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: releng/10.2/sys/netinet/ip_mroute.c 261208 2014-01-27 09:33:30Z glebius $");
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 #include <sys/counter.h>
97
98 #include <net/if.h>
99 #include <net/netisr.h>
100 #include <net/route.h>
101 #include <net/vnet.h>
102
103 #include <netinet/in.h>
104 #include <netinet/igmp.h>
105 #include <netinet/in_systm.h>
106 #include <netinet/in_var.h>
107 #include <netinet/ip.h>
108 #include <netinet/ip_encap.h>
109 #include <netinet/ip_mroute.h>
110 #include <netinet/ip_var.h>
111 #include <netinet/ip_options.h>
112 #include <netinet/pim.h>
113 #include <netinet/pim_var.h>
114 #include <netinet/udp.h>
115
116 #include <machine/in_cksum.h>
117
118 #ifndef KTR_IPMF
119 #define KTR_IPMF KTR_INET
120 #endif
121
122 #define VIFI_INVALID ((vifi_t) -1)
123 #define M_HASCL(m) ((m)->m_flags & M_EXT)
124
125 static VNET_DEFINE(uint32_t, last_tv_sec); /* last time we processed this */
126 #define V_last_tv_sec VNET(last_tv_sec)
127
128 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
129
130 /*
131 * Locking. We use two locks: one for the virtual interface table and
132 * one for the forwarding table. These locks may be nested in which case
133 * the VIF lock must always be taken first. Note that each lock is used
134 * to cover not only the specific data structure but also related data
135 * structures.
136 */
137
138 static struct mtx mrouter_mtx;
139 #define MROUTER_LOCK() mtx_lock(&mrouter_mtx)
140 #define MROUTER_UNLOCK() mtx_unlock(&mrouter_mtx)
141 #define MROUTER_LOCK_ASSERT() mtx_assert(&mrouter_mtx, MA_OWNED)
142 #define MROUTER_LOCK_INIT() \
143 mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
144 #define MROUTER_LOCK_DESTROY() mtx_destroy(&mrouter_mtx)
145
146 static int ip_mrouter_cnt; /* # of vnets with active mrouters */
147 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
148
149 static VNET_PCPUSTAT_DEFINE(struct mrtstat, mrtstat);
150 VNET_PCPUSTAT_SYSINIT(mrtstat);
151 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
152 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
153 mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
154 "netinet/ip_mroute.h)");
155
156 static VNET_DEFINE(u_long, mfchash);
157 #define V_mfchash VNET(mfchash)
158 #define MFCHASH(a, g) \
159 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
160 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
161 #define MFCHASHSIZE 256
162
163 static u_long mfchashsize; /* Hash size */
164 static VNET_DEFINE(u_char *, nexpire); /* 0..mfchashsize-1 */
165 #define V_nexpire VNET(nexpire)
166 static VNET_DEFINE(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
167 #define V_mfchashtbl VNET(mfchashtbl)
168
169 static struct mtx mfc_mtx;
170 #define MFC_LOCK() mtx_lock(&mfc_mtx)
171 #define MFC_UNLOCK() mtx_unlock(&mfc_mtx)
172 #define MFC_LOCK_ASSERT() mtx_assert(&mfc_mtx, MA_OWNED)
173 #define MFC_LOCK_INIT() \
174 mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
175 #define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx)
176
177 static VNET_DEFINE(vifi_t, numvifs);
178 #define V_numvifs VNET(numvifs)
179 static VNET_DEFINE(struct vif, viftable[MAXVIFS]);
180 #define V_viftable VNET(viftable)
181 SYSCTL_VNET_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
182 &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]",
183 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
184
185 static struct mtx vif_mtx;
186 #define VIF_LOCK() mtx_lock(&vif_mtx)
187 #define VIF_UNLOCK() mtx_unlock(&vif_mtx)
188 #define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED)
189 #define VIF_LOCK_INIT() \
190 mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
191 #define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx)
192
193 static eventhandler_tag if_detach_event_tag = NULL;
194
195 static VNET_DEFINE(struct callout, expire_upcalls_ch);
196 #define V_expire_upcalls_ch VNET(expire_upcalls_ch)
197
198 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
199 #define UPCALL_EXPIRE 6 /* number of timeouts */
200
201 /*
202 * Bandwidth meter variables and constants
203 */
204 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
205 /*
206 * Pending timeouts are stored in a hash table, the key being the
207 * expiration time. Periodically, the entries are analysed and processed.
208 */
209 #define BW_METER_BUCKETS 1024
210 static VNET_DEFINE(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]);
211 #define V_bw_meter_timers VNET(bw_meter_timers)
212 static VNET_DEFINE(struct callout, bw_meter_ch);
213 #define V_bw_meter_ch VNET(bw_meter_ch)
214 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
215
216 /*
217 * Pending upcalls are stored in a vector which is flushed when
218 * full, or periodically
219 */
220 static VNET_DEFINE(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]);
221 #define V_bw_upcalls VNET(bw_upcalls)
222 static VNET_DEFINE(u_int, bw_upcalls_n); /* # of pending upcalls */
223 #define V_bw_upcalls_n VNET(bw_upcalls_n)
224 static VNET_DEFINE(struct callout, bw_upcalls_ch);
225 #define V_bw_upcalls_ch VNET(bw_upcalls_ch)
226
227 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
228
229 static VNET_PCPUSTAT_DEFINE(struct pimstat, pimstat);
230 VNET_PCPUSTAT_SYSINIT(pimstat);
231 VNET_PCPUSTAT_SYSUNINIT(pimstat);
232
233 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
234 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
235 pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
236
237 static u_long pim_squelch_wholepkt = 0;
238 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
239 &pim_squelch_wholepkt, 0,
240 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
241
242 extern struct domain inetdomain;
243 static const struct protosw in_pim_protosw = {
244 .pr_type = SOCK_RAW,
245 .pr_domain = &inetdomain,
246 .pr_protocol = IPPROTO_PIM,
247 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
248 .pr_input = pim_input,
249 .pr_output = (pr_output_t*)rip_output,
250 .pr_ctloutput = rip_ctloutput,
251 .pr_usrreqs = &rip_usrreqs
252 };
253 static const struct encaptab *pim_encap_cookie;
254
255 static int pim_encapcheck(const struct mbuf *, int, int, void *);
256
257 /*
258 * Note: the PIM Register encapsulation adds the following in front of a
259 * data packet:
260 *
261 * struct pim_encap_hdr {
262 * struct ip ip;
263 * struct pim_encap_pimhdr pim;
264 * }
265 *
266 */
267
268 struct pim_encap_pimhdr {
269 struct pim pim;
270 uint32_t flags;
271 };
272 #define PIM_ENCAP_TTL 64
273
274 static struct ip pim_encap_iphdr = {
275 #if BYTE_ORDER == LITTLE_ENDIAN
276 sizeof(struct ip) >> 2,
277 IPVERSION,
278 #else
279 IPVERSION,
280 sizeof(struct ip) >> 2,
281 #endif
282 0, /* tos */
283 sizeof(struct ip), /* total length */
284 0, /* id */
285 0, /* frag offset */
286 PIM_ENCAP_TTL,
287 IPPROTO_PIM,
288 0, /* checksum */
289 };
290
291 static struct pim_encap_pimhdr pim_encap_pimhdr = {
292 {
293 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
294 0, /* reserved */
295 0, /* checksum */
296 },
297 0 /* flags */
298 };
299
300 static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID;
301 #define V_reg_vif_num VNET(reg_vif_num)
302 static VNET_DEFINE(struct ifnet, multicast_register_if);
303 #define V_multicast_register_if VNET(multicast_register_if)
304
305 /*
306 * Private variables.
307 */
308
309 static u_long X_ip_mcast_src(int);
310 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
311 struct ip_moptions *);
312 static int X_ip_mrouter_done(void);
313 static int X_ip_mrouter_get(struct socket *, struct sockopt *);
314 static int X_ip_mrouter_set(struct socket *, struct sockopt *);
315 static int X_legal_vif_num(int);
316 static int X_mrt_ioctl(u_long, caddr_t, int);
317
318 static int add_bw_upcall(struct bw_upcall *);
319 static int add_mfc(struct mfcctl2 *);
320 static int add_vif(struct vifctl *);
321 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
322 static void bw_meter_process(void);
323 static void bw_meter_receive_packet(struct bw_meter *, int,
324 struct timeval *);
325 static void bw_upcalls_send(void);
326 static int del_bw_upcall(struct bw_upcall *);
327 static int del_mfc(struct mfcctl2 *);
328 static int del_vif(vifi_t);
329 static int del_vif_locked(vifi_t);
330 static void expire_bw_meter_process(void *);
331 static void expire_bw_upcalls_send(void *);
332 static void expire_mfc(struct mfc *);
333 static void expire_upcalls(void *);
334 static void free_bw_list(struct bw_meter *);
335 static int get_sg_cnt(struct sioc_sg_req *);
336 static int get_vif_cnt(struct sioc_vif_req *);
337 static void if_detached_event(void *, struct ifnet *);
338 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
339 static int ip_mrouter_init(struct socket *, int);
340 static __inline struct mfc *
341 mfc_find(struct in_addr *, struct in_addr *);
342 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
343 static struct mbuf *
344 pim_register_prepare(struct ip *, struct mbuf *);
345 static int pim_register_send(struct ip *, struct vif *,
346 struct mbuf *, struct mfc *);
347 static int pim_register_send_rp(struct ip *, struct vif *,
348 struct mbuf *, struct mfc *);
349 static int pim_register_send_upcall(struct ip *, struct vif *,
350 struct mbuf *, struct mfc *);
351 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
352 static void send_packet(struct vif *, struct mbuf *);
353 static int set_api_config(uint32_t *);
354 static int set_assert(int);
355 static int socket_send(struct socket *, struct mbuf *,
356 struct sockaddr_in *);
357 static void unschedule_bw_meter(struct bw_meter *);
358
359 /*
360 * Kernel multicast forwarding API capabilities and setup.
