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
56 /*
57 * TODO: Prefix functions with ipmf_.
58 * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
59 * domain attachment (if_afdata) so we can track consumers of that service.
60 * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
61 * move it to socket options.
62 * TODO: Cleanup LSRR removal further.
63 * TODO: Push RSVP stubs into raw_ip.c.
64 * TODO: Use bitstring.h for vif set.
65 * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
66 * TODO: Sync ip6_mroute.c with this file.
67 */
68
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
71
72 #include "opt_inet.h"
73 #include "opt_mrouting.h"
74
75 #define _PIM_VT 1
76
77 #include <sys/param.h>
78 #include <sys/kernel.h>
79 #include <sys/stddef.h>
80 #include <sys/lock.h>
81 #include <sys/ktr.h>
82 #include <sys/malloc.h>
83 #include <sys/mbuf.h>
84 #include <sys/module.h>
85 #include <sys/priv.h>
86 #include <sys/protosw.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/socketvar.h>
90 #include <sys/sockio.h>
91 #include <sys/sx.h>
92 #include <sys/sysctl.h>
93 #include <sys/syslog.h>
94 #include <sys/systm.h>
95 #include <sys/time.h>
96
97 #include <net/if.h>
98 #include <net/netisr.h>
99 #include <net/route.h>
100 #include <net/vnet.h>
101
102 #include <netinet/in.h>
103 #include <netinet/igmp.h>
104 #include <netinet/in_systm.h>
105 #include <netinet/in_var.h>
106 #include <netinet/ip.h>
107 #include <netinet/ip_encap.h>
108 #include <netinet/ip_mroute.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/ip_options.h>
111 #include <netinet/pim.h>
112 #include <netinet/pim_var.h>
113 #include <netinet/udp.h>
114
115 #include <machine/in_cksum.h>
116
117 #include <security/mac/mac_framework.h>
118
119 #ifndef KTR_IPMF
120 #define KTR_IPMF KTR_INET
121 #endif
122
123 #define VIFI_INVALID ((vifi_t) -1)
124 #define M_HASCL(m) ((m)->m_flags & M_EXT)
125
126 static VNET_DEFINE(uint32_t, last_tv_sec); /* last time we processed this */
127 #define V_last_tv_sec VNET(last_tv_sec)
128
129 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
130
131 /*
132 * Locking. We use two locks: one for the virtual interface table and
133 * one for the forwarding table. These locks may be nested in which case
134 * the VIF lock must always be taken first. Note that each lock is used
135 * to cover not only the specific data structure but also related data
136 * structures.
137 */
138
139 static struct mtx mrouter_mtx;
140 #define MROUTER_LOCK() mtx_lock(&mrouter_mtx)
141 #define MROUTER_UNLOCK() mtx_unlock(&mrouter_mtx)
142 #define MROUTER_LOCK_ASSERT() mtx_assert(&mrouter_mtx, MA_OWNED)
143 #define MROUTER_LOCK_INIT() \
144 mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
145 #define MROUTER_LOCK_DESTROY() mtx_destroy(&mrouter_mtx)
146
147 static int ip_mrouter_cnt; /* # of vnets with active mrouters */
148 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
149
150 static VNET_DEFINE(struct mrtstat, mrtstat);
151 #define V_mrtstat VNET(mrtstat)
152 SYSCTL_VNET_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
153 &VNET_NAME(mrtstat), mrtstat,
154 "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
155 "netinet/ip_mroute.h)");
156
157 static VNET_DEFINE(u_long, mfchash);
158 #define V_mfchash VNET(mfchash)
159 #define MFCHASH(a, g) \
160 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
161 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
162 #define MFCHASHSIZE 256
163
164 static u_long mfchashsize; /* Hash size */
165 static VNET_DEFINE(u_char *, nexpire); /* 0..mfchashsize-1 */
166 #define V_nexpire VNET(nexpire)
167 static VNET_DEFINE(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
168 #define V_mfchashtbl VNET(mfchashtbl)
169
170 static struct mtx mfc_mtx;
171 #define MFC_LOCK() mtx_lock(&mfc_mtx)
172 #define MFC_UNLOCK() mtx_unlock(&mfc_mtx)
173 #define MFC_LOCK_ASSERT() mtx_assert(&mfc_mtx, MA_OWNED)
174 #define MFC_LOCK_INIT() \
175 mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
176 #define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx)
177
178 static VNET_DEFINE(vifi_t, numvifs);
179 #define V_numvifs VNET(numvifs)
180 static VNET_DEFINE(struct vif, viftable[MAXVIFS]);
181 #define V_viftable VNET(viftable)
182 SYSCTL_VNET_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
183 &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]",
184 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
185
186 static struct mtx vif_mtx;
187 #define VIF_LOCK() mtx_lock(&vif_mtx)
188 #define VIF_UNLOCK() mtx_unlock(&vif_mtx)
189 #define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED)
190 #define VIF_LOCK_INIT() \
191 mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
192 #define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx)
193
194 static eventhandler_tag if_detach_event_tag = NULL;
195
196 static VNET_DEFINE(struct callout, expire_upcalls_ch);
197 #define V_expire_upcalls_ch VNET(expire_upcalls_ch)
198
199 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
200 #define UPCALL_EXPIRE 6 /* number of timeouts */
201
202 /*
203 * Bandwidth meter variables and constants
204 */
205 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
206 /*
207 * Pending timeouts are stored in a hash table, the key being the
208 * expiration time. Periodically, the entries are analysed and processed.
209 */
210 #define BW_METER_BUCKETS 1024
211 static VNET_DEFINE(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]);
212 #define V_bw_meter_timers VNET(bw_meter_timers)
213 static VNET_DEFINE(struct callout, bw_meter_ch);
214 #define V_bw_meter_ch VNET(bw_meter_ch)
215 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
216
217 /*
218 * Pending upcalls are stored in a vector which is flushed when
219 * full, or periodically
220 */
221 static VNET_DEFINE(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]);
222 #define V_bw_upcalls VNET(bw_upcalls)
223 static VNET_DEFINE(u_int, bw_upcalls_n); /* # of pending upcalls */
224 #define V_bw_upcalls_n VNET(bw_upcalls_n)
225 static VNET_DEFINE(struct callout, bw_upcalls_ch);
226 #define V_bw_upcalls_ch VNET(bw_upcalls_ch)
227
228 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
229
230 static VNET_DEFINE(struct pimstat, pimstat);
231 #define V_pimstat VNET(pimstat)
232
233 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
234 SYSCTL_VNET_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
235 &VNET_NAME(pimstat), pimstat,
236 "PIM Statistics (struct pimstat, netinet/pim_var.h)");
237
238 static u_long pim_squelch_wholepkt = 0;
239 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
240 &pim_squelch_wholepkt, 0,
241 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
242
243 extern struct domain inetdomain;
244 static const struct protosw in_pim_protosw = {
245 .pr_type = SOCK_RAW,
246 .pr_domain = &inetdomain,
247 .pr_protocol = IPPROTO_PIM,
248 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
249 .pr_input = pim_input,
250 .pr_output = (pr_output_t*)rip_output,
251 .pr_ctloutput = rip_ctloutput,
252 .pr_usrreqs = &rip_usrreqs
253 };
254 static const struct encaptab *pim_encap_cookie;
255
256 static int pim_encapcheck(const struct mbuf *, int, int, void *);
257
258 /*
259 * Note: the PIM Register encapsulation adds the following in front of a
260 * data packet:
261 *
262 * struct pim_encap_hdr {
263 * struct ip ip;
264 * struct pim_encap_pimhdr pim;
265 * }
266 *
267 */
268
269 struct pim_encap_pimhdr {
270 struct pim pim;
271 uint32_t flags;
272 };
273 #define PIM_ENCAP_TTL 64
274
275 static struct ip pim_encap_iphdr = {
276 #if BYTE_ORDER == LITTLE_ENDIAN
277 sizeof(struct ip) >> 2,
278 IPVERSION,
279 #else
280 IPVERSION,
281 sizeof(struct ip) >> 2,
282 #endif
283 0, /* tos */
284 sizeof(struct ip), /* total length */
285 0, /* id */
286 0, /* frag offset */
287 PIM_ENCAP_TTL,
288 IPPROTO_PIM,
289 0, /* checksum */
290 };
291
292 static struct pim_encap_pimhdr pim_encap_pimhdr = {
293 {
294 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
295 0, /* reserved */
296 0, /* checksum */
297 },
298 0 /* flags */
299 };
300
301 static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID;
302 #define V_reg_vif_num VNET(reg_vif_num)
303 static VNET_DEFINE(struct ifnet, multicast_register_if);
304 #define V_multicast_register_if VNET(multicast_register_if)
305
306 /*
307 * Private variables.
308 */
309
310 static u_long X_ip_mcast_src(int);
311 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
312 struct ip_moptions *);
313 static int X_ip_mrouter_done(void);
314 static int X_ip_mrouter_get(struct socket *, struct sockopt *);
315 static int X_ip_mrouter_set(struct socket *, struct sockopt *);
316 static int X_legal_vif_num(int);
317 static int X_mrt_ioctl(u_long, caddr_t, int);
318
319 static int add_bw_upcall(struct bw_upcall *);
320 static int add_mfc(struct mfcctl2 *);
321 static int add_vif(struct vifctl *);
322 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
323 static void bw_meter_process(void);
324 static void bw_meter_receive_packet(struct bw_meter *, int,
325 struct timeval *);
326 static void bw_upcalls_send(void);
327 static int del_bw_upcall(struct bw_upcall *);
328 static int del_mfc(struct mfcctl2 *);
329 static int del_vif(vifi_t);
330 static int del_vif_locked(vifi_t);
331 static void expire_bw_meter_process(void *);
332 static void expire_bw_upcalls_send(void *);
333 static void expire_mfc(struct mfc *);
334 static void expire_upcalls(void *);
335 static void free_bw_list(struct bw_meter *);
336 static int get_sg_cnt(struct sioc_sg_req *);
337 static int get_vif_cnt(struct sioc_vif_req *);
338 static void if_detached_event(void *, struct ifnet *);
339 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
340 static int ip_mrouter_init(struct socket *, int);
341 static __inline struct mfc *
342 mfc_find(struct in_addr *, struct in_addr *);
343 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
344 static struct mbuf *
345 pim_register_prepare(struct ip *, struct mbuf *);
346 static int pim_register_send(struct ip *, struct vif *,
347 struct mbuf *, struct mfc *);
348 static int pim_register_send_rp(struct ip *, struct vif *,
349 struct mbuf *, struct mfc *);
350 static int pim_register_send_upcall(struct ip *, struct vif *,
351 struct mbuf *, struct mfc *);
352 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
353 static void send_packet(struct vif *, struct mbuf *);
354 static int set_api_config(uint32_t *);
355 static int set_assert(int);
356 static int socket_send(struct socket *, struct mbuf *,
357 struct sockaddr_in *);
358 static void unschedule_bw_meter(struct bw_meter *);
359
360 /*
361 * Kernel multicast forwarding API capabilities and setup.
