1 /*-
2 * Copyright (c) 1989 Stephen Deering
3 * Copyright (c) 1992, 1993
4 * The Regents of the University of California. All rights reserved.
5 *
6 * This code is derived from software contributed to Berkeley by
7 * Stephen Deering of Stanford University.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
34 */
35
36 /*
37 * IP multicast forwarding procedures
38 *
39 * Written by David Waitzman, BBN Labs, August 1988.
40 * Modified by Steve Deering, Stanford, February 1989.
41 * Modified by Mark J. Steiglitz, Stanford, May, 1991
42 * Modified by Van Jacobson, LBL, January 1993
43 * Modified by Ajit Thyagarajan, PARC, August 1993
44 * Modified by Bill Fenner, PARC, April 1995
45 * Modified by Ahmed Helmy, SGI, June 1996
46 * Modified by George Edmond Eddy (Rusty), ISI, February 1998
47 * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
48 * Modified by Hitoshi Asaeda, WIDE, August 2000
49 * Modified by Pavlin Radoslavov, ICSI, October 2002
50 *
51 * MROUTING Revision: 3.5
52 * and PIM-SMv2 and PIM-DM support, advanced API support,
53 * bandwidth metering and signaling
54 *
55 * $FreeBSD: releng/6.0/sys/netinet/ip_mroute.c 147549 2005-06-23 18:42:58Z imp $
56 */
57
58 #include "opt_mac.h"
59 #include "opt_mrouting.h"
60
61 #ifdef PIM
62 #define _PIM_VT 1
63 #endif
64
65 #include <sys/param.h>
66 #include <sys/kernel.h>
67 #include <sys/lock.h>
68 #include <sys/mac.h>
69 #include <sys/malloc.h>
70 #include <sys/mbuf.h>
71 #include <sys/module.h>
72 #include <sys/protosw.h>
73 #include <sys/signalvar.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/sockio.h>
77 #include <sys/sx.h>
78 #include <sys/sysctl.h>
79 #include <sys/syslog.h>
80 #include <sys/systm.h>
81 #include <sys/time.h>
82 #include <net/if.h>
83 #include <net/netisr.h>
84 #include <net/route.h>
85 #include <netinet/in.h>
86 #include <netinet/igmp.h>
87 #include <netinet/in_systm.h>
88 #include <netinet/in_var.h>
89 #include <netinet/ip.h>
90 #include <netinet/ip_encap.h>
91 #include <netinet/ip_mroute.h>
92 #include <netinet/ip_var.h>
93 #ifdef PIM
94 #include <netinet/pim.h>
95 #include <netinet/pim_var.h>
96 #endif
97 #include <netinet/udp.h>
98 #include <machine/in_cksum.h>
99
100 /*
101 * Control debugging code for rsvp and multicast routing code.
102 * Can only set them with the debugger.
103 */
104 static u_int rsvpdebug; /* non-zero enables debugging */
105
106 static u_int mrtdebug; /* any set of the flags below */
107 #define DEBUG_MFC 0x02
108 #define DEBUG_FORWARD 0x04
109 #define DEBUG_EXPIRE 0x08
110 #define DEBUG_XMIT 0x10
111 #define DEBUG_PIM 0x20
112
113 #define VIFI_INVALID ((vifi_t) -1)
114
115 #define M_HASCL(m) ((m)->m_flags & M_EXT)
116
117 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables");
118
119 /*
120 * Locking. We use two locks: one for the virtual interface table and
121 * one for the forwarding table. These locks may be nested in which case
122 * the VIF lock must always be taken first. Note that each lock is used
123 * to cover not only the specific data structure but also related data
124 * structures. It may be better to add more fine-grained locking later;
125 * it's not clear how performance-critical this code is.
126 */
127
128 static struct mrtstat mrtstat;
129 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
130 &mrtstat, mrtstat,
131 "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)");
132
133 static struct mfc *mfctable[MFCTBLSIZ];
134 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
135 &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]",
136 "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)");
137
138 static struct mtx mfc_mtx;
139 #define MFC_LOCK() mtx_lock(&mfc_mtx)
140 #define MFC_UNLOCK() mtx_unlock(&mfc_mtx)
141 #define MFC_LOCK_ASSERT() do { \
142 mtx_assert(&mfc_mtx, MA_OWNED); \
143 NET_ASSERT_GIANT(); \
144 } while (0)
145 #define MFC_LOCK_INIT() mtx_init(&mfc_mtx, "mroute mfc table", NULL, MTX_DEF)
146 #define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx)
147
148 static struct vif viftable[MAXVIFS];
149 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
150 &viftable, sizeof(viftable), "S,vif[MAXVIFS]",
151 "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
152
153 static struct mtx vif_mtx;
154 #define VIF_LOCK() mtx_lock(&vif_mtx)
155 #define VIF_UNLOCK() mtx_unlock(&vif_mtx)
156 #define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED)
157 #define VIF_LOCK_INIT() mtx_init(&vif_mtx, "mroute vif table", NULL, MTX_DEF)
158 #define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx)
159
160 static u_char nexpire[MFCTBLSIZ];
161
162 static struct callout expire_upcalls_ch;
163
164 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
165 #define UPCALL_EXPIRE 6 /* number of timeouts */
166
167 /*
168 * Define the token bucket filter structures
169 * tbftable -> each vif has one of these for storing info
170 */
171
172 static struct tbf tbftable[MAXVIFS];
173 #define TBF_REPROCESS (hz / 100) /* 100x / second */
174
175 /*
176 * 'Interfaces' associated with decapsulator (so we can tell
177 * packets that went through it from ones that get reflected
178 * by a broken gateway). These interfaces are never linked into
179 * the system ifnet list & no routes point to them. I.e., packets
180 * can't be sent this way. They only exist as a placeholder for
181 * multicast source verification.
182 */
183 static struct ifnet multicast_decap_if[MAXVIFS];
184
185 #define ENCAP_TTL 64
186 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */
187
188 /* prototype IP hdr for encapsulated packets */
189 static struct ip multicast_encap_iphdr = {
190 #if BYTE_ORDER == LITTLE_ENDIAN
191 sizeof(struct ip) >> 2, IPVERSION,
192 #else
193 IPVERSION, sizeof(struct ip) >> 2,
194 #endif
195 0, /* tos */
196 sizeof(struct ip), /* total length */
197 0, /* id */
198 0, /* frag offset */
199 ENCAP_TTL, ENCAP_PROTO,
200 0, /* checksum */
201 };
202
203 /*
204 * Bandwidth meter variables and constants
205 */
206 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
207 /*
208 * Pending timeouts are stored in a hash table, the key being the
209 * expiration time. Periodically, the entries are analysed and processed.
210 */
211 #define BW_METER_BUCKETS 1024
212 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
213 static struct callout bw_meter_ch;
214 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
215
216 /*
217 * Pending upcalls are stored in a vector which is flushed when
218 * full, or periodically
219 */
220 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX];
221 static u_int bw_upcalls_n; /* # of pending upcalls */
222 static struct callout bw_upcalls_ch;
223 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
224
225 #ifdef PIM
226 static struct pimstat pimstat;
227 SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
228 &pimstat, pimstat,
229 "PIM Statistics (struct pimstat, netinet/pim_var.h)");
230
231 /*
232 * Note: the PIM Register encapsulation adds the following in front of a
233 * data packet:
234 *
235 * struct pim_encap_hdr {
236 * struct ip ip;
237 * struct pim_encap_pimhdr pim;
238 * }
239 *
240 */
241
242 struct pim_encap_pimhdr {
243 struct pim pim;
244 uint32_t flags;
245 };
246
247 static struct ip pim_encap_iphdr = {
248 #if BYTE_ORDER == LITTLE_ENDIAN
249 sizeof(struct ip) >> 2,
250 IPVERSION,
251 #else
252 IPVERSION,
253 sizeof(struct ip) >> 2,
254 #endif
255 0, /* tos */
256 sizeof(struct ip), /* total length */
257 0, /* id */
258 0, /* frag offset */
259 ENCAP_TTL,
260 IPPROTO_PIM,
261 0, /* checksum */
262 };
263
264 static struct pim_encap_pimhdr pim_encap_pimhdr = {
265 {
266 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
267 0, /* reserved */
268 0, /* checksum */
269 },
270 0 /* flags */
271 };
272
273 static struct ifnet multicast_register_if;
274 static vifi_t reg_vif_num = VIFI_INVALID;
275 #endif /* PIM */
276
277 /*
278 * Private variables.
279 */
280 static vifi_t numvifs;
281 static const struct encaptab *encap_cookie;
282
283 /*
284 * one-back cache used by mroute_encapcheck to locate a tunnel's vif
285 * given a datagram's src ip address.
286 */
287 static u_long last_encap_src;
288 static struct vif *last_encap_vif;
289
290 /*
291 * Callout for queue processing.
292 */
293 static struct callout tbf_reprocess_ch;
294
295 static u_long X_ip_mcast_src(int vifi);
296 static int X_ip_mforward(struct ip *ip, struct ifnet *ifp,
297 struct mbuf *m, struct ip_moptions *imo);
298 static int X_ip_mrouter_done(void);
299 static int X_ip_mrouter_get(struct socket *so, struct sockopt *m);
300 static int X_ip_mrouter_set(struct socket *so, struct sockopt *m);
301 static int X_legal_vif_num(int vif);
302 static int X_mrt_ioctl(int cmd, caddr_t data);
303
304 static int get_sg_cnt(struct sioc_sg_req *);
305 static int get_vif_cnt(struct sioc_vif_req *);
306 static int ip_mrouter_init(struct socket *, int);
307 static int add_vif(struct vifctl *);
308 static int del_vif(vifi_t);
309 static int add_mfc(struct mfcctl2 *);
310 static int del_mfc(struct mfcctl2 *);
311 static int set_api_config(uint32_t *); /* chose API capabilities */
312 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
313 static int set_assert(int);
314 static void expire_upcalls(void *);
315 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
316 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
317 static void encap_send(struct ip *, struct vif *, struct mbuf *);
318 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long);
319 static void tbf_queue(struct vif *, struct mbuf *);
320 static void tbf_process_q(struct vif *);
321 static void tbf_reprocess_q(void *);
322 static int tbf_dq_sel(struct vif *, struct ip *);
323 static void tbf_send_packet(struct vif *, struct mbuf *);
324 static void tbf_update_tokens(struct vif *);
325 static int priority(struct vif *, struct ip *);
326
327 /*
328 * Bandwidth monitoring
329 */
330 static void free_bw_list(struct bw_meter *list);
331 static int add_bw_upcall(struct bw_upcall *);
332 static int del_bw_upcall(struct bw_upcall *);
333 static void bw_meter_receive_packet(struct bw_meter *x, int plen,
334 struct timeval *nowp);
335 static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp);
336 static void bw_upcalls_send(void);
337 static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp);
338 static void unschedule_bw_meter(struct bw_meter *x);
339 static void bw_meter_process(void);
340 static void expire_bw_upcalls_send(void *);
341 static void expire_bw_meter_process(void *);
342
343 #ifdef PIM
344 static int pim_register_send(struct ip *, struct vif *,
345 struct mbuf *, struct mfc *);
346 static int pim_register_send_rp(struct ip *, struct vif *,
347 struct mbuf *, struct mfc *);
348 static int pim_register_send_upcall(struct ip *, struct vif *,
349 struct mbuf *, struct mfc *);
350 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
351 #endif
352
353 /*
354 * whether or not special PIM assert processing is enabled.
355 */
356 static int pim_assert;
357 /*
358 * Rate limit for assert notification messages, in usec
359 */
360 #define ASSERT_MSG_TIME 3000000
361
362 /*
363 * Kernel multicast routing API capabilities and setup.
364 * If more API capabilities are added to the kernel, they should be
365 * recorded in `mrt_api_support'.
366 */
367 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
368 MRT_MFC_FLAGS_BORDER_VIF |
369 MRT_MFC_RP |
370 MRT_MFC_BW_UPCALL);
371 static uint32_t mrt_api_config = 0;
372
373 /*
374 * Hash function for a source, group entry
375 */
376 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
377 ((g) >> 20) ^ ((g) >> 10) ^ (g))
378
379 /*
380 * Find a route for a given origin IP address and Multicast group address
381 * Type of service parameter to be added in the future!!!
382 * Statistics are updated by the caller if needed
383 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
384 */
385 static struct mfc *
386 mfc_find(in_addr_t o, in_addr_t g)
387 {
388 struct mfc *rt;
389
390 MFC_LOCK_ASSERT();
391
392 for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next)
393 if ((rt->mfc_origin.s_addr == o) &&
394 (rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL))
395 break;
396 return rt;
397 }
398
399 /*
400 * Macros to compute elapsed time efficiently
401 * Borrowed from Van Jacobson's scheduling code
402 */
403 #define TV_DELTA(a, b, delta) { \
404 int xxs; \
405 delta = (a).tv_usec - (b).tv_usec; \
406 if ((xxs = (a).tv_sec - (b).tv_sec)) { \
407 switch (xxs) { \
408 case 2: \
409 delta += 1000000; \
410 /* FALLTHROUGH */ \
411 case 1: \
412 delta += 1000000; \
413 break; \
414 default: \
415 delta += (1000000 * xxs); \
416 } \
417 } \
418 }
419
420 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
421 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
422
423 /*
424 * Handle MRT setsockopt commands to modify the multicast routing tables.