361 * If more API capabilities are added to the kernel, they should be
362 * recorded in `mrt_api_support'.
363 */
364 #define MRT_API_VERSION 0x0305
365
366 static const int mrt_api_version = MRT_API_VERSION;
367 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
368 MRT_MFC_FLAGS_BORDER_VIF |
369 MRT_MFC_RP |
370 MRT_MFC_BW_UPCALL);
371 static VNET_DEFINE(uint32_t, mrt_api_config);
372 #define V_mrt_api_config VNET(mrt_api_config)
373 static VNET_DEFINE(int, pim_assert_enabled);
374 #define V_pim_assert_enabled VNET(pim_assert_enabled)
375 static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */
376
377 /*
378 * Find a route for a given origin IP address and multicast group address.
379 * Statistics must be updated by the caller.
380 */
381 static __inline struct mfc *
382 mfc_find(struct in_addr *o, struct in_addr *g)
383 {
384 struct mfc *rt;
385
386 MFC_LOCK_ASSERT();
387
388 LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
389 if (in_hosteq(rt->mfc_origin, *o) &&
390 in_hosteq(rt->mfc_mcastgrp, *g) &&
391 TAILQ_EMPTY(&rt->mfc_stall))
392 break;
393 }
394
395 return (rt);
396 }
397
398 /*
399 * Handle MRT setsockopt commands to modify the multicast forwarding tables.
400 */
401 static int
402 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
403 {
404 int error, optval;
405 vifi_t vifi;
406 struct vifctl vifc;
407 struct mfcctl2 mfc;
408 struct bw_upcall bw_upcall;
409 uint32_t i;
410
411 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
412 return EPERM;
413
414 error = 0;
415 switch (sopt->sopt_name) {
416 case MRT_INIT:
417 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
418 if (error)
419 break;
420 error = ip_mrouter_init(so, optval);
421 break;
422
423 case MRT_DONE:
424 error = ip_mrouter_done();
425 break;
426
427 case MRT_ADD_VIF:
428 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
429 if (error)
430 break;
431 error = add_vif(&vifc);
432 break;
433
434 case MRT_DEL_VIF:
435 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
436 if (error)
437 break;
438 error = del_vif(vifi);
439 break;
440
441 case MRT_ADD_MFC:
442 case MRT_DEL_MFC:
443 /*
444 * select data size depending on API version.
445 */
446 if (sopt->sopt_name == MRT_ADD_MFC &&
447 V_mrt_api_config & MRT_API_FLAGS_ALL) {
448 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
449 sizeof(struct mfcctl2));
450 } else {
451 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
452 sizeof(struct mfcctl));
453 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
454 sizeof(mfc) - sizeof(struct mfcctl));
455 }
456 if (error)
457 break;
458 if (sopt->sopt_name == MRT_ADD_MFC)
459 error = add_mfc(&mfc);
460 else
461 error = del_mfc(&mfc);
462 break;
463
464 case MRT_ASSERT:
465 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
466 if (error)
467 break;
468 set_assert(optval);
469 break;
470
471 case MRT_API_CONFIG:
472 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
473 if (!error)
474 error = set_api_config(&i);
475 if (!error)
476 error = sooptcopyout(sopt, &i, sizeof i);
477 break;
478
479 case MRT_ADD_BW_UPCALL:
480 case MRT_DEL_BW_UPCALL:
481 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
482 sizeof bw_upcall);
483 if (error)
484 break;
485 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
486 error = add_bw_upcall(&bw_upcall);
487 else
488 error = del_bw_upcall(&bw_upcall);
489 break;
490
491 default:
492 error = EOPNOTSUPP;
493 break;
494 }
495 return error;
496 }
497
498 /*
499 * Handle MRT getsockopt commands
500 */
501 static int
502 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
503 {
504 int error;
505
506 switch (sopt->sopt_name) {
507 case MRT_VERSION:
508 error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
509 break;
510
511 case MRT_ASSERT:
512 error = sooptcopyout(sopt, &V_pim_assert_enabled,
513 sizeof V_pim_assert_enabled);
514 break;
515
516 case MRT_API_SUPPORT:
517 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
518 break;
519
520 case MRT_API_CONFIG:
521 error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
522 break;
523
524 default:
525 error = EOPNOTSUPP;
526 break;
527 }
528 return error;
529 }
530
531 /*
532 * Handle ioctl commands to obtain information from the cache
533 */
534 static int
535 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
536 {
537 int error = 0;
538
539 /*
540 * Currently the only function calling this ioctl routine is rtioctl().
541 * Typically, only root can create the raw socket in order to execute
542 * this ioctl method, however the request might be coming from a prison
543 */
544 error = priv_check(curthread, PRIV_NETINET_MROUTE);
545 if (error)
546 return (error);
547 switch (cmd) {
548 case (SIOCGETVIFCNT):
549 error = get_vif_cnt((struct sioc_vif_req *)data);
550 break;
551
552 case (SIOCGETSGCNT):
553 error = get_sg_cnt((struct sioc_sg_req *)data);
554 break;
555
556 default:
557 error = EINVAL;
558 break;
559 }
560 return error;
561 }
562
563 /*
564 * returns the packet, byte, rpf-failure count for the source group provided
565 */
566 static int
567 get_sg_cnt(struct sioc_sg_req *req)
568 {
569 struct mfc *rt;
570
571 MFC_LOCK();
572 rt = mfc_find(&req->src, &req->grp);
573 if (rt == NULL) {
574 MFC_UNLOCK();
575 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
576 return EADDRNOTAVAIL;
577 }
578 req->pktcnt = rt->mfc_pkt_cnt;
579 req->bytecnt = rt->mfc_byte_cnt;
580 req->wrong_if = rt->mfc_wrong_if;
581 MFC_UNLOCK();
582 return 0;
583 }
584
585 /*
586 * returns the input and output packet and byte counts on the vif provided
587 */
588 static int
589 get_vif_cnt(struct sioc_vif_req *req)
590 {
591 vifi_t vifi = req->vifi;
592
593 VIF_LOCK();
594 if (vifi >= V_numvifs) {
595 VIF_UNLOCK();
596 return EINVAL;
597 }
598
599 req->icount = V_viftable[vifi].v_pkt_in;
600 req->ocount = V_viftable[vifi].v_pkt_out;
601 req->ibytes = V_viftable[vifi].v_bytes_in;
602 req->obytes = V_viftable[vifi].v_bytes_out;
603 VIF_UNLOCK();
604
605 return 0;
606 }
607
608 static void
609 if_detached_event(void *arg __unused, struct ifnet *ifp)
610 {
611 vifi_t vifi;
612 u_long i;
613
614 MROUTER_LOCK();
615
616 if (V_ip_mrouter == NULL) {
617 MROUTER_UNLOCK();
618 return;
619 }
620
621 VIF_LOCK();
622 MFC_LOCK();
623
624 /*
625 * Tear down multicast forwarder state associated with this ifnet.
626 * 1. Walk the vif list, matching vifs against this ifnet.
627 * 2. Walk the multicast forwarding cache (mfc) looking for
628 * inner matches with this vif's index.
629 * 3. Expire any matching multicast forwarding cache entries.
630 * 4. Free vif state. This should disable ALLMULTI on the interface.
631 */
632 for (vifi = 0; vifi < V_numvifs; vifi++) {
633 if (V_viftable[vifi].v_ifp != ifp)
634 continue;
635 for (i = 0; i < mfchashsize; i++) {
636 struct mfc *rt, *nrt;
637
638 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
639 if (rt->mfc_parent == vifi) {
640 expire_mfc(rt);
641 }
642 }
643 }
644 del_vif_locked(vifi);
645 }
646
647 MFC_UNLOCK();
648 VIF_UNLOCK();
649
650 MROUTER_UNLOCK();
651 }
652
653 /*
654 * Enable multicast forwarding.
