1 /* $NetBSD: ip_mroute.c,v 1.165 2022/03/15 21:39:59 andvar Exp $ */
2
3 /*
4 * Copyright (c) 1992, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Stephen Deering of Stanford University.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
35 */
36
37 /*
38 * Copyright (c) 1989 Stephen Deering
39 *
40 * This code is derived from software contributed to Berkeley by
41 * Stephen Deering of Stanford University.
42 *
43 * Redistribution and use in source and binary forms, with or without
44 * modification, are permitted provided that the following conditions
45 * are met:
46 * 1. Redistributions of source code must retain the above copyright
47 * notice, this list of conditions and the following disclaimer.
48 * 2. Redistributions in binary form must reproduce the above copyright
49 * notice, this list of conditions and the following disclaimer in the
50 * documentation and/or other materials provided with the distribution.
51 * 3. All advertising materials mentioning features or use of this software
52 * must display the following acknowledgement:
53 * This product includes software developed by the University of
54 * California, Berkeley and its contributors.
55 * 4. Neither the name of the University nor the names of its contributors
56 * may be used to endorse or promote products derived from this software
57 * without specific prior written permission.
58 *
59 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
60 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
61 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
62 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
63 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
64 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
65 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
66 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
67 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
68 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
69 * SUCH DAMAGE.
70 *
71 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
72 */
73
74 /*
75 * IP multicast forwarding procedures
76 *
77 * Written by David Waitzman, BBN Labs, August 1988.
78 * Modified by Steve Deering, Stanford, February 1989.
79 * Modified by Mark J. Steiglitz, Stanford, May, 1991
80 * Modified by Van Jacobson, LBL, January 1993
81 * Modified by Ajit Thyagarajan, PARC, August 1993
82 * Modified by Bill Fenner, PARC, April 1994
83 * Modified by Charles M. Hannum, NetBSD, May 1995.
84 * Modified by Ahmed Helmy, SGI, June 1996
85 * Modified by George Edmond Eddy (Rusty), ISI, February 1998
86 * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
87 * Modified by Hitoshi Asaeda, WIDE, August 2000
88 * Modified by Pavlin Radoslavov, ICSI, October 2002
89 *
90 * MROUTING Revision: 1.2
91 * and PIM-SMv2 and PIM-DM support, advanced API support,
92 * bandwidth metering and signaling
93 */
94
95 #include <sys/cdefs.h>
96 __KERNEL_RCSID(0, "$NetBSD: ip_mroute.c,v 1.165 2022/03/15 21:39:59 andvar Exp $");
97
98 #ifdef _KERNEL_OPT
99 #include "opt_inet.h"
100 #include "opt_ipsec.h"
101 #include "opt_pim.h"
102 #endif
103
104 #ifdef PIM
105 #define _PIM_VT 1
106 #endif
107
108 #include <sys/param.h>
109 #include <sys/systm.h>
110 #include <sys/callout.h>
111 #include <sys/mbuf.h>
112 #include <sys/socket.h>
113 #include <sys/socketvar.h>
114 #include <sys/errno.h>
115 #include <sys/time.h>
116 #include <sys/kernel.h>
117 #include <sys/kmem.h>
118 #include <sys/ioctl.h>
119 #include <sys/syslog.h>
120
121 #include <net/if.h>
122 #include <net/raw_cb.h>
123
124 #include <netinet/in.h>
125 #include <netinet/in_var.h>
126 #include <netinet/in_systm.h>
127 #include <netinet/in_offload.h>
128 #include <netinet/ip.h>
129 #include <netinet/ip_var.h>
130 #include <netinet/in_pcb.h>
131 #include <netinet/udp.h>
132 #include <netinet/igmp.h>
133 #include <netinet/igmp_var.h>
134 #include <netinet/ip_mroute.h>
135 #ifdef PIM
136 #include <netinet/pim.h>
137 #include <netinet/pim_var.h>
138 #endif
139 #include <netinet/ip_encap.h>
140
141 #ifdef IPSEC
142 #include <netipsec/ipsec.h>
143 #include <netipsec/key.h>
144 #endif
145
146 #define IP_MULTICASTOPTS 0
147 #define M_PULLUP(m, len) \
148 do { \
149 if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \
150 (m) = m_pullup((m), (len)); \
151 } while (/*CONSTCOND*/ 0)
152
153 /*
154 * Globals. All but ip_mrouter and ip_mrtproto could be static,
155 * except for netstat or debugging purposes.
156 */
157 struct socket *ip_mrouter = NULL;
158 int ip_mrtproto = IGMP_DVMRP; /* for netstat only */
159
160 #define MFCHASH(a, g) \
161 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
162 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
163 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
164 u_long mfchash;
165
166 u_char nexpire[MFCTBLSIZ];
167 struct vif viftable[MAXVIFS];
168 struct mrtstat mrtstat;
169 u_int mrtdebug = 0; /* debug level */
170 #define DEBUG_MFC 0x02
171 #define DEBUG_FORWARD 0x04
172 #define DEBUG_EXPIRE 0x08
173 #define DEBUG_XMIT 0x10
174 #define DEBUG_PIM 0x20
175
176 #define VIFI_INVALID ((vifi_t) -1)
177
178 u_int tbfdebug = 0; /* tbf debug level */
179
180 /* vif attachment using sys/netinet/ip_encap.c */
181 static void vif_input(struct mbuf *, int, int, void *);
182 static int vif_encapcheck(struct mbuf *, int, int, void *);
183
184 static const struct encapsw vif_encapsw = {
185 .encapsw4 = {
186 .pr_input = vif_input,
187 .pr_ctlinput = NULL,
188 }
189 };
190
191 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
192 #define UPCALL_EXPIRE 6 /* number of timeouts */
193
194 /*
195 * Define the token bucket filter structures
196 */
197
198 #define TBF_REPROCESS (hz / 100) /* 100x / second */
199
200 static int get_sg_cnt(struct sioc_sg_req *);
201 static int get_vif_cnt(struct sioc_vif_req *);
202 static int ip_mrouter_init(struct socket *, int);
203 static int set_assert(int);
204 static int add_vif(struct vifctl *);
205 static int del_vif(vifi_t *);
206 static void update_mfc_params(struct mfc *, struct mfcctl2 *);
207 static void init_mfc_params(struct mfc *, struct mfcctl2 *);
208 static void expire_mfc(struct mfc *);
209 static int add_mfc(struct sockopt *);
210 #ifdef UPCALL_TIMING
211 static void collate(struct timeval *);
212 #endif
213 static int del_mfc(struct sockopt *);
214 static int set_api_config(struct sockopt *); /* chose API capabilities */
215 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
216 static void expire_upcalls(void *);
217 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *);
218 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
219 static void encap_send(struct ip *, struct vif *, struct mbuf *);
220 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_int32_t);
221 static void tbf_queue(struct vif *, struct mbuf *);
222 static void tbf_process_q(struct vif *);
223 static void tbf_reprocess_q(void *);
224 static int tbf_dq_sel(struct vif *, struct ip *);
225 static void tbf_send_packet(struct vif *, struct mbuf *);
226 static void tbf_update_tokens(struct vif *);
227 static int priority(struct vif *, struct ip *);
228 static int ip_mforward_real(struct mbuf *, struct ifnet *);
229
230
231 /*
232 * Bandwidth monitoring
233 */
234 static void free_bw_list(struct bw_meter *);
235 static int add_bw_upcall(struct bw_upcall *);
236 static int del_bw_upcall(struct bw_upcall *);
237 static void bw_meter_receive_packet(struct bw_meter *, int , struct timeval *);
238 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
239 static void bw_upcalls_send(void);
240 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
241 static void unschedule_bw_meter(struct bw_meter *);
242 static void bw_meter_process(void);
243 static void expire_bw_upcalls_send(void *);
244 static void expire_bw_meter_process(void *);
245
246 #ifdef PIM
247 static int pim_register_send(struct ip *, struct vif *,
248 struct mbuf *, struct mfc *);
249 static int pim_register_send_rp(struct ip *, struct vif *,
250 struct mbuf *, struct mfc *);
251 static int pim_register_send_upcall(struct ip *, struct vif *,
252 struct mbuf *, struct mfc *);
253 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
254 #endif
255
256 #define ENCAP_TTL 64
257 #define ENCAP_PROTO IPPROTO_IPIP
258
259 /* prototype IP hdr for encapsulated packets */
260 static const struct ip multicast_encap_iphdr = {
261 .ip_hl = sizeof(struct ip) >> 2,
262 .ip_v = IPVERSION,
263 .ip_len = sizeof(struct ip),
264 .ip_ttl = ENCAP_TTL,
265 .ip_p = ENCAP_PROTO,
266 };
267
268 /*
269 * Bandwidth meter variables and constants
270 */
271
272 /*
273 * Pending timeouts are stored in a hash table, the key being the
274 * expiration time. Periodically, the entries are analysed and processed.
275 */
276 #define BW_METER_BUCKETS 1024
277 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
278 struct callout bw_meter_ch;
279 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
280
281 /*
282 * Pending upcalls are stored in a vector which is flushed when
283 * full, or periodically
284 */
285 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX];
286 static u_int bw_upcalls_n; /* # of pending upcalls */
287 struct callout bw_upcalls_ch;
288 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
289
290 #ifdef PIM
291 struct pimstat pimstat;
292
293 /*
294 * Note: the PIM Register encapsulation adds the following in front of a
295 * data packet:
296 *
297 * struct pim_encap_hdr {
298 * struct ip ip;
299 * struct pim_encap_pimhdr pim;
300 * }
301 */
302
303 struct pim_encap_pimhdr {
304 struct pim pim;
305 uint32_t flags;
306 };
307
308 static struct ip pim_encap_iphdr = {
309 .ip_v = IPVERSION,
310 .ip_hl = sizeof(struct ip) >> 2,
311 .ip_len = sizeof(struct ip),
312 .ip_ttl = ENCAP_TTL,
313 .ip_p = IPPROTO_PIM,
314 };
315
316 static struct pim_encap_pimhdr pim_encap_pimhdr = {
317 {
318 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
319 0, /* reserved */
320 0, /* checksum */
321 },
322 0 /* flags */
323 };
324
325 static struct ifnet multicast_register_if;
326 static vifi_t reg_vif_num = VIFI_INVALID;
327 #endif /* PIM */
328
329
330 /*
331 * Private variables.
332 */
333 static vifi_t numvifs = 0;
334
335 static struct callout expire_upcalls_ch;
336
337 /*
338 * whether or not special PIM assert processing is enabled.
339 */
340 static int pim_assert;
341 /*
342 * Rate limit for assert notification messages, in usec
343 */
344 #define ASSERT_MSG_TIME 3000000
345
346 /*
347 * Kernel multicast routing API capabilities and setup.
348 * If more API capabilities are added to the kernel, they should be
349 * recorded in `mrt_api_support'.
350 */
351 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
352 MRT_MFC_FLAGS_BORDER_VIF |
353 MRT_MFC_RP |
354 MRT_MFC_BW_UPCALL);
355 static u_int32_t mrt_api_config = 0;
356
357 /*
358 * Find a route for a given origin IP address and Multicast group address
359 * Type of service parameter to be added in the future!!!
360 * Statistics are updated by the caller if needed
361 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
362 */
363 static struct mfc *
mfc_find(struct in_addr * o,struct in_addr * g)364 mfc_find(struct in_addr *o, struct in_addr *g)
365 {
366 struct mfc *rt;
367
368 LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
369 if (in_hosteq(rt->mfc_origin, *o) &&
370 in_hosteq(rt->mfc_mcastgrp, *g) &&
371 (rt->mfc_stall == NULL))
372 break;
373 }
374
375 return rt;
376 }
377
378 /*
379 * Macros to compute elapsed time efficiently
380 * Borrowed from Van Jacobson's scheduling code
381 */
382 #define TV_DELTA(a, b, delta) do { \
383 int xxs; \
384 delta = (a).tv_usec - (b).tv_usec; \
385 xxs = (a).tv_sec - (b).tv_sec; \
386 switch (xxs) { \
387 case 2: \
388 delta += 1000000; \
389 /* fall through */ \
390 case 1: \
391 delta += 1000000; \
392 /* fall through */ \
393 case 0: \
394 break; \
395 default: \
396 delta += (1000000 * xxs); \
397 break; \
398 } \
399 } while (/*CONSTCOND*/ 0)
400
401 #ifdef UPCALL_TIMING
402 u_int32_t upcall_data[51];
403 #endif /* UPCALL_TIMING */
404
405 /*
406 * Handle MRT setsockopt commands to modify the multicast routing tables.