425 */
426 static int
427 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
428 {
429 int error, optval;
430 vifi_t vifi;
431 struct vifctl vifc;
432 struct mfcctl2 mfc;
433 struct bw_upcall bw_upcall;
434 uint32_t i;
435
436 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT)
437 return EPERM;
438
439 error = 0;
440 switch (sopt->sopt_name) {
441 case MRT_INIT:
442 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
443 if (error)
444 break;
445 error = ip_mrouter_init(so, optval);
446 break;
447
448 case MRT_DONE:
449 error = ip_mrouter_done();
450 break;
451
452 case MRT_ADD_VIF:
453 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
454 if (error)
455 break;
456 error = add_vif(&vifc);
457 break;
458
459 case MRT_DEL_VIF:
460 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
461 if (error)
462 break;
463 error = del_vif(vifi);
464 break;
465
466 case MRT_ADD_MFC:
467 case MRT_DEL_MFC:
468 /*
469 * select data size depending on API version.
470 */
471 if (sopt->sopt_name == MRT_ADD_MFC &&
472 mrt_api_config & MRT_API_FLAGS_ALL) {
473 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
474 sizeof(struct mfcctl2));
475 } else {
476 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
477 sizeof(struct mfcctl));
478 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
479 sizeof(mfc) - sizeof(struct mfcctl));
480 }
481 if (error)
482 break;
483 if (sopt->sopt_name == MRT_ADD_MFC)
484 error = add_mfc(&mfc);
485 else
486 error = del_mfc(&mfc);
487 break;
488
489 case MRT_ASSERT:
490 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
491 if (error)
492 break;
493 set_assert(optval);
494 break;
495
496 case MRT_API_CONFIG:
497 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
498 if (!error)
499 error = set_api_config(&i);
500 if (!error)
501 error = sooptcopyout(sopt, &i, sizeof i);
502 break;
503
504 case MRT_ADD_BW_UPCALL:
505 case MRT_DEL_BW_UPCALL:
506 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
507 sizeof bw_upcall);
508 if (error)
509 break;
510 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
511 error = add_bw_upcall(&bw_upcall);
512 else
513 error = del_bw_upcall(&bw_upcall);
514 break;
515
516 default:
517 error = EOPNOTSUPP;
518 break;
519 }
520 return error;
521 }
522
523 /*
524 * Handle MRT getsockopt commands
525 */
526 static int
527 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
528 {
529 int error;
530 static int version = 0x0305; /* !!! why is this here? XXX */
531
532 switch (sopt->sopt_name) {
533 case MRT_VERSION:
534 error = sooptcopyout(sopt, &version, sizeof version);
535 break;
536
537 case MRT_ASSERT:
538 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert);
539 break;
540
541 case MRT_API_SUPPORT:
542 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
543 break;
544
545 case MRT_API_CONFIG:
546 error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config);
547 break;
548
549 default:
550 error = EOPNOTSUPP;
551 break;
552 }
553 return error;
554 }
555
556 /*
557 * Handle ioctl commands to obtain information from the cache
558 */
559 static int
560 X_mrt_ioctl(int cmd, caddr_t data)
561 {
562 int error = 0;
563
564 /*
565 * Currently the only function calling this ioctl routine is rtioctl().
566 * Typically, only root can create the raw socket in order to execute
567 * this ioctl method, however the request might be coming from a prison
568 */
569 error = suser(curthread);
570 if (error)
571 return (error);
572 switch (cmd) {
573 case (SIOCGETVIFCNT):
574 error = get_vif_cnt((struct sioc_vif_req *)data);
575 break;
576
577 case (SIOCGETSGCNT):
578 error = get_sg_cnt((struct sioc_sg_req *)data);
579 break;
580
581 default:
582 error = EINVAL;
583 break;
584 }
585 return error;
586 }
587
588 /*
589 * returns the packet, byte, rpf-failure count for the source group provided
590 */
591 static int
592 get_sg_cnt(struct sioc_sg_req *req)
593 {
594 struct mfc *rt;
595
596 MFC_LOCK();
597 rt = mfc_find(req->src.s_addr, req->grp.s_addr);
598 if (rt == NULL) {
599 MFC_UNLOCK();
600 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
601 return EADDRNOTAVAIL;
602 }
603 req->pktcnt = rt->mfc_pkt_cnt;
604 req->bytecnt = rt->mfc_byte_cnt;
605 req->wrong_if = rt->mfc_wrong_if;
606 MFC_UNLOCK();
607 return 0;
608 }
609
610 /*
611 * returns the input and output packet and byte counts on the vif provided
612 */
613 static int
614 get_vif_cnt(struct sioc_vif_req *req)
615 {
616 vifi_t vifi = req->vifi;
617
618 VIF_LOCK();
619 if (vifi >= numvifs) {
620 VIF_UNLOCK();
621 return EINVAL;
622 }
623
624 req->icount = viftable[vifi].v_pkt_in;
625 req->ocount = viftable[vifi].v_pkt_out;
626 req->ibytes = viftable[vifi].v_bytes_in;
627 req->obytes = viftable[vifi].v_bytes_out;
628 VIF_UNLOCK();
629
630 return 0;
631 }
632
633 static void
634 ip_mrouter_reset(void)
635 {
636 bzero((caddr_t)mfctable, sizeof(mfctable));
637 bzero((caddr_t)nexpire, sizeof(nexpire));
638
639 pim_assert = 0;
640 mrt_api_config = 0;
641
642 callout_init(&expire_upcalls_ch, NET_CALLOUT_MPSAFE);
643
644 bw_upcalls_n = 0;
645 bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers));
646 callout_init(&bw_upcalls_ch, NET_CALLOUT_MPSAFE);
647 callout_init(&bw_meter_ch, NET_CALLOUT_MPSAFE);
648
649 callout_init(&tbf_reprocess_ch, NET_CALLOUT_MPSAFE);
650 }
651
652 static struct mtx mrouter_mtx; /* used to synch init/done work */
653
654 /*
655 * Enable multicast routing
656 */
657 static int
658 ip_mrouter_init(struct socket *so, int version)
659 {
660 if (mrtdebug)
661 log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
662 so->so_type, so->so_proto->pr_protocol);
663
664 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
665 return EOPNOTSUPP;
666
667 if (version != 1)
668 return ENOPROTOOPT;
669
670 mtx_lock(&mrouter_mtx);
671
672 if (ip_mrouter != NULL) {
673 mtx_unlock(&mrouter_mtx);
674 return EADDRINUSE;
675 }
676
677 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
678
679 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
680 expire_bw_upcalls_send, NULL);
681 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
682
683 ip_mrouter = so;
684
685 mtx_unlock(&mrouter_mtx);
686
687 if (mrtdebug)
688 log(LOG_DEBUG, "ip_mrouter_init\n");
689
690 return 0;
691 }
692
693 /*
694 * Disable multicast routing
695 */
696 static int
697 X_ip_mrouter_done(void)
698 {
699 vifi_t vifi;
700 int i;
701 struct ifnet *ifp;
702 struct ifreq ifr;
703 struct mfc *rt;
704 struct rtdetq *rte;
705
706 mtx_lock(&mrouter_mtx);
707
708 if (ip_mrouter == NULL) {
709 mtx_unlock(&mrouter_mtx);
710 return EINVAL;
711 }
712
713 /*
714 * Detach/disable hooks to the reset of the system.
715 */
716 ip_mrouter = NULL;
717 mrt_api_config = 0;
718
719 VIF_LOCK();
720 if (encap_cookie) {
721 const struct encaptab *c = encap_cookie;
722 encap_cookie = NULL;
723 encap_detach(c);
724 }
725 VIF_UNLOCK();
726
727 callout_stop(&tbf_reprocess_ch);
728
729 VIF_LOCK();
730 /*
731 * For each phyint in use, disable promiscuous reception of all IP
732 * multicasts.
733 */
734 for (vifi = 0; vifi < numvifs; vifi++) {
735 if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
736 !(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
737 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr);
738
739 so->sin_len = sizeof(struct sockaddr_in);
740 so->sin_family = AF_INET;
741 so->sin_addr.s_addr = INADDR_ANY;
742 ifp = viftable[vifi].v_ifp;
743 if_allmulti(ifp, 0);
744 }
745 }
746 bzero((caddr_t)tbftable, sizeof(tbftable));
747 bzero((caddr_t)viftable, sizeof(viftable));
748 numvifs = 0;
749 pim_assert = 0;
750 VIF_UNLOCK();
751
752 /*
753 * Free all multicast forwarding cache entries.
754 */
755 callout_stop(&expire_upcalls_ch);
756 callout_stop(&bw_upcalls_ch);
757 callout_stop(&bw_meter_ch);
758
759 MFC_LOCK();
760 for (i = 0; i < MFCTBLSIZ; i++) {
761 for (rt = mfctable[i]; rt != NULL; ) {
762 struct mfc *nr = rt->mfc_next;
763
764 for (rte = rt->mfc_stall; rte != NULL; ) {
765 struct rtdetq *n = rte->next;
766
767 m_freem(rte->m);
768 free(rte, M_MRTABLE);
769 rte = n;
770 }
771 free_bw_list(rt->mfc_bw_meter);
772 free(rt, M_MRTABLE);
773 rt = nr;
774 }
775 }
776 bzero((caddr_t)mfctable, sizeof(mfctable));
777 bzero((caddr_t)nexpire, sizeof(nexpire));
778 bw_upcalls_n = 0;
779 bzero(bw_meter_timers, sizeof(bw_meter_timers));
780 MFC_UNLOCK();
781
782 /*
783 * Reset de-encapsulation cache
784 */
785 last_encap_src = INADDR_ANY;
786 last_encap_vif = NULL;
787 #ifdef PIM
788 reg_vif_num = VIFI_INVALID;
789 #endif
790
791 mtx_unlock(&mrouter_mtx);
792
793 if (mrtdebug)
794 log(LOG_DEBUG, "ip_mrouter_done\n");
795
796 return 0;
797 }
798
799 /*
800 * Set PIM assert processing global
801 */
802 static int
803 set_assert(int i)
804 {
805 if ((i != 1) && (i != 0))
806 return EINVAL;
807
808 pim_assert = i;
809
810 return 0;
811 }
812
813 /*
814 * Configure API capabilities
815 */
816 int
817 set_api_config(uint32_t *apival)
818 {
819 int i;
820
821 /*
822 * We can set the API capabilities only if it is the first operation
823 * after MRT_INIT. I.e.:
824 * - there are no vifs installed
825 * - pim_assert is not enabled
826 * - the MFC table is empty
827 */
828 if (numvifs > 0) {
829 *apival = 0;
830 return EPERM;
831 }
832 if (pim_assert) {
833 *apival = 0;
834 return EPERM;
835 }
836 for (i = 0; i < MFCTBLSIZ; i++) {
837 if (mfctable[i] != NULL) {
838 *apival = 0;
839 return EPERM;
840 }
841 }
842
843 mrt_api_config = *apival & mrt_api_support;
844 *apival = mrt_api_config;
845
846 return 0;
847 }
848
849 /*
850 * Decide if a packet is from a tunnelled peer.
851 * Return 0 if not, 64 if so. XXX yuck.. 64 ???
852 */
853 static int
854 mroute_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
855 {
856 struct ip *ip = mtod(m, struct ip *);
857 int hlen = ip->ip_hl << 2;
858
859 /*
860 * don't claim the packet if it's not to a multicast destination or if
861 * we don't have an encapsulating tunnel with the source.
862 * Note: This code assumes that the remote site IP address
863 * uniquely identifies the tunnel (i.e., that this site has
864 * at most one tunnel with the remote site).
865 */
866 if (!IN_MULTICAST(ntohl(((struct ip *)((char *)ip+hlen))->ip_dst.s_addr)))
867 return 0;
868 if (ip->ip_src.s_addr != last_encap_src) {
869 struct vif *vifp = viftable;
870 struct vif *vife = vifp + numvifs;
871
872 last_encap_src = ip->ip_src.s_addr;
873 last_encap_vif = NULL;
874 for ( ; vifp < vife; ++vifp)
875 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
876 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) == VIFF_TUNNEL)
877 last_encap_vif = vifp;
878 break;
879 }
880 }
881 if (last_encap_vif == NULL) {
882 last_encap_src = INADDR_ANY;
883 return 0;
884 }
885 return 64;
886 }
887
888 /*
889 * De-encapsulate a packet and feed it back through ip input (this
890 * routine is called whenever IP gets a packet that mroute_encap_func()
891 * claimed).
892 */
893 static void
894 mroute_encap_input(struct mbuf *m, int off)
895 {
896 struct ip *ip = mtod(m, struct ip *);
897 int hlen = ip->ip_hl << 2;
898
899 if (hlen > sizeof(struct ip))
900 ip_stripoptions(m, (struct mbuf *) 0);
901 m->m_data += sizeof(struct ip);
902 m->m_len -= sizeof(struct ip);
903 m->m_pkthdr.len -= sizeof(struct ip);
904
905 m->m_pkthdr.rcvif = last_encap_vif->v_ifp;
906
907 netisr_queue(NETISR_IP, m); /* mbuf is free'd on failure. */
908 /*
909 * normally we would need a "schednetisr(NETISR_IP)"
910 * here but we were called by ip_input and it is going
911 * to loop back & try to dequeue the packet we just
912 * queued as soon as we return so we avoid the
913 * unnecessary software interrrupt.