362 * If more API capabilities are added to the kernel, they should be
363 * recorded in `mrt_api_support'.
364 */
365 #define MRT_API_VERSION 0x0305
366
367 static const int mrt_api_version = MRT_API_VERSION;
368 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
369 MRT_MFC_FLAGS_BORDER_VIF |
370 MRT_MFC_RP |
371 MRT_MFC_BW_UPCALL);
372 static VNET_DEFINE(uint32_t, mrt_api_config);
373 #define V_mrt_api_config VNET(mrt_api_config)
374 static VNET_DEFINE(int, pim_assert_enabled);
375 #define V_pim_assert_enabled VNET(pim_assert_enabled)
376 static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */
377
378 /*
379 * Find a route for a given origin IP address and multicast group address.
380 * Statistics must be updated by the caller.
381 */
382 static __inline struct mfc *
383 mfc_find(struct in_addr *o, struct in_addr *g)
384 {
385 struct mfc *rt;
386
387 MFC_LOCK_ASSERT();
388
389 LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
390 if (in_hosteq(rt->mfc_origin, *o) &&
391 in_hosteq(rt->mfc_mcastgrp, *g) &&
392 TAILQ_EMPTY(&rt->mfc_stall))
393 break;
394 }
395
396 return (rt);
397 }
398
399 /*
400 * Handle MRT setsockopt commands to modify the multicast forwarding tables.
401 */
402 static int
403 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
404 {
405 int error, optval;
406 vifi_t vifi;
407 struct vifctl vifc;
408 struct mfcctl2 mfc;
409 struct bw_upcall bw_upcall;
410 uint32_t i;
411
412 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
413 return EPERM;
414
415 error = 0;
416 switch (sopt->sopt_name) {
417 case MRT_INIT:
418 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
419 if (error)
420 break;
421 error = ip_mrouter_init(so, optval);
422 break;
423
424 case MRT_DONE:
425 error = ip_mrouter_done();
426 break;
427
428 case MRT_ADD_VIF:
429 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
430 if (error)
431 break;
432 error = add_vif(&vifc);
433 break;
434
435 case MRT_DEL_VIF:
436 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
437 if (error)
438 break;
439 error = del_vif(vifi);
440 break;
441
442 case MRT_ADD_MFC:
443 case MRT_DEL_MFC:
444 /*
445 * select data size depending on API version.
446 */
447 if (sopt->sopt_name == MRT_ADD_MFC &&
448 V_mrt_api_config & MRT_API_FLAGS_ALL) {
449 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
450 sizeof(struct mfcctl2));
451 } else {
452 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
453 sizeof(struct mfcctl));
454 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
455 sizeof(mfc) - sizeof(struct mfcctl));
456 }
457 if (error)
458 break;
459 if (sopt->sopt_name == MRT_ADD_MFC)
460 error = add_mfc(&mfc);
461 else
462 error = del_mfc(&mfc);
463 break;
464
465 case MRT_ASSERT:
466 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
467 if (error)
468 break;
469 set_assert(optval);
470 break;
471
472 case MRT_API_CONFIG:
473 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
474 if (!error)
475 error = set_api_config(&i);
476 if (!error)
477 error = sooptcopyout(sopt, &i, sizeof i);
478 break;
479
480 case MRT_ADD_BW_UPCALL:
481 case MRT_DEL_BW_UPCALL:
482 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
483 sizeof bw_upcall);
484 if (error)
485 break;
486 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
487 error = add_bw_upcall(&bw_upcall);
488 else
489 error = del_bw_upcall(&bw_upcall);
490 break;
491
492 default:
493 error = EOPNOTSUPP;
494 break;
495 }
496 return error;
497 }
498
499 /*
500 * Handle MRT getsockopt commands
501 */
502 static int
503 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
504 {
505 int error;
506
507 switch (sopt->sopt_name) {
508 case MRT_VERSION:
509 error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
510 break;
511
512 case MRT_ASSERT:
513 error = sooptcopyout(sopt, &V_pim_assert_enabled,
514 sizeof V_pim_assert_enabled);
515 break;
516
517 case MRT_API_SUPPORT:
518 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
519 break;
520
521 case MRT_API_CONFIG:
522 error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
523 break;
524
525 default:
526 error = EOPNOTSUPP;
527 break;
528 }
529 return error;
530 }
531
532 /*
533 * Handle ioctl commands to obtain information from the cache
534 */
535 static int
536 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
537 {
538 int error = 0;
539
540 /*
541 * Currently the only function calling this ioctl routine is rtioctl().
542 * Typically, only root can create the raw socket in order to execute
543 * this ioctl method, however the request might be coming from a prison
544 */
545 error = priv_check(curthread, PRIV_NETINET_MROUTE);
546 if (error)
547 return (error);
548 switch (cmd) {
549 case (SIOCGETVIFCNT):
550 error = get_vif_cnt((struct sioc_vif_req *)data);
551 break;
552
553 case (SIOCGETSGCNT):
554 error = get_sg_cnt((struct sioc_sg_req *)data);
555 break;
556
557 default:
558 error = EINVAL;
559 break;
560 }
561 return error;
562 }
563
564 /*
565 * returns the packet, byte, rpf-failure count for the source group provided
566 */
567 static int
568 get_sg_cnt(struct sioc_sg_req *req)
569 {
570 struct mfc *rt;
571
572 MFC_LOCK();
573 rt = mfc_find(&req->src, &req->grp);
574 if (rt == NULL) {
575 MFC_UNLOCK();
576 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
577 return EADDRNOTAVAIL;
578 }
579 req->pktcnt = rt->mfc_pkt_cnt;
580 req->bytecnt = rt->mfc_byte_cnt;
581 req->wrong_if = rt->mfc_wrong_if;
582 MFC_UNLOCK();
583 return 0;
584 }
585
586 /*
587 * returns the input and output packet and byte counts on the vif provided
588 */
589 static int
590 get_vif_cnt(struct sioc_vif_req *req)
591 {
592 vifi_t vifi = req->vifi;
593
594 VIF_LOCK();
595 if (vifi >= V_numvifs) {
596 VIF_UNLOCK();
597 return EINVAL;
598 }
599
600 req->icount = V_viftable[vifi].v_pkt_in;
601 req->ocount = V_viftable[vifi].v_pkt_out;
602 req->ibytes = V_viftable[vifi].v_bytes_in;
603 req->obytes = V_viftable[vifi].v_bytes_out;
604 VIF_UNLOCK();
605
606 return 0;
607 }
608
609 static void
610 if_detached_event(void *arg __unused, struct ifnet *ifp)
611 {
612 vifi_t vifi;
613 u_long i;
614
615 MROUTER_LOCK();
616
617 if (V_ip_mrouter == NULL) {
618 MROUTER_UNLOCK();
619 return;
620 }
621
622 VIF_LOCK();
623 MFC_LOCK();
624
625 /*
626 * Tear down multicast forwarder state associated with this ifnet.
627 * 1. Walk the vif list, matching vifs against this ifnet.
628 * 2. Walk the multicast forwarding cache (mfc) looking for
629 * inner matches with this vif's index.
630 * 3. Expire any matching multicast forwarding cache entries.
631 * 4. Free vif state. This should disable ALLMULTI on the interface.
632 */
633 for (vifi = 0; vifi < V_numvifs; vifi++) {
634 if (V_viftable[vifi].v_ifp != ifp)
635 continue;
636 for (i = 0; i < mfchashsize; i++) {
637 struct mfc *rt, *nrt;
638 for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
639 nrt = LIST_NEXT(rt, mfc_hash);
640 if (rt->mfc_parent == vifi) {
641 expire_mfc(rt);
642 }
643 }
644 }
645 del_vif_locked(vifi);
646 }
647
648 MFC_UNLOCK();
649 VIF_UNLOCK();
650
651 MROUTER_UNLOCK();
652 }
653
654 /*
655 * Enable multicast forwarding.
656 */
657 static int
658 ip_mrouter_init(struct socket *so, int version)
659 {
660
661 CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
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 MROUTER_LOCK();
671
672 if (ip_mrouter_unloading) {
673 MROUTER_UNLOCK();
674 return ENOPROTOOPT;
675 }
676
677 if (V_ip_mrouter != NULL) {
678 MROUTER_UNLOCK();
679 return EADDRINUSE;
680 }
681
682 V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
683 HASH_NOWAIT);
684
685 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
686 curvnet);
687 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
688 curvnet);
689 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
690 curvnet);
691
692 V_ip_mrouter = so;
693 ip_mrouter_cnt++;
694
695 MROUTER_UNLOCK();
696
697 CTR1(KTR_IPMF, "%s: done", __func__);
698
699 return 0;
700 }
701
702 /*
703 * Disable multicast forwarding.
704 */
705 static int
706 X_ip_mrouter_done(void)
707 {
708 struct ifnet *ifp;
709 u_long i;
710 vifi_t vifi;
711
712 MROUTER_LOCK();
713
714 if (V_ip_mrouter == NULL) {
715 MROUTER_UNLOCK();
716 return EINVAL;
717 }
718
719 /*
720 * Detach/disable hooks to the reset of the system.
721 */
722 V_ip_mrouter = NULL;
723 ip_mrouter_cnt--;
724 V_mrt_api_config = 0;
725
726 VIF_LOCK();
727
728 /*
729 * For each phyint in use, disable promiscuous reception of all IP
730 * multicasts.
731 */
732 for (vifi = 0; vifi < V_numvifs; vifi++) {
733 if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
734 !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
735 ifp = V_viftable[vifi].v_ifp;
736 if_allmulti(ifp, 0);
737 }
738 }
739 bzero((caddr_t)V_viftable, sizeof(V_viftable));
740 V_numvifs = 0;
741 V_pim_assert_enabled = 0;
742
743 VIF_UNLOCK();
744
745 callout_stop(&V_expire_upcalls_ch);
746 callout_stop(&V_bw_upcalls_ch);
747 callout_stop(&V_bw_meter_ch);
748
749 MFC_LOCK();
750
751 /*
752 * Free all multicast forwarding cache entries.
753 * Do not use hashdestroy(), as we must perform other cleanup.
