&
2699 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2700 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2701 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2702 break;
2703 }
2704 if (x != NULL) { /* Delete entry from the list for this MFC */
2705 if (prev != NULL)
2706 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
2707 else
2708 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2709
2710 unschedule_bw_meter(x);
2711 splx(s);
2712 /* Free the bw_meter entry */
2713 free(x, M_BWMETER);
2714 return 0;
2715 } else {
2716 splx(s);
2717 return EINVAL;
2718 }
2719 }
2720 /* NOTREACHED */
2721 }
2722
2723 /*
2724 * Perform bandwidth measurement processing that may result in an upcall
2725 */
2726 static void
2727 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2728 {
2729 struct timeval delta;
2730
2731 delta = *nowp;
2732 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2733
2734 if (x->bm_flags & BW_METER_GEQ) {
2735 /*
2736 * Processing for ">=" type of bw_meter entry
2737 */
2738 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2739 /* Reset the bw_meter entry */
2740 x->bm_start_time = *nowp;
2741 x->bm_measured.b_packets = 0;
2742 x->bm_measured.b_bytes = 0;
2743 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2744 }
2745
2746 /* Record that a packet is received */
2747 x->bm_measured.b_packets++;
2748 x->bm_measured.b_bytes += plen;
2749
2750 /*
2751 * Test if we should deliver an upcall
2752 */
2753 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2754 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2755 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2756 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2757 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2758 /* Prepare an upcall for delivery */
2759 bw_meter_prepare_upcall(x, nowp);
2760 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2761 }
2762 }
2763 } else if (x->bm_flags & BW_METER_LEQ) {
2764 /*
2765 * Processing for "<=" type of bw_meter entry
2766 */
2767 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2768 /*
2769 * We are behind time with the multicast forwarding table
2770 * scanning for "<=" type of bw_meter entries, so test now
2771 * if we should deliver an upcall.
2772 */
2773 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2774 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2775 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2776 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2777 /* Prepare an upcall for delivery */
2778 bw_meter_prepare_upcall(x, nowp);
2779 }
2780 /* Reschedule the bw_meter entry */
2781 unschedule_bw_meter(x);
2782 schedule_bw_meter(x, nowp);
2783 }
2784
2785 /* Record that a packet is received */
2786 x->bm_measured.b_packets++;
2787 x->bm_measured.b_bytes += plen;
2788
2789 /*
2790 * Test if we should restart the measuring interval
2791 */
2792 if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2793 x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2794 (x->bm_flags & BW_METER_UNIT_BYTES &&
2795 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2796 /* Don't restart the measuring interval */
2797 } else {
2798 /* Do restart the measuring interval */
2799 /*
2800 * XXX: note that we don't unschedule and schedule, because this
2801 * might be too much overhead per packet. Instead, when we process
2802 * all entries for a given timer hash bin, we check whether it is
2803 * really a timeout. If not, we reschedule at that time.
2804 */
2805 x->bm_start_time = *nowp;
2806 x->bm_measured.b_packets = 0;
2807 x->bm_measured.b_bytes = 0;
2808 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2809 }
2810 }
2811 }
2812
2813 /*
2814 * Prepare a bandwidth-related upcall
2815 */
2816 static void
2817 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2818 {
2819 struct timeval delta;
2820 struct bw_upcall *u;
2821
2822 /*
2823 * Compute the measured time interval
2824 */
2825 delta = *nowp;
2826 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2827
2828 /*
2829 * If there are too many pending upcalls, deliver them now
2830 */
2831 if (bw_upcalls_n >= BW_UPCALLS_MAX)
2832 bw_upcalls_send();
2833
2834 /*
2835 * Set the bw_upcall entry
2836 */
2837 u = &bw_upcalls[bw_upcalls_n++];
2838 u->bu_src = x->bm_mfc->mfc_origin;
2839 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2840 u->bu_threshold.b_time = x->bm_threshold.b_time;
2841 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2842 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2843 u->bu_measured.b_time = delta;
2844 u->bu_measured.b_packets = x->bm_measured.b_packets;
2845 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2846 u->bu_flags = 0;
2847 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2848 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2849 if (x->bm_flags & BW_METER_UNIT_BYTES)
2850 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2851 if (x->bm_flags & BW_METER_GEQ)
2852 u->bu_flags |= BW_UPCALL_GEQ;
2853 if (x->bm_flags & BW_METER_LEQ)
2854 u->bu_flags |= BW_UPCALL_LEQ;
2855 }
2856
2857 /*
2858 * Send the pending bandwidth-related upcalls
2859 */
2860 static void
2861 bw_upcalls_send(void)
2862 {
2863 struct mbuf *m;
2864 int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2865 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2866 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2867 0, /* unused2 */
2868 IGMPMSG_BW_UPCALL,/* im_msgtype */
2869 0, /* im_mbz */
2870 0, /* im_vif */
2871 0, /* unused3 */
2872 { 0 }, /* im_src */
2873 { 0 } }; /* im_dst */
2874
2875 if (bw_upcalls_n == 0)
2876 return; /* No pending upcalls */
2877
2878 bw_upcalls_n = 0;
2879
2880 /*
2881 * Allocate a new mbuf, initialize it with the header and
2882 * the payload for the pending calls.