655 */
656 static int
657 ip_mrouter_init(struct socket *so, int version)
658 {
659
660 CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
661 so->so_type, so->so_proto->pr_protocol);
662
663 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
664 return EOPNOTSUPP;
665
666 if (version != 1)
667 return ENOPROTOOPT;
668
669 MROUTER_LOCK();
670
671 if (ip_mrouter_unloading) {
672 MROUTER_UNLOCK();
673 return ENOPROTOOPT;
674 }
675
676 if (V_ip_mrouter != NULL) {
677 MROUTER_UNLOCK();
678 return EADDRINUSE;
679 }
680
681 V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
682 HASH_NOWAIT);
683
684 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
685 curvnet);
686 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
687 curvnet);
688 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
689 curvnet);
690
691 V_ip_mrouter = so;
692 ip_mrouter_cnt++;
693
694 MROUTER_UNLOCK();
695
696 CTR1(KTR_IPMF, "%s: done", __func__);
697
698 return 0;
699 }
700
701 /*
702 * Disable multicast forwarding.
703 */
704 static int
705 X_ip_mrouter_done(void)
706 {
707 struct ifnet *ifp;
708 u_long i;
709 vifi_t vifi;
710
711 MROUTER_LOCK();
712
713 if (V_ip_mrouter == NULL) {
714 MROUTER_UNLOCK();
715 return EINVAL;
716 }
717
718 /*
719 * Detach/disable hooks to the reset of the system.
720 */
721 V_ip_mrouter = NULL;
722 ip_mrouter_cnt--;
723 V_mrt_api_config = 0;
724
725 VIF_LOCK();
726
727 /*
728 * For each phyint in use, disable promiscuous reception of all IP
729 * multicasts.
730 */
731 for (vifi = 0; vifi < V_numvifs; vifi++) {
732 if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
733 !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
734 ifp = V_viftable[vifi].v_ifp;
735 if_allmulti(ifp, 0);
736 }
737 }
738 bzero((caddr_t)V_viftable, sizeof(V_viftable));
739 V_numvifs = 0;
740 V_pim_assert_enabled = 0;
741
742 VIF_UNLOCK();
743
744 callout_stop(&V_expire_upcalls_ch);
745 callout_stop(&V_bw_upcalls_ch);
746 callout_stop(&V_bw_meter_ch);
747
748 MFC_LOCK();
749
750 /*
751 * Free all multicast forwarding cache entries.
752 * Do not use hashdestroy(), as we must perform other cleanup.
753 */
754 for (i = 0; i < mfchashsize; i++) {
755 struct mfc *rt, *nrt;
756
757 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
758 expire_mfc(rt);
759 }
760 }
761 free(V_mfchashtbl, M_MRTABLE);
762 V_mfchashtbl = NULL;
763
764 bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
765
766 V_bw_upcalls_n = 0;
767 bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
768
769 MFC_UNLOCK();
770
771 V_reg_vif_num = VIFI_INVALID;
772
773 MROUTER_UNLOCK();
774
775 CTR1(KTR_IPMF, "%s: done", __func__);
776
777 return 0;
778 }
779
780 /*
781 * Set PIM assert processing global
782 */
783 static int
784 set_assert(int i)
785 {
786 if ((i != 1) && (i != 0))
787 return EINVAL;
788
789 V_pim_assert_enabled = i;
790
791 return 0;
792 }
793
794 /*
795 * Configure API capabilities
796 */
797 int
798 set_api_config(uint32_t *apival)
799 {
800 u_long i;
801
802 /*
803 * We can set the API capabilities only if it is the first operation
804 * after MRT_INIT. I.e.:
805 * - there are no vifs installed
806 * - pim_assert is not enabled
807 * - the MFC table is empty
808 */
809 if (V_numvifs > 0) {
810 *apival = 0;
811 return EPERM;
812 }
813 if (V_pim_assert_enabled) {
814 *apival = 0;
815 return EPERM;
816 }
817
818 MFC_LOCK();
819
820 for (i = 0; i < mfchashsize; i++) {
821 if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
822 MFC_UNLOCK();
823 *apival = 0;
824 return EPERM;
825 }
826 }
827
828 MFC_UNLOCK();
829
830 V_mrt_api_config = *apival & mrt_api_support;
831 *apival = V_mrt_api_config;
832
833 return 0;
834 }
835
836 /*
837 * Add a vif to the vif table
838 */
839 static int
840 add_vif(struct vifctl *vifcp)
841 {
842 struct vif *vifp = V_viftable + vifcp->vifc_vifi;
843 struct sockaddr_in sin = {sizeof sin, AF_INET};
844 struct ifaddr *ifa;
845 struct ifnet *ifp;
846 int error;
847
848 VIF_LOCK();
849 if (vifcp->vifc_vifi >= MAXVIFS) {
850 VIF_UNLOCK();
851 return EINVAL;
852 }
853 /* rate limiting is no longer supported by this code */
854 if (vifcp->vifc_rate_limit != 0) {
855 log(LOG_ERR, "rate limiting is no longer supported\n");
856 VIF_UNLOCK();
857 return EINVAL;
858 }
859 if (!in_nullhost(vifp->v_lcl_addr)) {
860 VIF_UNLOCK();
861 return EADDRINUSE;
862 }
863 if (in_nullhost(vifcp->vifc_lcl_addr)) {
864 VIF_UNLOCK();
865 return EADDRNOTAVAIL;
866 }
867
868 /* Find the interface with an address in AF_INET family */
869 if (vifcp->vifc_flags & VIFF_REGISTER) {
870 /*
871 * XXX: Because VIFF_REGISTER does not really need a valid
872 * local interface (e.g. it could be 127.0.0.2), we don't
873 * check its address.
874 */
875 ifp = NULL;
876 } else {
877 sin.sin_addr = vifcp->vifc_lcl_addr;
878 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
879 if (ifa == NULL) {
880 VIF_UNLOCK();
881 return EADDRNOTAVAIL;
882 }
883 ifp = ifa->ifa_ifp;
884 ifa_free(ifa);
885 }
886
887 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
888 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
889 VIF_UNLOCK();
890 return EOPNOTSUPP;
891 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
892 ifp = &V_multicast_register_if;
893 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
894 if (V_reg_vif_num == VIFI_INVALID) {
895 if_initname(&V_multicast_register_if, "register_vif", 0);
896 V_multicast_register_if.if_flags = IFF_LOOPBACK;
897 V_reg_vif_num = vifcp->vifc_vifi;
898 }
899 } else { /* Make sure the interface supports multicast */
900 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
901 VIF_UNLOCK();
902 return EOPNOTSUPP;
903 }
904
905 /* Enable promiscuous reception of all IP multicasts from the if */
906 error = if_allmulti(ifp, 1);
907 if (error) {
908 VIF_UNLOCK();
909 return error;
910 }
911 }
912
913 vifp->v_flags = vifcp->vifc_flags;
914 vifp->v_threshold = vifcp->vifc_threshold;
915 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
916 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
917 vifp->v_ifp = ifp;
918 /* initialize per vif pkt counters */
919 vifp->v_pkt_in = 0;
920 vifp->v_pkt_out = 0;
921 vifp->v_bytes_in = 0;
922 vifp->v_bytes_out = 0;
923
924 /* Adjust numvifs up if the vifi is higher than numvifs */
925 if (V_numvifs <= vifcp->vifc_vifi)
926 V_numvifs = vifcp->vifc_vifi + 1;
927
928 VIF_UNLOCK();
929
930 CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__,
931 (int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_addr),
932 (int)vifcp->vifc_threshold);
933
934 return 0;
935 }
936
937 /*
938 * Delete a vif from the vif table
939 */
940 static int
941 del_vif_locked(vifi_t vifi)
942 {
943 struct vif *vifp;
944
945 VIF_LOCK_ASSERT();
946
947 if (vifi >= V_numvifs) {
948 return EINVAL;
949 }
950 vifp = &V_viftable[vifi];
951 if (in_nullhost(vifp->v_lcl_addr)) {
952 return EADDRNOTAVAIL;
953 }
954
955 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
956 if_allmulti(vifp->v_ifp, 0);
957
958 if (vifp->v_flags & VIFF_REGISTER)
959 V_reg_vif_num = VIFI_INVALID;
960
961 bzero((caddr_t)vifp, sizeof (*vifp));
962
963 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
964
965 /* Adjust numvifs down */
966 for (vifi = V_numvifs; vifi > 0; vifi--)
967 if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
968 break;
969 V_numvifs = vifi;
970
971 return 0;
972 }
973
974 static int
975 del_vif(vifi_t vifi)
976 {
977 int cc;
978
979 VIF_LOCK();
980 cc = del_vif_locked(vifi);
981 VIF_UNLOCK();
982
983 return cc;
984 }
985
986 /*
987 * update an mfc entry without resetting counters and S,G addresses.
988 */
989 static void
990 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
991 {
992 int i;
993
994 rt->mfc_parent = mfccp->mfcc_parent;
995 for (i = 0; i < V_numvifs; i++) {
996 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
997 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
998 MRT_MFC_FLAGS_ALL;
999 }
1000 /* set the RP address */
1001 if (V_mrt_api_config & MRT_MFC_RP)
1002 rt->mfc_rp = mfccp->mfcc_rp;
1003 else
1004 rt->mfc_rp.s_addr = INADDR_ANY;
1005 }
1006
1007 /*
1008 * fully initialize an mfc entry from the parameter.