754 */
755 for (i = 0; i < mfchashsize; i++) {
756 struct mfc *rt, *nrt;
757 for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
758 nrt = LIST_NEXT(rt, mfc_hash);
759 expire_mfc(rt);
760 }
761 }
762 free(V_mfchashtbl, M_MRTABLE);
763 V_mfchashtbl = NULL;
764
765 bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
766
767 V_bw_upcalls_n = 0;
768 bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
769
770 MFC_UNLOCK();
771
772 V_reg_vif_num = VIFI_INVALID;
773
774 MROUTER_UNLOCK();
775
776 CTR1(KTR_IPMF, "%s: done", __func__);
777
778 return 0;
779 }
780
781 /*
782 * Set PIM assert processing global
783 */
784 static int
785 set_assert(int i)
786 {
787 if ((i != 1) && (i != 0))
788 return EINVAL;
789
790 V_pim_assert_enabled = i;
791
792 return 0;
793 }
794
795 /*
796 * Configure API capabilities
797 */
798 int
799 set_api_config(uint32_t *apival)
800 {
801 u_long i;
802
803 /*
804 * We can set the API capabilities only if it is the first operation
805 * after MRT_INIT. I.e.:
806 * - there are no vifs installed
807 * - pim_assert is not enabled
808 * - the MFC table is empty
809 */
810 if (V_numvifs > 0) {
811 *apival = 0;
812 return EPERM;
813 }
814 if (V_pim_assert_enabled) {
815 *apival = 0;
816 return EPERM;
817 }
818
819 MFC_LOCK();
820
821 for (i = 0; i < mfchashsize; i++) {
822 if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
823 MFC_UNLOCK();
824 *apival = 0;
825 return EPERM;
826 }
827 }
828
829 MFC_UNLOCK();
830
831 V_mrt_api_config = *apival & mrt_api_support;
832 *apival = V_mrt_api_config;
833
834 return 0;
835 }
836
837 /*
838 * Add a vif to the vif table
839 */
840 static int
841 add_vif(struct vifctl *vifcp)
842 {
843 struct vif *vifp = V_viftable + vifcp->vifc_vifi;
844 struct sockaddr_in sin = {sizeof sin, AF_INET};
845 struct ifaddr *ifa;
846 struct ifnet *ifp;
847 int error;
848
849 VIF_LOCK();
850 if (vifcp->vifc_vifi >= MAXVIFS) {
851 VIF_UNLOCK();
852 return EINVAL;
853 }
854 /* rate limiting is no longer supported by this code */
855 if (vifcp->vifc_rate_limit != 0) {
856 log(LOG_ERR, "rate limiting is no longer supported\n");
857 VIF_UNLOCK();
858 return EINVAL;
859 }
860 if (!in_nullhost(vifp->v_lcl_addr)) {
861 VIF_UNLOCK();
862 return EADDRINUSE;
863 }
864 if (in_nullhost(vifcp->vifc_lcl_addr)) {
865 VIF_UNLOCK();
866 return EADDRNOTAVAIL;
867 }
868
869 /* Find the interface with an address in AF_INET family */
870 if (vifcp->vifc_flags & VIFF_REGISTER) {
871 /*
872 * XXX: Because VIFF_REGISTER does not really need a valid
873 * local interface (e.g. it could be 127.0.0.2), we don't
874 * check its address.
875 */
876 ifp = NULL;
877 } else {
878 sin.sin_addr = vifcp->vifc_lcl_addr;
879 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
880 if (ifa == NULL) {
881 VIF_UNLOCK();
882 return EADDRNOTAVAIL;
883 }
884 ifp = ifa->ifa_ifp;
885 ifa_free(ifa);
886 }
887
888 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
889 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
890 VIF_UNLOCK();
891 return EOPNOTSUPP;
892 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
893 ifp = &V_multicast_register_if;
894 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
895 if (V_reg_vif_num == VIFI_INVALID) {
896 if_initname(&V_multicast_register_if, "register_vif", 0);
897 V_multicast_register_if.if_flags = IFF_LOOPBACK;
898 V_reg_vif_num = vifcp->vifc_vifi;
899 }
900 } else { /* Make sure the interface supports multicast */
901 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
902 VIF_UNLOCK();
903 return EOPNOTSUPP;
904 }
905
906 /* Enable promiscuous reception of all IP multicasts from the if */
907 error = if_allmulti(ifp, 1);
908 if (error) {
909 VIF_UNLOCK();
910 return error;
911 }
912 }
913
914 vifp->v_flags = vifcp->vifc_flags;
915 vifp->v_threshold = vifcp->vifc_threshold;
916 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
917 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
918 vifp->v_ifp = ifp;
919 /* initialize per vif pkt counters */
920 vifp->v_pkt_in = 0;
921 vifp->v_pkt_out = 0;
922 vifp->v_bytes_in = 0;
923 vifp->v_bytes_out = 0;
924 bzero(&vifp->v_route, sizeof(vifp->v_route));
925
926 /* Adjust numvifs up if the vifi is higher than numvifs */
927 if (V_numvifs <= vifcp->vifc_vifi)
928 V_numvifs = vifcp->vifc_vifi + 1;
929
930 VIF_UNLOCK();
931
932 CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__,
933 (int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_addr),
934 (int)vifcp->vifc_threshold);
935
936 return 0;
937 }
938
939 /*
940 * Delete a vif from the vif table
941 */
942 static int
943 del_vif_locked(vifi_t vifi)
944 {
945 struct vif *vifp;
946
947 VIF_LOCK_ASSERT();
948
949 if (vifi >= V_numvifs) {
950 return EINVAL;
951 }
952 vifp = &V_viftable[vifi];
953 if (in_nullhost(vifp->v_lcl_addr)) {
954 return EADDRNOTAVAIL;
955 }
956
957 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
958 if_allmulti(vifp->v_ifp, 0);
959
960 if (vifp->v_flags & VIFF_REGISTER)
961 V_reg_vif_num = VIFI_INVALID;
962
963 bzero((caddr_t)vifp, sizeof (*vifp));
964
965 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
966
967 /* Adjust numvifs down */
968 for (vifi = V_numvifs; vifi > 0; vifi--)
969 if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
970 break;
971 V_numvifs = vifi;
972
973 return 0;
974 }
975
976 static int
977 del_vif(vifi_t vifi)
978 {
979 int cc;
980
981 VIF_LOCK();
982 cc = del_vif_locked(vifi);
983 VIF_UNLOCK();
984
985 return cc;
986 }
987
988 /*
989 * update an mfc entry without resetting counters and S,G addresses.
990 */
991 static void
992 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
993 {
994 int i;
995
996 rt->mfc_parent = mfccp->mfcc_parent;
997 for (i = 0; i < V_numvifs; i++) {
998 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
999 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1000 MRT_MFC_FLAGS_ALL;
1001 }
1002 /* set the RP address */
1003 if (V_mrt_api_config & MRT_MFC_RP)
1004 rt->mfc_rp = mfccp->mfcc_rp;
1005 else
1006 rt->mfc_rp.s_addr = INADDR_ANY;
1007 }
1008
1009 /*
1010 * fully initialize an mfc entry from the parameter.
1011 */
1012 static void
1013 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1014 {
1015 rt->mfc_origin = mfccp->mfcc_origin;
1016 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1017
1018 update_mfc_params(rt, mfccp);
1019
1020 /* initialize pkt counters per src-grp */
1021 rt->mfc_pkt_cnt = 0;
1022 rt->mfc_byte_cnt = 0;
1023 rt->mfc_wrong_if = 0;
1024 timevalclear(&rt->mfc_last_assert);
1025 }
1026
1027 static void
1028 expire_mfc(struct mfc *rt)
1029 {
1030 struct rtdetq *rte, *nrte;
1031
1032 free_bw_list(rt->mfc_bw_meter);
1033
1034 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1035 m_freem(rte->m);
1036 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1037 free(rte, M_MRTABLE);
1038 }
1039
1040 LIST_REMOVE(rt, mfc_hash);
1041 free(rt, M_MRTABLE);
1042 }
1043
1044 /*
1045 * Add an mfc entry
1046 */
1047 static int
1048 add_mfc(struct mfcctl2 *mfccp)
1049 {
1050 struct mfc *rt;
1051 struct rtdetq *rte, *nrte;
1052 u_long hash = 0;
1053 u_short nstl;
1054
1055 VIF_LOCK();
1056 MFC_LOCK();
1057
1058 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1059
1060 /* If an entry already exists, just update the fields */
1061 if (rt) {
1062 CTR4(KTR_IPMF, "%s: update mfc orig %s group %lx parent %x",
1063 __func__, inet_ntoa(mfccp->mfcc_origin),
1064 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1065 mfccp->mfcc_parent);
1066 update_mfc_params(rt, mfccp);
1067 MFC_UNLOCK();
1068 VIF_UNLOCK();
1069 return (0);
1070 }
1071
1072 /*
1073 * Find the entry for which the upcall was made and update
1074 */
1075 nstl = 0;
1076 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1077 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1078 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1079 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1080 !TAILQ_EMPTY(&rt->mfc_stall)) {
1081 CTR5(KTR_IPMF,
1082 "%s: add mfc orig %s group %lx parent %x qh %p",
1083 __func__, inet_ntoa(mfccp->mfcc_origin),
1084 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1085 mfccp->mfcc_parent,
1086 TAILQ_FIRST(&rt->mfc_stall));
1087 if (nstl++)
1088 CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1089
1090 init_mfc_params(rt, mfccp);
1091 rt->mfc_expire = 0; /* Don't clean this guy up */
1092 V_nexpire[hash]--;
1093
1094 /* Free queued packets, but attempt to forward them first. */
1095 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1096 if (rte->ifp != NULL)
1097 ip_mdq(rte->m, rte->ifp, rt, -1);
1098 m_freem(rte->m);
1099 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1100 rt->mfc_nstall--;
1101 free(rte, M_MRTABLE);
1102 }
1103 }
1104 }
1105
1106 /*
1107 * It is possible that an entry is being inserted without an upcall
1108 */
1109 if (nstl == 0) {
1110 CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1111 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1112 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1113 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1114 init_mfc_params(rt, mfccp);
1115 if (rt->mfc_expire)
1116 V_nexpire[hash]--;
1117 rt->mfc_expire = 0;
1118 break; /* XXX */
1119 }
1120 }
1121
1122 if (rt == NULL) { /* no upcall, so make a new entry */
1123 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1124 if (rt == NULL) {
1125 MFC_UNLOCK();
1126 VIF_UNLOCK();
1127 return (ENOBUFS);
1128 }
1129
1130 init_mfc_params(rt, mfccp);
1131 TAILQ_INIT(&rt->mfc_stall);
1132 rt->mfc_nstall = 0;
1133
1134 rt->mfc_expire = 0;
1135 rt->mfc_bw_meter = NULL;
1136
1137 /* insert new entry at head of hash chain */
1138 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1139 }
1140 }
1141
1142 MFC_UNLOCK();
1143 VIF_UNLOCK();
1144
1145 return (0);
1146 }
1147
1148 /*
1149 * Delete an mfc entry
1150 */
1151 static int
1152 del_mfc(struct mfcctl2 *mfccp)
1153 {
1154 struct in_addr origin;
1155 struct in_addr mcastgrp;
1156 struct mfc *rt;
1157
1158 origin = mfccp->mfcc_origin;
1159 mcastgrp = mfccp->mfcc_mcastgrp;
1160
1161 CTR3(KTR_IPMF, "%s: delete mfc orig %s group %lx", __func__,
1162 inet_ntoa(origin), (u_long)ntohl(mcastgrp.s_addr));
1163
1164 MFC_LOCK();
1165
1166 rt = mfc_find(&origin, &mcastgrp);
1167 if (rt == NULL) {
1168 MFC_UNLOCK();
1169 return EADDRNOTAVAIL;
1170 }
1171
1172 /*
1173 * free the bw_meter entries
1174 */
1175 free_bw_list(rt->mfc_bw_meter);
1176 rt->mfc_bw_meter = NULL;
1177
1178 LIST_REMOVE(rt, mfc_hash);
1179 free(rt, M_MRTABLE);
1180
1181 MFC_UNLOCK();
1182
1183 return (0);
1184 }
1185
1186 /*
1187 * Send a message to the routing daemon on the multicast routing socket.