914 *
915 * XXX
916 * This no longer holds - we may have direct-dispatched the packet,
917 * or there may be a queue processing limit.
918 */
919 }
920
921 extern struct domain inetdomain;
922 static struct protosw mroute_encap_protosw =
923 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR,
924 mroute_encap_input, 0, 0, rip_ctloutput,
925 0,
926 0, 0, 0, 0,
927 &rip_usrreqs
928 };
929
930 /*
931 * Add a vif to the vif table
932 */
933 static int
934 add_vif(struct vifctl *vifcp)
935 {
936 struct vif *vifp = viftable + vifcp->vifc_vifi;
937 struct sockaddr_in sin = {sizeof sin, AF_INET};
938 struct ifaddr *ifa;
939 struct ifnet *ifp;
940 int error;
941 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
942
943 VIF_LOCK();
944 if (vifcp->vifc_vifi >= MAXVIFS) {
945 VIF_UNLOCK();
946 return EINVAL;
947 }
948 if (vifp->v_lcl_addr.s_addr != INADDR_ANY) {
949 VIF_UNLOCK();
950 return EADDRINUSE;
951 }
952 if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) {
953 VIF_UNLOCK();
954 return EADDRNOTAVAIL;
955 }
956
957 /* Find the interface with an address in AF_INET family */
958 #ifdef PIM
959 if (vifcp->vifc_flags & VIFF_REGISTER) {
960 /*
961 * XXX: Because VIFF_REGISTER does not really need a valid
962 * local interface (e.g. it could be 127.0.0.2), we don't
963 * check its address.
964 */
965 ifp = NULL;
966 } else
967 #endif
968 {
969 sin.sin_addr = vifcp->vifc_lcl_addr;
970 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
971 if (ifa == NULL) {
972 VIF_UNLOCK();
973 return EADDRNOTAVAIL;
974 }
975 ifp = ifa->ifa_ifp;
976 }
977
978 if (vifcp->vifc_flags & VIFF_TUNNEL) {
979 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
980 /*
981 * An encapsulating tunnel is wanted. Tell
982 * mroute_encap_input() to start paying attention
983 * to encapsulated packets.
984 */
985 if (encap_cookie == NULL) {
986 int i;
987
988 encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
989 mroute_encapcheck,
990 (struct protosw *)&mroute_encap_protosw, NULL);
991
992 if (encap_cookie == NULL) {
993 printf("ip_mroute: unable to attach encap\n");
994 VIF_UNLOCK();
995 return EIO; /* XXX */
996 }
997 for (i = 0; i < MAXVIFS; ++i) {
998 if_initname(&multicast_decap_if[i], "mdecap", i);
999 }
1000 }
1001 /*
1002 * Set interface to fake encapsulator interface
1003 */
1004 ifp = &multicast_decap_if[vifcp->vifc_vifi];
1005 /*
1006 * Prepare cached route entry
1007 */
1008 bzero(&vifp->v_route, sizeof(vifp->v_route));
1009 } else {
1010 log(LOG_ERR, "source routed tunnels not supported\n");
1011 VIF_UNLOCK();
1012 return EOPNOTSUPP;
1013 }
1014 #ifdef PIM
1015 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
1016 ifp = &multicast_register_if;
1017 if (mrtdebug)
1018 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
1019 (void *)&multicast_register_if);
1020 if (reg_vif_num == VIFI_INVALID) {
1021 if_initname(&multicast_register_if, "register_vif", 0);
1022 multicast_register_if.if_flags = IFF_LOOPBACK;
1023 bzero(&vifp->v_route, sizeof(vifp->v_route));
1024 reg_vif_num = vifcp->vifc_vifi;
1025 }
1026 #endif
1027 } else { /* Make sure the interface supports multicast */
1028 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
1029 VIF_UNLOCK();
1030 return EOPNOTSUPP;
1031 }
1032
1033 /* Enable promiscuous reception of all IP multicasts from the if */
1034 error = if_allmulti(ifp, 1);
1035 if (error) {
1036 VIF_UNLOCK();
1037 return error;
1038 }
1039 }
1040
1041 /* define parameters for the tbf structure */
1042 vifp->v_tbf = v_tbf;
1043 GET_TIME(vifp->v_tbf->tbf_last_pkt_t);
1044 vifp->v_tbf->tbf_n_tok = 0;
1045 vifp->v_tbf->tbf_q_len = 0;
1046 vifp->v_tbf->tbf_max_q_len = MAXQSIZE;
1047 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL;
1048
1049 vifp->v_flags = vifcp->vifc_flags;
1050 vifp->v_threshold = vifcp->vifc_threshold;
1051 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
1052 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
1053 vifp->v_ifp = ifp;
1054 /* scaling up here allows division by 1024 in critical code */
1055 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000;
1056 vifp->v_rsvp_on = 0;
1057 vifp->v_rsvpd = NULL;
1058 /* initialize per vif pkt counters */
1059 vifp->v_pkt_in = 0;
1060 vifp->v_pkt_out = 0;
1061 vifp->v_bytes_in = 0;
1062 vifp->v_bytes_out = 0;
1063
1064 /* Adjust numvifs up if the vifi is higher than numvifs */
1065 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
1066
1067 VIF_UNLOCK();
1068
1069 if (mrtdebug)
1070 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n",
1071 vifcp->vifc_vifi,
1072 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr),
1073 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
1074 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr),
1075 vifcp->vifc_threshold,
1076 vifcp->vifc_rate_limit);
1077
1078 return 0;
1079 }
1080
1081 /*
1082 * Delete a vif from the vif table
1083 */
1084 static int
1085 del_vif(vifi_t vifi)
1086 {
1087 struct vif *vifp;
1088
1089 VIF_LOCK();
1090
1091 if (vifi >= numvifs) {
1092 VIF_UNLOCK();
1093 return EINVAL;
1094 }
1095 vifp = &viftable[vifi];
1096 if (vifp->v_lcl_addr.s_addr == INADDR_ANY) {
1097 VIF_UNLOCK();
1098 return EADDRNOTAVAIL;
1099 }
1100
1101 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1102 if_allmulti(vifp->v_ifp, 0);
1103
1104 if (vifp == last_encap_vif) {
1105 last_encap_vif = NULL;
1106 last_encap_src = INADDR_ANY;
1107 }
1108
1109 /*
1110 * Free packets queued at the interface
1111 */
1112 while (vifp->v_tbf->tbf_q) {
1113 struct mbuf *m = vifp->v_tbf->tbf_q;
1114
1115 vifp->v_tbf->tbf_q = m->m_act;
1116 m_freem(m);
1117 }
1118
1119 #ifdef PIM
1120 if (vifp->v_flags & VIFF_REGISTER)
1121 reg_vif_num = VIFI_INVALID;
1122 #endif
1123
1124 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
1125 bzero((caddr_t)vifp, sizeof (*vifp));
1126
1127 if (mrtdebug)
1128 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs);
1129
1130 /* Adjust numvifs down */
1131 for (vifi = numvifs; vifi > 0; vifi--)
1132 if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY)
1133 break;
1134 numvifs = vifi;
1135
1136 VIF_UNLOCK();
1137
1138 return 0;
1139 }
1140
1141 /*
1142 * update an mfc entry without resetting counters and S,G addresses.
1143 */
1144 static void
1145 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1146 {
1147 int i;
1148
1149 rt->mfc_parent = mfccp->mfcc_parent;
1150 for (i = 0; i < numvifs; i++) {
1151 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1152 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1153 MRT_MFC_FLAGS_ALL;
1154 }
1155 /* set the RP address */
1156 if (mrt_api_config & MRT_MFC_RP)
1157 rt->mfc_rp = mfccp->mfcc_rp;
1158 else
1159 rt->mfc_rp.s_addr = INADDR_ANY;
1160 }
1161
1162 /*
1163 * fully initialize an mfc entry from the parameter.
1164 */
1165 static void
1166 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1167 {
1168 rt->mfc_origin = mfccp->mfcc_origin;
1169 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1170
1171 update_mfc_params(rt, mfccp);
1172
1173 /* initialize pkt counters per src-grp */
1174 rt->mfc_pkt_cnt = 0;
1175 rt->mfc_byte_cnt = 0;
1176 rt->mfc_wrong_if = 0;
1177 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
1178 }
1179
1180
1181 /*
1182 * Add an mfc entry
1183 */
1184 static int
1185 add_mfc(struct mfcctl2 *mfccp)
1186 {
1187 struct mfc *rt;
1188 u_long hash;
1189 struct rtdetq *rte;
1190 u_short nstl;
1191
1192 VIF_LOCK();
1193 MFC_LOCK();
1194
1195 rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
1196
1197 /* If an entry already exists, just update the fields */
1198 if (rt) {
1199 if (mrtdebug & DEBUG_MFC)
1200 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n",
1201 (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1202 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1203 mfccp->mfcc_parent);
1204
1205 update_mfc_params(rt, mfccp);
1206 MFC_UNLOCK();
1207 VIF_UNLOCK();
1208 return 0;
1209 }
1210
1211 /*
1212 * Find the entry for which the upcall was made and update
1213 */
1214 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
1215 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) {
1216
1217 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
1218 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
1219 (rt->mfc_stall != NULL)) {
1220
1221 if (nstl++)
1222 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n",
1223 "multiple kernel entries",
1224 (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1225 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1226 mfccp->mfcc_parent, (void *)rt->mfc_stall);
1227
1228 if (mrtdebug & DEBUG_MFC)
1229 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n",
1230 (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1231 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1232 mfccp->mfcc_parent, (void *)rt->mfc_stall);
1233
1234 init_mfc_params(rt, mfccp);
1235
1236 rt->mfc_expire = 0; /* Don't clean this guy up */
1237 nexpire[hash]--;
1238
1239 /* free packets Qed at the end of this entry */
1240 for (rte = rt->mfc_stall; rte != NULL; ) {
1241 struct rtdetq *n = rte->next;
1242
1243 ip_mdq(rte->m, rte->ifp, rt, -1);
1244 m_freem(rte->m);
1245 free(rte, M_MRTABLE);
1246 rte = n;
1247 }
1248 rt->mfc_stall = NULL;
1249 }
1250 }
1251
1252 /*
1253 * It is possible that an entry is being inserted without an upcall
1254 */
1255 if (nstl == 0) {
1256 if (mrtdebug & DEBUG_MFC)
1257 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n",
1258 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1259 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1260 mfccp->mfcc_parent);
1261
1262 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) {
1263 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
1264 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
1265 init_mfc_params(rt, mfccp);
1266 if (rt->mfc_expire)
1267 nexpire[hash]--;
1268 rt->mfc_expire = 0;
1269 break; /* XXX */
1270 }
1271 }
1272 if (rt == NULL) { /* no upcall, so make a new entry */
1273 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1274 if (rt == NULL) {
1275 MFC_UNLOCK();
1276 VIF_UNLOCK();
1277 return ENOBUFS;
1278 }
1279
1280 init_mfc_params(rt, mfccp);
1281 rt->mfc_expire = 0;
1282 rt->mfc_stall = NULL;
1283
1284 rt->mfc_bw_meter = NULL;
1285 /* insert new entry at head of hash chain */
1286 rt->mfc_next = mfctable[hash];
1287 mfctable[hash] = rt;
1288 }
1289 }
1290 MFC_UNLOCK();
1291 VIF_UNLOCK();
1292 return 0;
1293 }
1294
1295 /*
1296 * Delete an mfc entry
1297 */
1298 static int
1299 del_mfc(struct mfcctl2 *mfccp)
1300 {
1301 struct in_addr origin;
1302 struct in_addr mcastgrp;
1303 struct mfc *rt;
1304 struct mfc **nptr;
1305 u_long hash;
1306 struct bw_meter *list;
1307
1308 origin = mfccp->mfcc_origin;
1309 mcastgrp = mfccp->mfcc_mcastgrp;
1310
1311 if (mrtdebug & DEBUG_MFC)
1312 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n",
1313 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1314
1315 MFC_LOCK();
1316
1317 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1318 for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next)
1319 if (origin.s_addr == rt->mfc_origin.s_addr &&
1320 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1321 rt->mfc_stall == NULL)
1322 break;
1323 if (rt == NULL) {
1324 MFC_UNLOCK();
1325 return EADDRNOTAVAIL;
1326 }
1327
1328 *nptr = rt->mfc_next;
1329
1330 /*
1331 * free the bw_meter entries
1332 */
1333 list = rt->mfc_bw_meter;
1334 rt->mfc_bw_meter = NULL;
1335
1336 free(rt, M_MRTABLE);
1337
1338 free_bw_list(list);
1339
1340 MFC_UNLOCK();
1341
1342 return 0;
1343 }
1344
1345 /*
1346 * Send a message to mrouted on the multicast routing socket
1347 */
1348 static int
1349 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1350 {
1351 if (s) {
1352 SOCKBUF_LOCK(&s->so_rcv);
1353 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1354 NULL) != 0) {
1355 sorwakeup_locked(s);
1356 return 0;
1357 }
1358 SOCKBUF_UNLOCK(&s->so_rcv);
1359 }
1360 m_freem(mm);
1361 return -1;
1362 }
1363
1364 /*
1365 * IP multicast forwarding function. This function assumes that the packet
1366 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1367 * pointed to by "ifp", and the packet is to be relayed to other networks
1368 * that have members of the packet's destination IP multicast group.