1009 */
1010 static void
1011 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1012 {
1013 rt->mfc_origin = mfccp->mfcc_origin;
1014 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1015
1016 update_mfc_params(rt, mfccp);
1017
1018 /* initialize pkt counters per src-grp */
1019 rt->mfc_pkt_cnt = 0;
1020 rt->mfc_byte_cnt = 0;
1021 rt->mfc_wrong_if = 0;
1022 timevalclear(&rt->mfc_last_assert);
1023 }
1024
1025 static void
1026 expire_mfc(struct mfc *rt)
1027 {
1028 struct rtdetq *rte, *nrte;
1029
1030 MFC_LOCK_ASSERT();
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_NOWAIT);
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 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1449 if (TAILQ_EMPTY(&rt->mfc_stall))
1450 continue;
1451
1452 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1453 continue;
1454
1455 /*
1456 * free the bw_meter entries
1457 */
1458 while (rt->mfc_bw_meter != NULL) {
1459 struct bw_meter *x = rt->mfc_bw_meter;
1460
1461 rt->mfc_bw_meter = x->bm_mfc_next;
1462 free(x, M_BWMETER);
1463 }
1464
1465 MRTSTAT_INC(mrts_cache_cleanups);
1466 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1467 (u_long)ntohl(rt->mfc_origin.s_addr),
1468 (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1469
1470 expire_mfc(rt);
1471 }
1472 }
1473
1474 MFC_UNLOCK();
1475
1476 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1477 curvnet);
1478
1479 CURVNET_RESTORE();
1480 }
1481
1482 /*
1483 * Packet forwarding routine once entry in the cache is made
1484 */
1485 static int
1486 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1487 {
1488 struct ip *ip = mtod(m, struct ip *);
1489 vifi_t vifi;
1490 int plen = ntohs(ip->ip_len);
1491
1492 VIF_LOCK_ASSERT();
1493
1494 /*
1495 * If xmt_vif is not -1, send on only the requested vif.
1496 *
1497 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1498 */
1499 if (xmt_vif < V_numvifs) {
1500 if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1501 pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1502 else
1503 phyint_send(ip, V_viftable + xmt_vif, m);
1504 return 1;
1505 }
1506
1507 /*
1508 * Don't forward if it didn't arrive from the parent vif for its origin.
1509 */
1510 vifi = rt->mfc_parent;
1511 if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1512 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1513 __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1514 MRTSTAT_INC(mrts_wrong_if);
1515 ++rt->mfc_wrong_if;
1516 /*
1517 * If we are doing PIM assert processing, send a message
1518 * to the routing daemon.
1519 *
1520 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1521 * can complete the SPT switch, regardless of the type
1522 * of the iif (broadcast media, GRE tunnel, etc).
1523 */
1524 if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1525 V_viftable[vifi].v_ifp) {
1526
1527 if (ifp == &V_multicast_register_if)
1528 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1529
1530 /* Get vifi for the incoming packet */
1531 for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1532 vifi++)
1533 ;
1534 if (vifi >= V_numvifs)
1535 return 0; /* The iif is not found: ignore the packet. */
1536
1537 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1538 return 0; /* WRONGVIF disabled: ignore the packet */
1539
1540 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1541 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1542 struct igmpmsg *im;
1543 int hlen = ip->ip_hl << 2;
1544 struct mbuf *mm = m_copy(m, 0, hlen);
1545
1546 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1547 mm = m_pullup(mm, hlen);
1548 if (mm == NULL)
1549 return ENOBUFS;
1550
1551 im = mtod(mm, struct igmpmsg *);
1552 im->im_msgtype = IGMPMSG_WRONGVIF;
1553 im->im_mbz = 0;
1554 im->im_vif = vifi;
1555
1556 MRTSTAT_INC(mrts_upcalls);
1557
1558 k_igmpsrc.sin_addr = im->im_src;
1559 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1560 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1561 MRTSTAT_INC(mrts_upq_sockfull);
1562 return ENOBUFS;
1563 }
1564 }
1565 }
1566 return 0;
1567 }
1568
1569
1570 /* If I sourced this packet, it counts as output, else it was input. */
1571 if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1572 V_viftable[vifi].v_pkt_out++;
1573 V_viftable[vifi].v_bytes_out += plen;
1574 } else {
1575 V_viftable[vifi].v_pkt_in++;
1576 V_viftable[vifi].v_bytes_in += plen;
1577 }
1578 rt->mfc_pkt_cnt++;
1579 rt->mfc_byte_cnt += plen;
1580
1581 /*
1582 * For each vif, decide if a copy of the packet should be forwarded.
1583 * Forward if:
1584 * - the ttl exceeds the vif's threshold
1585 * - there are group members downstream on interface
1586 */
1587 for (vifi = 0; vifi < V_numvifs; vifi++)
1588 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1589 V_viftable[vifi].v_pkt_out++;
1590 V_viftable[vifi].v_bytes_out += plen;
1591 if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1592 pim_register_send(ip, V_viftable + vifi, m, rt);
1593 else
1594 phyint_send(ip, V_viftable + vifi, m);
1595 }
1596
1597 /*
1598 * Perform upcall-related bw measuring.
1599 */
1600 if (rt->mfc_bw_meter != NULL) {
1601 struct bw_meter *x;
1602 struct timeval now;
1603
1604 microtime(&now);
1605 MFC_LOCK_ASSERT();
1606 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1607 bw_meter_receive_packet(x, plen, &now);
1608 }
1609
1610 return 0;
1611 }
1612
1613 /*
1614 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1615 */
1616 static int
1617 X_legal_vif_num(int vif)
1618 {
1619 int ret;
1620
1621 ret = 0;
1622 if (vif < 0)
1623 return (ret);
1624
1625 VIF_LOCK();
1626 if (vif < V_numvifs)
1627 ret = 1;
1628 VIF_UNLOCK();
1629
1630 return (ret);
1631 }
1632
1633 /*
1634 * Return the local address used by this vif
1635 */
1636 static u_long
1637 X_ip_mcast_src(int vifi)
1638 {
1639 in_addr_t addr;
1640
1641 addr = INADDR_ANY;
1642 if (vifi < 0)
1643 return (addr);
1644
1645 VIF_LOCK();
1646 if (vifi < V_numvifs)
1647 addr = V_viftable[vifi].v_lcl_addr.s_addr;
1648 VIF_UNLOCK();
1649
1650 return (addr);
1651 }
1652
1653 static void
1654 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1655 {
1656 struct mbuf *mb_copy;
1657 int hlen = ip->ip_hl << 2;
1658
1659 VIF_LOCK_ASSERT();
1660
1661 /*
1662 * Make a new reference to the packet; make sure that
1663 * the IP header is actually copied, not just referenced,
1664 * so that ip_output() only scribbles on the copy.
1665 */
1666 mb_copy = m_copypacket(m, M_NOWAIT);
1667 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1668 mb_copy = m_pullup(mb_copy, hlen);
1669 if (mb_copy == NULL)
1670 return;
1671
1672 send_packet(vifp, mb_copy);
1673 }
1674
1675 static void
1676 send_packet(struct vif *vifp, struct mbuf *m)
1677 {
1678 struct ip_moptions imo;
1679 struct in_multi *imm[2];
1680 int error;
1681
1682 VIF_LOCK_ASSERT();
1683
1684 imo.imo_multicast_ifp = vifp->v_ifp;
1685 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1686 imo.imo_multicast_loop = 1;
1687 imo.imo_multicast_vif = -1;
1688 imo.imo_num_memberships = 0;
1689 imo.imo_max_memberships = 2;
1690 imo.imo_membership = &imm[0];
1691
1692 /*
1693 * Re-entrancy should not be a problem here, because
1694 * the packets that we send out and are looped back at us
1695 * should get rejected because they appear to come from
1696 * the loopback interface, thus preventing looping.
1697 */
1698 error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1699 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1700 (ptrdiff_t)(vifp - V_viftable), error);
1701 }
1702
1703 /*
1704 * Stubs for old RSVP socket shim implementation.