2883 */
2884 MGETHDR(m, M_DONTWAIT, MT_HEADER);
2885 if (m == NULL) {
2886 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2887 return;
2888 }
2889
2890 m->m_len = m->m_pkthdr.len = 0;
2891 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2892 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]);
2893
2894 /*
2895 * Send the upcalls
2896 * XXX do we need to set the address in k_igmpsrc ?
2897 */
2898 mrtstat.mrts_upcalls++;
2899 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2900 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2901 ++mrtstat.mrts_upq_sockfull;
2902 }
2903 }
2904
2905 /*
2906 * Compute the timeout hash value for the bw_meter entries
2907 */
2908 #define BW_METER_TIMEHASH(bw_meter, hash) \
2909 do { \
2910 struct timeval next_timeval = (bw_meter)->bm_start_time; \
2911 \
2912 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2913 (hash) = next_timeval.tv_sec; \
2914 if (next_timeval.tv_usec) \
2915 (hash)++; /* XXX: make sure we don't timeout early */ \
2916 (hash) %= BW_METER_BUCKETS; \
2917 } while (/*CONSTCOND*/ 0)
2918
2919 /*
2920 * Schedule a timer to process periodically bw_meter entry of type "<="
2921 * by linking the entry in the proper hash bucket.
2922 */
2923 static void
2924 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2925 {
2926 int time_hash;
2927
2928 if (!(x->bm_flags & BW_METER_LEQ))
2929 return; /* XXX: we schedule timers only for "<=" entries */
2930
2931 /*
2932 * Reset the bw_meter entry
2933 */
2934 x->bm_start_time = *nowp;
2935 x->bm_measured.b_packets = 0;
2936 x->bm_measured.b_bytes = 0;
2937 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2938
2939 /*
2940 * Compute the timeout hash value and insert the entry
2941 */
2942 BW_METER_TIMEHASH(x, time_hash);
2943 x->bm_time_next = bw_meter_timers[time_hash];
2944 bw_meter_timers[time_hash] = x;
2945 x->bm_time_hash = time_hash;
2946 }
2947
2948 /*
2949 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2950 * by removing the entry from the proper hash bucket.
2951 */
2952 static void
2953 unschedule_bw_meter(struct bw_meter *x)
2954 {
2955 int time_hash;
2956 struct bw_meter *prev, *tmp;
2957
2958 if (!(x->bm_flags & BW_METER_LEQ))
2959 return; /* XXX: we schedule timers only for "<=" entries */
2960
2961 /*
2962 * Compute the timeout hash value and delete the entry
2963 */
2964 time_hash = x->bm_time_hash;
2965 if (time_hash >= BW_METER_BUCKETS)
2966 return; /* Entry was not scheduled */
2967
2968 for (prev = NULL, tmp = bw_meter_timers[time_hash];
2969 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2970 if (tmp == x)
2971 break;
2972
2973 if (tmp == NULL)
2974 panic("unschedule_bw_meter: bw_meter entry not found");
2975
2976 if (prev != NULL)
2977 prev->bm_time_next = x->bm_time_next;
2978 else
2979 bw_meter_timers[time_hash] = x->bm_time_next;
2980
2981 x->bm_time_next = NULL;
2982 x->bm_time_hash = BW_METER_BUCKETS;
2983 }
2984
2985 /*
2986 * Process all "<=" type of bw_meter that should be processed now,
2987 * and for each entry prepare an upcall if necessary. Each processed
2988 * entry is rescheduled again for the (periodic) processing.