1369 *
1370 * The packet is returned unscathed to the caller, unless it is
1371 * erroneous, in which case a non-zero return value tells the caller to
1372 * discard it.
1373 */
1374
1375 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1376
1377 static int
1378 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1379 struct ip_moptions *imo)
1380 {
1381 struct mfc *rt;
1382 int error;
1383 vifi_t vifi;
1384
1385 if (mrtdebug & DEBUG_FORWARD)
1386 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n",
1387 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr),
1388 (void *)ifp);
1389
1390 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1391 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1392 /*
1393 * Packet arrived via a physical interface or
1394 * an encapsulated tunnel or a register_vif.
1395 */
1396 } else {
1397 /*
1398 * Packet arrived through a source-route tunnel.
1399 * Source-route tunnels are no longer supported.
1400 */
1401 static int last_log;
1402 if (last_log != time_second) {
1403 last_log = time_second;
1404 log(LOG_ERR,
1405 "ip_mforward: received source-routed packet from %lx\n",
1406 (u_long)ntohl(ip->ip_src.s_addr));
1407 }
1408 return 1;
1409 }
1410
1411 VIF_LOCK();
1412 MFC_LOCK();
1413 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1414 if (ip->ip_ttl < 255)
1415 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1416 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1417 struct vif *vifp = viftable + vifi;
1418
1419 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s)\n",
1420 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr),
1421 vifi,
1422 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1423 vifp->v_ifp->if_xname);
1424 }
1425 error = ip_mdq(m, ifp, NULL, vifi);
1426 MFC_UNLOCK();
1427 VIF_UNLOCK();
1428 return error;
1429 }
1430 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1431 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n",
1432 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr));
1433 if (!imo)
1434 printf("In fact, no options were specified at all\n");
1435 }
1436
1437 /*
1438 * Don't forward a packet with time-to-live of zero or one,
1439 * or a packet destined to a local-only group.
1440 */
1441 if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) {
1442 MFC_UNLOCK();
1443 VIF_UNLOCK();
1444 return 0;
1445 }
1446
1447 /*
1448 * Determine forwarding vifs from the forwarding cache table
1449 */
1450 ++mrtstat.mrts_mfc_lookups;
1451 rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1452
1453 /* Entry exists, so forward if necessary */
1454 if (rt != NULL) {
1455 error = ip_mdq(m, ifp, rt, -1);
1456 MFC_UNLOCK();
1457 VIF_UNLOCK();
1458 return error;
1459 } else {
1460 /*
1461 * If we don't have a route for packet's origin,
1462 * Make a copy of the packet & send message to routing daemon
1463 */
1464
1465 struct mbuf *mb0;
1466 struct rtdetq *rte;
1467 u_long hash;
1468 int hlen = ip->ip_hl << 2;
1469
1470 ++mrtstat.mrts_mfc_misses;
1471
1472 mrtstat.mrts_no_route++;
1473 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1474 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n",
1475 (u_long)ntohl(ip->ip_src.s_addr),
1476 (u_long)ntohl(ip->ip_dst.s_addr));
1477
1478 /*
1479 * Allocate mbufs early so that we don't do extra work if we are
1480 * just going to fail anyway. Make sure to pullup the header so
1481 * that other people can't step on it.
1482 */
1483 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT);
1484 if (rte == NULL) {
1485 MFC_UNLOCK();
1486 VIF_UNLOCK();
1487 return ENOBUFS;
1488 }
1489 mb0 = m_copypacket(m, M_DONTWAIT);
1490 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1491 mb0 = m_pullup(mb0, hlen);
1492 if (mb0 == NULL) {
1493 free(rte, M_MRTABLE);
1494 MFC_UNLOCK();
1495 VIF_UNLOCK();
1496 return ENOBUFS;
1497 }
1498
1499 /* is there an upcall waiting for this flow ? */
1500 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1501 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) {
1502 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1503 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1504 (rt->mfc_stall != NULL))
1505 break;
1506 }
1507
1508 if (rt == NULL) {
1509 int i;
1510 struct igmpmsg *im;
1511 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1512 struct mbuf *mm;
1513
1514 /*
1515 * Locate the vifi for the incoming interface for this packet.
1516 * If none found, drop packet.
1517 */
1518 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
1519 ;
1520 if (vifi >= numvifs) /* vif not found, drop packet */
1521 goto non_fatal;
1522
1523 /* no upcall, so make a new entry */
1524 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1525 if (rt == NULL)
1526 goto fail;
1527 /* Make a copy of the header to send to the user level process */
1528 mm = m_copy(mb0, 0, hlen);
1529 if (mm == NULL)
1530 goto fail1;
1531
1532 /*
1533 * Send message to routing daemon to install
1534 * a route into the kernel table
1535 */
1536
1537 im = mtod(mm, struct igmpmsg *);
1538 im->im_msgtype = IGMPMSG_NOCACHE;
1539 im->im_mbz = 0;
1540 im->im_vif = vifi;
1541
1542 mrtstat.mrts_upcalls++;
1543
1544 k_igmpsrc.sin_addr = ip->ip_src;
1545 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1546 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1547 ++mrtstat.mrts_upq_sockfull;
1548 fail1:
1549 free(rt, M_MRTABLE);
1550 fail:
1551 free(rte, M_MRTABLE);
1552 m_freem(mb0);
1553 MFC_UNLOCK();
1554 VIF_UNLOCK();
1555 return ENOBUFS;
1556 }
1557
1558 /* insert new entry at head of hash chain */
1559 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1560 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1561 rt->mfc_expire = UPCALL_EXPIRE;
1562 nexpire[hash]++;
1563 for (i = 0; i < numvifs; i++) {
1564 rt->mfc_ttls[i] = 0;
1565 rt->mfc_flags[i] = 0;
1566 }
1567 rt->mfc_parent = -1;
1568
1569 rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */
1570
1571 rt->mfc_bw_meter = NULL;
1572
1573 /* link into table */
1574 rt->mfc_next = mfctable[hash];
1575 mfctable[hash] = rt;
1576 rt->mfc_stall = rte;
1577
1578 } else {
1579 /* determine if q has overflowed */
1580 int npkts = 0;
1581 struct rtdetq **p;
1582
1583 /*
1584 * XXX ouch! we need to append to the list, but we
1585 * only have a pointer to the front, so we have to
1586 * scan the entire list every time.
1587 */
1588 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1589 npkts++;
1590
1591 if (npkts > MAX_UPQ) {
1592 mrtstat.mrts_upq_ovflw++;
1593 non_fatal:
1594 free(rte, M_MRTABLE);
1595 m_freem(mb0);
1596 MFC_UNLOCK();
1597 VIF_UNLOCK();
1598 return 0;
1599 }
1600
1601 /* Add this entry to the end of the queue */
1602 *p = rte;
1603 }
1604
1605 rte->m = mb0;
1606 rte->ifp = ifp;
1607 rte->next = NULL;
1608
1609 MFC_UNLOCK();
1610 VIF_UNLOCK();
1611
1612 return 0;
1613 }
1614 }
1615
1616 /*
1617 * Clean up the cache entry if upcall is not serviced
1618 */
1619 static void
1620 expire_upcalls(void *unused)
1621 {
1622 struct rtdetq *rte;
1623 struct mfc *mfc, **nptr;
1624 int i;
1625
1626 MFC_LOCK();
1627 for (i = 0; i < MFCTBLSIZ; i++) {
1628 if (nexpire[i] == 0)
1629 continue;
1630 nptr = &mfctable[i];
1631 for (mfc = *nptr; mfc != NULL; mfc = *nptr) {
1632 /*
1633 * Skip real cache entries
1634 * Make sure it wasn't marked to not expire (shouldn't happen)
1635 * If it expires now
1636 */
1637 if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 &&
1638 --mfc->mfc_expire == 0) {
1639 if (mrtdebug & DEBUG_EXPIRE)
1640 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n",
1641 (u_long)ntohl(mfc->mfc_origin.s_addr),
1642 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr));
1643 /*
1644 * drop all the packets
1645 * free the mbuf with the pkt, if, timing info
1646 */
1647 for (rte = mfc->mfc_stall; rte; ) {
1648 struct rtdetq *n = rte->next;
1649
1650 m_freem(rte->m);
1651 free(rte, M_MRTABLE);
1652 rte = n;
1653 }
1654 ++mrtstat.mrts_cache_cleanups;
1655 nexpire[i]--;
1656
1657 /*
1658 * free the bw_meter entries
1659 */
1660 while (mfc->mfc_bw_meter != NULL) {
1661 struct bw_meter *x = mfc->mfc_bw_meter;
1662
1663 mfc->mfc_bw_meter = x->bm_mfc_next;
1664 free(x, M_BWMETER);
1665 }
1666
1667 *nptr = mfc->mfc_next;
1668 free(mfc, M_MRTABLE);
1669 } else {
1670 nptr = &mfc->mfc_next;
1671 }
1672 }
1673 }
1674 MFC_UNLOCK();
1675
1676 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
1677 }
1678
1679 /*
1680 * Packet forwarding routine once entry in the cache is made
1681 */
1682 static int
1683 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1684 {
1685 struct ip *ip = mtod(m, struct ip *);
1686 vifi_t vifi;
1687 int plen = ip->ip_len;
1688
1689 VIF_LOCK_ASSERT();
1690 /*
1691 * Macro to send packet on vif. Since RSVP packets don't get counted on
1692 * input, they shouldn't get counted on output, so statistics keeping is
1693 * separate.
1694 */
1695 #define MC_SEND(ip,vifp,m) { \
1696 if ((vifp)->v_flags & VIFF_TUNNEL) \
1697 encap_send((ip), (vifp), (m)); \
1698 else \
1699 phyint_send((ip), (vifp), (m)); \
1700 }
1701
1702 /*
1703 * If xmt_vif is not -1, send on only the requested vif.
1704 *
1705 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1706 */
1707 if (xmt_vif < numvifs) {
1708 #ifdef PIM
1709 if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1710 pim_register_send(ip, viftable + xmt_vif, m, rt);
1711 else
1712 #endif
1713 MC_SEND(ip, viftable + xmt_vif, m);
1714 return 1;
1715 }
1716
1717 /*
1718 * Don't forward if it didn't arrive from the parent vif for its origin.
1719 */
1720 vifi = rt->mfc_parent;
1721 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1722 /* came in the wrong interface */
1723 if (mrtdebug & DEBUG_FORWARD)
1724 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1725 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp);
1726 ++mrtstat.mrts_wrong_if;
1727 ++rt->mfc_wrong_if;
1728 /*
1729 * If we are doing PIM assert processing, send a message
1730 * to the routing daemon.
1731 *
1732 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1733 * can complete the SPT switch, regardless of the type
1734 * of the iif (broadcast media, GRE tunnel, etc).
1735 */
1736 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1737 struct timeval now;
1738 u_long delta;
1739
1740 #ifdef PIM
1741 if (ifp == &multicast_register_if)
1742 pimstat.pims_rcv_registers_wrongiif++;
1743 #endif
1744
1745 /* Get vifi for the incoming packet */
1746 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
1747 ;
1748 if (vifi >= numvifs)
1749 return 0; /* The iif is not found: ignore the packet. */
1750
1751 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1752 return 0; /* WRONGVIF disabled: ignore the packet */
1753
1754 GET_TIME(now);
1755
1756 TV_DELTA(rt->mfc_last_assert, now, delta);
1757
1758 if (delta > ASSERT_MSG_TIME) {
1759 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1760 struct igmpmsg *im;
1761 int hlen = ip->ip_hl << 2;
1762 struct mbuf *mm = m_copy(m, 0, hlen);
1763
1764 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1765 mm = m_pullup(mm, hlen);
1766 if (mm == NULL)
1767 return ENOBUFS;
1768
1769 rt->mfc_last_assert = now;
1770
1771 im = mtod(mm, struct igmpmsg *);
1772 im->im_msgtype = IGMPMSG_WRONGVIF;
1773 im->im_mbz = 0;
1774 im->im_vif = vifi;
1775
1776 mrtstat.mrts_upcalls++;
1777
1778 k_igmpsrc.sin_addr = im->im_src;
1779 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1780 log(LOG_WARNING,
1781 "ip_mforward: ip_mrouter socket queue full\n");
1782 ++mrtstat.mrts_upq_sockfull;
1783 return ENOBUFS;
1784 }
1785 }
1786 }
1787 return 0;
1788 }
1789
1790 /* If I sourced this packet, it counts as output, else it was input. */
1791 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1792 viftable[vifi].v_pkt_out++;
1793 viftable[vifi].v_bytes_out += plen;
1794 } else {
1795 viftable[vifi].v_pkt_in++;
1796 viftable[vifi].v_bytes_in += plen;
1797 }
1798 rt->mfc_pkt_cnt++;
1799 rt->mfc_byte_cnt += plen;
1800
1801 /*
1802 * For each vif, decide if a copy of the packet should be forwarded.