1705 */
1706
1707 static int
1708 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1709 {
1710
1711 return (EOPNOTSUPP);
1712 }
1713
1714 static void
1715 X_ip_rsvp_force_done(struct socket *so __unused)
1716 {
1717
1718 }
1719
1720 static void
1721 X_rsvp_input(struct mbuf *m, int off __unused)
1722 {
1723
1724 if (!V_rsvp_on)
1725 m_freem(m);
1726 }
1727
1728 /*
1729 * Code for bandwidth monitors
1730 */
1731
1732 /*
1733 * Define common interface for timeval-related methods
1734 */
1735 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1736 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1737 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1738
1739 static uint32_t
1740 compute_bw_meter_flags(struct bw_upcall *req)
1741 {
1742 uint32_t flags = 0;
1743
1744 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1745 flags |= BW_METER_UNIT_PACKETS;
1746 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1747 flags |= BW_METER_UNIT_BYTES;
1748 if (req->bu_flags & BW_UPCALL_GEQ)
1749 flags |= BW_METER_GEQ;
1750 if (req->bu_flags & BW_UPCALL_LEQ)
1751 flags |= BW_METER_LEQ;
1752
1753 return flags;
1754 }
1755
1756 /*
1757 * Add a bw_meter entry
1758 */
1759 static int
1760 add_bw_upcall(struct bw_upcall *req)
1761 {
1762 struct mfc *mfc;
1763 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1764 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1765 struct timeval now;
1766 struct bw_meter *x;
1767 uint32_t flags;
1768
1769 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1770 return EOPNOTSUPP;
1771
1772 /* Test if the flags are valid */
1773 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1774 return EINVAL;
1775 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1776 return EINVAL;
1777 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1778 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1779 return EINVAL;
1780
1781 /* Test if the threshold time interval is valid */
1782 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1783 return EINVAL;
1784
1785 flags = compute_bw_meter_flags(req);
1786
1787 /*
1788 * Find if we have already same bw_meter entry
1789 */
1790 MFC_LOCK();
1791 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1792 if (mfc == NULL) {
1793 MFC_UNLOCK();
1794 return EADDRNOTAVAIL;
1795 }
1796 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1797 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1798 &req->bu_threshold.b_time, ==)) &&
1799 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1800 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1801 (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
1802 MFC_UNLOCK();
1803 return 0; /* XXX Already installed */
1804 }
1805 }
1806
1807 /* Allocate the new bw_meter entry */
1808 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1809 if (x == NULL) {
1810 MFC_UNLOCK();
1811 return ENOBUFS;
1812 }
1813
1814 /* Set the new bw_meter entry */
1815 x->bm_threshold.b_time = req->bu_threshold.b_time;
1816 microtime(&now);
1817 x->bm_start_time = now;
1818 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1819 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1820 x->bm_measured.b_packets = 0;
1821 x->bm_measured.b_bytes = 0;
1822 x->bm_flags = flags;
1823 x->bm_time_next = NULL;
1824 x->bm_time_hash = BW_METER_BUCKETS;
1825
1826 /* Add the new bw_meter entry to the front of entries for this MFC */
1827 x->bm_mfc = mfc;
1828 x->bm_mfc_next = mfc->mfc_bw_meter;
1829 mfc->mfc_bw_meter = x;
1830 schedule_bw_meter(x, &now);
1831 MFC_UNLOCK();
1832
1833 return 0;
1834 }
1835
1836 static void
1837 free_bw_list(struct bw_meter *list)
1838 {
1839 while (list != NULL) {
1840 struct bw_meter *x = list;
1841
1842 list = list->bm_mfc_next;
1843 unschedule_bw_meter(x);
1844 free(x, M_BWMETER);
1845 }
1846 }
1847
1848 /*
1849 * Delete one or multiple bw_meter entries
1850 */
1851 static int
1852 del_bw_upcall(struct bw_upcall *req)
1853 {
1854 struct mfc *mfc;
1855 struct bw_meter *x;
1856
1857 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1858 return EOPNOTSUPP;
1859
1860 MFC_LOCK();
1861
1862 /* Find the corresponding MFC entry */
1863 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1864 if (mfc == NULL) {
1865 MFC_UNLOCK();
1866 return EADDRNOTAVAIL;
1867 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1868 /*
1869 * Delete all bw_meter entries for this mfc
1870 */
1871 struct bw_meter *list;
1872
1873 list = mfc->mfc_bw_meter;
1874 mfc->mfc_bw_meter = NULL;
1875 free_bw_list(list);
1876 MFC_UNLOCK();
1877 return 0;
1878 } else { /* Delete a single bw_meter entry */
1879 struct bw_meter *prev;
1880 uint32_t flags = 0;
1881
1882 flags = compute_bw_meter_flags(req);
1883
1884 /* Find the bw_meter entry to delete */
1885 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1886 prev = x, x = x->bm_mfc_next) {
1887 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1888 &req->bu_threshold.b_time, ==)) &&
1889 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1890 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1891 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
1892 break;
1893 }
1894 if (x != NULL) { /* Delete entry from the list for this MFC */
1895 if (prev != NULL)
1896 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
1897 else
1898 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1899
1900 unschedule_bw_meter(x);
1901 MFC_UNLOCK();
1902 /* Free the bw_meter entry */
1903 free(x, M_BWMETER);
1904 return 0;
1905 } else {
1906 MFC_UNLOCK();
1907 return EINVAL;
1908 }
1909 }
1910 /* NOTREACHED */
1911 }
1912
1913 /*
1914 * Perform bandwidth measurement processing that may result in an upcall
1915 */
1916 static void
1917 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1918 {
1919 struct timeval delta;
1920
1921 MFC_LOCK_ASSERT();
1922
1923 delta = *nowp;
1924 BW_TIMEVALDECR(&delta, &x->bm_start_time);
1925
1926 if (x->bm_flags & BW_METER_GEQ) {
1927 /*
1928 * Processing for ">=" type of bw_meter entry
1929 */
1930 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1931 /* Reset the bw_meter entry */
1932 x->bm_start_time = *nowp;
1933 x->bm_measured.b_packets = 0;
1934 x->bm_measured.b_bytes = 0;
1935 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1936 }
1937
1938 /* Record that a packet is received */
1939 x->bm_measured.b_packets++;
1940 x->bm_measured.b_bytes += plen;
1941
1942 /*
1943 * Test if we should deliver an upcall
1944 */
1945 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1946 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1947 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1948 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1949 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1950 /* Prepare an upcall for delivery */
1951 bw_meter_prepare_upcall(x, nowp);
1952 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1953 }
1954 }
1955 } else if (x->bm_flags & BW_METER_LEQ) {
1956 /*
1957 * Processing for "<=" type of bw_meter entry
1958 */
1959 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1960 /*
1961 * We are behind time with the multicast forwarding table
1962 * scanning for "<=" type of bw_meter entries, so test now
1963 * if we should deliver an upcall.
1964 */
1965 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1966 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1967 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1968 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1969 /* Prepare an upcall for delivery */
1970 bw_meter_prepare_upcall(x, nowp);
1971 }
1972 /* Reschedule the bw_meter entry */
1973 unschedule_bw_meter(x);
1974 schedule_bw_meter(x, nowp);
1975 }
1976
1977 /* Record that a packet is received */
1978 x->bm_measured.b_packets++;
1979 x->bm_measured.b_bytes += plen;
1980
1981 /*
1982 * Test if we should restart the measuring interval
1983 */
1984 if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1985 x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1986 (x->bm_flags & BW_METER_UNIT_BYTES &&
1987 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1988 /* Don't restart the measuring interval */
1989 } else {
1990 /* Do restart the measuring interval */
1991 /*
1992 * XXX: note that we don't unschedule and schedule, because this
1993 * might be too much overhead per packet. Instead, when we process
1994 * all entries for a given timer hash bin, we check whether it is
1995 * really a timeout. If not, we reschedule at that time.
1996 */
1997 x->bm_start_time = *nowp;
1998 x->bm_measured.b_packets = 0;
1999 x->bm_measured.b_bytes = 0;
2000 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2001 }
2002 }
2003 }
2004
2005 /*
2006 * Prepare a bandwidth-related upcall
2007 */
2008 static void
2009 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2010 {
2011 struct timeval delta;
2012 struct bw_upcall *u;
2013
2014 MFC_LOCK_ASSERT();
2015
2016 /*
2017 * Compute the measured time interval
2018 */
2019 delta = *nowp;
2020 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2021
2022 /*
2023 * If there are too many pending upcalls, deliver them now
2024 */
2025 if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2026 bw_upcalls_send();
2027
2028 /*
2029 * Set the bw_upcall entry
2030 */
2031 u = &V_bw_upcalls[V_bw_upcalls_n++];
2032 u->bu_src = x->bm_mfc->mfc_origin;
2033 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2034 u->bu_threshold.b_time = x->bm_threshold.b_time;
2035 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2036 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2037 u->bu_measured.b_time = delta;
2038 u->bu_measured.b_packets = x->bm_measured.b_packets;
2039 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2040 u->bu_flags = 0;
2041 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2042 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2043 if (x->bm_flags & BW_METER_UNIT_BYTES)
2044 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2045 if (x->bm_flags & BW_METER_GEQ)
2046 u->bu_flags |= BW_UPCALL_GEQ;
2047 if (x->bm_flags & BW_METER_LEQ)
2048 u->bu_flags |= BW_UPCALL_LEQ;
2049 }
2050
2051 /*
2052 * Send the pending bandwidth-related upcalls
2053 */
2054 static void
2055 bw_upcalls_send(void)
2056 {
2057 struct mbuf *m;
2058 int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2059 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2060 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2061 0, /* unused2 */
2062 IGMPMSG_BW_UPCALL,/* im_msgtype */
2063 0, /* im_mbz */
2064 0, /* im_vif */
2065 0, /* unused3 */
2066 { 0 }, /* im_src */
2067 { 0 } }; /* im_dst */
2068
2069 MFC_LOCK_ASSERT();
2070
2071 if (V_bw_upcalls_n == 0)
2072 return; /* No pending upcalls */
2073
2074 V_bw_upcalls_n = 0;
2075
2076 /*
2077 * Allocate a new mbuf, initialize it with the header and
2078 * the payload for the pending calls.