2989 *
2990 * This is run periodically (once per second normally). On each round,
2991 * all the potentially matching entries are in the hash slot that we are
2992 * looking at.
2993 */
2994 static void
2995 bw_meter_process(void)
2996 {
2997 int s;
2998 static uint32_t last_tv_sec; /* last time we processed this */
2999
3000 uint32_t loops;
3001 int i;
3002 struct timeval now, process_endtime;
3003
3004 microtime(&now);
3005 if (last_tv_sec == now.tv_sec)
3006 return; /* nothing to do */
3007
3008 loops = now.tv_sec - last_tv_sec;
3009 last_tv_sec = now.tv_sec;
3010 if (loops > BW_METER_BUCKETS)
3011 loops = BW_METER_BUCKETS;
3012
3013 s = splsoftnet();
3014 /*
3015 * Process all bins of bw_meter entries from the one after the last
3016 * processed to the current one. On entry, i points to the last bucket
3017 * visited, so we need to increment i at the beginning of the loop.
3018 */
3019 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
3020 struct bw_meter *x, *tmp_list;
3021
3022 if (++i >= BW_METER_BUCKETS)
3023 i = 0;
3024
3025 /* Disconnect the list of bw_meter entries from the bin */
3026 tmp_list = bw_meter_timers[i];
3027 bw_meter_timers[i] = NULL;
3028
3029 /* Process the list of bw_meter entries */
3030 while (tmp_list != NULL) {
3031 x = tmp_list;
3032 tmp_list = tmp_list->bm_time_next;
3033
3034 /* Test if the time interval is over */
3035 process_endtime = x->bm_start_time;
3036 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
3037 if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
3038 /* Not yet: reschedule, but don't reset */
3039 int time_hash;
3040
3041 BW_METER_TIMEHASH(x, time_hash);
3042 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
3043 /*
3044 * XXX: somehow the bin processing is a bit ahead of time.
3045 * Put the entry in the next bin.
3046 */
3047 if (++time_hash >= BW_METER_BUCKETS)
3048 time_hash = 0;
3049 }
3050 x->bm_time_next = bw_meter_timers[time_hash];
3051 bw_meter_timers[time_hash] = x;
3052 x->bm_time_hash = time_hash;
3053
3054 continue;
3055 }
3056
3057 /*
3058 * Test if we should deliver an upcall
3059 */
3060 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
3061 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
3062 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
3063 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
3064 /* Prepare an upcall for delivery */
3065 bw_meter_prepare_upcall(x, &now);
3066 }
3067
3068 /*
3069 * Reschedule for next processing
3070 */
3071 schedule_bw_meter(x, &now);
3072 }
3073 }
3074
3075 /* Send all upcalls that are pending delivery */
3076 bw_upcalls_send();
3077
3078 splx(s);
3079 }
3080
3081 /*
3082 * A periodic function for sending all upcalls that are pending delivery
3083 */
3084 static void
3085 expire_bw_upcalls_send(void *unused)
3086 {
3087 int s;
3088
3089 s = splsoftnet();
3090 bw_upcalls_send();
3091 splx(s);
3092
3093 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
3094 expire_bw_upcalls_send, NULL);
3095 }
3096
3097 /*
3098 * A periodic function for periodic scanning of the multicast forwarding
3099 * table for processing all "<=" bw_meter entries.