407 */
408 int
ip_mrouter_set(struct socket * so,struct sockopt * sopt)409 ip_mrouter_set(struct socket *so, struct sockopt *sopt)
410 {
411 int error;
412 int optval;
413 struct vifctl vifc;
414 vifi_t vifi;
415 struct bw_upcall bwuc;
416
417 if (sopt->sopt_name != MRT_INIT && so != ip_mrouter)
418 error = ENOPROTOOPT;
419 else {
420 switch (sopt->sopt_name) {
421 case MRT_INIT:
422 error = sockopt_getint(sopt, &optval);
423 if (error)
424 break;
425
426 error = ip_mrouter_init(so, optval);
427 break;
428 case MRT_DONE:
429 error = ip_mrouter_done();
430 break;
431 case MRT_ADD_VIF:
432 error = sockopt_get(sopt, &vifc, sizeof(vifc));
433 if (error)
434 break;
435 error = add_vif(&vifc);
436 break;
437 case MRT_DEL_VIF:
438 error = sockopt_get(sopt, &vifi, sizeof(vifi));
439 if (error)
440 break;
441 error = del_vif(&vifi);
442 break;
443 case MRT_ADD_MFC:
444 error = add_mfc(sopt);
445 break;
446 case MRT_DEL_MFC:
447 error = del_mfc(sopt);
448 break;
449 case MRT_ASSERT:
450 error = sockopt_getint(sopt, &optval);
451 if (error)
452 break;
453 error = set_assert(optval);
454 break;
455 case MRT_API_CONFIG:
456 error = set_api_config(sopt);
457 break;
458 case MRT_ADD_BW_UPCALL:
459 error = sockopt_get(sopt, &bwuc, sizeof(bwuc));
460 if (error)
461 break;
462 error = add_bw_upcall(&bwuc);
463 break;
464 case MRT_DEL_BW_UPCALL:
465 error = sockopt_get(sopt, &bwuc, sizeof(bwuc));
466 if (error)
467 break;
468 error = del_bw_upcall(&bwuc);
469 break;
470 default:
471 error = ENOPROTOOPT;
472 break;
473 }
474 }
475 return error;
476 }
477
478 /*
479 * Handle MRT getsockopt commands
480 */
481 int
ip_mrouter_get(struct socket * so,struct sockopt * sopt)482 ip_mrouter_get(struct socket *so, struct sockopt *sopt)
483 {
484 int error;
485
486 if (so != ip_mrouter)
487 error = ENOPROTOOPT;
488 else {
489 switch (sopt->sopt_name) {
490 case MRT_VERSION:
491 error = sockopt_setint(sopt, 0x0305); /* XXX !!!! */
492 break;
493 case MRT_ASSERT:
494 error = sockopt_setint(sopt, pim_assert);
495 break;
496 case MRT_API_SUPPORT:
497 error = sockopt_set(sopt, &mrt_api_support,
498 sizeof(mrt_api_support));
499 break;
500 case MRT_API_CONFIG:
501 error = sockopt_set(sopt, &mrt_api_config,
502 sizeof(mrt_api_config));
503 break;
504 default:
505 error = ENOPROTOOPT;
506 break;
507 }
508 }
509 return error;
510 }
511
512 /*
513 * Handle ioctl commands to obtain information from the cache
514 */
515 int
mrt_ioctl(struct socket * so,u_long cmd,void * data)516 mrt_ioctl(struct socket *so, u_long cmd, void *data)
517 {
518 int error;
519
520 if (so != ip_mrouter)
521 error = EINVAL;
522 else
523 switch (cmd) {
524 case SIOCGETVIFCNT:
525 error = get_vif_cnt((struct sioc_vif_req *)data);
526 break;
527 case SIOCGETSGCNT:
528 error = get_sg_cnt((struct sioc_sg_req *)data);
529 break;
530 default:
531 error = EINVAL;
532 break;
533 }
534
535 return error;
536 }
537
538 /*
539 * returns the packet, byte, rpf-failure count for the source group provided
540 */
541 static int
get_sg_cnt(struct sioc_sg_req * req)542 get_sg_cnt(struct sioc_sg_req *req)
543 {
544 int s;
545 struct mfc *rt;
546
547 s = splsoftnet();
548 rt = mfc_find(&req->src, &req->grp);
549 if (rt == NULL) {
550 splx(s);
551 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
552 return EADDRNOTAVAIL;
553 }
554 req->pktcnt = rt->mfc_pkt_cnt;
555 req->bytecnt = rt->mfc_byte_cnt;
556 req->wrong_if = rt->mfc_wrong_if;
557 splx(s);
558
559 return 0;
560 }
561
562 /*
563 * returns the input and output packet and byte counts on the vif provided
564 */
565 static int
get_vif_cnt(struct sioc_vif_req * req)566 get_vif_cnt(struct sioc_vif_req *req)
567 {
568 vifi_t vifi = req->vifi;
569
570 if (vifi >= numvifs)
571 return EINVAL;
572
573 req->icount = viftable[vifi].v_pkt_in;
574 req->ocount = viftable[vifi].v_pkt_out;
575 req->ibytes = viftable[vifi].v_bytes_in;
576 req->obytes = viftable[vifi].v_bytes_out;
577
578 return 0;
579 }
580
581 /*
582 * Enable multicast routing
583 */
584 static int
ip_mrouter_init(struct socket * so,int v)585 ip_mrouter_init(struct socket *so, int v)
586 {
587 if (mrtdebug)
588 log(LOG_DEBUG,
589 "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
590 so->so_type, so->so_proto->pr_protocol);
591
592 if (so->so_type != SOCK_RAW ||
593 so->so_proto->pr_protocol != IPPROTO_IGMP)
594 return EOPNOTSUPP;
595
596 if (v != 1)
597 return EINVAL;
598
599 if (ip_mrouter != NULL)
600 return EADDRINUSE;
601
602 ip_mrouter = so;
603
604 mfchashtbl = hashinit(MFCTBLSIZ, HASH_LIST, true, &mfchash);
605 memset((void *)nexpire, 0, sizeof(nexpire));
606
607 pim_assert = 0;
608
609 callout_init(&expire_upcalls_ch, 0);
610 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
611 expire_upcalls, NULL);
612
613 callout_init(&bw_upcalls_ch, 0);
614 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
615 expire_bw_upcalls_send, NULL);
616
617 callout_init(&bw_meter_ch, 0);
618 callout_reset(&bw_meter_ch, BW_METER_PERIOD,
619 expire_bw_meter_process, NULL);
620
621 if (mrtdebug)
622 log(LOG_DEBUG, "ip_mrouter_init\n");
623
624 return 0;
625 }
626
627 /*
628 * Disable multicast routing
629 */
630 int
ip_mrouter_done(void)631 ip_mrouter_done(void)
632 {
633 vifi_t vifi;
634 struct vif *vifp;
635 int i;
636 int s;
637
638 s = splsoftnet();
639
640 /* Clear out all the vifs currently in use. */
641 for (vifi = 0; vifi < numvifs; vifi++) {
642 vifp = &viftable[vifi];
643 if (!in_nullhost(vifp->v_lcl_addr))
644 reset_vif(vifp);
645 }
646
647 numvifs = 0;
648 pim_assert = 0;
649 mrt_api_config = 0;
650
651 callout_stop(&expire_upcalls_ch);
652 callout_stop(&bw_upcalls_ch);
653 callout_stop(&bw_meter_ch);
654
655 /*
656 * Free all multicast forwarding cache entries.
657 */
658 for (i = 0; i < MFCTBLSIZ; i++) {
659 struct mfc *rt, *nrt;
660
661 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
662 nrt = LIST_NEXT(rt, mfc_hash);
663
664 expire_mfc(rt);
665 }
666 }
667
668 memset((void *)nexpire, 0, sizeof(nexpire));
669 hashdone(mfchashtbl, HASH_LIST, mfchash);
670 mfchashtbl = NULL;
671
672 bw_upcalls_n = 0;
673 memset(bw_meter_timers, 0, sizeof(bw_meter_timers));
674
675 /* Reset de-encapsulation cache. */
676
677 ip_mrouter = NULL;
678
679 splx(s);
680
681 if (mrtdebug)
682 log(LOG_DEBUG, "ip_mrouter_done\n");
683
684 return 0;
685 }
686
687 void
ip_mrouter_detach(struct ifnet * ifp)688 ip_mrouter_detach(struct ifnet *ifp)
689 {
690 int vifi, i;
691 struct vif *vifp;
692 struct mfc *rt;
693 struct rtdetq *rte;
694
695 /* XXX not sure about side effect to userland routing daemon */
696 for (vifi = 0; vifi < numvifs; vifi++) {
697 vifp = &viftable[vifi];
698 if (vifp->v_ifp == ifp)
699 reset_vif(vifp);
700 }
701 for (i = 0; i < MFCTBLSIZ; i++) {
702 if (nexpire[i] == 0)
703 continue;
704 LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
705 for (rte = rt->mfc_stall; rte; rte = rte->next) {
706 if (rte->ifp == ifp)
707 rte->ifp = NULL;
708 }
709 }
710 }
711 }
712
713 /*
714 * Set PIM assert processing global
715 */
716 static int
set_assert(int i)717 set_assert(int i)
718 {
719 pim_assert = !!i;
720 return 0;
721 }
722
723 /*
724 * Configure API capabilities
725 */
726 static int
set_api_config(struct sockopt * sopt)727 set_api_config(struct sockopt *sopt)
728 {
729 u_int32_t apival;
730 int i, error;
731
732 /*
733 * We can set the API capabilities only if it is the first operation
734 * after MRT_INIT. I.e.:
735 * - there are no vifs installed
736 * - pim_assert is not enabled
737 * - the MFC table is empty
738 */
739 error = sockopt_get(sopt, &apival, sizeof(apival));
740 if (error)
741 return error;
742 if (numvifs > 0)
743 return EPERM;
744 if (pim_assert)
745 return EPERM;
746 for (i = 0; i < MFCTBLSIZ; i++) {
747 if (LIST_FIRST(&mfchashtbl[i]) != NULL)
748 return EPERM;
749 }
750
751 mrt_api_config = apival & mrt_api_support;
752 return 0;
753 }
754
755 /*
756 * Add a vif to the vif table
757 */
758 static int
add_vif(struct vifctl * vifcp)759 add_vif(struct vifctl *vifcp)
760 {
761 struct vif *vifp;
762 struct ifnet *ifp;
763 int error, s;
764 struct sockaddr_in sin;
765
766 if (vifcp->vifc_vifi >= MAXVIFS)
767 return EINVAL;
768 if (in_nullhost(vifcp->vifc_lcl_addr))
769 return EADDRNOTAVAIL;
770
771 vifp = &viftable[vifcp->vifc_vifi];
772 if (!in_nullhost(vifp->v_lcl_addr))
773 return EADDRINUSE;
774
775 /* Find the interface with an address in AF_INET family. */
776 #ifdef PIM
777 if (vifcp->vifc_flags & VIFF_REGISTER) {
778 /*
779 * XXX: Because VIFF_REGISTER does not really need a valid
780 * local interface (e.g. it could be 127.0.0.2), we don't
781 * check its address.
782 */
783 ifp = NULL;
784 } else
785 #endif
786 {
787 struct ifaddr *ifa;
788
789 sockaddr_in_init(&sin, &vifcp->vifc_lcl_addr, 0);
790 s = pserialize_read_enter();
791 ifa = ifa_ifwithaddr(sintosa(&sin));
792 if (ifa == NULL) {
793 pserialize_read_exit(s);
794 return EADDRNOTAVAIL;
795 }
796 ifp = ifa->ifa_ifp;
797 /* FIXME NOMPSAFE */
798 pserialize_read_exit(s);
799 }
800
801 if (vifcp->vifc_flags & VIFF_TUNNEL) {
802 if (vifcp->vifc_flags & VIFF_SRCRT) {
803 log(LOG_ERR, "source routed tunnels not supported\n");
804 return EOPNOTSUPP;
805 }
806
807 /* attach this vif to decapsulator dispatch table */
808 /*
809 * XXX Use addresses in registration so that matching
810 * can be done with radix tree in decapsulator. But,
811 * we need to check inner header for multicast, so
812 * this requires both radix tree lookup and then a
813 * function to check, and this is not supported yet.