1803 * Forward if:
1804 * - the ttl exceeds the vif's threshold
1805 * - there are group members downstream on interface
1806 */
1807 for (vifi = 0; vifi < numvifs; vifi++)
1808 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1809 viftable[vifi].v_pkt_out++;
1810 viftable[vifi].v_bytes_out += plen;
1811 #ifdef PIM
1812 if (viftable[vifi].v_flags & VIFF_REGISTER)
1813 pim_register_send(ip, viftable + vifi, m, rt);
1814 else
1815 #endif
1816 MC_SEND(ip, viftable+vifi, m);
1817 }
1818
1819 /*
1820 * Perform upcall-related bw measuring.
1821 */
1822 if (rt->mfc_bw_meter != NULL) {
1823 struct bw_meter *x;
1824 struct timeval now;
1825
1826 GET_TIME(now);
1827 MFC_LOCK_ASSERT();
1828 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1829 bw_meter_receive_packet(x, plen, &now);
1830 }
1831
1832 return 0;
1833 }
1834
1835 /*
1836 * check if a vif number is legal/ok. This is used by ip_output.
1837 */
1838 static int
1839 X_legal_vif_num(int vif)
1840 {
1841 /* XXX unlocked, matter? */
1842 return (vif >= 0 && vif < numvifs);
1843 }
1844
1845 /*
1846 * Return the local address used by this vif
1847 */
1848 static u_long
1849 X_ip_mcast_src(int vifi)
1850 {
1851 /* XXX unlocked, matter? */
1852 if (vifi >= 0 && vifi < numvifs)
1853 return viftable[vifi].v_lcl_addr.s_addr;
1854 else
1855 return INADDR_ANY;
1856 }
1857
1858 static void
1859 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1860 {
1861 struct mbuf *mb_copy;
1862 int hlen = ip->ip_hl << 2;
1863
1864 VIF_LOCK_ASSERT();
1865
1866 /*
1867 * Make a new reference to the packet; make sure that
1868 * the IP header is actually copied, not just referenced,
1869 * so that ip_output() only scribbles on the copy.
1870 */
1871 mb_copy = m_copypacket(m, M_DONTWAIT);
1872 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1873 mb_copy = m_pullup(mb_copy, hlen);
1874 if (mb_copy == NULL)
1875 return;
1876
1877 if (vifp->v_rate_limit == 0)
1878 tbf_send_packet(vifp, mb_copy);
1879 else
1880 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1881 }
1882
1883 static void
1884 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1885 {
1886 struct mbuf *mb_copy;
1887 struct ip *ip_copy;
1888 int i, len = ip->ip_len;
1889
1890 VIF_LOCK_ASSERT();
1891
1892 /* Take care of delayed checksums */
1893 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1894 in_delayed_cksum(m);
1895 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1896 }
1897
1898 /*
1899 * copy the old packet & pullup its IP header into the
1900 * new mbuf so we can modify it. Try to fill the new
1901 * mbuf since if we don't the ethernet driver will.
1902 */
1903 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER);
1904 if (mb_copy == NULL)
1905 return;
1906 #ifdef MAC
1907 mac_create_mbuf_multicast_encap(m, vifp->v_ifp, mb_copy);
1908 #endif
1909 mb_copy->m_data += max_linkhdr;
1910 mb_copy->m_len = sizeof(multicast_encap_iphdr);
1911
1912 if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) {
1913 m_freem(mb_copy);
1914 return;
1915 }
1916 i = MHLEN - M_LEADINGSPACE(mb_copy);
1917 if (i > len)
1918 i = len;
1919 mb_copy = m_pullup(mb_copy, i);
1920 if (mb_copy == NULL)
1921 return;
1922 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1923
1924 /*
1925 * fill in the encapsulating IP header.
1926 */
1927 ip_copy = mtod(mb_copy, struct ip *);
1928 *ip_copy = multicast_encap_iphdr;
1929 ip_copy->ip_id = ip_newid();
1930 ip_copy->ip_len += len;
1931 ip_copy->ip_src = vifp->v_lcl_addr;
1932 ip_copy->ip_dst = vifp->v_rmt_addr;
1933
1934 /*
1935 * turn the encapsulated IP header back into a valid one.
1936 */
1937 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1938 --ip->ip_ttl;
1939 ip->ip_len = htons(ip->ip_len);
1940 ip->ip_off = htons(ip->ip_off);
1941 ip->ip_sum = 0;
1942 mb_copy->m_data += sizeof(multicast_encap_iphdr);
1943 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1944 mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1945
1946 if (vifp->v_rate_limit == 0)
1947 tbf_send_packet(vifp, mb_copy);
1948 else
1949 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1950 }
1951
1952 /*
1953 * Token bucket filter module
1954 */
1955
1956 static void
1957 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_long p_len)
1958 {
1959 struct tbf *t = vifp->v_tbf;
1960
1961 VIF_LOCK_ASSERT();
1962
1963 if (p_len > MAX_BKT_SIZE) { /* drop if packet is too large */
1964 mrtstat.mrts_pkt2large++;
1965 m_freem(m);
1966 return;
1967 }
1968
1969 tbf_update_tokens(vifp);
1970
1971 if (t->tbf_q_len == 0) { /* queue empty... */
1972 if (p_len <= t->tbf_n_tok) { /* send packet if enough tokens */
1973 t->tbf_n_tok -= p_len;
1974 tbf_send_packet(vifp, m);
1975 } else { /* no, queue packet and try later */
1976 tbf_queue(vifp, m);
1977 callout_reset(&tbf_reprocess_ch, TBF_REPROCESS,
1978 tbf_reprocess_q, vifp);
1979 }
1980 } else if (t->tbf_q_len < t->tbf_max_q_len) {
1981 /* finite queue length, so queue pkts and process queue */
1982 tbf_queue(vifp, m);
1983 tbf_process_q(vifp);
1984 } else {
1985 /* queue full, try to dq and queue and process */
1986 if (!tbf_dq_sel(vifp, ip)) {
1987 mrtstat.mrts_q_overflow++;
1988 m_freem(m);
1989 } else {
1990 tbf_queue(vifp, m);
1991 tbf_process_q(vifp);
1992 }
1993 }
1994 }
1995
1996 /*
1997 * adds a packet to the queue at the interface
1998 */
1999 static void
2000 tbf_queue(struct vif *vifp, struct mbuf *m)
2001 {
2002 struct tbf *t = vifp->v_tbf;
2003
2004 VIF_LOCK_ASSERT();
2005
2006 if (t->tbf_t == NULL) /* Queue was empty */
2007 t->tbf_q = m;
2008 else /* Insert at tail */
2009 t->tbf_t->m_act = m;
2010
2011 t->tbf_t = m; /* Set new tail pointer */
2012
2013 #ifdef DIAGNOSTIC
2014 /* Make sure we didn't get fed a bogus mbuf */
2015 if (m->m_act)
2016 panic("tbf_queue: m_act");
2017 #endif
2018 m->m_act = NULL;
2019
2020 t->tbf_q_len++;
2021 }
2022
2023 /*
2024 * processes the queue at the interface
2025 */
2026 static void
2027 tbf_process_q(struct vif *vifp)
2028 {
2029 struct tbf *t = vifp->v_tbf;
2030
2031 VIF_LOCK_ASSERT();
2032
2033 /* loop through the queue at the interface and send as many packets
2034 * as possible
2035 */
2036 while (t->tbf_q_len > 0) {
2037 struct mbuf *m = t->tbf_q;
2038 int len = mtod(m, struct ip *)->ip_len;
2039
2040 /* determine if the packet can be sent */
2041 if (len > t->tbf_n_tok) /* not enough tokens, we are done */
2042 break;
2043 /* ok, reduce no of tokens, dequeue and send the packet. */
2044 t->tbf_n_tok -= len;
2045
2046 t->tbf_q = m->m_act;
2047 if (--t->tbf_q_len == 0)
2048 t->tbf_t = NULL;
2049
2050 m->m_act = NULL;
2051 tbf_send_packet(vifp, m);
2052 }
2053 }
2054
2055 static void
2056 tbf_reprocess_q(void *xvifp)
2057 {
2058 struct vif *vifp = xvifp;
2059
2060 if (ip_mrouter == NULL)
2061 return;
2062 VIF_LOCK();
2063 tbf_update_tokens(vifp);
2064 tbf_process_q(vifp);
2065 if (vifp->v_tbf->tbf_q_len)
2066 callout_reset(&tbf_reprocess_ch, TBF_REPROCESS, tbf_reprocess_q, vifp);
2067 VIF_UNLOCK();
2068 }
2069
2070 /* function that will selectively discard a member of the queue
2071 * based on the precedence value and the priority
2072 */
2073 static int
2074 tbf_dq_sel(struct vif *vifp, struct ip *ip)
2075 {
2076 u_int p;
2077 struct mbuf *m, *last;
2078 struct mbuf **np;
2079 struct tbf *t = vifp->v_tbf;
2080
2081 VIF_LOCK_ASSERT();
2082
2083 p = priority(vifp, ip);
2084
2085 np = &t->tbf_q;
2086 last = NULL;
2087 while ((m = *np) != NULL) {
2088 if (p > priority(vifp, mtod(m, struct ip *))) {
2089 *np = m->m_act;
2090 /* If we're removing the last packet, fix the tail pointer */
2091 if (m == t->tbf_t)
2092 t->tbf_t = last;
2093 m_freem(m);
2094 /* It's impossible for the queue to be empty, but check anyways. */
2095 if (--t->tbf_q_len == 0)
2096 t->tbf_t = NULL;
2097 mrtstat.mrts_drop_sel++;
2098 return 1;
2099 }
2100 np = &m->m_act;
2101 last = m;
2102 }
2103 return 0;
2104 }
2105
2106 static void
2107 tbf_send_packet(struct vif *vifp, struct mbuf *m)
2108 {
2109 VIF_LOCK_ASSERT();
2110
2111 if (vifp->v_flags & VIFF_TUNNEL) /* If tunnel options */
2112 ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL);
2113 else {
2114 struct ip_moptions imo;
2115 int error;
2116 static struct route ro; /* XXX check this */
2117
2118 imo.imo_multicast_ifp = vifp->v_ifp;
2119 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2120 imo.imo_multicast_loop = 1;
2121 imo.imo_multicast_vif = -1;
2122
2123 /*
2124 * Re-entrancy should not be a problem here, because
2125 * the packets that we send out and are looped back at us
2126 * should get rejected because they appear to come from
2127 * the loopback interface, thus preventing looping.
2128 */
2129 error = ip_output(m, NULL, &ro, IP_FORWARDING, &imo, NULL);
2130
2131 if (mrtdebug & DEBUG_XMIT)
2132 log(LOG_DEBUG, "phyint_send on vif %d err %d\n",
2133 (int)(vifp - viftable), error);
2134 }
2135 }
2136
2137 /* determine the current time and then
2138 * the elapsed time (between the last time and time now)
2139 * in milliseconds & update the no. of tokens in the bucket
2140 */
2141 static void
2142 tbf_update_tokens(struct vif *vifp)
2143 {
2144 struct timeval tp;
2145 u_long tm;
2146 struct tbf *t = vifp->v_tbf;
2147
2148 VIF_LOCK_ASSERT();
2149
2150 GET_TIME(tp);
2151
2152 TV_DELTA(tp, t->tbf_last_pkt_t, tm);
2153
2154 /*
2155 * This formula is actually
2156 * "time in seconds" * "bytes/second".
2157 *
2158 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2159 *
2160 * The (1000/1024) was introduced in add_vif to optimize
2161 * this divide into a shift.
2162 */
2163 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8;
2164 t->tbf_last_pkt_t = tp;
2165
2166 if (t->tbf_n_tok > MAX_BKT_SIZE)
2167 t->tbf_n_tok = MAX_BKT_SIZE;
2168 }
2169
2170 static int
2171 priority(struct vif *vifp, struct ip *ip)
2172 {
2173 int prio = 50; /* the lowest priority -- default case */
2174
2175 /* temporary hack; may add general packet classifier some day */
2176
2177 /*
2178 * The UDP port space is divided up into four priority ranges:
2179 * [0, 16384) : unclassified - lowest priority
2180 * [16384, 32768) : audio - highest priority
2181 * [32768, 49152) : whiteboard - medium priority
2182 * [49152, 65536) : video - low priority
2183 *
2184 * Everything else gets lowest priority.
2185 */
2186 if (ip->ip_p == IPPROTO_UDP) {
2187 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2188 switch (ntohs(udp->uh_dport) & 0xc000) {
2189 case 0x4000:
2190 prio = 70;
2191 break;
2192 case 0x8000:
2193 prio = 60;
2194 break;
2195 case 0xc000:
2196 prio = 55;
2197 break;
2198 }
2199 }
2200 return prio;
2201 }
2202
2203 /*
2204 * End of token bucket filter modifications
2205 */
2206
2207 static int
2208 X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt)
2209 {
2210 int error, vifi;
2211
2212 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2213 return EOPNOTSUPP;
2214
2215 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
2216 if (error)
2217 return error;
2218
2219 VIF_LOCK();
2220
2221 if (vifi < 0 || vifi >= numvifs) { /* Error if vif is invalid */
2222 VIF_UNLOCK();
2223 return EADDRNOTAVAIL;
2224 }
2225
2226 if (sopt->sopt_name == IP_RSVP_VIF_ON) {
2227 /* Check if socket is available. */
2228 if (viftable[vifi].v_rsvpd != NULL) {
2229 VIF_UNLOCK();
2230 return EADDRINUSE;
2231 }
2232
2233 viftable[vifi].v_rsvpd = so;
2234 /* This may seem silly, but we need to be sure we don't over-increment
2235 * the RSVP counter, in case something slips up.