3100 */
3101 static void
3102 expire_bw_meter_process(void *unused)
3103 {
3104 if (mrt_api_config & MRT_MFC_BW_UPCALL)
3105 bw_meter_process();
3106
3107 callout_reset(&bw_meter_ch, BW_METER_PERIOD,
3108 expire_bw_meter_process, NULL);
3109 }
3110
3111 /*
3112 * End of bandwidth monitoring code
3113 */
3114
3115 #ifdef PIM
3116 /*
3117 * Send the packet up to the user daemon, or eventually do kernel encapsulation
3118 */
3119 static int
3120 pim_register_send(struct ip *ip, struct vif *vifp,
3121 struct mbuf *m, struct mfc *rt)
3122 {
3123 struct mbuf *mb_copy, *mm;
3124
3125 if (mrtdebug & DEBUG_PIM)
3126 log(LOG_DEBUG, "pim_register_send: ");
3127
3128 mb_copy = pim_register_prepare(ip, m);
3129 if (mb_copy == NULL)
3130 return ENOBUFS;
3131
3132 /*
3133 * Send all the fragments. Note that the mbuf for each fragment
3134 * is freed by the sending machinery.
3135 */
3136 for (mm = mb_copy; mm; mm = mb_copy) {
3137 mb_copy = mm->m_nextpkt;
3138 mm->m_nextpkt = NULL;
3139 mm = m_pullup(mm, sizeof(struct ip));
3140 if (mm != NULL) {
3141 ip = mtod(mm, struct ip *);
3142 if ((mrt_api_config & MRT_MFC_RP) &&
3143 !in_nullhost(rt->mfc_rp)) {
3144 pim_register_send_rp(ip, vifp, mm, rt);
3145 } else {
3146 pim_register_send_upcall(ip, vifp, mm, rt);
3147 }
3148 }
3149 }
3150
3151 return 0;
3152 }
3153
3154 /*
3155 * Return a copy of the data packet that is ready for PIM Register
3156 * encapsulation.
3157 * XXX: Note that in the returned copy the IP header is a valid one.
3158 */
3159 static struct mbuf *
3160 pim_register_prepare(struct ip *ip, struct mbuf *m)
3161 {
3162 struct mbuf *mb_copy = NULL;
3163 int mtu;
3164
3165 /* Take care of delayed checksums */
3166 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3167 in_delayed_cksum(m);
3168 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3169 }
3170
3171 /*
3172 * Copy the old packet & pullup its IP header into the
3173 * new mbuf so we can modify it.
3174 */
3175 mb_copy = m_copy(m, 0, M_COPYALL);
3176 if (mb_copy == NULL)
3177 return NULL;
3178 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3179 if (mb_copy == NULL)
3180 return NULL;
3181
3182 /* take care of the TTL */
3183 ip = mtod(mb_copy, struct ip *);
3184 --ip->ip_ttl;
3185
3186 /* Compute the MTU after the PIM Register encapsulation */
3187 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3188
3189 if (ntohs(ip->ip_len) <= mtu) {
3190 /* Turn the IP header into a valid one */
3191 ip->ip_sum = 0;
3192 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3193 } else {
3194 /* Fragment the packet */
3195 if (ip_fragment(mb_copy, NULL, mtu) != 0) {
3196 /* XXX: mb_copy was freed by ip_fragment() */
3197 return NULL;
3198 }
3199 }
3200 return mb_copy;
3201 }
3202
3203 /*
3204 * Send an upcall with the data packet to the user-level process.
3205 */
3206 static int
3207 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3208 struct mbuf *mb_copy, struct mfc *rt)
3209 {
3210 struct mbuf *mb_first;
3211 int len = ntohs(ip->ip_len);
3212 struct igmpmsg *im;
3213 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
3214
3215 /*
3216 * Add a new mbuf with an upcall header
3217 */
3218 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3219 if (mb_first == NULL) {
3220 m_freem(mb_copy);
3221 return ENOBUFS;
3222 }
3223 mb_first->m_data += max_linkhdr;
3224 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3225 mb_first->m_len = sizeof(struct igmpmsg);
3226 mb_first->m_next = mb_copy;
3227
3228 /* Send message to routing daemon */
3229 im = mtod(mb_first, struct igmpmsg *);
3230 im->im_msgtype = IGMPMSG_WHOLEPKT;
3231 im->im_mbz = 0;
3232 im->im_vif = vifp - viftable;
3233 im->im_src = ip->ip_src;
3234 im->im_dst = ip->ip_dst;
3235
3236 k_igmpsrc.sin_addr = ip->ip_src;
3237
3238 mrtstat.mrts_upcalls++;
3239
3240 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3241 if (mrtdebug & DEBUG_PIM)
3242 log(LOG_WARNING,
3243 "mcast: pim_register_send_upcall: ip_mrouter socket queue full");
3244 ++mrtstat.mrts_upq_sockfull;
3245 return ENOBUFS;
3246 }
3247
3248 /* Keep statistics */
3249 pimstat.pims_snd_registers_msgs++;
3250 pimstat.pims_snd_registers_bytes += len;
3251
3252 return 0;
3253 }
3254
3255 /*
3256 * Encapsulate the data packet in PIM Register message and send it to the RP.