2079 */
2080 m = m_gethdr(M_NOWAIT, MT_DATA);
2081 if (m == NULL) {
2082 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2083 return;
2084 }
2085
2086 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2087 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2088
2089 /*
2090 * Send the upcalls
2091 * XXX do we need to set the address in k_igmpsrc ?
2092 */
2093 MRTSTAT_INC(mrts_upcalls);
2094 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2095 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2096 MRTSTAT_INC(mrts_upq_sockfull);
2097 }
2098 }
2099
2100 /*
2101 * Compute the timeout hash value for the bw_meter entries
2102 */
2103 #define BW_METER_TIMEHASH(bw_meter, hash) \
2104 do { \
2105 struct timeval next_timeval = (bw_meter)->bm_start_time; \
2106 \
2107 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2108 (hash) = next_timeval.tv_sec; \
2109 if (next_timeval.tv_usec) \
2110 (hash)++; /* XXX: make sure we don't timeout early */ \
2111 (hash) %= BW_METER_BUCKETS; \
2112 } while (0)
2113
2114 /*
2115 * Schedule a timer to process periodically bw_meter entry of type "<="
2116 * by linking the entry in the proper hash bucket.
2117 */
2118 static void
2119 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2120 {
2121 int time_hash;
2122
2123 MFC_LOCK_ASSERT();
2124
2125 if (!(x->bm_flags & BW_METER_LEQ))
2126 return; /* XXX: we schedule timers only for "<=" entries */
2127
2128 /*
2129 * Reset the bw_meter entry
2130 */
2131 x->bm_start_time = *nowp;
2132 x->bm_measured.b_packets = 0;
2133 x->bm_measured.b_bytes = 0;
2134 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2135
2136 /*
2137 * Compute the timeout hash value and insert the entry
2138 */
2139 BW_METER_TIMEHASH(x, time_hash);
2140 x->bm_time_next = V_bw_meter_timers[time_hash];
2141 V_bw_meter_timers[time_hash] = x;
2142 x->bm_time_hash = time_hash;
2143 }
2144
2145 /*
2146 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2147 * by removing the entry from the proper hash bucket.
2148 */
2149 static void
2150 unschedule_bw_meter(struct bw_meter *x)
2151 {
2152 int time_hash;
2153 struct bw_meter *prev, *tmp;
2154
2155 MFC_LOCK_ASSERT();
2156
2157 if (!(x->bm_flags & BW_METER_LEQ))
2158 return; /* XXX: we schedule timers only for "<=" entries */
2159
2160 /*
2161 * Compute the timeout hash value and delete the entry
2162 */
2163 time_hash = x->bm_time_hash;
2164 if (time_hash >= BW_METER_BUCKETS)
2165 return; /* Entry was not scheduled */
2166
2167 for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2168 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2169 if (tmp == x)
2170 break;
2171
2172 if (tmp == NULL)
2173 panic("unschedule_bw_meter: bw_meter entry not found");
2174
2175 if (prev != NULL)
2176 prev->bm_time_next = x->bm_time_next;
2177 else
2178 V_bw_meter_timers[time_hash] = x->bm_time_next;
2179
2180 x->bm_time_next = NULL;
2181 x->bm_time_hash = BW_METER_BUCKETS;
2182 }
2183
2184
2185 /*
2186 * Process all "<=" type of bw_meter that should be processed now,
2187 * and for each entry prepare an upcall if necessary. Each processed
2188 * entry is rescheduled again for the (periodic) processing.
2189 *
2190 * This is run periodically (once per second normally). On each round,
2191 * all the potentially matching entries are in the hash slot that we are
2192 * looking at.
2193 */
2194 static void
2195 bw_meter_process()
2196 {
2197 uint32_t loops;
2198 int i;
2199 struct timeval now, process_endtime;
2200
2201 microtime(&now);
2202 if (V_last_tv_sec == now.tv_sec)
2203 return; /* nothing to do */
2204
2205 loops = now.tv_sec - V_last_tv_sec;
2206 V_last_tv_sec = now.tv_sec;
2207 if (loops > BW_METER_BUCKETS)
2208 loops = BW_METER_BUCKETS;
2209
2210 MFC_LOCK();
2211 /*
2212 * Process all bins of bw_meter entries from the one after the last
2213 * processed to the current one. On entry, i points to the last bucket
2214 * visited, so we need to increment i at the beginning of the loop.
2215 */
2216 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2217 struct bw_meter *x, *tmp_list;
2218
2219 if (++i >= BW_METER_BUCKETS)
2220 i = 0;
2221
2222 /* Disconnect the list of bw_meter entries from the bin */
2223 tmp_list = V_bw_meter_timers[i];
2224 V_bw_meter_timers[i] = NULL;
2225
2226 /* Process the list of bw_meter entries */
2227 while (tmp_list != NULL) {
2228 x = tmp_list;
2229 tmp_list = tmp_list->bm_time_next;
2230
2231 /* Test if the time interval is over */
2232 process_endtime = x->bm_start_time;
2233 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2234 if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2235 /* Not yet: reschedule, but don't reset */
2236 int time_hash;
2237
2238 BW_METER_TIMEHASH(x, time_hash);
2239 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2240 /*
2241 * XXX: somehow the bin processing is a bit ahead of time.
2242 * Put the entry in the next bin.
2243 */
2244 if (++time_hash >= BW_METER_BUCKETS)
2245 time_hash = 0;
2246 }
2247 x->bm_time_next = V_bw_meter_timers[time_hash];
2248 V_bw_meter_timers[time_hash] = x;
2249 x->bm_time_hash = time_hash;
2250
2251 continue;
2252 }
2253
2254 /*
2255 * Test if we should deliver an upcall
2256 */
2257 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2258 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2259 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2260 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2261 /* Prepare an upcall for delivery */
2262 bw_meter_prepare_upcall(x, &now);
2263 }
2264
2265 /*
2266 * Reschedule for next processing
2267 */
2268 schedule_bw_meter(x, &now);
2269 }
2270 }
2271
2272 /* Send all upcalls that are pending delivery */
2273 bw_upcalls_send();
2274
2275 MFC_UNLOCK();
2276 }
2277
2278 /*
2279 * A periodic function for sending all upcalls that are pending delivery
2280 */
2281 static void
2282 expire_bw_upcalls_send(void *arg)
2283 {
2284 CURVNET_SET((struct vnet *) arg);
2285
2286 MFC_LOCK();
2287 bw_upcalls_send();
2288 MFC_UNLOCK();
2289
2290 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2291 curvnet);
2292 CURVNET_RESTORE();
2293 }
2294
2295 /*
2296 * A periodic function for periodic scanning of the multicast forwarding
2297 * table for processing all "<=" bw_meter entries.
2298 */
2299 static void
2300 expire_bw_meter_process(void *arg)
2301 {
2302 CURVNET_SET((struct vnet *) arg);
2303
2304 if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2305 bw_meter_process();
2306
2307 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2308 curvnet);
2309 CURVNET_RESTORE();
2310 }
2311
2312 /*
2313 * End of bandwidth monitoring code
2314 */
2315
2316 /*
2317 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2318 *
2319 */
2320 static int
2321 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2322 struct mfc *rt)
2323 {
2324 struct mbuf *mb_copy, *mm;
2325
2326 /*
2327 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2328 * rendezvous point was unspecified, and we were told not to.
2329 */
2330 if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2331 in_nullhost(rt->mfc_rp))
2332 return 0;
2333
2334 mb_copy = pim_register_prepare(ip, m);
2335 if (mb_copy == NULL)
2336 return ENOBUFS;
2337
2338 /*
2339 * Send all the fragments. Note that the mbuf for each fragment
2340 * is freed by the sending machinery.
2341 */
2342 for (mm = mb_copy; mm; mm = mb_copy) {
2343 mb_copy = mm->m_nextpkt;
2344 mm->m_nextpkt = 0;
2345 mm = m_pullup(mm, sizeof(struct ip));
2346 if (mm != NULL) {
2347 ip = mtod(mm, struct ip *);
2348 if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2349 pim_register_send_rp(ip, vifp, mm, rt);
2350 } else {
2351 pim_register_send_upcall(ip, vifp, mm, rt);
2352 }
2353 }
2354 }
2355
2356 return 0;
2357 }
2358
2359 /*
2360 * Return a copy of the data packet that is ready for PIM Register
2361 * encapsulation.
2362 * XXX: Note that in the returned copy the IP header is a valid one.