1188 */
1189 static int
1190 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1191 {
1192 if (s) {
1193 SOCKBUF_LOCK(&s->so_rcv);
1194 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1195 NULL) != 0) {
1196 sorwakeup_locked(s);
1197 return 0;
1198 }
1199 SOCKBUF_UNLOCK(&s->so_rcv);
1200 }
1201 m_freem(mm);
1202 return -1;
1203 }
1204
1205 /*
1206 * IP multicast forwarding function. This function assumes that the packet
1207 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1208 * pointed to by "ifp", and the packet is to be relayed to other networks
1209 * that have members of the packet's destination IP multicast group.
1210 *
1211 * The packet is returned unscathed to the caller, unless it is
1212 * erroneous, in which case a non-zero return value tells the caller to
1213 * discard it.
1214 */
1215
1216 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1217
1218 static int
1219 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1220 struct ip_moptions *imo)
1221 {
1222 struct mfc *rt;
1223 int error;
1224 vifi_t vifi;
1225
1226 CTR3(KTR_IPMF, "ip_mforward: delete mfc orig %s group %lx ifp %p",
1227 inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1228
1229 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1230 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1231 /*
1232 * Packet arrived via a physical interface or
1233 * an encapsulated tunnel or a register_vif.
1234 */
1235 } else {
1236 /*
1237 * Packet arrived through a source-route tunnel.
1238 * Source-route tunnels are no longer supported.
1239 */
1240 return (1);
1241 }
1242
1243 VIF_LOCK();
1244 MFC_LOCK();
1245 if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1246 if (ip->ip_ttl < MAXTTL)
1247 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1248 error = ip_mdq(m, ifp, NULL, vifi);
1249 MFC_UNLOCK();
1250 VIF_UNLOCK();
1251 return error;
1252 }
1253
1254 /*
1255 * Don't forward a packet with time-to-live of zero or one,
1256 * or a packet destined to a local-only group.
1257 */
1258 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1259 MFC_UNLOCK();
1260 VIF_UNLOCK();
1261 return 0;
1262 }
1263
1264 /*
1265 * Determine forwarding vifs from the forwarding cache table
1266 */
1267 MRTSTAT_INC(mrts_mfc_lookups);
1268 rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1269
1270 /* Entry exists, so forward if necessary */
1271 if (rt != NULL) {
1272 error = ip_mdq(m, ifp, rt, -1);
1273 MFC_UNLOCK();
1274 VIF_UNLOCK();
1275 return error;
1276 } else {
1277 /*
1278 * If we don't have a route for packet's origin,
1279 * Make a copy of the packet & send message to routing daemon
1280 */
1281
1282 struct mbuf *mb0;
1283 struct rtdetq *rte;
1284 u_long hash;
1285 int hlen = ip->ip_hl << 2;
1286
1287 MRTSTAT_INC(mrts_mfc_misses);
1288 MRTSTAT_INC(mrts_no_route);
1289 CTR2(KTR_IPMF, "ip_mforward: no mfc for (%s,%lx)",
1290 inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr));
1291
1292 /*
1293 * Allocate mbufs early so that we don't do extra work if we are
1294 * just going to fail anyway. Make sure to pullup the header so
1295 * that other people can't step on it.
1296 */
1297 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1298 M_NOWAIT|M_ZERO);
1299 if (rte == NULL) {
1300 MFC_UNLOCK();
1301 VIF_UNLOCK();
1302 return ENOBUFS;
1303 }
1304
1305 mb0 = m_copypacket(m, M_DONTWAIT);
1306 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1307 mb0 = m_pullup(mb0, hlen);
1308 if (mb0 == NULL) {
1309 free(rte, M_MRTABLE);
1310 MFC_UNLOCK();
1311 VIF_UNLOCK();
1312 return ENOBUFS;
1313 }
1314
1315 /* is there an upcall waiting for this flow ? */
1316 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1317 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1318 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1319 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1320 !TAILQ_EMPTY(&rt->mfc_stall))
1321 break;
1322 }
1323
1324 if (rt == NULL) {
1325 int i;
1326 struct igmpmsg *im;
1327 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1328 struct mbuf *mm;
1329
1330 /*
1331 * Locate the vifi for the incoming interface for this packet.
1332 * If none found, drop packet.
1333 */
1334 for (vifi = 0; vifi < V_numvifs &&
1335 V_viftable[vifi].v_ifp != ifp; vifi++)
1336 ;
1337 if (vifi >= V_numvifs) /* vif not found, drop packet */
1338 goto non_fatal;
1339
1340 /* no upcall, so make a new entry */
1341 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1342 if (rt == NULL)
1343 goto fail;
1344
1345 /* Make a copy of the header to send to the user level process */
1346 mm = m_copy(mb0, 0, hlen);
1347 if (mm == NULL)
1348 goto fail1;
1349
1350 /*
1351 * Send message to routing daemon to install
1352 * a route into the kernel table
1353 */
1354
1355 im = mtod(mm, struct igmpmsg *);
1356 im->im_msgtype = IGMPMSG_NOCACHE;
1357 im->im_mbz = 0;
1358 im->im_vif = vifi;
1359
1360 MRTSTAT_INC(mrts_upcalls);
1361
1362 k_igmpsrc.sin_addr = ip->ip_src;
1363 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1364 CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1365 MRTSTAT_INC(mrts_upq_sockfull);
1366 fail1:
1367 free(rt, M_MRTABLE);
1368 fail:
1369 free(rte, M_MRTABLE);
1370 m_freem(mb0);
1371 MFC_UNLOCK();
1372 VIF_UNLOCK();
1373 return ENOBUFS;
1374 }
1375
1376 /* insert new entry at head of hash chain */
1377 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1378 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1379 rt->mfc_expire = UPCALL_EXPIRE;
1380 V_nexpire[hash]++;
1381 for (i = 0; i < V_numvifs; i++) {
1382 rt->mfc_ttls[i] = 0;
1383 rt->mfc_flags[i] = 0;
1384 }
1385 rt->mfc_parent = -1;
1386
1387 /* clear the RP address */
1388 rt->mfc_rp.s_addr = INADDR_ANY;
1389 rt->mfc_bw_meter = NULL;
1390
1391 /* initialize pkt counters per src-grp */
1392 rt->mfc_pkt_cnt = 0;
1393 rt->mfc_byte_cnt = 0;
1394 rt->mfc_wrong_if = 0;
1395 timevalclear(&rt->mfc_last_assert);
1396
1397 TAILQ_INIT(&rt->mfc_stall);
1398 rt->mfc_nstall = 0;
1399
1400 /* link into table */
1401 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1402 TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1403 rt->mfc_nstall++;
1404
1405 } else {
1406 /* determine if queue has overflowed */
1407 if (rt->mfc_nstall > MAX_UPQ) {
1408 MRTSTAT_INC(mrts_upq_ovflw);
1409 non_fatal:
1410 free(rte, M_MRTABLE);
1411 m_freem(mb0);
1412 MFC_UNLOCK();
1413 VIF_UNLOCK();
1414 return (0);
1415 }
1416 TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1417 rt->mfc_nstall++;
1418 }
1419
1420 rte->m = mb0;
1421 rte->ifp = ifp;
1422
1423 MFC_UNLOCK();
1424 VIF_UNLOCK();
1425
1426 return 0;
1427 }
1428 }
1429
1430 /*
1431 * Clean up the cache entry if upcall is not serviced
1432 */
1433 static void
1434 expire_upcalls(void *arg)
1435 {
1436 u_long i;
1437
1438 CURVNET_SET((struct vnet *) arg);
1439
1440 MFC_LOCK();
1441
1442 for (i = 0; i < mfchashsize; i++) {
1443 struct mfc *rt, *nrt;
1444
1445 if (V_nexpire[i] == 0)
1446 continue;
1447
1448 for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
1449 nrt = LIST_NEXT(rt, mfc_hash);
1450
1451 if (TAILQ_EMPTY(&rt->mfc_stall))
1452 continue;
1453
1454 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1455 continue;
1456
1457 /*
1458 * free the bw_meter entries
1459 */
1460 while (rt->mfc_bw_meter != NULL) {
1461 struct bw_meter *x = rt->mfc_bw_meter;
1462
1463 rt->mfc_bw_meter = x->bm_mfc_next;
1464 free(x, M_BWMETER);
1465 }
1466
1467 MRTSTAT_INC(mrts_cache_cleanups);
1468 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1469 (u_long)ntohl(rt->mfc_origin.s_addr),
1470 (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1471
1472 expire_mfc(rt);
1473 }
1474 }
1475
1476 MFC_UNLOCK();
1477
1478 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1479 curvnet);
1480
1481 CURVNET_RESTORE();
1482 }
1483
1484 /*
1485 * Packet forwarding routine once entry in the cache is made
1486 */
1487 static int
1488 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1489 {
1490 struct ip *ip = mtod(m, struct ip *);
1491 vifi_t vifi;
1492 int plen = ip->ip_len;
1493
1494 VIF_LOCK_ASSERT();
1495
1496 /*
1497 * If xmt_vif is not -1, send on only the requested vif.
1498 *
1499 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1500 */
1501 if (xmt_vif < V_numvifs) {
1502 if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1503 pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1504 else
1505 phyint_send(ip, V_viftable + xmt_vif, m);
1506 return 1;
1507 }
1508
1509 /*
1510 * Don't forward if it didn't arrive from the parent vif for its origin.