3257 */
3258 static int
3259 pim_register_send_rp(struct ip *ip, struct vif *vifp,
3260 struct mbuf *mb_copy, struct mfc *rt)
3261 {
3262 struct mbuf *mb_first;
3263 struct ip *ip_outer;
3264 struct pim_encap_pimhdr *pimhdr;
3265 int len = ntohs(ip->ip_len);
3266 vifi_t vifi = rt->mfc_parent;
3267
3268 if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
3269 m_freem(mb_copy);
3270 return EADDRNOTAVAIL; /* The iif vif is invalid */
3271 }
3272
3273 /*
3274 * Add a new mbuf with the encapsulating header
3275 */
3276 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3277 if (mb_first == NULL) {
3278 m_freem(mb_copy);
3279 return ENOBUFS;
3280 }
3281 mb_first->m_data += max_linkhdr;
3282 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3283 mb_first->m_next = mb_copy;
3284
3285 mb_first->m_pkthdr.len = len + mb_first->m_len;
3286
3287 /*
3288 * Fill in the encapsulating IP and PIM header
3289 */
3290 ip_outer = mtod(mb_first, struct ip *);
3291 *ip_outer = pim_encap_iphdr;
3292 ip_outer->ip_id = ip_newid();
3293 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
3294 sizeof(pim_encap_pimhdr));
3295 ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3296 ip_outer->ip_dst = rt->mfc_rp;
3297 /*
3298 * Copy the inner header TOS to the outer header, and take care of the
3299 * IP_DF bit.
3300 */
3301 ip_outer->ip_tos = ip->ip_tos;
3302 if (ntohs(ip->ip_off) & IP_DF)
3303 ip_outer->ip_off |= IP_DF;
3304 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
3305 + sizeof(pim_encap_iphdr));
3306 *pimhdr = pim_encap_pimhdr;
3307 /* If the iif crosses a border, set the Border-bit */
3308 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3309 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3310
3311 mb_first->m_data += sizeof(pim_encap_iphdr);
3312 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3313 mb_first->m_data -= sizeof(pim_encap_iphdr);
3314
3315 if (vifp->v_rate_limit == 0)
3316 tbf_send_packet(vifp, mb_first);
3317 else
3318 tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len));
3319
3320 /* Keep statistics */
3321 pimstat.pims_snd_registers_msgs++;
3322 pimstat.pims_snd_registers_bytes += len;
3323
3324 return 0;
3325 }
3326
3327 /*
3328 * PIM-SMv2 and PIM-DM messages processing.
3329 * Receives and verifies the PIM control messages, and passes them
3330 * up to the listening socket, using rip_input().
3331 * The only message with special processing is the PIM_REGISTER message
3332 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3333 * is passed to if_simloop().
3334 */
3335 void
3336 pim_input(struct mbuf *m, ...)