2363 */
2364 static struct mbuf *
2365 pim_register_prepare(struct ip *ip, struct mbuf *m)
2366 {
2367 struct mbuf *mb_copy = NULL;
2368 int mtu;
2369
2370 /* Take care of delayed checksums */
2371 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2372 in_delayed_cksum(m);
2373 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2374 }
2375
2376 /*
2377 * Copy the old packet & pullup its IP header into the
2378 * new mbuf so we can modify it.
2379 */
2380 mb_copy = m_copypacket(m, M_NOWAIT);
2381 if (mb_copy == NULL)
2382 return NULL;
2383 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2384 if (mb_copy == NULL)
2385 return NULL;
2386
2387 /* take care of the TTL */
2388 ip = mtod(mb_copy, struct ip *);
2389 --ip->ip_ttl;
2390
2391 /* Compute the MTU after the PIM Register encapsulation */
2392 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2393
2394 if (ntohs(ip->ip_len) <= mtu) {
2395 /* Turn the IP header into a valid one */
2396 ip->ip_sum = 0;
2397 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2398 } else {
2399 /* Fragment the packet */
2400 mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2401 if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2402 m_freem(mb_copy);
2403 return NULL;
2404 }
2405 }
2406 return mb_copy;
2407 }
2408
2409 /*
2410 * Send an upcall with the data packet to the user-level process.
2411 */
2412 static int
2413 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2414 struct mbuf *mb_copy, struct mfc *rt)
2415 {
2416 struct mbuf *mb_first;
2417 int len = ntohs(ip->ip_len);
2418 struct igmpmsg *im;
2419 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2420
2421 VIF_LOCK_ASSERT();
2422
2423 /*
2424 * Add a new mbuf with an upcall header
2425 */
2426 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2427 if (mb_first == NULL) {
2428 m_freem(mb_copy);
2429 return ENOBUFS;
2430 }
2431 mb_first->m_data += max_linkhdr;
2432 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2433 mb_first->m_len = sizeof(struct igmpmsg);
2434 mb_first->m_next = mb_copy;
2435
2436 /* Send message to routing daemon */
2437 im = mtod(mb_first, struct igmpmsg *);
2438 im->im_msgtype = IGMPMSG_WHOLEPKT;
2439 im->im_mbz = 0;
2440 im->im_vif = vifp - V_viftable;
2441 im->im_src = ip->ip_src;
2442 im->im_dst = ip->ip_dst;
2443
2444 k_igmpsrc.sin_addr = ip->ip_src;
2445
2446 MRTSTAT_INC(mrts_upcalls);
2447
2448 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2449 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2450 MRTSTAT_INC(mrts_upq_sockfull);
2451 return ENOBUFS;
2452 }
2453
2454 /* Keep statistics */
2455 PIMSTAT_INC(pims_snd_registers_msgs);
2456 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2457
2458 return 0;
2459 }
2460
2461 /*
2462 * Encapsulate the data packet in PIM Register message and send it to the RP.
2463 */
2464 static int
2465 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2466 struct mfc *rt)
2467 {
2468 struct mbuf *mb_first;
2469 struct ip *ip_outer;
2470 struct pim_encap_pimhdr *pimhdr;
2471 int len = ntohs(ip->ip_len);
2472 vifi_t vifi = rt->mfc_parent;
2473
2474 VIF_LOCK_ASSERT();
2475
2476 if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2477 m_freem(mb_copy);
2478 return EADDRNOTAVAIL; /* The iif vif is invalid */
2479 }
2480
2481 /*
2482 * Add a new mbuf with the encapsulating header
2483 */
2484 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2485 if (mb_first == NULL) {
2486 m_freem(mb_copy);
2487 return ENOBUFS;
2488 }
2489 mb_first->m_data += max_linkhdr;
2490 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2491 mb_first->m_next = mb_copy;
2492
2493 mb_first->m_pkthdr.len = len + mb_first->m_len;
2494
2495 /*
2496 * Fill in the encapsulating IP and PIM header
2497 */
2498 ip_outer = mtod(mb_first, struct ip *);
2499 *ip_outer = pim_encap_iphdr;
2500 ip_outer->ip_id = ip_newid();
2501 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2502 sizeof(pim_encap_pimhdr));
2503 ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2504 ip_outer->ip_dst = rt->mfc_rp;
2505 /*
2506 * Copy the inner header TOS to the outer header, and take care of the
2507 * IP_DF bit.
2508 */
2509 ip_outer->ip_tos = ip->ip_tos;
2510 if (ip->ip_off & htons(IP_DF))
2511 ip_outer->ip_off |= htons(IP_DF);
2512 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2513 + sizeof(pim_encap_iphdr));
2514 *pimhdr = pim_encap_pimhdr;
2515 /* If the iif crosses a border, set the Border-bit */
2516 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2517 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2518
2519 mb_first->m_data += sizeof(pim_encap_iphdr);
2520 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2521 mb_first->m_data -= sizeof(pim_encap_iphdr);
2522
2523 send_packet(vifp, mb_first);
2524
2525 /* Keep statistics */
2526 PIMSTAT_INC(pims_snd_registers_msgs);
2527 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2528
2529 return 0;
2530 }
2531
2532 /*
2533 * pim_encapcheck() is called by the encap4_input() path at runtime to
2534 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2535 * into the kernel.
2536 */
2537 static int
2538 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2539 {
2540
2541 #ifdef DIAGNOSTIC
2542 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2543 #endif
2544 if (proto != IPPROTO_PIM)
2545 return 0; /* not for us; reject the datagram. */
2546
2547 return 64; /* claim the datagram. */
2548 }
2549
2550 /*
2551 * PIM-SMv2 and PIM-DM messages processing.
2552 * Receives and verifies the PIM control messages, and passes them
2553 * up to the listening socket, using rip_input().
2554 * The only message with special processing is the PIM_REGISTER message
2555 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2556 * is passed to if_simloop().
2557 */
2558 void
2559 pim_input(struct mbuf *m, int iphlen)
2560 {
2561 struct ip *ip = mtod(m, struct ip *);
2562 struct pim *pim;
2563 int minlen;
2564 int datalen = ntohs(ip->ip_len) - iphlen;
2565 int ip_tos;
2566
2567 /* Keep statistics */
2568 PIMSTAT_INC(pims_rcv_total_msgs);
2569 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2570
2571 /*
2572 * Validate lengths
2573 */
2574 if (datalen < PIM_MINLEN) {
2575 PIMSTAT_INC(pims_rcv_tooshort);
2576 CTR3(KTR_IPMF, "%s: short packet (%d) from %s",
2577 __func__, datalen, inet_ntoa(ip->ip_src));
2578 m_freem(m);
2579 return;
2580 }
2581
2582 /*
2583 * If the packet is at least as big as a REGISTER, go agead
2584 * and grab the PIM REGISTER header size, to avoid another
2585 * possible m_pullup() later.
2586 *
2587 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2588 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2589 */
2590 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2591 /*
2592 * Get the IP and PIM headers in contiguous memory, and
2593 * possibly the PIM REGISTER header.
2594 */
2595 if (m->m_len < minlen && (m = m_pullup(m, minlen)) == 0) {
2596 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2597 return;
2598 }
2599
2600 /* m_pullup() may have given us a new mbuf so reset ip. */
2601 ip = mtod(m, struct ip *);
2602 ip_tos = ip->ip_tos;
2603
2604 /* adjust mbuf to point to the PIM header */
2605 m->m_data += iphlen;
2606 m->m_len -= iphlen;
2607 pim = mtod(m, struct pim *);
2608
2609 /*
2610 * Validate checksum. If PIM REGISTER, exclude the data packet.
2611 *
2612 * XXX: some older PIMv2 implementations don't make this distinction,
2613 * so for compatibility reason perform the checksum over part of the
2614 * message, and if error, then over the whole message.
2615 */
2616 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2617 /* do nothing, checksum okay */
2618 } else if (in_cksum(m, datalen)) {
2619 PIMSTAT_INC(pims_rcv_badsum);
2620 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2621 m_freem(m);
2622 return;
2623 }
2624
2625 /* PIM version check */
2626 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2627 PIMSTAT_INC(pims_rcv_badversion);
2628 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2629 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2630 m_freem(m);
2631 return;
2632 }
2633
2634 /* restore mbuf back to the outer IP */
2635 m->m_data -= iphlen;
2636 m->m_len += iphlen;
2637
2638 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2639 /*
2640 * Since this is a REGISTER, we'll make a copy of the register
2641 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2642 * routing daemon.