2236 */
2237 if (!viftable[vifi].v_rsvp_on) {
2238 viftable[vifi].v_rsvp_on = 1;
2239 rsvp_on++;
2240 }
2241 } else { /* must be VIF_OFF */
2242 /*
2243 * XXX as an additional consistency check, one could make sure
2244 * that viftable[vifi].v_rsvpd == so, otherwise passing so as
2245 * first parameter is pretty useless.
2246 */
2247 viftable[vifi].v_rsvpd = NULL;
2248 /*
2249 * This may seem silly, but we need to be sure we don't over-decrement
2250 * the RSVP counter, in case something slips up.
2251 */
2252 if (viftable[vifi].v_rsvp_on) {
2253 viftable[vifi].v_rsvp_on = 0;
2254 rsvp_on--;
2255 }
2256 }
2257 VIF_UNLOCK();
2258 return 0;
2259 }
2260
2261 static void
2262 X_ip_rsvp_force_done(struct socket *so)
2263 {
2264 int vifi;
2265
2266 /* Don't bother if it is not the right type of socket. */
2267 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2268 return;
2269
2270 VIF_LOCK();
2271
2272 /* The socket may be attached to more than one vif...this
2273 * is perfectly legal.
2274 */
2275 for (vifi = 0; vifi < numvifs; vifi++) {
2276 if (viftable[vifi].v_rsvpd == so) {
2277 viftable[vifi].v_rsvpd = NULL;
2278 /* This may seem silly, but we need to be sure we don't
2279 * over-decrement the RSVP counter, in case something slips up.
2280 */
2281 if (viftable[vifi].v_rsvp_on) {
2282 viftable[vifi].v_rsvp_on = 0;
2283 rsvp_on--;
2284 }
2285 }
2286 }
2287
2288 VIF_UNLOCK();
2289 }
2290
2291 static void
2292 X_rsvp_input(struct mbuf *m, int off)
2293 {
2294 int vifi;
2295 struct ip *ip = mtod(m, struct ip *);
2296 struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET };
2297 struct ifnet *ifp;
2298
2299 if (rsvpdebug)
2300 printf("rsvp_input: rsvp_on %d\n",rsvp_on);
2301
2302 /* Can still get packets with rsvp_on = 0 if there is a local member
2303 * of the group to which the RSVP packet is addressed. But in this
2304 * case we want to throw the packet away.
2305 */
2306 if (!rsvp_on) {
2307 m_freem(m);
2308 return;
2309 }
2310
2311 if (rsvpdebug)
2312 printf("rsvp_input: check vifs\n");
2313
2314 #ifdef DIAGNOSTIC
2315 M_ASSERTPKTHDR(m);
2316 #endif
2317
2318 ifp = m->m_pkthdr.rcvif;
2319
2320 VIF_LOCK();
2321 /* Find which vif the packet arrived on. */
2322 for (vifi = 0; vifi < numvifs; vifi++)
2323 if (viftable[vifi].v_ifp == ifp)
2324 break;
2325
2326 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) {
2327 /*
2328 * Drop the lock here to avoid holding it across rip_input.
2329 * This could make rsvpdebug printfs wrong. If you care,
2330 * record the state of stuff before dropping the lock.
2331 */
2332 VIF_UNLOCK();
2333 /*
2334 * If the old-style non-vif-associated socket is set,
2335 * then use it. Otherwise, drop packet since there
2336 * is no specific socket for this vif.
2337 */
2338 if (ip_rsvpd != NULL) {
2339 if (rsvpdebug)
2340 printf("rsvp_input: Sending packet up old-style socket\n");
2341 rip_input(m, off); /* xxx */
2342 } else {
2343 if (rsvpdebug && vifi == numvifs)
2344 printf("rsvp_input: Can't find vif for packet.\n");
2345 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL)
2346 printf("rsvp_input: No socket defined for vif %d\n",vifi);
2347 m_freem(m);
2348 }
2349 return;
2350 }
2351 rsvp_src.sin_addr = ip->ip_src;
2352
2353 if (rsvpdebug && m)
2354 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n",
2355 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv)));
2356
2357 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) {
2358 if (rsvpdebug)
2359 printf("rsvp_input: Failed to append to socket\n");
2360 } else {
2361 if (rsvpdebug)
2362 printf("rsvp_input: send packet up\n");
2363 }
2364 VIF_UNLOCK();
2365 }
2366
2367 /*
2368 * Code for bandwidth monitors
2369 */
2370
2371 /*
2372 * Define common interface for timeval-related methods
2373 */
2374 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
2375 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
2376 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
2377
2378 static uint32_t
2379 compute_bw_meter_flags(struct bw_upcall *req)
2380 {
2381 uint32_t flags = 0;
2382
2383 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2384 flags |= BW_METER_UNIT_PACKETS;
2385 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2386 flags |= BW_METER_UNIT_BYTES;
2387 if (req->bu_flags & BW_UPCALL_GEQ)
2388 flags |= BW_METER_GEQ;
2389 if (req->bu_flags & BW_UPCALL_LEQ)
2390 flags |= BW_METER_LEQ;
2391
2392 return flags;
2393 }
2394
2395 /*
2396 * Add a bw_meter entry
2397 */
2398 static int
2399 add_bw_upcall(struct bw_upcall *req)
2400 {
2401 struct mfc *mfc;
2402 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2403 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2404 struct timeval now;
2405 struct bw_meter *x;
2406 uint32_t flags;
2407
2408 if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2409 return EOPNOTSUPP;
2410
2411 /* Test if the flags are valid */
2412 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2413 return EINVAL;
2414 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2415 return EINVAL;
2416 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2417 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2418 return EINVAL;
2419
2420 /* Test if the threshold time interval is valid */
2421 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2422 return EINVAL;
2423
2424 flags = compute_bw_meter_flags(req);
2425
2426 /*
2427 * Find if we have already same bw_meter entry
2428 */
2429 MFC_LOCK();
2430 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr);
2431 if (mfc == NULL) {
2432 MFC_UNLOCK();
2433 return EADDRNOTAVAIL;
2434 }
2435 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2436 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2437 &req->bu_threshold.b_time, ==)) &&
2438 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2439 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2440 (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
2441 MFC_UNLOCK();
2442 return 0; /* XXX Already installed */
2443 }
2444 }
2445
2446 /* Allocate the new bw_meter entry */
2447 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
2448 if (x == NULL) {
2449 MFC_UNLOCK();
2450 return ENOBUFS;
2451 }
2452
2453 /* Set the new bw_meter entry */
2454 x->bm_threshold.b_time = req->bu_threshold.b_time;
2455 GET_TIME(now);
2456 x->bm_start_time = now;
2457 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2458 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2459 x->bm_measured.b_packets = 0;
2460 x->bm_measured.b_bytes = 0;
2461 x->bm_flags = flags;
2462 x->bm_time_next = NULL;
2463 x->bm_time_hash = BW_METER_BUCKETS;
2464
2465 /* Add the new bw_meter entry to the front of entries for this MFC */
2466 x->bm_mfc = mfc;
2467 x->bm_mfc_next = mfc->mfc_bw_meter;
2468 mfc->mfc_bw_meter = x;
2469 schedule_bw_meter(x, &now);
2470 MFC_UNLOCK();
2471
2472 return 0;
2473 }
2474
2475 static void
2476 free_bw_list(struct bw_meter *list)
2477 {
2478 while (list != NULL) {
2479 struct bw_meter *x = list;
2480
2481 list = list->bm_mfc_next;
2482 unschedule_bw_meter(x);
2483 free(x, M_BWMETER);
2484 }
2485 }
2486
2487 /*
2488 * Delete one or multiple bw_meter entries
2489 */
2490 static int
2491 del_bw_upcall(struct bw_upcall *req)
2492 {
2493 struct mfc *mfc;
2494 struct bw_meter *x;
2495
2496 if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2497 return EOPNOTSUPP;
2498
2499 MFC_LOCK();
2500 /* Find the corresponding MFC entry */
2501 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr);
2502 if (mfc == NULL) {
2503 MFC_UNLOCK();
2504 return EADDRNOTAVAIL;
2505 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2506 /*
2507 * Delete all bw_meter entries for this mfc
2508 */
2509 struct bw_meter *list;
2510
2511 list = mfc->mfc_bw_meter;
2512 mfc->mfc_bw_meter = NULL;
2513 free_bw_list(list);
2514 MFC_UNLOCK();
2515 return 0;
2516 } else { /* Delete a single bw_meter entry */
2517 struct bw_meter *prev;
2518 uint32_t flags = 0;
2519
2520 flags = compute_bw_meter_flags(req);
2521
2522 /* Find the bw_meter entry to delete */
2523 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2524 prev = x, x = x->bm_mfc_next) {
2525 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2526 &req->bu_threshold.b_time, ==)) &&
2527 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2528 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2529 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2530 break;
2531 }
2532 if (x != NULL) { /* Delete entry from the list for this MFC */
2533 if (prev != NULL)
2534 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
2535 else
2536 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2537
2538 unschedule_bw_meter(x);
2539 MFC_UNLOCK();
2540 /* Free the bw_meter entry */
2541 free(x, M_BWMETER);
2542 return 0;
2543 } else {
2544 MFC_UNLOCK();
2545 return EINVAL;
2546 }
2547 }
2548 /* NOTREACHED */
2549 }
2550
2551 /*
2552 * Perform bandwidth measurement processing that may result in an upcall
2553 */
2554 static void
2555 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2556 {
2557 struct timeval delta;
2558
2559 MFC_LOCK_ASSERT();
2560
2561 delta = *nowp;
2562 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2563
2564 if (x->bm_flags & BW_METER_GEQ) {
2565 /*
2566 * Processing for ">=" type of bw_meter entry
2567 */
2568 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2569 /* Reset the bw_meter entry */
2570 x->bm_start_time = *nowp;
2571 x->bm_measured.b_packets = 0;
2572 x->bm_measured.b_bytes = 0;
2573 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2574 }
2575
2576 /* Record that a packet is received */
2577 x->bm_measured.b_packets++;
2578 x->bm_measured.b_bytes += plen;
2579
2580 /*
2581 * Test if we should deliver an upcall
2582 */
2583 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2584 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2585 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2586 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2587 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2588 /* Prepare an upcall for delivery */
2589 bw_meter_prepare_upcall(x, nowp);
2590 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2591 }
2592 }
2593 } else if (x->bm_flags & BW_METER_LEQ) {
2594 /*
2595 * Processing for "<=" type of bw_meter entry
2596 */
2597 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2598 /*
2599 * We are behind time with the multicast forwarding table
2600 * scanning for "<=" type of bw_meter entries, so test now
2601 * if we should deliver an upcall.
2602 */
2603 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2604 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2605 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2606 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2607 /* Prepare an upcall for delivery */
2608 bw_meter_prepare_upcall(x, nowp);
2609 }
2610 /* Reschedule the bw_meter entry */
2611 unschedule_bw_meter(x);
2612 schedule_bw_meter(x, nowp);
2613 }
2614
2615 /* Record that a packet is received */
2616 x->bm_measured.b_packets++;
2617 x->bm_measured.b_bytes += plen;
2618
2619 /*
2620 * Test if we should restart the measuring interval
2621 */
2622 if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2623 x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2624 (x->bm_flags & BW_METER_UNIT_BYTES &&
2625 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2626 /* Don't restart the measuring interval */
2627 } else {
2628 /* Do restart the measuring interval */
2629 /*
2630 * XXX: note that we don't unschedule and schedule, because this
2631 * might be too much overhead per packet. Instead, when we process
2632 * all entries for a given timer hash bin, we check whether it is
2633 * really a timeout. If not, we reschedule at that time.