3337 {
3338 struct ip *ip = mtod(m, struct ip *);
3339 struct pim *pim;
3340 int minlen;
3341 int datalen;
3342 int ip_tos;
3343 int proto;
3344 int iphlen;
3345 va_list ap;
3346
3347 va_start(ap, m);
3348 iphlen = va_arg(ap, int);
3349 proto = va_arg(ap, int);
3350 va_end(ap);
3351
3352 datalen = ntohs(ip->ip_len) - iphlen;
3353
3354 /* Keep statistics */
3355 pimstat.pims_rcv_total_msgs++;
3356 pimstat.pims_rcv_total_bytes += datalen;
3357
3358 /*
3359 * Validate lengths
3360 */
3361 if (datalen < PIM_MINLEN) {
3362 pimstat.pims_rcv_tooshort++;
3363 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3364 datalen, (u_long)ip->ip_src.s_addr);
3365 m_freem(m);
3366 return;
3367 }
3368
3369 /*
3370 * If the packet is at least as big as a REGISTER, go agead
3371 * and grab the PIM REGISTER header size, to avoid another
3372 * possible m_pullup() later.
3373 *
3374 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
3375 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3376 */
3377 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3378 /*
3379 * Get the IP and PIM headers in contiguous memory, and
3380 * possibly the PIM REGISTER header.
3381 */
3382 if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3383 (m = m_pullup(m, minlen)) == NULL) {
3384 log(LOG_ERR, "pim_input: m_pullup failure\n");
3385 return;
3386 }
3387 /* m_pullup() may have given us a new mbuf so reset ip. */
3388 ip = mtod(m, struct ip *);
3389 ip_tos = ip->ip_tos;
3390
3391 /* adjust mbuf to point to the PIM header */
3392 m->m_data += iphlen;
3393 m->m_len -= iphlen;
3394 pim = mtod(m, struct pim *);
3395
3396 /*
3397 * Validate checksum. If PIM REGISTER, exclude the data packet.
3398 *
3399 * XXX: some older PIMv2 implementations don't make this distinction,
3400 * so for compatibility reason perform the checksum over part of the
3401 * message, and if error, then over the whole message.
3402 */
3403 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3404 /* do nothing, checksum okay */
3405 } else if (in_cksum(m, datalen)) {
3406 pimstat.pims_rcv_badsum++;
3407 if (mrtdebug & DEBUG_PIM)
3408 log(LOG_DEBUG, "pim_input: invalid checksum");
3409 m_freem(m);
3410 return;
3411 }
3412
3413 /* PIM version check */
3414 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3415 pimstat.pims_rcv_badversion++;
3416 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3417 PIM_VT_V(pim->pim_vt), PIM_VERSION);
3418 m_freem(m);
3419 return;
3420 }
3421
3422 /* restore mbuf back to the outer IP */
3423 m->m_data -= iphlen;
3424 m->m_len += iphlen;
3425
3426 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3427 /*
3428 * Since this is a REGISTER, we'll make a copy of the register
3429 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3430 * routing daemon.
3431 */
3432 int s;
3433 struct sockaddr_in dst = { sizeof(dst), AF_INET };
3434 struct mbuf *mcp;
3435 struct ip *encap_ip;
3436 u_int32_t *reghdr;
3437 struct ifnet *vifp;
3438
3439 s = splsoftnet();
3440 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3441 splx(s);
3442 if (mrtdebug & DEBUG_PIM)
3443 log(LOG_DEBUG,
3444 "pim_input: register vif not set: %d\n", reg_vif_num);
3445 m_freem(m);
3446 return;
3447 }
3448 /* XXX need refcnt? */
3449 vifp = viftable[reg_vif_num].v_ifp;
3450 splx(s);
3451
3452 /*
3453 * Validate length
3454 */
3455 if (datalen < PIM_REG_MINLEN) {
3456 pimstat.pims_rcv_tooshort++;
3457 pimstat.pims_rcv_badregisters++;
3458 log(LOG_ERR,
3459 "pim_input: register packet size too small %d from %lx\n",
3460 datalen, (u_long)ip->ip_src.s_addr);
3461 m_freem(m);
3462 return;
3463 }
3464
3465 reghdr = (u_int32_t *)(pim + 1);
3466 encap_ip = (struct ip *)(reghdr + 1);
3467
3468 if (mrtdebug & DEBUG_PIM) {
3469 log(LOG_DEBUG,
3470 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3471 (u_long)ntohl(encap_ip->ip_src.s_addr),
3472 (u_long)ntohl(encap_ip->ip_dst.s_addr),
3473 ntohs(encap_ip->ip_len));
3474 }
3475
3476 /* verify the version number of the inner packet */
3477 if (encap_ip->ip_v != IPVERSION) {
3478 pimstat.pims_rcv_badregisters++;
3479 if (mrtdebug & DEBUG_PIM) {
3480 log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3481 "of the inner packet\n", encap_ip->ip_v);
3482 }
3483 m_freem(m);
3484 return;
3485 }
3486
3487 /* verify the inner packet is destined to a mcast group */
3488 if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) {
3489 pimstat.pims_rcv_badregisters++;
3490 if (mrtdebug & DEBUG_PIM)
3491 log(LOG_DEBUG,
3492 "pim_input: inner packet of register is not "
3493 "multicast %lx\n",
3494 (u_long)ntohl(encap_ip->ip_dst.s_addr));
3495 m_freem(m);
3496 return;
3497 }
3498
3499 /* If a NULL_REGISTER, pass it to the daemon */
3500 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3501 goto pim_input_to_daemon;
3502
3503 /*
3504 * Copy the TOS from the outer IP header to the inner IP header.