814 */
815 error = encap_lock_enter();
816 if (error)
817 return error;
818 vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
819 vif_encapcheck, &vif_encapsw, vifp);
820 encap_lock_exit();
821 if (!vifp->v_encap_cookie)
822 return EINVAL;
823
824 /* Create a fake encapsulation interface. */
825 ifp = malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK|M_ZERO);
826 snprintf(ifp->if_xname, sizeof(ifp->if_xname),
827 "mdecap%d", vifcp->vifc_vifi);
828
829 /* Prepare cached route entry. */
830 memset(&vifp->v_route, 0, sizeof(vifp->v_route));
831 #ifdef PIM
832 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
833 ifp = &multicast_register_if;
834 if (mrtdebug)
835 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
836 (void *)ifp);
837 if (reg_vif_num == VIFI_INVALID) {
838 memset(ifp, 0, sizeof(*ifp));
839 snprintf(ifp->if_xname, sizeof(ifp->if_xname),
840 "register_vif");
841 ifp->if_flags = IFF_LOOPBACK;
842 memset(&vifp->v_route, 0, sizeof(vifp->v_route));
843 reg_vif_num = vifcp->vifc_vifi;
844 }
845 #endif
846 } else {
847 /* Make sure the interface supports multicast. */
848 if ((ifp->if_flags & IFF_MULTICAST) == 0)
849 return EOPNOTSUPP;
850
851 /* Enable promiscuous reception of all IP multicasts. */
852 sockaddr_in_init(&sin, &zeroin_addr, 0);
853 error = if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin));
854 if (error)
855 return error;
856 }
857
858 s = splsoftnet();
859
860 /* Define parameters for the tbf structure. */
861 vifp->tbf_q = NULL;
862 vifp->tbf_t = &vifp->tbf_q;
863 microtime(&vifp->tbf_last_pkt_t);
864 vifp->tbf_n_tok = 0;
865 vifp->tbf_q_len = 0;
866 vifp->tbf_max_q_len = MAXQSIZE;
867
868 vifp->v_flags = vifcp->vifc_flags;
869 vifp->v_threshold = vifcp->vifc_threshold;
870 /* scaling up here allows division by 1024 in critical code */
871 vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
872 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
873 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
874 vifp->v_ifp = ifp;
875 /* Initialize per vif pkt counters. */
876 vifp->v_pkt_in = 0;
877 vifp->v_pkt_out = 0;
878 vifp->v_bytes_in = 0;
879 vifp->v_bytes_out = 0;
880
881 callout_init(&vifp->v_repq_ch, 0);
882
883 splx(s);
884
885 /* Adjust numvifs up if the vifi is higher than numvifs. */
886 if (numvifs <= vifcp->vifc_vifi)
887 numvifs = vifcp->vifc_vifi + 1;
888
889 if (mrtdebug)
890 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
891 vifcp->vifc_vifi,
892 ntohl(vifcp->vifc_lcl_addr.s_addr),
893 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
894 ntohl(vifcp->vifc_rmt_addr.s_addr),
895 vifcp->vifc_threshold,
896 vifcp->vifc_rate_limit);
897
898 return 0;
899 }
900
901 void
reset_vif(struct vif * vifp)902 reset_vif(struct vif *vifp)
903 {
904 struct mbuf *m, *n;
905 struct ifnet *ifp;
906 struct sockaddr_in sin;
907
908 callout_stop(&vifp->v_repq_ch);
909
910 /* detach this vif from decapsulator dispatch table */
911 encap_lock_enter();
912 encap_detach(vifp->v_encap_cookie);
913 encap_lock_exit();
914 vifp->v_encap_cookie = NULL;
915
916 /*
917 * Free packets queued at the interface
918 */
919 for (m = vifp->tbf_q; m != NULL; m = n) {
920 n = m->m_nextpkt;
921 m_freem(m);
922 }
923
924 if (vifp->v_flags & VIFF_TUNNEL)
925 free(vifp->v_ifp, M_MRTABLE);
926 else if (vifp->v_flags & VIFF_REGISTER) {
927 #ifdef PIM
928 reg_vif_num = VIFI_INVALID;
929 #endif
930 } else {
931 sockaddr_in_init(&sin, &zeroin_addr, 0);
932 ifp = vifp->v_ifp;
933 if_mcast_op(ifp, SIOCDELMULTI, sintosa(&sin));
934 }
935 memset((void *)vifp, 0, sizeof(*vifp));
936 }
937
938 /*
939 * Delete a vif from the vif table
940 */
941 static int
del_vif(vifi_t * vifip)942 del_vif(vifi_t *vifip)
943 {
944 struct vif *vifp;
945 vifi_t vifi;
946 int s;
947
948 if (*vifip >= numvifs)
949 return EINVAL;
950
951 vifp = &viftable[*vifip];
952 if (in_nullhost(vifp->v_lcl_addr))
953 return EADDRNOTAVAIL;
954
955 s = splsoftnet();
956
957 reset_vif(vifp);
958
959 /* Adjust numvifs down */
960 for (vifi = numvifs; vifi > 0; vifi--)
961 if (!in_nullhost(viftable[vifi - 1].v_lcl_addr))
962 break;
963 numvifs = vifi;
964
965 splx(s);
966
967 if (mrtdebug)
968 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
969
970 return 0;
971 }
972
973 /*
974 * update an mfc entry without resetting counters and S,G addresses.
975 */
976 static void
update_mfc_params(struct mfc * rt,struct mfcctl2 * mfccp)977 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
978 {
979 int i;
980
981 rt->mfc_parent = mfccp->mfcc_parent;
982 for (i = 0; i < numvifs; i++) {
983 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
984 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
985 MRT_MFC_FLAGS_ALL;
986 }
987 /* set the RP address */
988 if (mrt_api_config & MRT_MFC_RP)
989 rt->mfc_rp = mfccp->mfcc_rp;
990 else
991 rt->mfc_rp = zeroin_addr;
992 }
993
994 /*
995 * fully initialize an mfc entry from the parameter.
996 */
997 static void
init_mfc_params(struct mfc * rt,struct mfcctl2 * mfccp)998 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
999 {
1000 rt->mfc_origin = mfccp->mfcc_origin;
1001 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1002
1003 update_mfc_params(rt, mfccp);
1004
1005 /* initialize pkt counters per src-grp */
1006 rt->mfc_pkt_cnt = 0;
1007 rt->mfc_byte_cnt = 0;
1008 rt->mfc_wrong_if = 0;
1009 timerclear(&rt->mfc_last_assert);
1010 }
1011
1012 static void
expire_mfc(struct mfc * rt)1013 expire_mfc(struct mfc *rt)
1014 {
1015 struct rtdetq *rte, *nrte;
1016
1017 free_bw_list(rt->mfc_bw_meter);
1018
1019 for (rte = rt->mfc_stall; rte != NULL; rte = nrte) {
1020 nrte = rte->next;
1021 m_freem(rte->m);
1022 free(rte, M_MRTABLE);
1023 }
1024
1025 LIST_REMOVE(rt, mfc_hash);
1026 free(rt, M_MRTABLE);
1027 }
1028
1029 /*
1030 * Add an mfc entry
1031 */
1032 static int
add_mfc(struct sockopt * sopt)1033 add_mfc(struct sockopt *sopt)
1034 {
1035 struct mfcctl2 mfcctl2;
1036 struct mfcctl2 *mfccp;
1037 struct mfc *rt;
1038 u_int32_t hash = 0;
1039 struct rtdetq *rte, *nrte;
1040 u_short nstl;
1041 int s;
1042 int error;
1043
1044 /*
1045 * select data size depending on API version.
1046 */
1047 mfccp = &mfcctl2;
1048 memset(&mfcctl2, 0, sizeof(mfcctl2));
1049
1050 if (mrt_api_config & MRT_API_FLAGS_ALL)
1051 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1052 else
1053 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1054
1055 if (error)
1056 return error;
1057
1058 s = splsoftnet();
1059 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1060
1061 /* If an entry already exists, just update the fields */
1062 if (rt) {
1063 if (mrtdebug & DEBUG_MFC)
1064 log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n",
1065 ntohl(mfccp->mfcc_origin.s_addr),
1066 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1067 mfccp->mfcc_parent);
1068
1069 update_mfc_params(rt, mfccp);
1070
1071 splx(s);
1072 return 0;
1073 }
1074
1075 /*
1076 * Find the entry for which the upcall was made and update
1077 */
1078 nstl = 0;
1079 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1080 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1081 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1082 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1083 rt->mfc_stall != NULL) {
1084 if (nstl++)
1085 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n",
1086 "multiple kernel entries",
1087 ntohl(mfccp->mfcc_origin.s_addr),
1088 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1089 mfccp->mfcc_parent, rt->mfc_stall);
1090
1091 if (mrtdebug & DEBUG_MFC)
1092 log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n",
1093 ntohl(mfccp->mfcc_origin.s_addr),
1094 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1095 mfccp->mfcc_parent, rt->mfc_stall);
1096
1097 rte = rt->mfc_stall;
1098 init_mfc_params(rt, mfccp);
1099 rt->mfc_stall = NULL;
1100
1101 rt->mfc_expire = 0; /* Don't clean this guy up */
1102 nexpire[hash]--;
1103
1104 /* free packets Qed at the end of this entry */
1105 for (; rte != NULL; rte = nrte) {
1106 nrte = rte->next;
1107 if (rte->ifp) {
1108 ip_mdq(rte->m, rte->ifp, rt);
1109 }
1110 m_freem(rte->m);
1111 #ifdef UPCALL_TIMING
1112 collate(&rte->t);
1113 #endif /* UPCALL_TIMING */
1114 free(rte, M_MRTABLE);
1115 }
1116 }
1117 }
1118
1119 /*
1120 * It is possible that an entry is being inserted without an upcall
1121 */
1122 if (nstl == 0) {
1123 /*
1124 * No mfc; make a new one
1125 */
1126 if (mrtdebug & DEBUG_MFC)
1127 log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n",
1128 ntohl(mfccp->mfcc_origin.s_addr),
1129 ntohl(mfccp->mfcc_mcastgrp.s_addr),
1130 mfccp->mfcc_parent);
1131
1132 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1133 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1134 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1135 init_mfc_params(rt, mfccp);
1136 if (rt->mfc_expire)
1137 nexpire[hash]--;
1138 rt->mfc_expire = 0;
1139 break; /* XXX */
1140 }
1141 }
1142 if (rt == NULL) { /* no upcall, so make a new entry */
1143 rt = malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1144 if (rt == NULL) {
1145 splx(s);
1146 return ENOBUFS;
1147 }
1148
1149 init_mfc_params(rt, mfccp);
1150 rt->mfc_expire = 0;
1151 rt->mfc_stall = NULL;
1152 rt->mfc_bw_meter = NULL;
1153
1154 /* insert new entry at head of hash chain */
1155 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1156 }
1157 }
1158
1159 splx(s);
1160 return 0;
1161 }
1162
1163 #ifdef UPCALL_TIMING
1164 /*
1165 * collect delay statistics on the upcalls
1166 */
1167 static void
collate(struct timeval * t)1168 collate(struct timeval *t)
1169 {
1170 u_int32_t d;
1171 struct timeval tp;
1172 u_int32_t delta;
1173
1174 microtime(&tp);
1175
1176 if (timercmp(t, &tp, <)) {
1177 TV_DELTA(tp, *t, delta);
1178
1179 d = delta >> 10;
1180 if (d > 50)
1181 d = 50;
1182
1183 ++upcall_data[d];
1184 }
1185 }
1186 #endif /* UPCALL_TIMING */
1187
1188 /*
1189 * Delete an mfc entry
1190 */
1191 static int
del_mfc(struct sockopt * sopt)1192 del_mfc(struct sockopt *sopt)
1193 {
1194 struct mfcctl2 mfcctl2;
1195 struct mfcctl2 *mfccp;
1196 struct mfc *rt;
1197 int s;
1198 int error;
1199
1200 /*
1201 * XXX: for deleting MFC entries the information in entries
1202 * of size "struct mfcctl" is sufficient.
1203 */
1204
1205 mfccp = &mfcctl2;
1206 memset(&mfcctl2, 0, sizeof(mfcctl2));
1207
1208 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl));
1209 if (error) {
1210 /* Try with the size of mfcctl2. */
1211 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2));
1212 if (error)
1213 return error;
1214 }
1215
1216 if (mrtdebug & DEBUG_MFC)
1217 log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
1218 ntohl(mfccp->mfcc_origin.s_addr),
1219 ntohl(mfccp->mfcc_mcastgrp.s_addr));
1220
1221 s = splsoftnet();
1222
1223 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1224 if (rt == NULL) {
1225 splx(s);
1226 return EADDRNOTAVAIL;
1227 }
1228
1229 /*
1230 * free the bw_meter entries
1231 */
1232 free_bw_list(rt->mfc_bw_meter);
1233 rt->mfc_bw_meter = NULL;
1234
1235 LIST_REMOVE(rt, mfc_hash);
1236 free(rt, M_MRTABLE);
1237
1238 splx(s);
1239 return 0;
1240 }
1241
1242 static int
socket_send(struct socket * s,struct mbuf * mm,struct sockaddr_in * src)1243 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1244 {
1245 if (s) {
1246 if (sbappendaddr(&s->so_rcv, sintosa(src), mm, NULL) != 0) {
1247 sorwakeup(s);
1248 return 0;
1249 }
1250 soroverflow(s);
1251 }
1252 m_freem(mm);
1253 return -1;
1254 }
1255
1256 /*
1257 * IP multicast forwarding function. This function assumes that the packet
1258 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1259 * pointed to by "ifp", and the packet is to be relayed to other networks
1260 * that have members of the packet's destination IP multicast group.