2634 */
2635 x->bm_start_time = *nowp;
2636 x->bm_measured.b_packets = 0;
2637 x->bm_measured.b_bytes = 0;
2638 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2639 }
2640 }
2641 }
2642
2643 /*
2644 * Prepare a bandwidth-related upcall
2645 */
2646 static void
2647 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2648 {
2649 struct timeval delta;
2650 struct bw_upcall *u;
2651
2652 MFC_LOCK_ASSERT();
2653
2654 /*
2655 * Compute the measured time interval
2656 */
2657 delta = *nowp;
2658 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2659
2660 /*
2661 * If there are too many pending upcalls, deliver them now
2662 */
2663 if (bw_upcalls_n >= BW_UPCALLS_MAX)
2664 bw_upcalls_send();
2665
2666 /*
2667 * Set the bw_upcall entry
2668 */
2669 u = &bw_upcalls[bw_upcalls_n++];
2670 u->bu_src = x->bm_mfc->mfc_origin;
2671 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2672 u->bu_threshold.b_time = x->bm_threshold.b_time;
2673 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2674 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2675 u->bu_measured.b_time = delta;
2676 u->bu_measured.b_packets = x->bm_measured.b_packets;
2677 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2678 u->bu_flags = 0;
2679 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2680 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2681 if (x->bm_flags & BW_METER_UNIT_BYTES)
2682 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2683 if (x->bm_flags & BW_METER_GEQ)
2684 u->bu_flags |= BW_UPCALL_GEQ;
2685 if (x->bm_flags & BW_METER_LEQ)
2686 u->bu_flags |= BW_UPCALL_LEQ;
2687 }
2688
2689 /*
2690 * Send the pending bandwidth-related upcalls
2691 */
2692 static void
2693 bw_upcalls_send(void)
2694 {
2695 struct mbuf *m;
2696 int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2697 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2698 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2699 0, /* unused2 */
2700 IGMPMSG_BW_UPCALL,/* im_msgtype */
2701 0, /* im_mbz */
2702 0, /* im_vif */
2703 0, /* unused3 */
2704 { 0 }, /* im_src */
2705 { 0 } }; /* im_dst */
2706
2707 MFC_LOCK_ASSERT();
2708
2709 if (bw_upcalls_n == 0)
2710 return; /* No pending upcalls */
2711
2712 bw_upcalls_n = 0;
2713
2714 /*
2715 * Allocate a new mbuf, initialize it with the header and
2716 * the payload for the pending calls.
2717 */
2718 MGETHDR(m, M_DONTWAIT, MT_HEADER);
2719 if (m == NULL) {
2720 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2721 return;
2722 }
2723
2724 m->m_len = m->m_pkthdr.len = 0;
2725 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2726 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]);
2727
2728 /*
2729 * Send the upcalls
2730 * XXX do we need to set the address in k_igmpsrc ?
2731 */
2732 mrtstat.mrts_upcalls++;
2733 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2734 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2735 ++mrtstat.mrts_upq_sockfull;
2736 }
2737 }
2738
2739 /*
2740 * Compute the timeout hash value for the bw_meter entries
2741 */
2742 #define BW_METER_TIMEHASH(bw_meter, hash) \
2743 do { \
2744 struct timeval next_timeval = (bw_meter)->bm_start_time; \
2745 \
2746 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2747 (hash) = next_timeval.tv_sec; \
2748 if (next_timeval.tv_usec) \
2749 (hash)++; /* XXX: make sure we don't timeout early */ \
2750 (hash) %= BW_METER_BUCKETS; \
2751 } while (0)
2752
2753 /*
2754 * Schedule a timer to process periodically bw_meter entry of type "<="
2755 * by linking the entry in the proper hash bucket.
2756 */
2757 static void
2758 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2759 {
2760 int time_hash;
2761
2762 MFC_LOCK_ASSERT();
2763
2764 if (!(x->bm_flags & BW_METER_LEQ))
2765 return; /* XXX: we schedule timers only for "<=" entries */
2766
2767 /*
2768 * Reset the bw_meter entry
2769 */
2770 x->bm_start_time = *nowp;
2771 x->bm_measured.b_packets = 0;
2772 x->bm_measured.b_bytes = 0;
2773 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2774
2775 /*
2776 * Compute the timeout hash value and insert the entry
2777 */
2778 BW_METER_TIMEHASH(x, time_hash);
2779 x->bm_time_next = bw_meter_timers[time_hash];
2780 bw_meter_timers[time_hash] = x;
2781 x->bm_time_hash = time_hash;
2782 }
2783
2784 /*
2785 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2786 * by removing the entry from the proper hash bucket.
2787 */
2788 static void
2789 unschedule_bw_meter(struct bw_meter *x)
2790 {
2791 int time_hash;
2792 struct bw_meter *prev, *tmp;
2793
2794 MFC_LOCK_ASSERT();
2795
2796 if (!(x->bm_flags & BW_METER_LEQ))
2797 return; /* XXX: we schedule timers only for "<=" entries */
2798
2799 /*
2800 * Compute the timeout hash value and delete the entry
2801 */
2802 time_hash = x->bm_time_hash;
2803 if (time_hash >= BW_METER_BUCKETS)
2804 return; /* Entry was not scheduled */
2805
2806 for (prev = NULL, tmp = bw_meter_timers[time_hash];
2807 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2808 if (tmp == x)
2809 break;
2810
2811 if (tmp == NULL)
2812 panic("unschedule_bw_meter: bw_meter entry not found");
2813
2814 if (prev != NULL)
2815 prev->bm_time_next = x->bm_time_next;
2816 else
2817 bw_meter_timers[time_hash] = x->bm_time_next;
2818
2819 x->bm_time_next = NULL;
2820 x->bm_time_hash = BW_METER_BUCKETS;
2821 }
2822
2823
2824 /*
2825 * Process all "<=" type of bw_meter that should be processed now,
2826 * and for each entry prepare an upcall if necessary. Each processed
2827 * entry is rescheduled again for the (periodic) processing.
2828 *
2829 * This is run periodically (once per second normally). On each round,
2830 * all the potentially matching entries are in the hash slot that we are
2831 * looking at.
2832 */
2833 static void
2834 bw_meter_process()
2835 {
2836 static uint32_t last_tv_sec; /* last time we processed this */
2837
2838 uint32_t loops;
2839 int i;
2840 struct timeval now, process_endtime;
2841
2842 GET_TIME(now);
2843 if (last_tv_sec == now.tv_sec)
2844 return; /* nothing to do */
2845
2846 loops = now.tv_sec - last_tv_sec;
2847 last_tv_sec = now.tv_sec;
2848 if (loops > BW_METER_BUCKETS)
2849 loops = BW_METER_BUCKETS;
2850
2851 MFC_LOCK();
2852 /*
2853 * Process all bins of bw_meter entries from the one after the last
2854 * processed to the current one. On entry, i points to the last bucket
2855 * visited, so we need to increment i at the beginning of the loop.
2856 */
2857 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2858 struct bw_meter *x, *tmp_list;
2859
2860 if (++i >= BW_METER_BUCKETS)
2861 i = 0;
2862
2863 /* Disconnect the list of bw_meter entries from the bin */
2864 tmp_list = bw_meter_timers[i];
2865 bw_meter_timers[i] = NULL;
2866
2867 /* Process the list of bw_meter entries */
2868 while (tmp_list != NULL) {
2869 x = tmp_list;
2870 tmp_list = tmp_list->bm_time_next;
2871
2872 /* Test if the time interval is over */
2873 process_endtime = x->bm_start_time;
2874 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2875 if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2876 /* Not yet: reschedule, but don't reset */
2877 int time_hash;
2878
2879 BW_METER_TIMEHASH(x, time_hash);
2880 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2881 /*
2882 * XXX: somehow the bin processing is a bit ahead of time.
2883 * Put the entry in the next bin.
2884 */
2885 if (++time_hash >= BW_METER_BUCKETS)
2886 time_hash = 0;
2887 }
2888 x->bm_time_next = bw_meter_timers[time_hash];
2889 bw_meter_timers[time_hash] = x;
2890 x->bm_time_hash = time_hash;
2891
2892 continue;
2893 }
2894
2895 /*
2896 * Test if we should deliver an upcall
2897 */
2898 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2899 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2900 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2901 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2902 /* Prepare an upcall for delivery */
2903 bw_meter_prepare_upcall(x, &now);
2904 }
2905
2906 /*
2907 * Reschedule for next processing
2908 */
2909 schedule_bw_meter(x, &now);
2910 }
2911 }
2912
2913 /* Send all upcalls that are pending delivery */
2914 bw_upcalls_send();
2915
2916 MFC_UNLOCK();
2917 }
2918
2919 /*
2920 * A periodic function for sending all upcalls that are pending delivery
2921 */
2922 static void
2923 expire_bw_upcalls_send(void *unused)
2924 {
2925 MFC_LOCK();
2926 bw_upcalls_send();
2927 MFC_UNLOCK();
2928
2929 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
2930 expire_bw_upcalls_send, NULL);
2931 }
2932
2933 /*
2934 * A periodic function for periodic scanning of the multicast forwarding
2935 * table for processing all "<=" bw_meter entries.
2936 */
2937 static void
2938 expire_bw_meter_process(void *unused)
2939 {
2940 if (mrt_api_config & MRT_MFC_BW_UPCALL)
2941 bw_meter_process();
2942
2943 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
2944 }
2945
2946 /*
2947 * End of bandwidth monitoring code
2948 */
2949
2950 #ifdef PIM
2951 /*
2952 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2953 *
2954 */
2955 static int
2956 pim_register_send(struct ip *ip, struct vif *vifp,
2957 struct mbuf *m, struct mfc *rt)
2958 {
2959 struct mbuf *mb_copy, *mm;
2960
2961 if (mrtdebug & DEBUG_PIM)
2962 log(LOG_DEBUG, "pim_register_send: ");
2963
2964 mb_copy = pim_register_prepare(ip, m);
2965 if (mb_copy == NULL)
2966 return ENOBUFS;
2967
2968 /*
2969 * Send all the fragments. Note that the mbuf for each fragment
2970 * is freed by the sending machinery.
2971 */
2972 for (mm = mb_copy; mm; mm = mb_copy) {
2973 mb_copy = mm->m_nextpkt;
2974 mm->m_nextpkt = 0;
2975 mm = m_pullup(mm, sizeof(struct ip));
2976 if (mm != NULL) {
2977 ip = mtod(mm, struct ip *);
2978 if ((mrt_api_config & MRT_MFC_RP) &&
2979 (rt->mfc_rp.s_addr != INADDR_ANY)) {
2980 pim_register_send_rp(ip, vifp, mm, rt);
2981 } else {
2982 pim_register_send_upcall(ip, vifp, mm, rt);
2983 }
2984 }
2985 }
2986
2987 return 0;
2988 }
2989
2990 /*
2991 * Return a copy of the data packet that is ready for PIM Register
2992 * encapsulation.
2993 * XXX: Note that in the returned copy the IP header is a valid one.
2994 */
2995 static struct mbuf *
2996 pim_register_prepare(struct ip *ip, struct mbuf *m)
2997 {
2998 struct mbuf *mb_copy = NULL;
2999 int mtu;
3000
3001 /* Take care of delayed checksums */
3002 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3003 in_delayed_cksum(m);
3004 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3005 }
3006
3007 /*
3008 * Copy the old packet & pullup its IP header into the
3009 * new mbuf so we can modify it.
3010 */
3011 mb_copy = m_copypacket(m, M_DONTWAIT);
3012 if (mb_copy == NULL)
3013 return NULL;
3014 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3015 if (mb_copy == NULL)
3016 return NULL;
3017
3018 /* take care of the TTL */
3019 ip = mtod(mb_copy, struct ip *);
3020 --ip->ip_ttl;
3021
3022 /* Compute the MTU after the PIM Register encapsulation */
3023 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3024
3025 if (ip->ip_len <= mtu) {
3026 /* Turn the IP header into a valid one */
3027 ip->ip_len = htons(ip->ip_len);
3028 ip->ip_off = htons(ip->ip_off);
3029 ip->ip_sum = 0;
3030 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3031 } else {
3032 /* Fragment the packet */
3033 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) {
3034 m_freem(mb_copy);
3035 return NULL;
3036 }
3037 }
3038 return mb_copy;
3039 }
3040
3041 /*
3042 * Send an upcall with the data packet to the user-level process.
3043 */
3044 static int
3045 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3046 struct mbuf *mb_copy, struct mfc *rt)
3047 {
3048 struct mbuf *mb_first;
3049 int len = ntohs(ip->ip_len);
3050 struct igmpmsg *im;
3051 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
3052
3053 VIF_LOCK_ASSERT();
3054
3055 /*
3056 * Add a new mbuf with an upcall header
3057 */
3058 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3059 if (mb_first == NULL) {
3060 m_freem(mb_copy);
3061 return ENOBUFS;
3062 }
3063 mb_first->m_data += max_linkhdr;
3064 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3065 mb_first->m_len = sizeof(struct igmpmsg);
3066 mb_first->m_next = mb_copy;
3067
3068 /* Send message to routing daemon */
3069 im = mtod(mb_first, struct igmpmsg *);
3070 im->im_msgtype = IGMPMSG_WHOLEPKT;
3071 im->im_mbz = 0;
3072 im->im_vif = vifp - viftable;
3073 im->im_src = ip->ip_src;
3074 im->im_dst = ip->ip_dst;
3075
3076 k_igmpsrc.sin_addr = ip->ip_src;
3077
3078 mrtstat.mrts_upcalls++;
3079
3080 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3081 if (mrtdebug & DEBUG_PIM)
3082 log(LOG_WARNING,
3083 "mcast: pim_register_send_upcall: ip_mrouter socket queue full");
3084 ++mrtstat.mrts_upq_sockfull;
3085 return ENOBUFS;
3086 }
3087
3088 /* Keep statistics */
3089 pimstat.pims_snd_registers_msgs++;
3090 pimstat.pims_snd_registers_bytes += len;
3091
3092 return 0;
3093 }
3094
3095 /*
3096 * Encapsulate the data packet in PIM Register message and send it to the RP.