1511 */
1512 vifi = rt->mfc_parent;
1513 if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1514 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1515 __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1516 MRTSTAT_INC(mrts_wrong_if);
1517 ++rt->mfc_wrong_if;
1518 /*
1519 * If we are doing PIM assert processing, send a message
1520 * to the routing daemon.
1521 *
1522 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1523 * can complete the SPT switch, regardless of the type
1524 * of the iif (broadcast media, GRE tunnel, etc).
1525 */
1526 if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1527 V_viftable[vifi].v_ifp) {
1528
1529 if (ifp == &V_multicast_register_if)
1530 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1531
1532 /* Get vifi for the incoming packet */
1533 for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1534 vifi++)
1535 ;
1536 if (vifi >= V_numvifs)
1537 return 0; /* The iif is not found: ignore the packet. */
1538
1539 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1540 return 0; /* WRONGVIF disabled: ignore the packet */
1541
1542 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1543 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1544 struct igmpmsg *im;
1545 int hlen = ip->ip_hl << 2;
1546 struct mbuf *mm = m_copy(m, 0, hlen);
1547
1548 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1549 mm = m_pullup(mm, hlen);
1550 if (mm == NULL)
1551 return ENOBUFS;
1552
1553 im = mtod(mm, struct igmpmsg *);
1554 im->im_msgtype = IGMPMSG_WRONGVIF;
1555 im->im_mbz = 0;
1556 im->im_vif = vifi;
1557
1558 MRTSTAT_INC(mrts_upcalls);
1559
1560 k_igmpsrc.sin_addr = im->im_src;
1561 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1562 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1563 MRTSTAT_INC(mrts_upq_sockfull);
1564 return ENOBUFS;
1565 }
1566 }
1567 }
1568 return 0;
1569 }
1570
1571
1572 /* If I sourced this packet, it counts as output, else it was input. */
1573 if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1574 V_viftable[vifi].v_pkt_out++;
1575 V_viftable[vifi].v_bytes_out += plen;
1576 } else {
1577 V_viftable[vifi].v_pkt_in++;
1578 V_viftable[vifi].v_bytes_in += plen;
1579 }
1580 rt->mfc_pkt_cnt++;
1581 rt->mfc_byte_cnt += plen;
1582
1583 /*
1584 * For each vif, decide if a copy of the packet should be forwarded.
1585 * Forward if:
1586 * - the ttl exceeds the vif's threshold
1587 * - there are group members downstream on interface
1588 */
1589 for (vifi = 0; vifi < V_numvifs; vifi++)
1590 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1591 V_viftable[vifi].v_pkt_out++;
1592 V_viftable[vifi].v_bytes_out += plen;
1593 if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1594 pim_register_send(ip, V_viftable + vifi, m, rt);
1595 else
1596 phyint_send(ip, V_viftable + vifi, m);
1597 }
1598
1599 /*
1600 * Perform upcall-related bw measuring.
1601 */
1602 if (rt->mfc_bw_meter != NULL) {
1603 struct bw_meter *x;
1604 struct timeval now;
1605
1606 microtime(&now);
1607 MFC_LOCK_ASSERT();
1608 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1609 bw_meter_receive_packet(x, plen, &now);
1610 }
1611
1612 return 0;
1613 }
1614
1615 /*
1616 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1617 */
1618 static int
1619 X_legal_vif_num(int vif)
1620 {
1621 int ret;
1622
1623 ret = 0;
1624 if (vif < 0)
1625 return (ret);
1626
1627 VIF_LOCK();
1628 if (vif < V_numvifs)
1629 ret = 1;
1630 VIF_UNLOCK();
1631
1632 return (ret);
1633 }
1634
1635 /*
1636 * Return the local address used by this vif
1637 */
1638 static u_long
1639 X_ip_mcast_src(int vifi)
1640 {
1641 in_addr_t addr;
1642
1643 addr = INADDR_ANY;
1644 if (vifi < 0)
1645 return (addr);
1646
1647 VIF_LOCK();
1648 if (vifi < V_numvifs)
1649 addr = V_viftable[vifi].v_lcl_addr.s_addr;
1650 VIF_UNLOCK();
1651
1652 return (addr);
1653 }
1654
1655 static void
1656 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1657 {
1658 struct mbuf *mb_copy;
1659 int hlen = ip->ip_hl << 2;
1660
1661 VIF_LOCK_ASSERT();
1662
1663 /*
1664 * Make a new reference to the packet; make sure that
1665 * the IP header is actually copied, not just referenced,
1666 * so that ip_output() only scribbles on the copy.
1667 */
1668 mb_copy = m_copypacket(m, M_DONTWAIT);
1669 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1670 mb_copy = m_pullup(mb_copy, hlen);
1671 if (mb_copy == NULL)
1672 return;
1673
1674 send_packet(vifp, mb_copy);
1675 }
1676
1677 static void
1678 send_packet(struct vif *vifp, struct mbuf *m)
1679 {
1680 struct ip_moptions imo;
1681 struct in_multi *imm[2];
1682 int error;
1683
1684 VIF_LOCK_ASSERT();
1685
1686 imo.imo_multicast_ifp = vifp->v_ifp;
1687 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1688 imo.imo_multicast_loop = 1;
1689 imo.imo_multicast_vif = -1;
1690 imo.imo_num_memberships = 0;
1691 imo.imo_max_memberships = 2;
1692 imo.imo_membership = &imm[0];
1693
1694 /*
1695 * Re-entrancy should not be a problem here, because
1696 * the packets that we send out and are looped back at us
1697 * should get rejected because they appear to come from
1698 * the loopback interface, thus preventing looping.
1699 */
1700 error = ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, &imo, NULL);
1701 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1702 (ptrdiff_t)(vifp - V_viftable), error);
1703 }
1704
1705 /*
1706 * Stubs for old RSVP socket shim implementation.
1707 */
1708
1709 static int
1710 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1711 {
1712
1713 return (EOPNOTSUPP);
1714 }
1715
1716 static void
1717 X_ip_rsvp_force_done(struct socket *so __unused)
1718 {
1719
1720 }
1721
1722 static void
1723 X_rsvp_input(struct mbuf *m, int off __unused)
1724 {
1725
1726 if (!V_rsvp_on)
1727 m_freem(m);
1728 }
1729
1730 /*
1731 * Code for bandwidth monitors
1732 */
1733
1734 /*
1735 * Define common interface for timeval-related methods
1736 */
1737 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1738 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1739 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1740
1741 static uint32_t
1742 compute_bw_meter_flags(struct bw_upcall *req)
1743 {
1744 uint32_t flags = 0;
1745
1746 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1747 flags |= BW_METER_UNIT_PACKETS;
1748 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1749 flags |= BW_METER_UNIT_BYTES;
1750 if (req->bu_flags & BW_UPCALL_GEQ)
1751 flags |= BW_METER_GEQ;
1752 if (req->bu_flags & BW_UPCALL_LEQ)
1753 flags |= BW_METER_LEQ;
1754
1755 return flags;
1756 }
1757
1758 /*
1759 * Add a bw_meter entry
1760 */
1761 static int
1762 add_bw_upcall(struct bw_upcall *req)
1763 {
1764 struct mfc *mfc;
1765 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1766 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1767 struct timeval now;
1768 struct bw_meter *x;
1769 uint32_t flags;
1770
1771 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1772 return EOPNOTSUPP;
1773
1774 /* Test if the flags are valid */
1775 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1776 return EINVAL;
1777 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1778 return EINVAL;
1779 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1780 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1781 return EINVAL;
1782
1783 /* Test if the threshold time interval is valid */
1784 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1785 return EINVAL;
1786
1787 flags = compute_bw_meter_flags(req);
1788
1789 /*
1790 * Find if we have already same bw_meter entry
1791 */
1792 MFC_LOCK();
1793 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1794 if (mfc == NULL) {
1795 MFC_UNLOCK();
1796 return EADDRNOTAVAIL;
1797 }
1798 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1799 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1800 &req->bu_threshold.b_time, ==)) &&
1801 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1802 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1803 (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
1804 MFC_UNLOCK();
1805 return 0; /* XXX Already installed */
1806 }
1807 }
1808
1809 /* Allocate the new bw_meter entry */
1810 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1811 if (x == NULL) {
1812 MFC_UNLOCK();
1813 return ENOBUFS;
1814 }
1815
1816 /* Set the new bw_meter entry */
1817 x->bm_threshold.b_time = req->bu_threshold.b_time;
1818 microtime(&now);
1819 x->bm_start_time = now;
1820 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1821 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1822 x->bm_measured.b_packets = 0;
1823 x->bm_measured.b_bytes = 0;
1824 x->bm_flags = flags;
1825 x->bm_time_next = NULL;
1826 x->bm_time_hash = BW_METER_BUCKETS;
1827
1828 /* Add the new bw_meter entry to the front of entries for this MFC */
1829 x->bm_mfc = mfc;
1830 x->bm_mfc_next = mfc->mfc_bw_meter;
1831 mfc->mfc_bw_meter = x;
1832 schedule_bw_meter(x, &now);
1833 MFC_UNLOCK();
1834
1835 return 0;
1836 }
1837
1838 static void
1839 free_bw_list(struct bw_meter *list)
1840 {
1841 while (list != NULL) {
1842 struct bw_meter *x = list;
1843
1844 list = list->bm_mfc_next;
1845 unschedule_bw_meter(x);
1846 free(x, M_BWMETER);
1847 }
1848 }
1849
1850 /*
1851 * Delete one or multiple bw_meter entries
1852 */
1853 static int
1854 del_bw_upcall(struct bw_upcall *req)
1855 {
1856 struct mfc *mfc;
1857 struct bw_meter *x;
1858
1859 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1860 return EOPNOTSUPP;
1861
1862 MFC_LOCK();
1863
1864 /* Find the corresponding MFC entry */
1865 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1866 if (mfc == NULL) {
1867 MFC_UNLOCK();
1868 return EADDRNOTAVAIL;
1869 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1870 /*
1871 * Delete all bw_meter entries for this mfc
1872 */
1873 struct bw_meter *list;
1874
1875 list = mfc->mfc_bw_meter;
1876 mfc->mfc_bw_meter = NULL;
1877 free_bw_list(list);
1878 MFC_UNLOCK();
1879 return 0;
1880 } else { /* Delete a single bw_meter entry */
1881 struct bw_meter *prev;
1882 uint32_t flags = 0;
1883
1884 flags = compute_bw_meter_flags(req);
1885
1886 /* Find the bw_meter entry to delete */
1887 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1888 prev = x, x = x->bm_mfc_next) {
1889 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1890 &req->bu_threshold.b_time, ==)) &&
1891 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1892 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1893 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
1894 break;
1895 }
1896 if (x != NULL) { /* Delete entry from the list for this MFC */
1897 if (prev != NULL)
1898 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
1899 else
1900 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1901
1902 unschedule_bw_meter(x);
1903 MFC_UNLOCK();
1904 /* Free the bw_meter entry */
1905 free(x, M_BWMETER);
1906 return 0;
1907 } else {
1908 MFC_UNLOCK();
1909 return EINVAL;
1910 }
1911 }
1912 /* NOTREACHED */
1913 }
1914
1915 /*
1916 * Perform bandwidth measurement processing that may result in an upcall
1917 */
1918 static void
1919 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1920 {
1921 struct timeval delta;
1922
1923 MFC_LOCK_ASSERT();
1924
1925 delta = *nowp;
1926 BW_TIMEVALDECR(&delta, &x->bm_start_time);
1927
1928 if (x->bm_flags & BW_METER_GEQ) {
1929 /*
1930 * Processing for ">=" type of bw_meter entry
1931 */
1932 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1933 /* Reset the bw_meter entry */
1934 x->bm_start_time = *nowp;
1935 x->bm_measured.b_packets = 0;
1936 x->bm_measured.b_bytes = 0;
1937 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1938 }
1939
1940 /* Record that a packet is received */
1941 x->bm_measured.b_packets++;
1942 x->bm_measured.b_bytes += plen;
1943
1944 /*
1945 * Test if we should deliver an upcall
1946 */
1947 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1948 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1949 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1950 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1951 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1952 /* Prepare an upcall for delivery */
1953 bw_meter_prepare_upcall(x, nowp);
1954 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1955 }
1956 }
1957 } else if (x->bm_flags & BW_METER_LEQ) {
1958 /*
1959 * Processing for "<=" type of bw_meter entry
1960 */
1961 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1962 /*
1963 * We are behind time with the multicast forwarding table
1964 * scanning for "<=" type of bw_meter entries, so test now
1965 * if we should deliver an upcall.