1261 *
1262 * The packet is returned unscathed to the caller, unless it is
1263 * erroneous, in which case a non-zero return value tells the caller to
1264 * discard it.
1265 */
1266
1267 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */
1268 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1269
1270 int
ip_mforward(struct mbuf * m,struct ifnet * ifp)1271 ip_mforward(struct mbuf *m, struct ifnet *ifp)
1272 {
1273 int rc;
1274 /*
1275 * save csum_flags to uphold the
1276 * "unscathed" guarantee.
1277 * ip_output() relies on that and
1278 * without it we send out
1279 * multicast packets with an invalid
1280 * checksum
1281 *
1282 * see PR kern/55779
1283 */
1284 int csum_flags = m->m_pkthdr.csum_flags;
1285
1286 /*
1287 * Temporarily clear any in-bound checksum flags for this packet.
1288 */
1289 m->m_pkthdr.csum_flags = 0;
1290
1291 rc = ip_mforward_real(m, ifp);
1292
1293 m->m_pkthdr.csum_flags = csum_flags;
1294
1295 return rc;
1296 }
1297
1298 static int
ip_mforward_real(struct mbuf * m,struct ifnet * ifp)1299 ip_mforward_real(struct mbuf *m, struct ifnet *ifp)
1300 {
1301 struct ip *ip = mtod(m, struct ip *);
1302 struct mfc *rt;
1303 static int srctun = 0;
1304 struct mbuf *mm;
1305 struct sockaddr_in sin;
1306 int s;
1307 vifi_t vifi;
1308
1309 if (mrtdebug & DEBUG_FORWARD)
1310 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1311 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1312
1313 /*
1314 * XXX XXX: Why do we check [1] against IPOPT_LSRR? Because we
1315 * expect [0] to be IPOPT_NOP, maybe? In all cases that doesn't
1316 * make a lot of sense, a forged packet can just put two IPOPT_NOPs
1317 * followed by one IPOPT_LSRR, and bypass the check.
1318 */
1319 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1320 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1321 /*
1322 * Packet arrived via a physical interface or
1323 * an encapsulated tunnel or a register_vif.
1324 */
1325 } else {
1326 /*
1327 * Packet arrived through a source-route tunnel.
1328 * Source-route tunnels are no longer supported.
1329 */
1330 if ((srctun++ % 1000) == 0)
1331 log(LOG_ERR,
1332 "ip_mforward: received source-routed packet from %x\n",
1333 ntohl(ip->ip_src.s_addr));
1334 return EOPNOTSUPP;
1335 }
1336
1337 /*
1338 * Don't forward a packet with time-to-live of zero or one,
1339 * or a packet destined to a local-only group.
1340 */
1341 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1342 return 0;
1343
1344 /*
1345 * Determine forwarding vifs from the forwarding cache table
1346 */
1347 s = splsoftnet();
1348 ++mrtstat.mrts_mfc_lookups;
1349 rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1350
1351 /* Entry exists, so forward if necessary */
1352 if (rt != NULL) {
1353 splx(s);
1354 return ip_mdq(m, ifp, rt);
1355 } else {
1356 /*
1357 * If we don't have a route for packet's origin, make a copy
1358 * of the packet and send message to routing daemon.
1359 */
1360
1361 struct mbuf *mb0;
1362 struct rtdetq *rte;
1363 u_int32_t hash;
1364 const int hlen = ip->ip_hl << 2;
1365 #ifdef UPCALL_TIMING
1366 struct timeval tp;
1367 microtime(&tp);
1368 #endif
1369
1370 ++mrtstat.mrts_mfc_misses;
1371
1372 mrtstat.mrts_no_route++;
1373 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1374 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1375 ntohl(ip->ip_src.s_addr),
1376 ntohl(ip->ip_dst.s_addr));
1377
1378 /*
1379 * Allocate mbufs early so that we don't do extra work if we are
1380 * just going to fail anyway. Make sure to pullup the header so
1381 * that other people can't step on it.
1382 */
1383 rte = malloc(sizeof(*rte), M_MRTABLE, M_NOWAIT);
1384 if (rte == NULL) {
1385 splx(s);
1386 return ENOBUFS;
1387 }
1388 mb0 = m_copypacket(m, M_DONTWAIT);
1389 M_PULLUP(mb0, hlen);
1390 if (mb0 == NULL) {
1391 free(rte, M_MRTABLE);
1392 splx(s);
1393 return ENOBUFS;
1394 }
1395
1396 /* is there an upcall waiting for this flow? */
1397 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1398 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1399 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1400 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1401 rt->mfc_stall != NULL)
1402 break;
1403 }
1404
1405 if (rt == NULL) {
1406 int i;
1407 struct igmpmsg *im;
1408
1409 /*
1410 * Locate the vifi for the incoming interface for
1411 * this packet.
1412 * If none found, drop packet.
1413 */
1414 for (vifi = 0; vifi < numvifs &&
1415 viftable[vifi].v_ifp != ifp; vifi++)
1416 ;
1417 if (vifi >= numvifs) /* vif not found, drop packet */
1418 goto non_fatal;
1419
1420 /* no upcall, so make a new entry */
1421 rt = malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1422 if (rt == NULL)
1423 goto fail;
1424
1425 /*
1426 * Make a copy of the header to send to the user level
1427 * process
1428 */
1429 mm = m_copym(m, 0, hlen, M_DONTWAIT);
1430 M_PULLUP(mm, hlen);
1431 if (mm == NULL)
1432 goto fail1;
1433
1434 /*
1435 * Send message to routing daemon to install
1436 * a route into the kernel table
1437 */
1438
1439 im = mtod(mm, struct igmpmsg *);
1440 im->im_msgtype = IGMPMSG_NOCACHE;
1441 im->im_mbz = 0;
1442 im->im_vif = vifi;
1443
1444 mrtstat.mrts_upcalls++;
1445
1446 sockaddr_in_init(&sin, &ip->ip_src, 0);
1447 if (socket_send(ip_mrouter, mm, &sin) < 0) {
1448 log(LOG_WARNING,
1449 "ip_mforward: ip_mrouter socket queue full\n");
1450 ++mrtstat.mrts_upq_sockfull;
1451 fail1:
1452 free(rt, M_MRTABLE);
1453 fail:
1454 free(rte, M_MRTABLE);
1455 m_freem(mb0);
1456 splx(s);
1457 return ENOBUFS;
1458 }
1459
1460 /* insert new entry at head of hash chain */
1461 rt->mfc_origin = ip->ip_src;
1462 rt->mfc_mcastgrp = ip->ip_dst;
1463 rt->mfc_pkt_cnt = 0;
1464 rt->mfc_byte_cnt = 0;
1465 rt->mfc_wrong_if = 0;
1466 rt->mfc_expire = UPCALL_EXPIRE;
1467 nexpire[hash]++;
1468 for (i = 0; i < numvifs; i++) {
1469 rt->mfc_ttls[i] = 0;
1470 rt->mfc_flags[i] = 0;
1471 }
1472 rt->mfc_parent = -1;
1473
1474 /* clear the RP address */
1475 rt->mfc_rp = zeroin_addr;
1476
1477 rt->mfc_bw_meter = NULL;
1478
1479 /* link into table */
1480 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1481 /* Add this entry to the end of the queue */
1482 rt->mfc_stall = rte;
1483 } else {
1484 /* determine if q has overflowed */
1485 struct rtdetq **p;
1486 int npkts = 0;
1487
1488 /*
1489 * XXX ouch! we need to append to the list, but we
1490 * only have a pointer to the front, so we have to
1491 * scan the entire list every time.
1492 */
1493 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1494 if (++npkts > MAX_UPQ) {
1495 mrtstat.mrts_upq_ovflw++;
1496 non_fatal:
1497 free(rte, M_MRTABLE);
1498 m_freem(mb0);
1499 splx(s);
1500 return 0;
1501 }
1502
1503 /* Add this entry to the end of the queue */
1504 *p = rte;
1505 }
1506
1507 rte->next = NULL;
1508 rte->m = mb0;
1509 rte->ifp = ifp;
1510 #ifdef UPCALL_TIMING
1511 rte->t = tp;
1512 #endif
1513
1514 splx(s);
1515
1516 return 0;
1517 }
1518 }
1519
1520 /*ARGSUSED*/
1521 static void
expire_upcalls(void * v)1522 expire_upcalls(void *v)
1523 {
1524 int i;
1525
1526 /* XXX NOMPSAFE still need softnet_lock */
1527 mutex_enter(softnet_lock);
1528 KERNEL_LOCK(1, NULL);
1529
1530 for (i = 0; i < MFCTBLSIZ; i++) {
1531 struct mfc *rt, *nrt;
1532
1533 if (nexpire[i] == 0)
1534 continue;
1535
1536 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1537 nrt = LIST_NEXT(rt, mfc_hash);
1538
1539 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1540 continue;
1541 nexpire[i]--;
1542
1543 /*
1544 * free the bw_meter entries
1545 */
1546 while (rt->mfc_bw_meter != NULL) {
1547 struct bw_meter *x = rt->mfc_bw_meter;
1548
1549 rt->mfc_bw_meter = x->bm_mfc_next;
1550 kmem_intr_free(x, sizeof(*x));
1551 }
1552
1553 ++mrtstat.mrts_cache_cleanups;
1554 if (mrtdebug & DEBUG_EXPIRE)
1555 log(LOG_DEBUG,
1556 "expire_upcalls: expiring (%x %x)\n",
1557 ntohl(rt->mfc_origin.s_addr),
1558 ntohl(rt->mfc_mcastgrp.s_addr));
1559
1560 expire_mfc(rt);
1561 }
1562 }
1563
1564 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1565 expire_upcalls, NULL);
1566
1567 KERNEL_UNLOCK_ONE(NULL);
1568 mutex_exit(softnet_lock);
1569 }
1570
1571 /*
1572 * Macro to send packet on vif.
1573 */
1574 #define MC_SEND(ip, vifp, m) do { \
1575 if ((vifp)->v_flags & VIFF_TUNNEL) \
1576 encap_send((ip), (vifp), (m)); \
1577 else \
1578 phyint_send((ip), (vifp), (m)); \
1579 } while (/*CONSTCOND*/ 0)
1580
1581 /*
1582 * Packet forwarding routine once entry in the cache is made
1583 */
1584 static int
ip_mdq(struct mbuf * m,struct ifnet * ifp,struct mfc * rt)1585 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt)
1586 {
1587 struct ip *ip = mtod(m, struct ip *);
1588 vifi_t vifi;
1589 struct vif *vifp;
1590 struct sockaddr_in sin;
1591 const int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1592
1593 /*
1594 * Don't forward if it didn't arrive from the parent vif for its origin.
1595 */
1596 vifi = rt->mfc_parent;
1597 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1598 /* came in the wrong interface */
1599 if (mrtdebug & DEBUG_FORWARD)
1600 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1601 ifp, vifi,
1602 vifi >= numvifs ? 0 : viftable[vifi].v_ifp);
1603 ++mrtstat.mrts_wrong_if;
1604 ++rt->mfc_wrong_if;
1605
1606 /*
1607 * If we are doing PIM assert processing, send a message
1608 * to the routing daemon.
1609 *
1610 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1611 * can complete the SPT switch, regardless of the type
1612 * of the iif (broadcast media, GRE tunnel, etc).
