1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1989 Stephen Deering
5 * Copyright (c) 1992, 1993
6 * The Regents of the University of California. All rights reserved.
7 *
8 * This code is derived from software contributed to Berkeley by
9 * Stephen Deering of Stanford University.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
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 * Modified by Wojciech Macek, Semihalf, May 2021
51 *
52 * MROUTING Revision: 3.5
53 * and PIM-SMv2 and PIM-DM support, advanced API support,
54 * bandwidth metering and signaling
55 */
56
57 /*
58 * TODO: Prefix functions with ipmf_.
59 * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
60 * domain attachment (if_afdata) so we can track consumers of that service.
61 * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
62 * move it to socket options.
63 * TODO: Cleanup LSRR removal further.
64 * TODO: Push RSVP stubs into raw_ip.c.
65 * TODO: Use bitstring.h for vif set.
66 * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
67 * TODO: Sync ip6_mroute.c with this file.
68 */
69
70 #include "opt_inet.h"
71 #include "opt_mrouting.h"
72
73 #define _PIM_VT 1
74
75 #include <sys/types.h>
76 #include <sys/param.h>
77 #include <sys/kernel.h>
78 #include <sys/stddef.h>
79 #include <sys/condvar.h>
80 #include <sys/eventhandler.h>
81 #include <sys/lock.h>
82 #include <sys/kthread.h>
83 #include <sys/ktr.h>
84 #include <sys/malloc.h>
85 #include <sys/mbuf.h>
86 #include <sys/module.h>
87 #include <sys/priv.h>
88 #include <sys/proc.h>
89 #include <sys/protosw.h>
90 #include <sys/signalvar.h>
91 #include <sys/socket.h>
92 #include <sys/socketvar.h>
93 #include <sys/sockio.h>
94 #include <sys/sx.h>
95 #include <sys/sysctl.h>
96 #include <sys/syslog.h>
97 #include <sys/systm.h>
98 #include <sys/taskqueue.h>
99 #include <sys/time.h>
100 #include <sys/counter.h>
101 #include <machine/atomic.h>
102
103 #include <net/if.h>
104 #include <net/if_var.h>
105 #include <net/if_private.h>
106 #include <net/if_types.h>
107 #include <net/netisr.h>
108 #include <net/route.h>
109 #include <net/vnet.h>
110
111 #include <netinet/in.h>
112 #include <netinet/igmp.h>
113 #include <netinet/in_systm.h>
114 #include <netinet/in_var.h>
115 #include <netinet/ip.h>
116 #include <netinet/ip_encap.h>
117 #include <netinet/ip_mroute.h>
118 #include <netinet/ip_var.h>
119 #include <netinet/ip_options.h>
120 #include <netinet/pim.h>
121 #include <netinet/pim_var.h>
122 #include <netinet/udp.h>
123
124 #include <machine/in_cksum.h>
125
126 #ifndef KTR_IPMF
127 #define KTR_IPMF KTR_INET
128 #endif
129
130 #define VIFI_INVALID ((vifi_t) -1)
131
132 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
133
134 /*
135 * Locking. We use two locks: one for the virtual interface table and
136 * one for the forwarding table. These locks may be nested in which case
137 * the VIF lock must always be taken first. Note that each lock is used
138 * to cover not only the specific data structure but also related data
139 * structures.