1966 */
1967 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1968 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1969 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1970 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1971 /* Prepare an upcall for delivery */
1972 bw_meter_prepare_upcall(x, nowp);
1973 }
1974 /* Reschedule the bw_meter entry */
1975 unschedule_bw_meter(x);
1976 schedule_bw_meter(x, nowp);
1977 }
1978
1979 /* Record that a packet is received */
1980 x->bm_measured.b_packets++;
1981 x->bm_measured.b_bytes += plen;
1982
1983 /*
1984 * Test if we should restart the measuring interval
1985 */
1986 if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1987 x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1988 (x->bm_flags & BW_METER_UNIT_BYTES &&
1989 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1990 /* Don't restart the measuring interval */
1991 } else {
1992 /* Do restart the measuring interval */
1993 /*
1994 * XXX: note that we don't unschedule and schedule, because this
1995 * might be too much overhead per packet. Instead, when we process
1996 * all entries for a given timer hash bin, we check whether it is
1997 * really a timeout. If not, we reschedule at that time.
1998 */
1999 x->bm_start_time = *nowp;
2000 x->bm_measured.b_packets = 0;
2001 x->bm_measured.b_bytes = 0;
2002 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2003 }
2004 }
2005 }
2006
2007 /*
2008 * Prepare a bandwidth-related upcall
2009 */
2010 static void
2011 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2012 {
2013 struct timeval delta;
2014 struct bw_upcall *u;
2015
2016 MFC_LOCK_ASSERT();
2017
2018 /*
2019 * Compute the measured time interval
2020 */
2021 delta = *nowp;
2022 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2023
2024 /*
2025 * If there are too many pending upcalls, deliver them now
2026 */
2027 if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2028 bw_upcalls_send();
2029
2030 /*
2031 * Set the bw_upcall entry
2032 */
2033 u = &V_bw_upcalls[V_bw_upcalls_n++];
2034 u->bu_src = x->bm_mfc->mfc_origin;
2035 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2036 u->bu_threshold.b_time = x->bm_threshold.b_time;
2037 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2038 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2039 u->bu_measured.b_time = delta;
2040 u->bu_measured.b_packets = x->bm_measured.b_packets;
2041 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2042 u->bu_flags = 0;
2043 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2044 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2045 if (x->bm_flags & BW_METER_UNIT_BYTES)
2046 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2047 if (x->bm_flags & BW_METER_GEQ)
2048 u->bu_flags |= BW_UPCALL_GEQ;
2049 if (x->bm_flags & BW_METER_LEQ)
2050 u->bu_flags |= BW_UPCALL_LEQ;
2051 }
2052
2053 /*
2054 * Send the pending bandwidth-related upcalls
2055 */
2056 static void
2057 bw_upcalls_send(void)
2058 {
2059 struct mbuf *m;
2060 int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2061 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2062 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2063 0, /* unused2 */
2064 IGMPMSG_BW_UPCALL,/* im_msgtype */
2065 0, /* im_mbz */
2066 0, /* im_vif */
2067 0, /* unused3 */
2068 { 0 }, /* im_src */
2069 { 0 } }; /* im_dst */
2070
2071 MFC_LOCK_ASSERT();
2072
2073 if (V_bw_upcalls_n == 0)
2074 return; /* No pending upcalls */
2075
2076 V_bw_upcalls_n = 0;
2077
2078 /*
2079 * Allocate a new mbuf, initialize it with the header and
2080 * the payload for the pending calls.
2081 */
2082 MGETHDR(m, M_DONTWAIT, MT_DATA);
2083 if (m == NULL) {
2084 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2085 return;
2086 }
2087
2088 m->m_len = m->m_pkthdr.len = 0;
2089 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2090 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2091
2092 /*
2093 * Send the upcalls
2094 * XXX do we need to set the address in k_igmpsrc ?
2095 */
2096 MRTSTAT_INC(mrts_upcalls);
2097 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2098 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2099 MRTSTAT_INC(mrts_upq_sockfull);
2100 }
2101 }
2102
2103 /*
2104 * Compute the timeout hash value for the bw_meter entries
2105 */
2106 #define BW_METER_TIMEHASH(bw_meter, hash) \
2107 do { \
2108 struct timeval next_timeval = (bw_meter)->bm_start_time; \
2109 \
2110 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2111 (hash) = next_timeval.tv_sec; \
2112 if (next_timeval.tv_usec) \
2113 (hash)++; /* XXX: make sure we don't timeout early */ \
2114 (hash) %= BW_METER_BUCKETS; \
2115 } while (0)
2116
2117 /*
2118 * Schedule a timer to process periodically bw_meter entry of type "<="
2119 * by linking the entry in the proper hash bucket.
2120 */
2121 static void
2122 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2123 {
2124 int time_hash;
2125
2126 MFC_LOCK_ASSERT();
2127
2128 if (!(x->bm_flags & BW_METER_LEQ))
2129 return; /* XXX: we schedule timers only for "<=" entries */
2130
2131 /*
2132 * Reset the bw_meter entry
2133 */
2134 x->bm_start_time = *nowp;
2135 x->bm_measured.b_packets = 0;
2136 x->bm_measured.b_bytes = 0;
2137 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2138
2139 /*
2140 * Compute the timeout hash value and insert the entry
2141 */
2142 BW_METER_TIMEHASH(x, time_hash);
2143 x->bm_time_next = V_bw_meter_timers[time_hash];
2144 V_bw_meter_timers[time_hash] = x;
2145 x->bm_time_hash = time_hash;
2146 }
2147
2148 /*
2149 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2150 * by removing the entry from the proper hash bucket.
2151 */
2152 static void
2153 unschedule_bw_meter(struct bw_meter *x)
2154 {
2155 int time_hash;
2156 struct bw_meter *prev, *tmp;
2157
2158 MFC_LOCK_ASSERT();
2159
2160 if (!(x->bm_flags & BW_METER_LEQ))
2161 return; /* XXX: we schedule timers only for "<=" entries */
2162
2163 /*
2164 * Compute the timeout hash value and delete the entry
2165 */
2166 time_hash = x->bm_time_hash;
2167 if (time_hash >= BW_METER_BUCKETS)
2168 return; /* Entry was not scheduled */
2169
2170 for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2171 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2172 if (tmp == x)
2173 break;
2174
2175 if (tmp == NULL)
2176 panic("unschedule_bw_meter: bw_meter entry not found");
2177
2178 if (prev != NULL)
2179 prev->bm_time_next = x->bm_time_next;
2180 else
2181 V_bw_meter_timers[time_hash] = x->bm_time_next;
2182
2183 x->bm_time_next = NULL;
2184 x->bm_time_hash = BW_METER_BUCKETS;
2185 }
2186
2187
2188 /*
2189 * Process all "<=" type of bw_meter that should be processed now,
2190 * and for each entry prepare an upcall if necessary. Each processed
2191 * entry is rescheduled again for the (periodic) processing.
2192 *
2193 * This is run periodically (once per second normally). On each round,
2194 * all the potentially matching entries are in the hash slot that we are
2195 * looking at.
2196 */
2197 static void
2198 bw_meter_process()
2199 {
2200 uint32_t loops;
2201 int i;
2202 struct timeval now, process_endtime;
2203
2204 microtime(&now);
2205 if (V_last_tv_sec == now.tv_sec)
2206 return; /* nothing to do */
2207
2208 loops = now.tv_sec - V_last_tv_sec;
2209 V_last_tv_sec = now.tv_sec;
2210 if (loops > BW_METER_BUCKETS)
2211 loops = BW_METER_BUCKETS;
2212
2213 MFC_LOCK();
2214 /*
2215 * Process all bins of bw_meter entries from the one after the last
2216 * processed to the current one. On entry, i points to the last bucket
2217 * visited, so we need to increment i at the beginning of the loop.
2218 */
2219 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2220 struct bw_meter *x, *tmp_list;
2221
2222 if (++i >= BW_METER_BUCKETS)
2223 i = 0;
2224
2225 /* Disconnect the list of bw_meter entries from the bin */
2226 tmp_list = V_bw_meter_timers[i];
2227 V_bw_meter_timers[i] = NULL;
2228
2229 /* Process the list of bw_meter entries */
2230 while (tmp_list != NULL) {
2231 x = tmp_list;
2232 tmp_list = tmp_list->bm_time_next;
2233
2234 /* Test if the time interval is over */
2235 process_endtime = x->bm_start_time;
2236 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2237 if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2238 /* Not yet: reschedule, but don't reset */
2239 int time_hash;
2240
2241 BW_METER_TIMEHASH(x, time_hash);
2242 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2243 /*
2244 * XXX: somehow the bin processing is a bit ahead of time.
2245 * Put the entry in the next bin.