1613 */
1614 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1615 struct timeval now;
1616 u_int32_t delta;
1617
1618 #ifdef PIM
1619 if (ifp == &multicast_register_if)
1620 pimstat.pims_rcv_registers_wrongiif++;
1621 #endif
1622
1623 /* Get vifi for the incoming packet */
1624 for (vifi = 0;
1625 vifi < numvifs && viftable[vifi].v_ifp != ifp;
1626 vifi++)
1627 ;
1628 if (vifi >= numvifs) {
1629 /* The iif is not found: ignore the packet. */
1630 return 0;
1631 }
1632
1633 if (rt->mfc_flags[vifi] &
1634 MRT_MFC_FLAGS_DISABLE_WRONGVIF) {
1635 /* WRONGVIF disabled: ignore the packet */
1636 return 0;
1637 }
1638
1639 microtime(&now);
1640
1641 TV_DELTA(rt->mfc_last_assert, now, delta);
1642
1643 if (delta > ASSERT_MSG_TIME) {
1644 struct igmpmsg *im;
1645 const int hlen = ip->ip_hl << 2;
1646 struct mbuf *mm =
1647 m_copym(m, 0, hlen, M_DONTWAIT);
1648
1649 M_PULLUP(mm, hlen);
1650 if (mm == NULL)
1651 return ENOBUFS;
1652
1653 rt->mfc_last_assert = now;
1654
1655 im = mtod(mm, struct igmpmsg *);
1656 im->im_msgtype = IGMPMSG_WRONGVIF;
1657 im->im_mbz = 0;
1658 im->im_vif = vifi;
1659
1660 mrtstat.mrts_upcalls++;
1661
1662 sockaddr_in_init(&sin, &im->im_src, 0);
1663 if (socket_send(ip_mrouter, mm, &sin) < 0) {
1664 log(LOG_WARNING,
1665 "ip_mforward: ip_mrouter socket queue full\n");
1666 ++mrtstat.mrts_upq_sockfull;
1667 return ENOBUFS;
1668 }
1669 }
1670 }
1671 return 0;
1672 }
1673
1674 /* If I sourced this packet, it counts as output, else it was input. */
1675 if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1676 viftable[vifi].v_pkt_out++;
1677 viftable[vifi].v_bytes_out += plen;
1678 } else {
1679 viftable[vifi].v_pkt_in++;
1680 viftable[vifi].v_bytes_in += plen;
1681 }
1682 rt->mfc_pkt_cnt++;
1683 rt->mfc_byte_cnt += plen;
1684
1685 /*
1686 * For each vif, decide if a copy of the packet should be forwarded.
1687 * Forward if:
1688 * - the ttl exceeds the vif's threshold
1689 * - there are group members downstream on interface
1690 */
1691 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) {
1692 if ((rt->mfc_ttls[vifi] > 0) &&
1693 (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1694 vifp->v_pkt_out++;
1695 vifp->v_bytes_out += plen;
1696 #ifdef PIM
1697 if (vifp->v_flags & VIFF_REGISTER)
1698 pim_register_send(ip, vifp, m, rt);
1699 else
1700 #endif
1701 MC_SEND(ip, vifp, m);
1702 }
1703 }
1704
1705 /*
1706 * Perform upcall-related bw measuring.
1707 */
1708 if (rt->mfc_bw_meter != NULL) {
1709 struct bw_meter *x;
1710 struct timeval now;
1711
1712 microtime(&now);
1713 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1714 bw_meter_receive_packet(x, plen, &now);
1715 }
1716
1717 return 0;
1718 }
1719
1720 static void
phyint_send(struct ip * ip,struct vif * vifp,struct mbuf * m)1721 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1722 {
1723 struct mbuf *mb_copy;
1724 const int hlen = ip->ip_hl << 2;
1725
1726 /*
1727 * Make a new reference to the packet; make sure that
1728 * the IP header is actually copied, not just referenced,
1729 * so that ip_output() only scribbles on the copy.
1730 */
1731 mb_copy = m_copypacket(m, M_DONTWAIT);
1732 M_PULLUP(mb_copy, hlen);
1733 if (mb_copy == NULL)
1734 return;
1735
1736 if (vifp->v_rate_limit <= 0)
1737 tbf_send_packet(vifp, mb_copy);
1738 else
1739 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *),
1740 ntohs(ip->ip_len));
1741 }
1742
1743 static void
encap_send(struct ip * ip,struct vif * vifp,struct mbuf * m)1744 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1745 {
1746 struct mbuf *mb_copy;
1747 struct ip *ip_copy;
1748 int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr);
1749
1750 /* Take care of delayed checksums */
1751 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1752 in_undefer_cksum_tcpudp(m);
1753 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1754 }
1755
1756 /*
1757 * copy the old packet & pullup its IP header into the
1758 * new mbuf so we can modify it. Try to fill the new
1759 * mbuf since if we don't the ethernet driver will.
1760 */
1761 MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1762 if (mb_copy == NULL)
1763 return;
1764 mb_copy->m_data += max_linkhdr;
1765 mb_copy->m_pkthdr.len = len;
1766 mb_copy->m_len = sizeof(multicast_encap_iphdr);
1767
1768 if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) {
1769 m_freem(mb_copy);
1770 return;
1771 }
1772 i = MHLEN - max_linkhdr;
1773 if (i > len)
1774 i = len;
1775 mb_copy = m_pullup(mb_copy, i);
1776 if (mb_copy == NULL)
1777 return;
1778
1779 /*
1780 * fill in the encapsulating IP header.
1781 */
1782 ip_copy = mtod(mb_copy, struct ip *);
1783 *ip_copy = multicast_encap_iphdr;
1784 if (len < IP_MINFRAGSIZE)
1785 ip_copy->ip_id = 0;
1786 else
1787 ip_copy->ip_id = ip_newid(NULL);
1788 ip_copy->ip_len = htons(len);
1789 ip_copy->ip_src = vifp->v_lcl_addr;
1790 ip_copy->ip_dst = vifp->v_rmt_addr;
1791
1792 /*
1793 * turn the encapsulated IP header back into a valid one.
1794 */
1795 ip = (struct ip *)((char *)ip_copy + sizeof(multicast_encap_iphdr));
1796 --ip->ip_ttl;
1797 ip->ip_sum = 0;
1798 mb_copy->m_data += sizeof(multicast_encap_iphdr);
1799 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1800 mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1801
1802 if (vifp->v_rate_limit <= 0)
1803 tbf_send_packet(vifp, mb_copy);
1804 else
1805 tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len));
1806 }
1807
1808 /*
1809 * De-encapsulate a packet and feed it back through ip input.
1810 */
1811 static void
vif_input(struct mbuf * m,int off,int proto,void * eparg)1812 vif_input(struct mbuf *m, int off, int proto, void *eparg)
1813 {
1814 struct vif *vifp = eparg;
1815
1816 KASSERT(vifp != NULL);
1817
1818 if (proto != ENCAP_PROTO) {
1819 m_freem(m);
1820 mrtstat.mrts_bad_tunnel++;
1821 return;
1822 }
1823
1824 m_adj(m, off);
1825 m_set_rcvif(m, vifp->v_ifp);
1826
1827 if (__predict_false(!pktq_enqueue(ip_pktq, m, 0))) {
1828 m_freem(m);
1829 }
1830 }
1831
1832 /*
1833 * Check if the packet should be received on the vif denoted by arg.
1834 * (The encap selection code will call this once per vif since each is
1835 * registered separately.)
1836 */
1837 static int
vif_encapcheck(struct mbuf * m,int off,int proto,void * arg)1838 vif_encapcheck(struct mbuf *m, int off, int proto, void *arg)
1839 {
1840 struct vif *vifp;
1841 struct ip ip;
1842
1843 #ifdef DIAGNOSTIC
1844 if (!arg || proto != IPPROTO_IPV4)
1845 panic("unexpected arg in vif_encapcheck");
1846 #endif
1847
1848 /*
1849 * Accept the packet only if the inner header is multicast
1850 * and the outer header matches a tunnel-mode vif. Order
1851 * checks in the hope that common non-matching packets will be
1852 * rejected quickly. Assume that unicast IPv4 traffic in a
1853 * parallel tunnel (e.g. gif(4)) is unlikely.
1854 */
1855
1856 /* Obtain the outer IP header and the vif pointer. */
1857 m_copydata(m, 0, sizeof(ip), (void *)&ip);
1858 vifp = (struct vif *)arg;
1859
1860 /*
1861 * The outer source must match the vif's remote peer address.
1862 * For a multicast router with several tunnels, this is the
1863 * only check that will fail on packets in other tunnels,
1864 * assuming the local address is the same.
1865 */
1866 if (!in_hosteq(vifp->v_rmt_addr, ip.ip_src))
1867 return 0;
1868
1869 /* The outer destination must match the vif's local address. */
1870 if (!in_hosteq(vifp->v_lcl_addr, ip.ip_dst))
1871 return 0;
1872
1873 /* The vif must be of tunnel type. */
1874 if ((vifp->v_flags & VIFF_TUNNEL) == 0)
1875 return 0;
1876
1877 /* Check that the inner destination is multicast. */
1878 if (off + sizeof(ip) > m->m_pkthdr.len)
1879 return 0;
1880 m_copydata(m, off, sizeof(ip), (void *)&ip);
1881 if (!IN_MULTICAST(ip.ip_dst.s_addr))
1882 return 0;
1883
1884 /*
1885 * We have checked that both the outer src and dst addresses
1886 * match the vif, and that the inner destination is multicast
1887 * (224/5). By claiming more than 64, we intend to
1888 * preferentially take packets that also match a parallel
1889 * gif(4).
1890 */
1891 return 32 + 32 + 5;
1892 }
1893
1894 /*
1895 * Token bucket filter module
1896 */
1897 static void
tbf_control(struct vif * vifp,struct mbuf * m,struct ip * ip,u_int32_t len)1898 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len)
1899 {
1900
1901 if (len > MAX_BKT_SIZE) {
1902 /* drop if packet is too large */
1903 mrtstat.mrts_pkt2large++;
1904 m_freem(m);
1905 return;
1906 }
1907
1908 tbf_update_tokens(vifp);
1909
1910 /*
1911 * If there are enough tokens, and the queue is empty, send this packet
1912 * out immediately. Otherwise, try to insert it on this vif's queue.
1913 */
1914 if (vifp->tbf_q_len == 0) {
1915 if (len <= vifp->tbf_n_tok) {
1916 vifp->tbf_n_tok -= len;
1917 tbf_send_packet(vifp, m);
1918 } else {
1919 /* queue packet and timeout till later */
1920 tbf_queue(vifp, m);
1921 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
1922 tbf_reprocess_q, vifp);
1923 }
1924 } else {
1925 if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
1926 !tbf_dq_sel(vifp, ip)) {
1927 /* queue full, and couldn't make room */
1928 mrtstat.mrts_q_overflow++;
1929 m_freem(m);
1930 } else {
1931 /* queue length low enough, or made room */
1932 tbf_queue(vifp, m);
1933 tbf_process_q(vifp);
1934 }
1935 }
1936 }
1937
1938 /*
1939 * adds a packet to the queue at the interface
1940 */
1941 static void
tbf_queue(struct vif * vifp,struct mbuf * m)1942 tbf_queue(struct vif *vifp, struct mbuf *m)
1943 {
1944 int s = splsoftnet();
1945
1946 /* insert at tail */
1947 *vifp->tbf_t = m;
1948 vifp->tbf_t = &m->m_nextpkt;
1949 vifp->tbf_q_len++;
1950
1951 splx(s);
1952 }
1953
1954 /*
1955 * processes the queue at the interface
1956 */
1957 static void
tbf_process_q(struct vif * vifp)1958 tbf_process_q(struct vif *vifp)
1959 {
1960 struct mbuf *m;
1961 int len;
1962 int s = splsoftnet();
1963
1964 /*
1965 * Loop through the queue at the interface and send as many packets
1966 * as possible.
1967 */
1968 for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) {
1969 len = ntohs(mtod(m, struct ip *)->ip_len);
1970
1971 /* determine if the packet can be sent */
1972 if (len <= vifp->tbf_n_tok) {
1973 /* if so,
1974 * reduce no of tokens, dequeue the packet,
1975 * send the packet.
1976 */
1977 if ((vifp->tbf_q = m->m_nextpkt) == NULL)
1978 vifp->tbf_t = &vifp->tbf_q;
1979 --vifp->tbf_q_len;
1980
1981 m->m_nextpkt = NULL;
1982 vifp->tbf_n_tok -= len;
1983 tbf_send_packet(vifp, m);
1984 } else
1985 break;
1986 }
1987 splx(s);
1988 }
1989
1990 static void
tbf_reprocess_q(void * arg)1991 tbf_reprocess_q(void *arg)
1992 {
1993 struct vif *vifp = arg;
1994
1995 if (ip_mrouter == NULL)
1996 return;
1997
1998 tbf_update_tokens(vifp);
1999 tbf_process_q(vifp);
2000
2001 if (vifp->tbf_q_len != 0)
2002 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS,
2003 tbf_reprocess_q, vifp);
2004 }
2005
2006 /* function that will selectively discard a member of the queue
2007 * based on the precedence value and the priority
2008 */
2009 static int
tbf_dq_sel(struct vif * vifp,struct ip * ip)2010 tbf_dq_sel(struct vif *vifp, struct ip *ip)
2011 {
2012 u_int p;
2013 struct mbuf **mp, *m;
2014 int s = splsoftnet();
2015
2016 p = priority(vifp, ip);
2017
2018 for (mp = &vifp->tbf_q, m = *mp;
2019 m != NULL;
2020 mp = &m->m_nextpkt, m = *mp) {
2021 if (p > priority(vifp, mtod(m, struct ip *))) {
2022 if ((*mp = m->m_nextpkt) == NULL)
2023 vifp->tbf_t = mp;
2024 --vifp->tbf_q_len;
2025
2026 m_freem(m);
2027 mrtstat.mrts_drop_sel++;
2028 splx(s);
2029 return 1;
2030 }
2031 }
2032 splx(s);
2033 return 0;
2034 }
2035
2036 static void
tbf_send_packet(struct vif * vifp,struct mbuf * m)2037 tbf_send_packet(struct vif *vifp, struct mbuf *m)
2038 {
2039 int error;
2040 int s = splsoftnet();
2041
2042 if (vifp->v_flags & VIFF_TUNNEL) {
2043 /* If tunnel options */
2044 ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL);
2045 } else {
2046 /* if physical interface option, extract the options and then send */
2047 struct ip_moptions imo;
2048
2049 imo.imo_multicast_if_index = if_get_index(vifp->v_ifp);
2050 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2051 imo.imo_multicast_loop = 1;
2052
2053 error = ip_output(m, NULL, NULL, IP_FORWARDING|IP_MULTICASTOPTS,
2054 &imo, NULL);
2055
2056 if (mrtdebug & DEBUG_XMIT)
2057 log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
2058 (long)(vifp - viftable), error);
2059 }
2060 splx(s);
2061 }
2062
2063 /* determine the current time and then
2064 * the elapsed time (between the last time and time now)
2065 * in milliseconds & update the no. of tokens in the bucket
2066 */
2067 static void
tbf_update_tokens(struct vif * vifp)2068 tbf_update_tokens(struct vif *vifp)
2069 {
2070 struct timeval tp;
2071 u_int32_t tm;
2072 int s = splsoftnet();
2073
2074 microtime(&tp);
2075
2076 TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
2077
2078 /*
2079 * This formula is actually
2080 * "time in seconds" * "bytes/second".