3505 */
3506 if (encap_ip->ip_tos != ip_tos) {
3507 /* Outer TOS -> inner TOS */
3508 encap_ip->ip_tos = ip_tos;
3509 /* Recompute the inner header checksum. Sigh... */
3510
3511 /* adjust mbuf to point to the inner IP header */
3512 m->m_data += (iphlen + PIM_MINLEN);
3513 m->m_len -= (iphlen + PIM_MINLEN);
3514
3515 encap_ip->ip_sum = 0;
3516 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3517
3518 /* restore mbuf to point back to the outer IP header */
3519 m->m_data -= (iphlen + PIM_MINLEN);
3520 m->m_len += (iphlen + PIM_MINLEN);
3521 }
3522
3523 /*
3524 * Decapsulate the inner IP packet and loopback to forward it
3525 * as a normal multicast packet. Also, make a copy of the
3526 * outer_iphdr + pimhdr + reghdr + encap_iphdr
3527 * to pass to the daemon later, so it can take the appropriate
3528 * actions (e.g., send back PIM_REGISTER_STOP).
3529 * XXX: here m->m_data points to the outer IP header.
3530 */
3531 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
3532 if (mcp == NULL) {
3533 log(LOG_ERR,
3534 "pim_input: pim register: could not copy register head\n");
3535 m_freem(m);
3536 return;
3537 }
3538
3539 /* Keep statistics */
3540 /* XXX: registers_bytes include only the encap. mcast pkt */
3541 pimstat.pims_rcv_registers_msgs++;
3542 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3543
3544 /*
3545 * forward the inner ip packet; point m_data at the inner ip.
3546 */
3547 m_adj(m, iphlen + PIM_MINLEN);
3548
3549 if (mrtdebug & DEBUG_PIM) {
3550 log(LOG_DEBUG,
3551 "pim_input: forwarding decapsulated register: "
3552 "src %lx, dst %lx, vif %d\n",
3553 (u_long)ntohl(encap_ip->ip_src.s_addr),
3554 (u_long)ntohl(encap_ip->ip_dst.s_addr),
3555 reg_vif_num);
3556 }
3557 /* NB: vifp was collected above; can it change on us? */
3558 looutput(vifp, m, (struct sockaddr *)&dst, (struct rtentry *)NULL);
3559
3560 /* prepare the register head to send to the mrouting daemon */
3561 m = mcp;
3562 }
3563
3564 pim_input_to_daemon:
3565 /*
3566 * Pass the PIM message up to the daemon; if it is a Register message,
3567 * pass the 'head' only up to the daemon. This includes the
3568 * outer IP header, PIM header, PIM-Register header and the
3569 * inner IP header.
3570 * XXX: the outer IP header pkt size of a Register is not adjust to
3571 * reflect the fact that the inner multicast data is truncated.
3572 */
3573 rip_input(m, iphlen, proto);
3574
3575 return;
3576 }
3577 #endif /* PIM */
Cache object: 96ba10bf00d88440cc673d72e5d3ccf4
|