3097 */
3098 static int
3099 pim_register_send_rp(struct ip *ip, struct vif *vifp,
3100 struct mbuf *mb_copy, struct mfc *rt)
3101 {
3102 struct mbuf *mb_first;
3103 struct ip *ip_outer;
3104 struct pim_encap_pimhdr *pimhdr;
3105 int len = ntohs(ip->ip_len);
3106 vifi_t vifi = rt->mfc_parent;
3107
3108 VIF_LOCK_ASSERT();
3109
3110 if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) {
3111 m_freem(mb_copy);
3112 return EADDRNOTAVAIL; /* The iif vif is invalid */
3113 }
3114
3115 /*
3116 * Add a new mbuf with the encapsulating header
3117 */
3118 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3119 if (mb_first == NULL) {
3120 m_freem(mb_copy);
3121 return ENOBUFS;
3122 }
3123 mb_first->m_data += max_linkhdr;
3124 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3125 mb_first->m_next = mb_copy;
3126
3127 mb_first->m_pkthdr.len = len + mb_first->m_len;
3128
3129 /*
3130 * Fill in the encapsulating IP and PIM header
3131 */
3132 ip_outer = mtod(mb_first, struct ip *);
3133 *ip_outer = pim_encap_iphdr;
3134 ip_outer->ip_id = ip_newid();
3135 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3136 ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3137 ip_outer->ip_dst = rt->mfc_rp;
3138 /*
3139 * Copy the inner header TOS to the outer header, and take care of the
3140 * IP_DF bit.
3141 */
3142 ip_outer->ip_tos = ip->ip_tos;
3143 if (ntohs(ip->ip_off) & IP_DF)
3144 ip_outer->ip_off |= IP_DF;
3145 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
3146 + sizeof(pim_encap_iphdr));
3147 *pimhdr = pim_encap_pimhdr;
3148 /* If the iif crosses a border, set the Border-bit */
3149 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3150 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3151
3152 mb_first->m_data += sizeof(pim_encap_iphdr);
3153 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3154 mb_first->m_data -= sizeof(pim_encap_iphdr);
3155
3156 if (vifp->v_rate_limit == 0)
3157 tbf_send_packet(vifp, mb_first);
3158 else
3159 tbf_control(vifp, mb_first, ip, ip_outer->ip_len);
3160
3161 /* Keep statistics */
3162 pimstat.pims_snd_registers_msgs++;
3163 pimstat.pims_snd_registers_bytes += len;
3164
3165 return 0;
3166 }
3167
3168 /*
3169 * PIM-SMv2 and PIM-DM messages processing.
3170 * Receives and verifies the PIM control messages, and passes them
3171 * up to the listening socket, using rip_input().
3172 * The only message with special processing is the PIM_REGISTER message
3173 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3174 * is passed to if_simloop().
3175 */
3176 void
3177 pim_input(struct mbuf *m, int off)
3178 {
3179 struct ip *ip = mtod(m, struct ip *);
3180 struct pim *pim;
3181 int minlen;
3182 int datalen = ip->ip_len;
3183 int ip_tos;
3184 int iphlen = off;
3185
3186 /* Keep statistics */
3187 pimstat.pims_rcv_total_msgs++;
3188 pimstat.pims_rcv_total_bytes += datalen;
3189
3190 /*
3191 * Validate lengths
3192 */
3193 if (datalen < PIM_MINLEN) {
3194 pimstat.pims_rcv_tooshort++;
3195 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3196 datalen, (u_long)ip->ip_src.s_addr);
3197 m_freem(m);
3198 return;
3199 }
3200
3201 /*
3202 * If the packet is at least as big as a REGISTER, go agead
3203 * and grab the PIM REGISTER header size, to avoid another
3204 * possible m_pullup() later.
3205 *
3206 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
3207 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3208 */
3209 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3210 /*
3211 * Get the IP and PIM headers in contiguous memory, and
3212 * possibly the PIM REGISTER header.
3213 */
3214 if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3215 (m = m_pullup(m, minlen)) == 0) {
3216 log(LOG_ERR, "pim_input: m_pullup failure\n");
3217 return;
3218 }
3219 /* m_pullup() may have given us a new mbuf so reset ip. */
3220 ip = mtod(m, struct ip *);
3221 ip_tos = ip->ip_tos;
3222
3223 /* adjust mbuf to point to the PIM header */
3224 m->m_data += iphlen;
3225 m->m_len -= iphlen;
3226 pim = mtod(m, struct pim *);
3227
3228 /*
3229 * Validate checksum. If PIM REGISTER, exclude the data packet.
3230 *
3231 * XXX: some older PIMv2 implementations don't make this distinction,
3232 * so for compatibility reason perform the checksum over part of the
3233 * message, and if error, then over the whole message.
3234 */
3235 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3236 /* do nothing, checksum okay */
3237 } else if (in_cksum(m, datalen)) {
3238 pimstat.pims_rcv_badsum++;
3239 if (mrtdebug & DEBUG_PIM)
3240 log(LOG_DEBUG, "pim_input: invalid checksum");
3241 m_freem(m);
3242 return;
3243 }
3244
3245 /* PIM version check */
3246 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3247 pimstat.pims_rcv_badversion++;
3248 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3249 PIM_VT_V(pim->pim_vt), PIM_VERSION);
3250 m_freem(m);
3251 return;
3252 }
3253
3254 /* restore mbuf back to the outer IP */
3255 m->m_data -= iphlen;
3256 m->m_len += iphlen;
3257
3258 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3259 /*
3260 * Since this is a REGISTER, we'll make a copy of the register
3261 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3262 * routing daemon.
3263 */
3264 struct sockaddr_in dst = { sizeof(dst), AF_INET };
3265 struct mbuf *mcp;
3266 struct ip *encap_ip;
3267 u_int32_t *reghdr;
3268 struct ifnet *vifp;
3269
3270 VIF_LOCK();
3271 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3272 VIF_UNLOCK();
3273 if (mrtdebug & DEBUG_PIM)
3274 log(LOG_DEBUG,
3275 "pim_input: register vif not set: %d\n", reg_vif_num);
3276 m_freem(m);
3277 return;
3278 }
3279 /* XXX need refcnt? */
3280 vifp = viftable[reg_vif_num].v_ifp;
3281 VIF_UNLOCK();
3282
3283 /*
3284 * Validate length
3285 */
3286 if (datalen < PIM_REG_MINLEN) {
3287 pimstat.pims_rcv_tooshort++;
3288 pimstat.pims_rcv_badregisters++;
3289 log(LOG_ERR,
3290 "pim_input: register packet size too small %d from %lx\n",
3291 datalen, (u_long)ip->ip_src.s_addr);
3292 m_freem(m);
3293 return;
3294 }
3295
3296 reghdr = (u_int32_t *)(pim + 1);
3297 encap_ip = (struct ip *)(reghdr + 1);
3298
3299 if (mrtdebug & DEBUG_PIM) {
3300 log(LOG_DEBUG,
3301 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3302 (u_long)ntohl(encap_ip->ip_src.s_addr),
3303 (u_long)ntohl(encap_ip->ip_dst.s_addr),
3304 ntohs(encap_ip->ip_len));
3305 }
3306
3307 /* verify the version number of the inner packet */
3308 if (encap_ip->ip_v != IPVERSION) {
3309 pimstat.pims_rcv_badregisters++;
3310 if (mrtdebug & DEBUG_PIM) {
3311 log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3312 "of the inner packet\n", encap_ip->ip_v);
3313 }
3314 m_freem(m);
3315 return;
3316 }
3317
3318 /* verify the inner packet is destined to a mcast group */
3319 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
3320 pimstat.pims_rcv_badregisters++;
3321 if (mrtdebug & DEBUG_PIM)
3322 log(LOG_DEBUG,
3323 "pim_input: inner packet of register is not "
3324 "multicast %lx\n",
3325 (u_long)ntohl(encap_ip->ip_dst.s_addr));
3326 m_freem(m);
3327 return;
3328 }
3329
3330 /* If a NULL_REGISTER, pass it to the daemon */
3331 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3332 goto pim_input_to_daemon;
3333
3334 /*
3335 * Copy the TOS from the outer IP header to the inner IP header.
3336 */
3337 if (encap_ip->ip_tos != ip_tos) {
3338 /* Outer TOS -> inner TOS */
3339 encap_ip->ip_tos = ip_tos;
3340 /* Recompute the inner header checksum. Sigh... */
3341
3342 /* adjust mbuf to point to the inner IP header */
3343 m->m_data += (iphlen + PIM_MINLEN);
3344 m->m_len -= (iphlen + PIM_MINLEN);
3345
3346 encap_ip->ip_sum = 0;
3347 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3348
3349 /* restore mbuf to point back to the outer IP header */
3350 m->m_data -= (iphlen + PIM_MINLEN);
3351 m->m_len += (iphlen + PIM_MINLEN);
3352 }
3353
3354 /*
3355 * Decapsulate the inner IP packet and loopback to forward it
3356 * as a normal multicast packet. Also, make a copy of the
3357 * outer_iphdr + pimhdr + reghdr + encap_iphdr
3358 * to pass to the daemon later, so it can take the appropriate
3359 * actions (e.g., send back PIM_REGISTER_STOP).
3360 * XXX: here m->m_data points to the outer IP header.
3361 */
3362 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
3363 if (mcp == NULL) {
3364 log(LOG_ERR,
3365 "pim_input: pim register: could not copy register head\n");
3366 m_freem(m);
3367 return;
3368 }
3369
3370 /* Keep statistics */
3371 /* XXX: registers_bytes include only the encap. mcast pkt */
3372 pimstat.pims_rcv_registers_msgs++;
3373 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3374
3375 /*
3376 * forward the inner ip packet; point m_data at the inner ip.
3377 */
3378 m_adj(m, iphlen + PIM_MINLEN);
3379
3380 if (mrtdebug & DEBUG_PIM) {
3381 log(LOG_DEBUG,
3382 "pim_input: forwarding decapsulated register: "
3383 "src %lx, dst %lx, vif %d\n",
3384 (u_long)ntohl(encap_ip->ip_src.s_addr),
3385 (u_long)ntohl(encap_ip->ip_dst.s_addr),
3386 reg_vif_num);
3387 }
3388 /* NB: vifp was collected above; can it change on us? */
3389 if_simloop(vifp, m, dst.sin_family, 0);
3390
3391 /* prepare the register head to send to the mrouting daemon */
3392 m = mcp;
3393 }
3394
3395 pim_input_to_daemon:
3396 /*
3397 * Pass the PIM message up to the daemon; if it is a Register message,
3398 * pass the 'head' only up to the daemon. This includes the
3399 * outer IP header, PIM header, PIM-Register header and the
3400 * inner IP header.
3401 * XXX: the outer IP header pkt size of a Register is not adjust to
3402 * reflect the fact that the inner multicast data is truncated.
3403 */
3404 rip_input(m, iphlen);
3405
3406 return;
3407 }
3408 #endif /* PIM */
3409
3410 static int
3411 ip_mroute_modevent(module_t mod, int type, void *unused)
3412 {
3413 switch (type) {
3414 case MOD_LOAD:
3415 mtx_init(&mrouter_mtx, "mrouter initialization", NULL, MTX_DEF);
3416 MFC_LOCK_INIT();
3417 VIF_LOCK_INIT();
3418 ip_mrouter_reset();
3419 ip_mcast_src = X_ip_mcast_src;
3420 ip_mforward = X_ip_mforward;
3421 ip_mrouter_done = X_ip_mrouter_done;
3422 ip_mrouter_get = X_ip_mrouter_get;
3423 ip_mrouter_set = X_ip_mrouter_set;
3424 ip_rsvp_force_done = X_ip_rsvp_force_done;
3425 ip_rsvp_vif = X_ip_rsvp_vif;
3426 legal_vif_num = X_legal_vif_num;
3427 mrt_ioctl = X_mrt_ioctl;
3428 rsvp_input_p = X_rsvp_input;
3429 break;
3430
3431 case MOD_UNLOAD:
3432 /*
3433 * Typically module unload happens after the user-level
3434 * process has shutdown the kernel services (the check
3435 * below insures someone can't just yank the module out
3436 * from under a running process). But if the module is
3437 * just loaded and then unloaded w/o starting up a user
3438 * process we still need to cleanup.
3439 */
3440 if (ip_mrouter)
3441 return EINVAL;
3442
3443 X_ip_mrouter_done();
3444 ip_mcast_src = NULL;
3445 ip_mforward = NULL;
3446 ip_mrouter_done = NULL;
3447 ip_mrouter_get = NULL;
3448 ip_mrouter_set = NULL;
3449 ip_rsvp_force_done = NULL;
3450 ip_rsvp_vif = NULL;
3451 legal_vif_num = NULL;
3452 mrt_ioctl = NULL;
3453 rsvp_input_p = NULL;
3454 VIF_LOCK_DESTROY();
3455 MFC_LOCK_DESTROY();
3456 mtx_destroy(&mrouter_mtx);
3457 break;
3458 default:
3459 return EOPNOTSUPP;
3460 }
3461 return 0;
3462 }
3463
3464 static moduledata_t ip_mroutemod = {
3465 "ip_mroute",
3466 ip_mroute_modevent,
3467 0
3468 };
3469 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
Cache object: bbb58439d5125fb503f69040bd995b5f
|