2081 *
2082 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2083 *
2084 * The (1000/1024) was introduced in add_vif to optimize
2085 * this divide into a shift.
2086 */
2087 vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
2088 vifp->tbf_last_pkt_t = tp;
2089
2090 if (vifp->tbf_n_tok > MAX_BKT_SIZE)
2091 vifp->tbf_n_tok = MAX_BKT_SIZE;
2092
2093 splx(s);
2094 }
2095
2096 static int
priority(struct vif * vifp,struct ip * ip)2097 priority(struct vif *vifp, struct ip *ip)
2098 {
2099 int prio = 50; /* the lowest priority -- default case */
2100
2101 /* temporary hack; may add general packet classifier some day */
2102
2103 /*
2104 * XXX XXX: We're reading the UDP header, but we didn't ensure
2105 * it was present in the packet.
2106 */
2107
2108 /*
2109 * The UDP port space is divided up into four priority ranges:
2110 * [0, 16384) : unclassified - lowest priority
2111 * [16384, 32768) : audio - highest priority
2112 * [32768, 49152) : whiteboard - medium priority
2113 * [49152, 65536) : video - low priority
2114 */
2115 if (ip->ip_p == IPPROTO_UDP) {
2116 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2117
2118 switch (ntohs(udp->uh_dport) & 0xc000) {
2119 case 0x4000:
2120 prio = 70;
2121 break;
2122 case 0x8000:
2123 prio = 60;
2124 break;
2125 case 0xc000:
2126 prio = 55;
2127 break;
2128 }
2129
2130 if (tbfdebug > 1)
2131 log(LOG_DEBUG, "port %x prio %d\n",
2132 ntohs(udp->uh_dport), prio);
2133 }
2134
2135 return prio;
2136 }
2137
2138 /*
2139 * Code for bandwidth monitors
2140 */
2141
2142 /*
2143 * Define common interface for timeval-related methods
2144 */
2145 #define BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp)
2146 #define BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp))
2147 #define BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp))
2148
2149 static uint32_t
compute_bw_meter_flags(struct bw_upcall * req)2150 compute_bw_meter_flags(struct bw_upcall *req)
2151 {
2152 uint32_t flags = 0;
2153
2154 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2155 flags |= BW_METER_UNIT_PACKETS;
2156 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2157 flags |= BW_METER_UNIT_BYTES;
2158 if (req->bu_flags & BW_UPCALL_GEQ)
2159 flags |= BW_METER_GEQ;
2160 if (req->bu_flags & BW_UPCALL_LEQ)
2161 flags |= BW_METER_LEQ;
2162
2163 return flags;
2164 }
2165
2166 /*
2167 * Add a bw_meter entry
2168 */
2169 static int
add_bw_upcall(struct bw_upcall * req)2170 add_bw_upcall(struct bw_upcall *req)
2171 {
2172 int s;
2173 struct mfc *mfc;
2174 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2175 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2176 struct timeval now;
2177 struct bw_meter *x;
2178 uint32_t flags;
2179
2180 if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2181 return EOPNOTSUPP;
2182
2183 /* Test if the flags are valid */
2184 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2185 return EINVAL;
2186 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2187 return EINVAL;
2188 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2189 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2190 return EINVAL;
2191
2192 /* Test if the threshold time interval is valid */
2193 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2194 return EINVAL;
2195
2196 flags = compute_bw_meter_flags(req);
2197
2198 /*
2199 * Find if we have already same bw_meter entry
2200 */
2201 s = splsoftnet();
2202 mfc = mfc_find(&req->bu_src, &req->bu_dst);
2203 if (mfc == NULL) {
2204 splx(s);
2205 return EADDRNOTAVAIL;
2206 }
2207 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2208 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2209 &req->bu_threshold.b_time, ==)) &&
2210 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2211 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2212 (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
2213 splx(s);
2214 return 0; /* XXX Already installed */
2215 }
2216 }
2217
2218 /* Allocate the new bw_meter entry */
2219 x = kmem_intr_alloc(sizeof(*x), KM_NOSLEEP);
2220 if (x == NULL) {
2221 splx(s);
2222 return ENOBUFS;
2223 }
2224
2225 /* Set the new bw_meter entry */
2226 x->bm_threshold.b_time = req->bu_threshold.b_time;
2227 microtime(&now);
2228 x->bm_start_time = now;
2229 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2230 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2231 x->bm_measured.b_packets = 0;
2232 x->bm_measured.b_bytes = 0;
2233 x->bm_flags = flags;
2234 x->bm_time_next = NULL;
2235 x->bm_time_hash = BW_METER_BUCKETS;
2236
2237 /* Add the new bw_meter entry to the front of entries for this MFC */
2238 x->bm_mfc = mfc;
2239 x->bm_mfc_next = mfc->mfc_bw_meter;
2240 mfc->mfc_bw_meter = x;
2241 schedule_bw_meter(x, &now);
2242 splx(s);
2243
2244 return 0;
2245 }
2246
2247 static void
free_bw_list(struct bw_meter * list)2248 free_bw_list(struct bw_meter *list)
2249 {
2250 while (list != NULL) {
2251 struct bw_meter *x = list;
2252
2253 list = list->bm_mfc_next;
2254 unschedule_bw_meter(x);
2255 kmem_intr_free(x, sizeof(*x));
2256 }
2257 }
2258
2259 /*
2260 * Delete one or multiple bw_meter entries
2261 */
2262 static int
del_bw_upcall(struct bw_upcall * req)2263 del_bw_upcall(struct bw_upcall *req)
2264 {
2265 int s;
2266 struct mfc *mfc;
2267 struct bw_meter *x;
2268
2269 if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2270 return EOPNOTSUPP;
2271
2272 s = splsoftnet();
2273 /* Find the corresponding MFC entry */
2274 mfc = mfc_find(&req->bu_src, &req->bu_dst);
2275 if (mfc == NULL) {
2276 splx(s);
2277 return EADDRNOTAVAIL;
2278 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2279 /*
2280 * Delete all bw_meter entries for this mfc
2281 */
2282 struct bw_meter *list;
2283
2284 list = mfc->mfc_bw_meter;
2285 mfc->mfc_bw_meter = NULL;
2286 free_bw_list(list);
2287 splx(s);
2288 return 0;
2289 } else { /* Delete a single bw_meter entry */
2290 struct bw_meter *prev;
2291 uint32_t flags = 0;
2292
2293 flags = compute_bw_meter_flags(req);
2294
2295 /* Find the bw_meter entry to delete */
2296 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2297 prev = x, x = x->bm_mfc_next) {
2298 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2299 &req->bu_threshold.b_time, ==)) &&
2300 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2301 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2302 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2303 break;
2304 }
2305 if (x != NULL) { /* Delete entry from the list for this MFC */
2306 if (prev != NULL)
2307 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
2308 else
2309 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2310
2311 unschedule_bw_meter(x);
2312 splx(s);
2313 /* Free the bw_meter entry */
2314 kmem_intr_free(x, sizeof(*x));
2315 return 0;
2316 } else {
2317 splx(s);
2318 return EINVAL;
2319 }
2320 }
2321 /* NOTREACHED */
2322 }
2323
2324 /*
2325 * Perform bandwidth measurement processing that may result in an upcall
2326 */
2327 static void
bw_meter_receive_packet(struct bw_meter * x,int plen,struct timeval * nowp)2328 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2329 {
2330 struct timeval delta;
2331
2332 delta = *nowp;
2333 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2334
2335 if (x->bm_flags & BW_METER_GEQ) {
2336 /*
2337 * Processing for ">=" type of bw_meter entry
2338 */
2339 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2340 /* Reset the bw_meter entry */
2341 x->bm_start_time = *nowp;
2342 x->bm_measured.b_packets = 0;
2343 x->bm_measured.b_bytes = 0;
2344 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2345 }
2346
2347 /* Record that a packet is received */
2348 x->bm_measured.b_packets++;
2349 x->bm_measured.b_bytes += plen;
2350
2351 /*
2352 * Test if we should deliver an upcall
2353 */
2354 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2355 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2356 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2357 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2358 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2359 /* Prepare an upcall for delivery */
2360 bw_meter_prepare_upcall(x, nowp);
2361 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2362 }
2363 }
2364 } else if (x->bm_flags & BW_METER_LEQ) {
2365 /*
2366 * Processing for "<=" type of bw_meter entry
2367 */
2368 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2369 /*
2370 * We are behind time with the multicast forwarding table
2371 * scanning for "<=" type of bw_meter entries, so test now
2372 * if we should deliver an upcall.
2373 */
2374 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2375 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2376 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2377 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2378 /* Prepare an upcall for delivery */
2379 bw_meter_prepare_upcall(x, nowp);
2380 }
2381 /* Reschedule the bw_meter entry */
2382 unschedule_bw_meter(x);
2383 schedule_bw_meter(x, nowp);
2384 }
2385
2386 /* Record that a packet is received */
2387 x->bm_measured.b_packets++;
2388 x->bm_measured.b_bytes += plen;
2389
2390 /*
2391 * Test if we should restart the measuring interval
2392 */
2393 if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2394 x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2395 (x->bm_flags & BW_METER_UNIT_BYTES &&
2396 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2397 /* Don't restart the measuring interval */
2398 } else {
2399 /* Do restart the measuring interval */
2400 /*
2401 * XXX: note that we don't unschedule and schedule, because this
2402 * might be too much overhead per packet. Instead, when we process
2403 * all entries for a given timer hash bin, we check whether it is
2404 * really a timeout. If not, we reschedule at that time.
2405 */
2406 x->bm_start_time = *nowp;
2407 x->bm_measured.b_packets = 0;
2408 x->bm_measured.b_bytes = 0;
2409 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2410 }
2411 }
2412 }
2413
2414 /*
2415 * Prepare a bandwidth-related upcall
2416 */
2417 static void
bw_meter_prepare_upcall(struct bw_meter * x,struct timeval * nowp)2418 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2419 {
2420 struct timeval delta;
2421 struct bw_upcall *u;
2422
2423 /*
2424 * Compute the measured time interval
2425 */
2426 delta = *nowp;
2427 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2428
2429 /*
2430 * If there are too many pending upcalls, deliver them now
2431 */
2432 if (bw_upcalls_n >= BW_UPCALLS_MAX)
2433 bw_upcalls_send();
2434
2435 /*
2436 * Set the bw_upcall entry
2437 */
2438 u = &bw_upcalls[bw_upcalls_n++];
2439 u->bu_src = x->bm_mfc->mfc_origin;
2440 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2441 u->bu_threshold.b_time = x->bm_threshold.b_time;
2442 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2443 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2444 u->bu_measured.b_time = delta;
2445 u->bu_measured.b_packets = x->bm_measured.b_packets;
2446 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2447 u->bu_flags = 0;
2448 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2449 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2450 if (x->bm_flags & BW_METER_UNIT_BYTES)
2451 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2452 if (x->bm_flags & BW_METER_GEQ)
2453 u->bu_flags |= BW_UPCALL_GEQ;
2454 if (x->bm_flags & BW_METER_LEQ)
2455 u->bu_flags |= BW_UPCALL_LEQ;
2456 }
2457
2458 /*
2459 * Send the pending bandwidth-related upcalls
2460 */
2461 static void
bw_upcalls_send(void)2462 bw_upcalls_send(void)
2463 {
2464 struct mbuf *m;
2465 int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2466 struct sockaddr_in k_igmpsrc = {
2467 .sin_len = sizeof(k_igmpsrc),
2468 .sin_family = AF_INET,
2469 };
2470 static struct igmpmsg igmpmsg = {
2471 0, /* unused1 */
2472 0, /* unused2 */
2473 IGMPMSG_BW_UPCALL,/* im_msgtype */
2474 0, /* im_mbz */
2475 0, /* im_vif */
2476 0, /* unused3 */
2477 { 0 }, /* im_src */
2478 { 0 } /* im_dst */
2479 };
2480
2481 if (bw_upcalls_n == 0)
2482 return; /* No pending upcalls */
2483
2484 bw_upcalls_n = 0;
2485
2486 /*
2487 * Allocate a new mbuf, initialize it with the header and
2488 * the payload for the pending calls.