140 */
141
142 static struct sx __exclusive_cache_line mrouter_teardown;
143 #define MRW_TEARDOWN_WLOCK() sx_xlock(&mrouter_teardown)
144 #define MRW_TEARDOWN_WUNLOCK() sx_xunlock(&mrouter_teardown)
145 #define MRW_TEARDOWN_LOCK_INIT() \
146 sx_init(&mrouter_teardown, "IPv4 multicast forwarding teardown")
147 #define MRW_TEARDOWN_LOCK_DESTROY() sx_destroy(&mrouter_teardown)
148
149 static struct rwlock mrouter_lock;
150 #define MRW_RLOCK() rw_rlock(&mrouter_lock)
151 #define MRW_WLOCK() rw_wlock(&mrouter_lock)
152 #define MRW_RUNLOCK() rw_runlock(&mrouter_lock)
153 #define MRW_WUNLOCK() rw_wunlock(&mrouter_lock)
154 #define MRW_UNLOCK() rw_unlock(&mrouter_lock)
155 #define MRW_LOCK_ASSERT() rw_assert(&mrouter_lock, RA_LOCKED)
156 #define MRW_WLOCK_ASSERT() rw_assert(&mrouter_lock, RA_WLOCKED)
157 #define MRW_LOCK_TRY_UPGRADE() rw_try_upgrade(&mrouter_lock)
158 #define MRW_WOWNED() rw_wowned(&mrouter_lock)
159 #define MRW_LOCK_INIT() \
160 rw_init(&mrouter_lock, "IPv4 multicast forwarding")
161 #define MRW_LOCK_DESTROY() rw_destroy(&mrouter_lock)
162
163 static int ip_mrouter_cnt; /* # of vnets with active mrouters */
164 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
165
166 VNET_PCPUSTAT_DEFINE_STATIC(struct mrtstat, mrtstat);
167 VNET_PCPUSTAT_SYSINIT(mrtstat);
168 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
169 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
170 mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
171 "netinet/ip_mroute.h)");
172
173 struct mfctable {
174 struct socket *router;
175 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
176 u_char *nexpire;
177 vifi_t numvifs;
178 struct vif viftable[MAXVIFS];
179
180 struct buf_ring *bw_upcalls;
181 struct mtx bw_upcalls_mtx;
182
183 struct ifnet *register_if;
184 vifi_t register_vif;
185
186 uint32_t api_config;
187 int pim_assert_enabled;
188 struct timeval pim_assert_interval;
189 };
190
191 VNET_DEFINE_STATIC(struct mfctable *, mfctables);
192 #define V_mfctables VNET(mfctables)
193 VNET_DEFINE_STATIC(uint32_t, nmfctables);
194 #define V_nmfctables VNET(nmfctables)
195
196 VNET_DEFINE_STATIC(u_long, mfchash);
197 #define V_mfchash VNET(mfchash)
198 #define MFCHASH(a, g) \
199 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
200 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
201 #define MFCHASHSIZE 256
202
203 static u_long mfchashsize = MFCHASHSIZE; /* Hash size */
204 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, mfchashsize, CTLFLAG_RDTUN,
205 &mfchashsize, 0, "IPv4 Multicast Forwarding Table hash size");
206
207 VNET_DEFINE_STATIC(struct taskqueue *, task_queue);
208 #define V_task_queue VNET(task_queue)
209 VNET_DEFINE_STATIC(struct task, task);
210 #define V_task VNET(task)
211
212 static eventhandler_tag if_detach_event_tag;
213 static eventhandler_tag rtnumfibs_change_tag;
214
215 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
216 #define V_expire_upcalls_ch VNET(expire_upcalls_ch)
217
218 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
219 #define UPCALL_EXPIRE 6 /* number of timeouts */
220
221 /*
222 * Bandwidth meter variables and constants
223 */
224 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
225
226 /*
227 * Pending upcalls are stored in a ring which is flushed when
228 * full, or periodically
229 */
230 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
231 #define V_bw_upcalls_ch VNET(bw_upcalls_ch)
232
233 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
234
235 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
236 VNET_PCPUSTAT_SYSINIT(pimstat);
237 VNET_PCPUSTAT_SYSUNINIT(pimstat);
238
239 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
240 "PIM");
241 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
242 pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
243
244 static u_long pim_squelch_wholepkt = 0;
245 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RWTUN,
246 &pim_squelch_wholepkt, 0,