2246 */
2247 if (++time_hash >= BW_METER_BUCKETS)
2248 time_hash = 0;
2249 }
2250 x->bm_time_next = V_bw_meter_timers[time_hash];
2251 V_bw_meter_timers[time_hash] = x;
2252 x->bm_time_hash = time_hash;
2253
2254 continue;
2255 }
2256
2257 /*
2258 * Test if we should deliver an upcall
2259 */
2260 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2261 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2262 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2263 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2264 /* Prepare an upcall for delivery */
2265 bw_meter_prepare_upcall(x, &now);
2266 }
2267
2268 /*
2269 * Reschedule for next processing
2270 */
2271 schedule_bw_meter(x, &now);
2272 }
2273 }
2274
2275 /* Send all upcalls that are pending delivery */
2276 bw_upcalls_send();
2277
2278 MFC_UNLOCK();
2279 }
2280
2281 /*
2282 * A periodic function for sending all upcalls that are pending delivery
2283 */
2284 static void
2285 expire_bw_upcalls_send(void *arg)
2286 {
2287 CURVNET_SET((struct vnet *) arg);
2288
2289 MFC_LOCK();
2290 bw_upcalls_send();
2291 MFC_UNLOCK();
2292
2293 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2294 curvnet);
2295 CURVNET_RESTORE();
2296 }
2297
2298 /*
2299 * A periodic function for periodic scanning of the multicast forwarding
2300 * table for processing all "<=" bw_meter entries.
2301 */
2302 static void
2303 expire_bw_meter_process(void *arg)
2304 {
2305 CURVNET_SET((struct vnet *) arg);
2306
2307 if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2308 bw_meter_process();
2309
2310 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2311 curvnet);
2312 CURVNET_RESTORE();
2313 }
2314
2315 /*
2316 * End of bandwidth monitoring code
2317 */
2318
2319 /*
2320 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2321 *
2322 */
2323 static int
2324 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2325 struct mfc *rt)
2326 {
2327 struct mbuf *mb_copy, *mm;
2328
2329 /*
2330 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2331 * rendezvous point was unspecified, and we were told not to.
2332 */
2333 if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2334 in_nullhost(rt->mfc_rp))
2335 return 0;
2336
2337 mb_copy = pim_register_prepare(ip, m);
2338 if (mb_copy == NULL)
2339 return ENOBUFS;
2340
2341 /*
2342 * Send all the fragments. Note that the mbuf for each fragment
2343 * is freed by the sending machinery.
2344 */
2345 for (mm = mb_copy; mm; mm = mb_copy) {
2346 mb_copy = mm->m_nextpkt;
2347 mm->m_nextpkt = 0;
2348 mm = m_pullup(mm, sizeof(struct ip));
2349 if (mm != NULL) {
2350 ip = mtod(mm, struct ip *);
2351 if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2352 pim_register_send_rp(ip, vifp, mm, rt);
2353 } else {
2354 pim_register_send_upcall(ip, vifp, mm, rt);
2355 }
2356 }
2357 }
2358
2359 return 0;
2360 }
2361
2362 /*
2363 * Return a copy of the data packet that is ready for PIM Register
2364 * encapsulation.
2365 * XXX: Note that in the returned copy the IP header is a valid one.
2366 */
2367 static struct mbuf *
2368 pim_register_prepare(struct ip *ip, struct mbuf *m)
2369 {
2370 struct mbuf *mb_copy = NULL;
2371 int mtu;
2372
2373 /* Take care of delayed checksums */
2374 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2375 in_delayed_cksum(m);
2376 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2377 }
2378
2379 /*
2380 * Copy the old packet & pullup its IP header into the
2381 * new mbuf so we can modify it.
2382 */
2383 mb_copy = m_copypacket(m, M_DONTWAIT);
2384 if (mb_copy == NULL)
2385 return NULL;
2386 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2387 if (mb_copy == NULL)
2388 return NULL;
2389
2390 /* take care of the TTL */
2391 ip = mtod(mb_copy, struct ip *);
2392 --ip->ip_ttl;
2393
2394 /* Compute the MTU after the PIM Register encapsulation */
2395 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2396
2397 if (ip->ip_len <= mtu) {
2398 /* Turn the IP header into a valid one */
2399 ip->ip_len = htons(ip->ip_len);
2400 ip->ip_off = htons(ip->ip_off);
2401 ip->ip_sum = 0;
2402 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2403 } else {
2404 /* Fragment the packet */
2405 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) {
2406 m_freem(mb_copy);
2407 return NULL;
2408 }
2409 }
2410 return mb_copy;
2411 }
2412
2413 /*
2414 * Send an upcall with the data packet to the user-level process.
2415 */
2416 static int
2417 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2418 struct mbuf *mb_copy, struct mfc *rt)
2419 {
2420 struct mbuf *mb_first;
2421 int len = ntohs(ip->ip_len);
2422 struct igmpmsg *im;
2423 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2424
2425 VIF_LOCK_ASSERT();
2426
2427 /*
2428 * Add a new mbuf with an upcall header
2429 */
2430 MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2431 if (mb_first == NULL) {
2432 m_freem(mb_copy);
2433 return ENOBUFS;
2434 }
2435 mb_first->m_data += max_linkhdr;
2436 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2437 mb_first->m_len = sizeof(struct igmpmsg);
2438 mb_first->m_next = mb_copy;
2439
2440 /* Send message to routing daemon */
2441 im = mtod(mb_first, struct igmpmsg *);
2442 im->im_msgtype = IGMPMSG_WHOLEPKT;
2443 im->im_mbz = 0;
2444 im->im_vif = vifp - V_viftable;
2445 im->im_src = ip->ip_src;
2446 im->im_dst = ip->ip_dst;
2447
2448 k_igmpsrc.sin_addr = ip->ip_src;
2449
2450 MRTSTAT_INC(mrts_upcalls);
2451
2452 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2453 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2454 MRTSTAT_INC(mrts_upq_sockfull);
2455 return ENOBUFS;
2456 }
2457
2458 /* Keep statistics */
2459 PIMSTAT_INC(pims_snd_registers_msgs);
2460 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2461
2462 return 0;
2463 }
2464
2465 /*
2466 * Encapsulate the data packet in PIM Register message and send it to the RP.
2467 */
2468 static int
2469 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2470 struct mfc *rt)
2471 {
2472 struct mbuf *mb_first;
2473 struct ip *ip_outer;
2474 struct pim_encap_pimhdr *pimhdr;
2475 int len = ntohs(ip->ip_len);
2476 vifi_t vifi = rt->mfc_parent;
2477
2478 VIF_LOCK_ASSERT();
2479
2480 if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2481 m_freem(mb_copy);
2482 return EADDRNOTAVAIL; /* The iif vif is invalid */
2483 }
2484
2485 /*
2486 * Add a new mbuf with the encapsulating header
2487 */
2488 MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2489 if (mb_first == NULL) {
2490 m_freem(mb_copy);
2491 return ENOBUFS;
2492 }
2493 mb_first->m_data += max_linkhdr;
2494 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2495 mb_first->m_next = mb_copy;
2496
2497 mb_first->m_pkthdr.len = len + mb_first->m_len;
2498
2499 /*
2500 * Fill in the encapsulating IP and PIM header
2501 */
2502 ip_outer = mtod(mb_first, struct ip *);
2503 *ip_outer = pim_encap_iphdr;
2504 ip_outer->ip_id = ip_newid();
2505 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2506 ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2507 ip_outer->ip_dst = rt->mfc_rp;
2508 /*
2509 * Copy the inner header TOS to the outer header, and take care of the
2510 * IP_DF bit.
2511 */
2512 ip_outer->ip_tos = ip->ip_tos;
2513 if (ntohs(ip->ip_off) & IP_DF)
2514 ip_outer->ip_off |= IP_DF;
2515 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2516 + sizeof(pim_encap_iphdr));
2517 *pimhdr = pim_encap_pimhdr;
2518 /* If the iif crosses a border, set the Border-bit */
2519 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2520 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2521
2522 mb_first->m_data += sizeof(pim_encap_iphdr);
2523 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2524 mb_first->m_data -= sizeof(pim_encap_iphdr);
2525
2526 send_packet(vifp, mb_first);
2527
2528 /* Keep statistics */
2529 PIMSTAT_INC(pims_snd_registers_msgs);
2530 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2531
2532 return 0;
2533 }
2534
2535 /*
2536 * pim_encapcheck() is called by the encap4_input() path at runtime to
2537 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2538 * into the kernel.
2539 */
2540 static int
2541 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2542 {
2543
2544 #ifdef DIAGNOSTIC
2545 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2546 #endif
2547 if (proto != IPPROTO_PIM)
2548 return 0; /* not for us; reject the datagram. */
2549
2550 return 64; /* claim the datagram. */
2551 }
2552
2553 /*
2554 * PIM-SMv2 and PIM-DM messages processing.
2555 * Receives and verifies the PIM control messages, and passes them
2556 * up to the listening socket, using rip_input().
2557 * The only message with special processing is the PIM_REGISTER message
2558 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2559 * is passed to if_simloop().
2560 */
2561 void
2562 pim_input(struct mbuf *m, int off)
2563 {
2564 struct ip *ip = mtod(m, struct ip *);
2565 struct pim *pim;
2566 int minlen;
2567 int datalen = ip->ip_len;
2568 int ip_tos;
2569 int iphlen = off;
2570
2571 /* Keep statistics */
2572 PIMSTAT_INC(pims_rcv_total_msgs);
2573 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2574
2575 /*
2576 * Validate lengths
2577 */
2578 if (datalen < PIM_MINLEN) {
2579 PIMSTAT_INC(pims_rcv_tooshort);
2580 CTR3(KTR_IPMF, "%s: short packet (%d) from %s",
2581 __func__, datalen, inet_ntoa(ip->ip_src));
2582 m_freem(m);
2583 return;
2584 }
2585
2586 /*
2587 * If the packet is at least as big as a REGISTER, go agead
2588 * and grab the PIM REGISTER header size, to avoid another
2589 * possible m_pullup() later.
2590 *
2591 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2592 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2593 */
2594 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2595 /*
2596 * Get the IP and PIM headers in contiguous memory, and
2597 * possibly the PIM REGISTER header.
2598 */
2599 if ((m->m_flags & M_EXT || m->m_len < minlen) &&
2600 (m = m_pullup(m, minlen)) == 0) {
2601 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2602 return;
2603 }
2604
2605 /* m_pullup() may have given us a new mbuf so reset ip. */
2606 ip = mtod(m, struct ip *);
2607 ip_tos = ip->ip_tos;
2608
2609 /* adjust mbuf to point to the PIM header */
2610 m->m_data += iphlen;
2611 m->m_len -= iphlen;
2612 pim = mtod(m, struct pim *);
2613
2614 /*
2615 * Validate checksum. If PIM REGISTER, exclude the data packet.