2489 */
2490 MGETHDR(m, M_DONTWAIT, MT_HEADER);
2491 if (m == NULL) {
2492 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2493 return;
2494 }
2495
2496 m->m_len = m->m_pkthdr.len = 0;
2497 m_copyback(m, 0, sizeof(struct igmpmsg), (void *)&igmpmsg);
2498 m_copyback(m, sizeof(struct igmpmsg), len, (void *)&bw_upcalls[0]);
2499
2500 /*
2501 * Send the upcalls
2502 * XXX do we need to set the address in k_igmpsrc ?
2503 */
2504 mrtstat.mrts_upcalls++;
2505 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2506 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2507 ++mrtstat.mrts_upq_sockfull;
2508 }
2509 }
2510
2511 /*
2512 * Compute the timeout hash value for the bw_meter entries
2513 */
2514 #define BW_METER_TIMEHASH(bw_meter, hash) \
2515 do { \
2516 struct timeval next_timeval = (bw_meter)->bm_start_time; \
2517 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2518 (hash) = next_timeval.tv_sec; \
2519 if (next_timeval.tv_usec) \
2520 (hash)++; /* XXX: make sure we don't timeout early */ \
2521 (hash) %= BW_METER_BUCKETS; \
2522 } while (/*CONSTCOND*/ 0)
2523
2524 /*
2525 * Schedule a timer to process periodically bw_meter entry of type "<="
2526 * by linking the entry in the proper hash bucket.
2527 */
2528 static void
schedule_bw_meter(struct bw_meter * x,struct timeval * nowp)2529 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2530 {
2531 int time_hash;
2532
2533 if (!(x->bm_flags & BW_METER_LEQ))
2534 return; /* XXX: we schedule timers only for "<=" entries */
2535
2536 /*
2537 * Reset the bw_meter entry
2538 */
2539 x->bm_start_time = *nowp;
2540 x->bm_measured.b_packets = 0;
2541 x->bm_measured.b_bytes = 0;
2542 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2543
2544 /*
2545 * Compute the timeout hash value and insert the entry
2546 */
2547 BW_METER_TIMEHASH(x, time_hash);
2548 x->bm_time_next = bw_meter_timers[time_hash];
2549 bw_meter_timers[time_hash] = x;
2550 x->bm_time_hash = time_hash;
2551 }
2552
2553 /*
2554 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2555 * by removing the entry from the proper hash bucket.
2556 */
2557 static void
unschedule_bw_meter(struct bw_meter * x)2558 unschedule_bw_meter(struct bw_meter *x)
2559 {
2560 int time_hash;
2561 struct bw_meter *prev, *tmp;
2562
2563 if (!(x->bm_flags & BW_METER_LEQ))
2564 return; /* XXX: we schedule timers only for "<=" entries */
2565
2566 /*
2567 * Compute the timeout hash value and delete the entry
2568 */
2569 time_hash = x->bm_time_hash;
2570 if (time_hash >= BW_METER_BUCKETS)
2571 return; /* Entry was not scheduled */
2572
2573 for (prev = NULL, tmp = bw_meter_timers[time_hash];
2574 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2575 if (tmp == x)
2576 break;
2577
2578 if (tmp == NULL)
2579 panic("unschedule_bw_meter: bw_meter entry not found");
2580
2581 if (prev != NULL)
2582 prev->bm_time_next = x->bm_time_next;
2583 else
2584 bw_meter_timers[time_hash] = x->bm_time_next;
2585
2586 x->bm_time_next = NULL;
2587 x->bm_time_hash = BW_METER_BUCKETS;
2588 }
2589
2590 /*
2591 * Process all "<=" type of bw_meter that should be processed now,
2592 * and for each entry prepare an upcall if necessary. Each processed
2593 * entry is rescheduled again for the (periodic) processing.
2594 *
2595 * This is run periodically (once per second normally). On each round,
2596 * all the potentially matching entries are in the hash slot that we are
2597 * looking at.
2598 */
2599 static void
bw_meter_process(void)2600 bw_meter_process(void)
2601 {
2602 int s;
2603 static uint32_t last_tv_sec; /* last time we processed this */
2604
2605 uint32_t loops;
2606 int i;
2607 struct timeval now, process_endtime;
2608
2609 microtime(&now);
2610 if (last_tv_sec == now.tv_sec)
2611 return; /* nothing to do */
2612
2613 loops = now.tv_sec - last_tv_sec;
2614 last_tv_sec = now.tv_sec;
2615 if (loops > BW_METER_BUCKETS)
2616 loops = BW_METER_BUCKETS;
2617
2618 s = splsoftnet();
2619 /*
2620 * Process all bins of bw_meter entries from the one after the last
2621 * processed to the current one. On entry, i points to the last bucket
2622 * visited, so we need to increment i at the beginning of the loop.
2623 */
2624 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2625 struct bw_meter *x, *tmp_list;
2626
2627 if (++i >= BW_METER_BUCKETS)
2628 i = 0;
2629
2630 /* Disconnect the list of bw_meter entries from the bin */
2631 tmp_list = bw_meter_timers[i];
2632 bw_meter_timers[i] = NULL;
2633
2634 /* Process the list of bw_meter entries */
2635 while (tmp_list != NULL) {
2636 x = tmp_list;
2637 tmp_list = tmp_list->bm_time_next;
2638
2639 /* Test if the time interval is over */
2640 process_endtime = x->bm_start_time;
2641 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2642 if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2643 /* Not yet: reschedule, but don't reset */
2644 int time_hash;
2645
2646 BW_METER_TIMEHASH(x, time_hash);
2647 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2648 /*
2649 * XXX: somehow the bin processing is a bit ahead of time.
2650 * Put the entry in the next bin.
2651 */
2652 if (++time_hash >= BW_METER_BUCKETS)
2653 time_hash = 0;
2654 }
2655 x->bm_time_next = bw_meter_timers[time_hash];
2656 bw_meter_timers[time_hash] = x;
2657 x->bm_time_hash = time_hash;
2658
2659 continue;
2660 }
2661
2662 /*
2663 * Test if we should deliver an upcall
2664 */
2665 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2666 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2667 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2668 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2669 /* Prepare an upcall for delivery */
2670 bw_meter_prepare_upcall(x, &now);
2671 }
2672
2673 /*
2674 * Reschedule for next processing
2675 */
2676 schedule_bw_meter(x, &now);
2677 }
2678 }
2679
2680 /* Send all upcalls that are pending delivery */
2681 bw_upcalls_send();
2682
2683 splx(s);
2684 }
2685
2686 /*
2687 * A periodic function for sending all upcalls that are pending delivery
2688 */
2689 static void
expire_bw_upcalls_send(void * unused)2690 expire_bw_upcalls_send(void *unused)
2691 {
2692 int s;
2693
2694 s = splsoftnet();
2695 bw_upcalls_send();
2696 splx(s);
2697
2698 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
2699 expire_bw_upcalls_send, NULL);
2700 }
2701
2702 /*
2703 * A periodic function for periodic scanning of the multicast forwarding
2704 * table for processing all "<=" bw_meter entries.
2705 */
2706 static void
expire_bw_meter_process(void * unused)2707 expire_bw_meter_process(void *unused)
2708 {
2709 if (mrt_api_config & MRT_MFC_BW_UPCALL)
2710 bw_meter_process();
2711
2712 callout_reset(&bw_meter_ch, BW_METER_PERIOD,
2713 expire_bw_meter_process, NULL);
2714 }
2715
2716 /*
2717 * End of bandwidth monitoring code
2718 */
2719
2720 #ifdef PIM
2721 /*
2722 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2723 */
2724 static int
pim_register_send(struct ip * ip,struct vif * vifp,struct mbuf * m,struct mfc * rt)2725 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2726 struct mfc *rt)
2727 {
2728 struct mbuf *mb_copy, *mm;
2729
2730 if (mrtdebug & DEBUG_PIM)
2731 log(LOG_DEBUG, "pim_register_send: \n");
2732
2733 mb_copy = pim_register_prepare(ip, m);
2734 if (mb_copy == NULL)
2735 return ENOBUFS;
2736
2737 /*
2738 * Send all the fragments. Note that the mbuf for each fragment
2739 * is freed by the sending machinery.
2740 */
2741 for (mm = mb_copy; mm; mm = mb_copy) {
2742 mb_copy = mm->m_nextpkt;
2743 mm->m_nextpkt = NULL;
2744 mm = m_pullup(mm, sizeof(struct ip));
2745 if (mm != NULL) {
2746 ip = mtod(mm, struct ip *);
2747 if ((mrt_api_config & MRT_MFC_RP) &&
2748 !in_nullhost(rt->mfc_rp)) {
2749 pim_register_send_rp(ip, vifp, mm, rt);
2750 } else {
2751 pim_register_send_upcall(ip, vifp, mm, rt);
2752 }
2753 }
2754 }
2755
2756 return 0;
2757 }
2758
2759 /*
2760 * Return a copy of the data packet that is ready for PIM Register
2761 * encapsulation.
2762 * XXX: Note that in the returned copy the IP header is a valid one.
2763 */
2764 static struct mbuf *
pim_register_prepare(struct ip * ip,struct mbuf * m)2765 pim_register_prepare(struct ip *ip, struct mbuf *m)
2766 {
2767 struct mbuf *mb_copy = NULL;
2768 int mtu;
2769
2770 /* Take care of delayed checksums */
2771 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
2772 in_undefer_cksum_tcpudp(m);
2773 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
2774 }
2775
2776 /*
2777 * Copy the old packet & pullup its IP header into the
2778 * new mbuf so we can modify it.
2779 */
2780 mb_copy = m_copypacket(m, M_DONTWAIT);
2781 if (mb_copy == NULL)
2782 return NULL;
2783 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2784 if (mb_copy == NULL)
2785 return NULL;
2786
2787 /* take care of the TTL */
2788 ip = mtod(mb_copy, struct ip *);
2789 --ip->ip_ttl;
2790
2791 /* Compute the MTU after the PIM Register encapsulation */
2792 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2793
2794 if (ntohs(ip->ip_len) <= mtu) {
2795 /* Turn the IP header into a valid one */
2796 ip->ip_sum = 0;
2797 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2798 } else {
2799 /* Fragment the packet */
2800 if (ip_fragment(mb_copy, NULL, mtu) != 0) {
2801 /* XXX: mb_copy was freed by ip_fragment() */
2802 return NULL;
2803 }
2804 }
2805 return mb_copy;
2806 }
2807
2808 /*
2809 * Send an upcall with the data packet to the user-level process.
2810 */
2811 static int
pim_register_send_upcall(struct ip * ip,struct vif * vifp,struct mbuf * mb_copy,struct mfc * rt)2812 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2813 struct mbuf *mb_copy, struct mfc *rt)
2814 {
2815 struct mbuf *mb_first;
2816 int len = ntohs(ip->ip_len);
2817 struct igmpmsg *im;
2818 struct sockaddr_in k_igmpsrc = {
2819 .sin_len = sizeof(k_igmpsrc),
2820 .sin_family = AF_INET,
2821 };
2822
2823 /*
2824 * Add a new mbuf with an upcall header
2825 */
2826 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
2827 if (mb_first == NULL) {
2828 m_freem(mb_copy);
2829 return ENOBUFS;
2830 }
2831 mb_first->m_data += max_linkhdr;
2832 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2833 mb_first->m_len = sizeof(struct igmpmsg);
2834 mb_first->m_next = mb_copy;
2835
2836 /* Send message to routing daemon */
2837 im = mtod(mb_first, struct igmpmsg *);
2838 im->im_msgtype = IGMPMSG_WHOLEPKT;
2839 im->im_mbz = 0;
2840 im->im_vif = vifp - viftable;
2841 im->im_src = ip->ip_src;
2842 im->im_dst = ip->ip_dst;
2843
2844 k_igmpsrc.sin_addr = ip->ip_src;
2845
2846 mrtstat.mrts_upcalls++;
2847
2848 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2849 if (mrtdebug & DEBUG_PIM)
2850 log(LOG_WARNING,
2851 "mcast: pim_register_send_upcall: ip_mrouter socket queue full\n");
2852 ++mrtstat.mrts_upq_sockfull;
2853 return ENOBUFS;
2854 }
2855
2856 /* Keep statistics */
2857 pimstat.pims_snd_registers_msgs++;
2858 pimstat.pims_snd_registers_bytes += len;
2859
2860 return 0;
2861 }
2862
2863 /*
2864 * Encapsulate the data packet in PIM Register message and send it to the RP.