247 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
248
249 static const struct encaptab *pim_encap_cookie;
250 static int pim_encapcheck(const struct mbuf *, int, int, void *);
251 static int pim_input(struct mbuf *, int, int, void *);
252
253 extern int in_mcast_loop;
254
255 static const struct encap_config ipv4_encap_cfg = {
256 .proto = IPPROTO_PIM,
257 .min_length = sizeof(struct ip) + PIM_MINLEN,
258 .exact_match = 8,
259 .check = pim_encapcheck,
260 .input = pim_input
261 };
262
263 /*
264 * Note: the PIM Register encapsulation adds the following in front of a
265 * data packet:
266 *
267 * struct pim_encap_hdr {
268 * struct ip ip;
269 * struct pim_encap_pimhdr pim;
270 * }
271 *
272 */
273
274 struct pim_encap_pimhdr {
275 struct pim pim;
276 uint32_t flags;
277 };
278 #define PIM_ENCAP_TTL 64
279
280 static struct ip pim_encap_iphdr = {
281 #if BYTE_ORDER == LITTLE_ENDIAN
282 sizeof(struct ip) >> 2,
283 IPVERSION,
284 #else
285 IPVERSION,
286 sizeof(struct ip) >> 2,
287 #endif
288 0, /* tos */
289 sizeof(struct ip), /* total length */
290 0, /* id */
291 0, /* frag offset */
292 PIM_ENCAP_TTL,
293 IPPROTO_PIM,
294 0, /* checksum */
295 };
296
297 static struct pim_encap_pimhdr pim_encap_pimhdr = {
298 {
299 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
300 0, /* reserved */
301 0, /* checksum */
302 },
303 0 /* flags */
304 };
305
306 /*
307 * Private variables.
308 */
309
310 static u_long X_ip_mcast_src(int, int);
311 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
312 struct ip_moptions *);
313 static void X_ip_mrouter_done(struct socket *);
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, int);
317 static int X_mrt_ioctl(u_long, caddr_t, int);
318
319 static int add_bw_upcall(struct mfctable *, struct bw_upcall *);
320 static int add_mfc(struct mfctable *, struct mfcctl2 *);
321 static int add_vif(struct mfctable *, int, struct vifctl *);
322 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
323 static void bw_meter_geq_receive_packet(struct bw_meter *, int,
324 struct timeval *);
325 static void bw_upcalls_send(struct mfctable *);
326 static void bw_upcalls_send_all(void);
327 static int del_bw_upcall(struct mfctable *, struct bw_upcall *);
328 static int del_mfc(struct mfctable *, struct mfcctl2 *);
329 static int del_vif(struct mfctable *, vifi_t);
330 static int del_vif_locked(struct mfctable *, vifi_t, struct ifnet **,
331 struct ifnet **);
332 static void expire_bw_upcalls_send(void *);
333 static void expire_mfc(struct mfc *);
334 static void expire_upcalls(struct mfctable *);
335 static void expire_upcalls_all(void *);
336 static void free_bw_list(struct bw_meter *);
337 static int get_sg_cnt(struct mfctable *, struct sioc_sg_req *);
338 static int get_vif_cnt(struct mfctable *, struct sioc_vif_req *);
339 static void if_detached_event(void *, struct ifnet *);
340 static int ip_mdq(struct mfctable *, struct mbuf *, struct ifnet *,
341 struct mfc *, vifi_t);
342 static int ip_mrouter_init(struct socket *, int);
343 static __inline struct mfc *
344 mfc_find(const struct mfctable *mfct, const struct in_addr *,
345 const struct in_addr *);
346 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
347 static struct mbuf *
348 pim_register_prepare(struct ip *, struct mbuf *);
349 static int pim_register_send(struct mfctable *, struct ip *, struct vif *,
350 struct mbuf *, struct mfc *);
351 static int pim_register_send_rp(struct mfctable *, struct ip *,
352 struct vif *, struct mbuf *, struct mfc *);
353 static int pim_register_send_upcall(struct mfctable *, struct ip *,
354 struct vif *, struct mbuf *, struct mfc *);
355 static void send_packet(struct vif *, struct mbuf *);
356 static int set_api_config(struct mfctable *, uint32_t *);
357 static int set_assert(struct mfctable *, int);
358 static int socket_send(struct socket *, struct mbuf *,
359 struct sockaddr_in *);
360
361 /*
362 * Kernel multicast forwarding API capabilities and setup.