2616 *
2617 * XXX: some older PIMv2 implementations don't make this distinction,
2618 * so for compatibility reason perform the checksum over part of the
2619 * message, and if error, then over the whole message.
2620 */
2621 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2622 /* do nothing, checksum okay */
2623 } else if (in_cksum(m, datalen)) {
2624 PIMSTAT_INC(pims_rcv_badsum);
2625 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2626 m_freem(m);
2627 return;
2628 }
2629
2630 /* PIM version check */
2631 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2632 PIMSTAT_INC(pims_rcv_badversion);
2633 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2634 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2635 m_freem(m);
2636 return;
2637 }
2638
2639 /* restore mbuf back to the outer IP */
2640 m->m_data -= iphlen;
2641 m->m_len += iphlen;
2642
2643 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2644 /*
2645 * Since this is a REGISTER, we'll make a copy of the register
2646 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2647 * routing daemon.
2648 */
2649 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2650 struct mbuf *mcp;
2651 struct ip *encap_ip;
2652 u_int32_t *reghdr;
2653 struct ifnet *vifp;
2654
2655 VIF_LOCK();
2656 if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2657 VIF_UNLOCK();
2658 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2659 (int)V_reg_vif_num);
2660 m_freem(m);
2661 return;
2662 }
2663 /* XXX need refcnt? */
2664 vifp = V_viftable[V_reg_vif_num].v_ifp;
2665 VIF_UNLOCK();
2666
2667 /*
2668 * Validate length
2669 */
2670 if (datalen < PIM_REG_MINLEN) {
2671 PIMSTAT_INC(pims_rcv_tooshort);
2672 PIMSTAT_INC(pims_rcv_badregisters);
2673 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2674 m_freem(m);
2675 return;
2676 }
2677
2678 reghdr = (u_int32_t *)(pim + 1);
2679 encap_ip = (struct ip *)(reghdr + 1);
2680
2681 CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d",
2682 __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len));
2683
2684 /* verify the version number of the inner packet */
2685 if (encap_ip->ip_v != IPVERSION) {
2686 PIMSTAT_INC(pims_rcv_badregisters);
2687 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2688 m_freem(m);
2689 return;
2690 }
2691
2692 /* verify the inner packet is destined to a mcast group */
2693 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2694 PIMSTAT_INC(pims_rcv_badregisters);
2695 CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__,
2696 inet_ntoa(encap_ip->ip_dst));
2697 m_freem(m);
2698 return;
2699 }
2700
2701 /* If a NULL_REGISTER, pass it to the daemon */
2702 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2703 goto pim_input_to_daemon;
2704
2705 /*
2706 * Copy the TOS from the outer IP header to the inner IP header.
2707 */
2708 if (encap_ip->ip_tos != ip_tos) {
2709 /* Outer TOS -> inner TOS */
2710 encap_ip->ip_tos = ip_tos;
2711 /* Recompute the inner header checksum. Sigh... */
2712
2713 /* adjust mbuf to point to the inner IP header */
2714 m->m_data += (iphlen + PIM_MINLEN);
2715 m->m_len -= (iphlen + PIM_MINLEN);
2716
2717 encap_ip->ip_sum = 0;
2718 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2719
2720 /* restore mbuf to point back to the outer IP header */
2721 m->m_data -= (iphlen + PIM_MINLEN);
2722 m->m_len += (iphlen + PIM_MINLEN);
2723 }
2724
2725 /*
2726 * Decapsulate the inner IP packet and loopback to forward it
2727 * as a normal multicast packet. Also, make a copy of the
2728 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2729 * to pass to the daemon later, so it can take the appropriate
2730 * actions (e.g., send back PIM_REGISTER_STOP).
2731 * XXX: here m->m_data points to the outer IP header.
2732 */
2733 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
2734 if (mcp == NULL) {
2735 CTR1(KTR_IPMF, "%s: m_copy() failed", __func__);
2736 m_freem(m);
2737 return;
2738 }
2739
2740 /* Keep statistics */
2741 /* XXX: registers_bytes include only the encap. mcast pkt */
2742 PIMSTAT_INC(pims_rcv_registers_msgs);
2743 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2744
2745 /*
2746 * forward the inner ip packet; point m_data at the inner ip.
2747 */
2748 m_adj(m, iphlen + PIM_MINLEN);
2749
2750 CTR4(KTR_IPMF,
2751 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2752 __func__,
2753 (u_long)ntohl(encap_ip->ip_src.s_addr),
2754 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2755 (int)V_reg_vif_num);
2756
2757 /* NB: vifp was collected above; can it change on us? */
2758 if_simloop(vifp, m, dst.sin_family, 0);
2759
2760 /* prepare the register head to send to the mrouting daemon */
2761 m = mcp;
2762 }
2763
2764 pim_input_to_daemon:
2765 /*
2766 * Pass the PIM message up to the daemon; if it is a Register message,
2767 * pass the 'head' only up to the daemon. This includes the
2768 * outer IP header, PIM header, PIM-Register header and the
2769 * inner IP header.
2770 * XXX: the outer IP header pkt size of a Register is not adjust to
2771 * reflect the fact that the inner multicast data is truncated.
2772 */
2773 rip_input(m, iphlen);
2774
2775 return;
2776 }
2777
2778 static int
2779 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2780 {
2781 struct mfc *rt;
2782 int error, i;
2783
2784 if (req->newptr)
2785 return (EPERM);
2786 if (V_mfchashtbl == NULL) /* XXX unlocked */
2787 return (0);
2788 error = sysctl_wire_old_buffer(req, 0);
2789 if (error)
2790 return (error);
2791
2792 MFC_LOCK();
2793 for (i = 0; i < mfchashsize; i++) {
2794 LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2795 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2796 if (error)
2797 goto out_locked;
2798 }
2799 }
2800 out_locked:
2801 MFC_UNLOCK();
2802 return (error);
2803 }
2804
2805 SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, sysctl_mfctable,
2806 "IPv4 Multicast Forwarding Table (struct *mfc[mfchashsize], "
2807 "netinet/ip_mroute.h)");
2808
2809 static void
2810 vnet_mroute_init(const void *unused __unused)
2811 {
2812
2813 MALLOC(V_nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2814 bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2815 callout_init(&V_expire_upcalls_ch, CALLOUT_MPSAFE);
2816 callout_init(&V_bw_upcalls_ch, CALLOUT_MPSAFE);
2817 callout_init(&V_bw_meter_ch, CALLOUT_MPSAFE);
2818 }
2819
2820 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mroute_init,
2821 NULL);
2822
2823 static void
2824 vnet_mroute_uninit(const void *unused __unused)
2825 {
2826
2827 FREE(V_nexpire, M_MRTABLE);
2828 V_nexpire = NULL;
2829 }
2830
2831 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE,
2832 vnet_mroute_uninit, NULL);
2833
2834 static int
2835 ip_mroute_modevent(module_t mod, int type, void *unused)
2836 {
2837
2838 switch (type) {
2839 case MOD_LOAD:
2840 MROUTER_LOCK_INIT();
2841
2842 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2843 if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2844 if (if_detach_event_tag == NULL) {
2845 printf("ip_mroute: unable to register "
2846 "ifnet_departure_event handler\n");
2847 MROUTER_LOCK_DESTROY();
2848 return (EINVAL);
2849 }
2850
2851 MFC_LOCK_INIT();
2852 VIF_LOCK_INIT();
2853
2854 mfchashsize = MFCHASHSIZE;
2855 if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2856 !powerof2(mfchashsize)) {
2857 printf("WARNING: %s not a power of 2; using default\n",
2858 "net.inet.ip.mfchashsize");
2859 mfchashsize = MFCHASHSIZE;
2860 }
2861
2862 pim_squelch_wholepkt = 0;
2863 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2864 &pim_squelch_wholepkt);
2865
2866 pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2867 pim_encapcheck, &in_pim_protosw, NULL);
2868 if (pim_encap_cookie == NULL) {
2869 printf("ip_mroute: unable to attach pim encap\n");
2870 VIF_LOCK_DESTROY();
2871 MFC_LOCK_DESTROY();
2872 MROUTER_LOCK_DESTROY();
2873 return (EINVAL);
2874 }
2875
2876 ip_mcast_src = X_ip_mcast_src;
2877 ip_mforward = X_ip_mforward;
2878 ip_mrouter_done = X_ip_mrouter_done;
2879 ip_mrouter_get = X_ip_mrouter_get;
2880 ip_mrouter_set = X_ip_mrouter_set;
2881
2882 ip_rsvp_force_done = X_ip_rsvp_force_done;
2883 ip_rsvp_vif = X_ip_rsvp_vif;
2884
2885 legal_vif_num = X_legal_vif_num;
2886 mrt_ioctl = X_mrt_ioctl;
2887 rsvp_input_p = X_rsvp_input;
2888 break;
2889
2890 case MOD_UNLOAD:
2891 /*
2892 * Typically module unload happens after the user-level
2893 * process has shutdown the kernel services (the check
2894 * below insures someone can't just yank the module out
2895 * from under a running process). But if the module is
2896 * just loaded and then unloaded w/o starting up a user
2897 * process we still need to cleanup.
2898 */
2899 MROUTER_LOCK();
2900 if (ip_mrouter_cnt != 0) {
2901 MROUTER_UNLOCK();
2902 return (EINVAL);
2903 }
2904 ip_mrouter_unloading = 1;
2905 MROUTER_UNLOCK();
2906
2907 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2908
2909 if (pim_encap_cookie) {
2910 encap_detach(pim_encap_cookie);
2911 pim_encap_cookie = NULL;
2912 }
2913
2914 ip_mcast_src = NULL;
2915 ip_mforward = NULL;
2916 ip_mrouter_done = NULL;
2917 ip_mrouter_get = NULL;
2918 ip_mrouter_set = NULL;
2919
2920 ip_rsvp_force_done = NULL;
2921 ip_rsvp_vif = NULL;
2922
2923 legal_vif_num = NULL;
2924 mrt_ioctl = NULL;
2925 rsvp_input_p = NULL;
2926
2927 VIF_LOCK_DESTROY();
2928 MFC_LOCK_DESTROY();
2929 MROUTER_LOCK_DESTROY();
2930 break;
2931
2932 default:
2933 return EOPNOTSUPP;
2934 }
2935 return 0;
2936 }
2937
2938 static moduledata_t ip_mroutemod = {
2939 "ip_mroute",
2940 ip_mroute_modevent,
2941 0
2942 };
2943
2944 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_MIDDLE);
Cache object: f51f7a16e50b492a537cf801bdf7c2f4
|