2643 */
2644 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2645 struct mbuf *mcp;
2646 struct ip *encap_ip;
2647 u_int32_t *reghdr;
2648 struct ifnet *vifp;
2649
2650 VIF_LOCK();
2651 if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2652 VIF_UNLOCK();
2653 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2654 (int)V_reg_vif_num);
2655 m_freem(m);
2656 return;
2657 }
2658 /* XXX need refcnt? */
2659 vifp = V_viftable[V_reg_vif_num].v_ifp;
2660 VIF_UNLOCK();
2661
2662 /*
2663 * Validate length
2664 */
2665 if (datalen < PIM_REG_MINLEN) {
2666 PIMSTAT_INC(pims_rcv_tooshort);
2667 PIMSTAT_INC(pims_rcv_badregisters);
2668 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2669 m_freem(m);
2670 return;
2671 }
2672
2673 reghdr = (u_int32_t *)(pim + 1);
2674 encap_ip = (struct ip *)(reghdr + 1);
2675
2676 CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d",
2677 __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len));
2678
2679 /* verify the version number of the inner packet */
2680 if (encap_ip->ip_v != IPVERSION) {
2681 PIMSTAT_INC(pims_rcv_badregisters);
2682 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2683 m_freem(m);
2684 return;
2685 }
2686
2687 /* verify the inner packet is destined to a mcast group */
2688 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2689 PIMSTAT_INC(pims_rcv_badregisters);
2690 CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__,
2691 inet_ntoa(encap_ip->ip_dst));
2692 m_freem(m);
2693 return;
2694 }
2695
2696 /* If a NULL_REGISTER, pass it to the daemon */
2697 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2698 goto pim_input_to_daemon;
2699
2700 /*
2701 * Copy the TOS from the outer IP header to the inner IP header.
2702 */
2703 if (encap_ip->ip_tos != ip_tos) {
2704 /* Outer TOS -> inner TOS */
2705 encap_ip->ip_tos = ip_tos;
2706 /* Recompute the inner header checksum. Sigh... */
2707
2708 /* adjust mbuf to point to the inner IP header */
2709 m->m_data += (iphlen + PIM_MINLEN);
2710 m->m_len -= (iphlen + PIM_MINLEN);
2711
2712 encap_ip->ip_sum = 0;
2713 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2714
2715 /* restore mbuf to point back to the outer IP header */
2716 m->m_data -= (iphlen + PIM_MINLEN);
2717 m->m_len += (iphlen + PIM_MINLEN);
2718 }
2719
2720 /*
2721 * Decapsulate the inner IP packet and loopback to forward it
2722 * as a normal multicast packet. Also, make a copy of the
2723 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2724 * to pass to the daemon later, so it can take the appropriate
2725 * actions (e.g., send back PIM_REGISTER_STOP).
2726 * XXX: here m->m_data points to the outer IP header.
2727 */
2728 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
2729 if (mcp == NULL) {
2730 CTR1(KTR_IPMF, "%s: m_copy() failed", __func__);
2731 m_freem(m);
2732 return;
2733 }
2734
2735 /* Keep statistics */
2736 /* XXX: registers_bytes include only the encap. mcast pkt */
2737 PIMSTAT_INC(pims_rcv_registers_msgs);
2738 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2739
2740 /*
2741 * forward the inner ip packet; point m_data at the inner ip.
2742 */
2743 m_adj(m, iphlen + PIM_MINLEN);
2744
2745 CTR4(KTR_IPMF,
2746 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2747 __func__,
2748 (u_long)ntohl(encap_ip->ip_src.s_addr),
2749 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2750 (int)V_reg_vif_num);
2751
2752 /* NB: vifp was collected above; can it change on us? */
2753 if_simloop(vifp, m, dst.sin_family, 0);
2754
2755 /* prepare the register head to send to the mrouting daemon */
2756 m = mcp;
2757 }
2758
2759 pim_input_to_daemon:
2760 /*
2761 * Pass the PIM message up to the daemon; if it is a Register message,
2762 * pass the 'head' only up to the daemon. This includes the
2763 * outer IP header, PIM header, PIM-Register header and the
2764 * inner IP header.
2765 * XXX: the outer IP header pkt size of a Register is not adjust to
2766 * reflect the fact that the inner multicast data is truncated.
2767 */
2768 rip_input(m, iphlen);
2769
2770 return;
2771 }
2772
2773 static int
2774 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2775 {
2776 struct mfc *rt;
2777 int error, i;
2778
2779 if (req->newptr)
2780 return (EPERM);
2781 if (V_mfchashtbl == NULL) /* XXX unlocked */
2782 return (0);
2783 error = sysctl_wire_old_buffer(req, 0);
2784 if (error)
2785 return (error);
2786
2787 MFC_LOCK();
2788 for (i = 0; i < mfchashsize; i++) {
2789 LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2790 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2791 if (error)
2792 goto out_locked;
2793 }
2794 }
2795 out_locked:
2796 MFC_UNLOCK();
2797 return (error);
2798 }
2799
2800 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2801 sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2802 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2803
2804 static void
2805 vnet_mroute_init(const void *unused __unused)
2806 {
2807
2808 MALLOC(V_nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2809 bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2810 callout_init(&V_expire_upcalls_ch, CALLOUT_MPSAFE);
2811 callout_init(&V_bw_upcalls_ch, CALLOUT_MPSAFE);
2812 callout_init(&V_bw_meter_ch, CALLOUT_MPSAFE);
2813 }
2814
2815 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mroute_init,
2816 NULL);
2817
2818 static void
2819 vnet_mroute_uninit(const void *unused __unused)
2820 {
2821
2822 FREE(V_nexpire, M_MRTABLE);
2823 V_nexpire = NULL;
2824 }
2825
2826 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE,
2827 vnet_mroute_uninit, NULL);
2828
2829 static int
2830 ip_mroute_modevent(module_t mod, int type, void *unused)
2831 {
2832
2833 switch (type) {
2834 case MOD_LOAD:
2835 MROUTER_LOCK_INIT();
2836
2837 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2838 if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2839 if (if_detach_event_tag == NULL) {
2840 printf("ip_mroute: unable to register "
2841 "ifnet_departure_event handler\n");
2842 MROUTER_LOCK_DESTROY();
2843 return (EINVAL);
2844 }
2845
2846 MFC_LOCK_INIT();
2847 VIF_LOCK_INIT();
2848
2849 mfchashsize = MFCHASHSIZE;
2850 if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2851 !powerof2(mfchashsize)) {
2852 printf("WARNING: %s not a power of 2; using default\n",
2853 "net.inet.ip.mfchashsize");
2854 mfchashsize = MFCHASHSIZE;
2855 }
2856
2857 pim_squelch_wholepkt = 0;
2858 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2859 &pim_squelch_wholepkt);
2860
2861 pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2862 pim_encapcheck, &in_pim_protosw, NULL);
2863 if (pim_encap_cookie == NULL) {
2864 printf("ip_mroute: unable to attach pim encap\n");
2865 VIF_LOCK_DESTROY();
2866 MFC_LOCK_DESTROY();
2867 MROUTER_LOCK_DESTROY();
2868 return (EINVAL);
2869 }
2870
2871 ip_mcast_src = X_ip_mcast_src;
2872 ip_mforward = X_ip_mforward;
2873 ip_mrouter_done = X_ip_mrouter_done;
2874 ip_mrouter_get = X_ip_mrouter_get;
2875 ip_mrouter_set = X_ip_mrouter_set;
2876
2877 ip_rsvp_force_done = X_ip_rsvp_force_done;
2878 ip_rsvp_vif = X_ip_rsvp_vif;
2879
2880 legal_vif_num = X_legal_vif_num;
2881 mrt_ioctl = X_mrt_ioctl;
2882 rsvp_input_p = X_rsvp_input;
2883 break;
2884
2885 case MOD_UNLOAD:
2886 /*
2887 * Typically module unload happens after the user-level
2888 * process has shutdown the kernel services (the check
2889 * below insures someone can't just yank the module out
2890 * from under a running process). But if the module is
2891 * just loaded and then unloaded w/o starting up a user
2892 * process we still need to cleanup.
2893 */
2894 MROUTER_LOCK();
2895 if (ip_mrouter_cnt != 0) {
2896 MROUTER_UNLOCK();
2897 return (EINVAL);
2898 }
2899 ip_mrouter_unloading = 1;
2900 MROUTER_UNLOCK();
2901
2902 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2903
2904 if (pim_encap_cookie) {
2905 encap_detach(pim_encap_cookie);
2906 pim_encap_cookie = NULL;
2907 }
2908
2909 ip_mcast_src = NULL;
2910 ip_mforward = NULL;
2911 ip_mrouter_done = NULL;
2912 ip_mrouter_get = NULL;
2913 ip_mrouter_set = NULL;
2914
2915 ip_rsvp_force_done = NULL;
2916 ip_rsvp_vif = NULL;
2917
2918 legal_vif_num = NULL;
2919 mrt_ioctl = NULL;
2920 rsvp_input_p = NULL;
2921
2922 VIF_LOCK_DESTROY();
2923 MFC_LOCK_DESTROY();
2924 MROUTER_LOCK_DESTROY();
2925 break;
2926
2927 default:
2928 return EOPNOTSUPP;
2929 }
2930 return 0;
2931 }
2932
2933 static moduledata_t ip_mroutemod = {
2934 "ip_mroute",
2935 ip_mroute_modevent,
2936 0
2937 };
2938
2939 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_MIDDLE);
Cache object: bc2f42f3e570a30b08f93d3dd6c74a9c
|