363 * If more API capabilities are added to the kernel, they should be
364 * recorded in `mrt_api_support'.
365 */
366 #define MRT_API_VERSION 0x0305
367
368 static const int mrt_api_version = MRT_API_VERSION;
369 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
370 MRT_MFC_FLAGS_BORDER_VIF |
371 MRT_MFC_RP |
372 MRT_MFC_BW_UPCALL);
373
374 /*
375 * Find a route for a given origin IP address and multicast group address.
376 * Statistics must be updated by the caller.
377 */
378 static struct mfc *
mfc_find(const struct mfctable * mfct,const struct in_addr * o,const struct in_addr * g)379 mfc_find(const struct mfctable *mfct, const struct in_addr *o,
380 const struct in_addr *g)
381 {
382 struct mfc *rt;
383
384 MRW_LOCK_ASSERT();
385
386 LIST_FOREACH(rt, &mfct->mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
387 if (in_hosteq(rt->mfc_origin, *o) &&
388 in_hosteq(rt->mfc_mcastgrp, *g) &&
389 buf_ring_empty(rt->mfc_stall_ring))
390 break;
391 }
392
393 return (rt);
394 }
395
396 static __inline struct mfc *
mfc_alloc(void)397 mfc_alloc(void)
398 {
399 struct mfc *rt;
400 rt = malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT | M_ZERO);
401 if (rt == NULL)
402 return rt;
403
404 rt->mfc_stall_ring = buf_ring_alloc(MAX_UPQ, M_MRTABLE, M_NOWAIT, NULL);
405 if (rt->mfc_stall_ring == NULL) {
406 free(rt, M_MRTABLE);
407 return NULL;
408 }
409
410 return rt;
411 }
412
413 static struct mfctable *
somfctable(struct socket * so)414 somfctable(struct socket *so)
415 {
416 int fib;
417
418 fib = atomic_load_int(&so->so_fibnum);
419 KASSERT(fib >= 0 && fib < V_nmfctables,
420 ("%s: so_fibnum %d out of range", __func__, fib));
421 return (&V_mfctables[fib]);
422 }
423
424 /*
425 * Handle MRT setsockopt commands to modify the multicast forwarding tables.
426 */
427 static int
X_ip_mrouter_set(struct socket * so,struct sockopt * sopt)428 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
429 {
430 struct mfctable *mfct;
431 int error, optval;
432 vifi_t vifi;
433 struct vifctl vifc;
434 struct mfcctl2 mfc;
435 struct bw_upcall bw_upcall;
436 uint32_t i;
437
438 mfct = somfctable(so);
439 if (so != mfct->router && sopt->sopt_name != MRT_INIT)
440 return EPERM;
441
442 error = 0;
443 switch (sopt->sopt_name) {
444 case MRT_INIT:
445 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
446 if (error)
447 break;
448 error = ip_mrouter_init(so, optval);
449 break;
450 case MRT_DONE:
451 ip_mrouter_done(so);
452 break;
453 case MRT_ADD_VIF:
454 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
455 if (error)
456 break;
457 error = add_vif(mfct, so->so_fibnum, &vifc);
458 break;
459 case MRT_DEL_VIF:
460 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
461 if (error)
462 break;
463 error = del_vif(mfct, vifi);
464 break;
465 case MRT_ADD_MFC:
466 case MRT_DEL_MFC:
467 /*
468 * select data size depending on API version.