2865 */
2866 static int
pim_register_send_rp(struct ip * ip,struct vif * vifp,struct mbuf * mb_copy,struct mfc * rt)2867 pim_register_send_rp(struct ip *ip, struct vif *vifp,
2868 struct mbuf *mb_copy, struct mfc *rt)
2869 {
2870 struct mbuf *mb_first;
2871 struct ip *ip_outer;
2872 struct pim_encap_pimhdr *pimhdr;
2873 int len = ntohs(ip->ip_len);
2874 vifi_t vifi = rt->mfc_parent;
2875
2876 if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
2877 m_freem(mb_copy);
2878 return EADDRNOTAVAIL; /* The iif vif is invalid */
2879 }
2880
2881 /*
2882 * Add a new mbuf with the encapsulating header
2883 */
2884 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
2885 if (mb_first == NULL) {
2886 m_freem(mb_copy);
2887 return ENOBUFS;
2888 }
2889 mb_first->m_data += max_linkhdr;
2890 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2891 mb_first->m_next = mb_copy;
2892
2893 mb_first->m_pkthdr.len = len + mb_first->m_len;
2894
2895 /*
2896 * Fill in the encapsulating IP and PIM header
2897 */
2898 ip_outer = mtod(mb_first, struct ip *);
2899 *ip_outer = pim_encap_iphdr;
2900 if (mb_first->m_pkthdr.len < IP_MINFRAGSIZE)
2901 ip_outer->ip_id = 0;
2902 else
2903 ip_outer->ip_id = ip_newid(NULL);
2904 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2905 sizeof(pim_encap_pimhdr));
2906 ip_outer->ip_src = viftable[vifi].v_lcl_addr;
2907 ip_outer->ip_dst = rt->mfc_rp;
2908 /*
2909 * Copy the inner header TOS to the outer header, and take care of the
2910 * IP_DF bit.
2911 */
2912 ip_outer->ip_tos = ip->ip_tos;
2913 if (ntohs(ip->ip_off) & IP_DF)
2914 ip_outer->ip_off |= htons(IP_DF);
2915 pimhdr = (struct pim_encap_pimhdr *)((char *)ip_outer
2916 + sizeof(pim_encap_iphdr));
2917 *pimhdr = pim_encap_pimhdr;
2918 /* If the iif crosses a border, set the Border-bit */
2919 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
2920 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2921
2922 mb_first->m_data += sizeof(pim_encap_iphdr);
2923 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2924 mb_first->m_data -= sizeof(pim_encap_iphdr);
2925
2926 if (vifp->v_rate_limit == 0)
2927 tbf_send_packet(vifp, mb_first);
2928 else
2929 tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len));
2930
2931 /* Keep statistics */
2932 pimstat.pims_snd_registers_msgs++;
2933 pimstat.pims_snd_registers_bytes += len;
2934
2935 return 0;
2936 }
2937
2938 /*
2939 * PIM-SMv2 and PIM-DM messages processing.
2940 * Receives and verifies the PIM control messages, and passes them
2941 * up to the listening socket, using rip_input().
2942 * The only message with special processing is the PIM_REGISTER message
2943 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2944 * is passed to if_simloop().
2945 */
2946 void
pim_input(struct mbuf * m,int off,int proto)2947 pim_input(struct mbuf *m, int off, int proto)
2948 {
2949 struct ip *ip = mtod(m, struct ip *);
2950 struct pim *pim;
2951 int minlen;
2952 int datalen;
2953 int ip_tos;
2954 int iphlen;
2955
2956 iphlen = off;
2957 datalen = ntohs(ip->ip_len) - iphlen;
2958
2959 /* Keep statistics */
2960 pimstat.pims_rcv_total_msgs++;
2961 pimstat.pims_rcv_total_bytes += datalen;
2962
2963 /*
2964 * Validate lengths
2965 */
2966 if (datalen < PIM_MINLEN) {
2967 pimstat.pims_rcv_tooshort++;
2968 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
2969 datalen, (u_long)ip->ip_src.s_addr);
2970 m_freem(m);
2971 return;
2972 }
2973
2974 /*
2975 * If the packet is at least as big as a REGISTER, go ahead
2976 * and grab the PIM REGISTER header size, to avoid another
2977 * possible m_pullup() later.
2978 *
2979 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2980 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2981 */
2982 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2983
2984 /*
2985 * Get the IP and PIM headers in contiguous memory, and
2986 * possibly the PIM REGISTER header.
2987 */
2988 if ((m->m_flags & M_EXT || m->m_len < minlen) &&
2989 (m = m_pullup(m, minlen)) == NULL) {
2990 log(LOG_ERR, "pim_input: m_pullup failure\n");
2991 return;
2992 }
2993 ip = mtod(m, struct ip *);
2994 ip_tos = ip->ip_tos;
2995
2996 /* adjust mbuf to point to the PIM header */
2997 m->m_data += iphlen;
2998 m->m_len -= iphlen;
2999 pim = mtod(m, struct pim *);
3000
3001 /*
3002 * Validate checksum. If PIM REGISTER, exclude the data packet.
3003 *
3004 * XXX: some older PIMv2 implementations don't make this distinction,
3005 * so for compatibility reason perform the checksum over part of the
3006 * message, and if error, then over the whole message.
3007 */
3008 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3009 /* do nothing, checksum okay */
3010 } else if (in_cksum(m, datalen)) {
3011 pimstat.pims_rcv_badsum++;
3012 if (mrtdebug & DEBUG_PIM)
3013 log(LOG_DEBUG, "pim_input: invalid checksum\n");
3014 m_freem(m);
3015 return;
3016 }
3017
3018 /* PIM version check */
3019 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3020 pimstat.pims_rcv_badversion++;
3021 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3022 PIM_VT_V(pim->pim_vt), PIM_VERSION);
3023 m_freem(m);
3024 return;
3025 }
3026
3027 /* restore mbuf back to the outer IP */
3028 m->m_data -= iphlen;
3029 m->m_len += iphlen;
3030
3031 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3032 /*
3033 * Since this is a REGISTER, we'll make a copy of the register
3034 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3035 * routing daemon.
3036 */
3037 int s;
3038 struct sockaddr_in dst = {
3039 .sin_len = sizeof(dst),
3040 .sin_family = AF_INET,
3041 };
3042 struct mbuf *mcp;
3043 struct ip *encap_ip;
3044 u_int32_t *reghdr;
3045 struct ifnet *vifp;
3046
3047 s = splsoftnet();
3048 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3049 splx(s);
3050 if (mrtdebug & DEBUG_PIM)
3051 log(LOG_DEBUG,
3052 "pim_input: register vif not set: %d\n", reg_vif_num);
3053 m_freem(m);
3054 return;
3055 }
3056 /* XXX need refcnt? */
3057 vifp = viftable[reg_vif_num].v_ifp;
3058 splx(s);
3059
3060 /*
3061 * Validate length
3062 */
3063 if (datalen < PIM_REG_MINLEN) {
3064 pimstat.pims_rcv_tooshort++;
3065 pimstat.pims_rcv_badregisters++;
3066 log(LOG_ERR,
3067 "pim_input: register packet size too small %d from %lx\n",
3068 datalen, (u_long)ip->ip_src.s_addr);
3069 m_freem(m);
3070 return;
3071 }
3072
3073 reghdr = (u_int32_t *)(pim + 1);
3074 encap_ip = (struct ip *)(reghdr + 1);
3075
3076 if (mrtdebug & DEBUG_PIM) {
3077 log(LOG_DEBUG,
3078 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3079 (u_long)ntohl(encap_ip->ip_src.s_addr),
3080 (u_long)ntohl(encap_ip->ip_dst.s_addr),
3081 ntohs(encap_ip->ip_len));
3082 }
3083
3084 /* verify the version number of the inner packet */
3085 if (encap_ip->ip_v != IPVERSION) {
3086 pimstat.pims_rcv_badregisters++;
3087 if (mrtdebug & DEBUG_PIM) {
3088 log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3089 "of the inner packet\n", encap_ip->ip_v);
3090 }
3091 m_freem(m);
3092 return;
3093 }
3094
3095 /* verify the inner packet doesn't have options */
3096 if (encap_ip->ip_hl != (sizeof(struct ip) >> 2)) {
3097 pimstat.pims_rcv_badregisters++;
3098 m_freem(m);
3099 return;
3100 }
3101
3102 /* verify the inner packet is destined to a mcast group */
3103 if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) {
3104 pimstat.pims_rcv_badregisters++;
3105 if (mrtdebug & DEBUG_PIM)
3106 log(LOG_DEBUG,
3107 "pim_input: inner packet of register is not "
3108 "multicast %lx\n",
3109 (u_long)ntohl(encap_ip->ip_dst.s_addr));
3110 m_freem(m);
3111 return;
3112 }
3113
3114 /* If a NULL_REGISTER, pass it to the daemon */
3115 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3116 goto pim_input_to_daemon;
3117
3118 /*
3119 * Copy the TOS from the outer IP header to the inner IP header.
3120 */
3121 if (encap_ip->ip_tos != ip_tos) {
3122 /* Outer TOS -> inner TOS */
3123 encap_ip->ip_tos = ip_tos;
3124 /* Recompute the inner header checksum. Sigh... */
3125
3126 /* adjust mbuf to point to the inner IP header */
3127 m->m_data += (iphlen + PIM_MINLEN);
3128 m->m_len -= (iphlen + PIM_MINLEN);
3129
3130 encap_ip->ip_sum = 0;
3131 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3132
3133 /* restore mbuf to point back to the outer IP header */
3134 m->m_data -= (iphlen + PIM_MINLEN);
3135 m->m_len += (iphlen + PIM_MINLEN);
3136 }
3137
3138 /*
3139 * Decapsulate the inner IP packet and loopback to forward it
3140 * as a normal multicast packet. Also, make a copy of the
3141 * outer_iphdr + pimhdr + reghdr + encap_iphdr
3142 * to pass to the daemon later, so it can take the appropriate
3143 * actions (e.g., send back PIM_REGISTER_STOP).
3144 * XXX: here m->m_data points to the outer IP header.
3145 */
3146 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_DONTWAIT);
3147 if (mcp == NULL) {
3148 log(LOG_ERR,
3149 "pim_input: pim register: could not copy register head\n");
3150 m_freem(m);
3151 return;
3152 }
3153
3154 /* Keep statistics */
3155 /* XXX: registers_bytes include only the encap. mcast pkt */
3156 pimstat.pims_rcv_registers_msgs++;
3157 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3158
3159 /*
3160 * forward the inner ip packet; point m_data at the inner ip.
3161 */
3162 m_adj(m, iphlen + PIM_MINLEN);
3163
3164 if (mrtdebug & DEBUG_PIM) {
3165 log(LOG_DEBUG,
3166 "pim_input: forwarding decapsulated register: "
3167 "src %lx, dst %lx, vif %d\n",
3168 (u_long)ntohl(encap_ip->ip_src.s_addr),
3169 (u_long)ntohl(encap_ip->ip_dst.s_addr),
3170 reg_vif_num);
3171 }
3172 /* NB: vifp was collected above; can it change on us? */
3173 looutput(vifp, m, (struct sockaddr *)&dst, NULL);
3174
3175 /* prepare the register head to send to the mrouting daemon */
3176 m = mcp;
3177 }
3178
3179 pim_input_to_daemon:
3180 /*
3181 * Pass the PIM message up to the daemon; if it is a Register message,
3182 * pass the 'head' only up to the daemon. This includes the
3183 * outer IP header, PIM header, PIM-Register header and the
3184 * inner IP header.
3185 * XXX: the outer IP header pkt size of a Register is not adjust to
3186 * reflect the fact that the inner multicast data is truncated.
3187 */
3188 /*
3189 * Currently, pim_input() is always called holding softnet_lock
3190 * by ipintr()(!NET_MPSAFE) or PR_INPUT_WRAP()(NET_MPSAFE).
3191 */
3192 KASSERT(mutex_owned(softnet_lock));
3193 rip_input(m, iphlen, proto);
3194
3195 return;
3196 }
3197 #endif /* PIM */
3198