469 */
470 if (sopt->sopt_name == MRT_ADD_MFC &&
471 (mfct->api_config & MRT_API_FLAGS_ALL) != 0) {
472 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
473 sizeof(struct mfcctl2));
474 } else {
475 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
476 sizeof(struct mfcctl));
477 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
478 sizeof(mfc) - sizeof(struct mfcctl));
479 }
480 if (error)
481 break;
482 if (sopt->sopt_name == MRT_ADD_MFC)
483 error = add_mfc(mfct, &mfc);
484 else
485 error = del_mfc(mfct, &mfc);
486 break;
487
488 case MRT_ASSERT:
489 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
490 if (error)
491 break;
492 set_assert(mfct, optval);
493 break;
494
495 case MRT_API_CONFIG:
496 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
497 if (!error)
498 error = set_api_config(mfct, &i);
499 if (!error)
500 error = sooptcopyout(sopt, &i, sizeof i);
501 break;
502
503 case MRT_ADD_BW_UPCALL:
504 case MRT_DEL_BW_UPCALL:
505 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
506 sizeof bw_upcall);
507 if (error)
508 break;
509 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
510 error = add_bw_upcall(mfct, &bw_upcall);
511 else
512 error = del_bw_upcall(mfct, &bw_upcall);
513 break;
514
515 default:
516 error = EOPNOTSUPP;
517 break;
518 }
519 return error;
520 }
521
522 /*
523 * Handle MRT getsockopt commands
524 */
525 static int
X_ip_mrouter_get(struct socket * so,struct sockopt * sopt)526 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
527 {
528 struct mfctable *mfct;
529 int error;
530
531 mfct = somfctable(so);
532 switch (sopt->sopt_name) {
533 case MRT_VERSION:
534 error = sooptcopyout(sopt, &mrt_api_version,
535 sizeof mrt_api_version);
536 break;
537 case MRT_ASSERT:
538 error = sooptcopyout(sopt, &mfct->pim_assert_enabled,
539 sizeof(mfct->pim_assert_enabled));
540 break;
541 case MRT_API_SUPPORT:
542 error = sooptcopyout(sopt, &mrt_api_support,
543 sizeof mrt_api_support);
544 break;
545 case MRT_API_CONFIG:
546 error = sooptcopyout(sopt, &mfct->api_config,
547 sizeof(mfct->api_config));
548 break;
549 default:
550 error = EOPNOTSUPP;
551 break;
552 }
553 return error;
554 }
555
556 /*
557 * Handle ioctl commands to obtain information from the cache
558 */
559 static int
X_mrt_ioctl(u_long cmd,caddr_t data,int fibnum)560 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum)
561 {
562 struct mfctable *mfct;
563 int error;
564
565 error = priv_check(curthread, PRIV_NETINET_MROUTE);
566 if (error)
567 return (error);
568
569 mfct = &V_mfctables[fibnum];
570 switch (cmd) {
571 case SIOCGETVIFCNT:
572 error = get_vif_cnt(mfct, (struct sioc_vif_req *)data);
573 break;
574
575 case SIOCGETSGCNT:
576 error = get_sg_cnt(mfct, (struct sioc_sg_req *)data);
577 break;
578
579 default:
580 error = EINVAL;
581 break;
582 }
583 return error;
584 }
585
586 /*
587 * returns the packet, byte, rpf-failure count for the source group provided
588 */
589 static int
get_sg_cnt(struct mfctable * mfct,struct sioc_sg_req * req)590 get_sg_cnt(struct mfctable *mfct, struct sioc_sg_req *req)
591 {
592 struct mfc *rt;
593
594 MRW_RLOCK();
595 rt = mfc_find(mfct, &req->src, &req->grp);
596 if (rt == NULL) {
597 MRW_RUNLOCK();
598 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
599 return EADDRNOTAVAIL;
600 }
601 req->pktcnt = rt->mfc_pkt_cnt;
602 req->bytecnt = rt->mfc_byte_cnt;
603 req->wrong_if = rt->mfc_wrong_if;
604 MRW_RUNLOCK();
605 return 0;
606 }
607
608 /*
609 * returns the input and output packet and byte counts on the vif provided
610 */
611 static int
get_vif_cnt(struct mfctable * mfct,struct sioc_vif_req * req)612 get_vif_cnt(struct mfctable *mfct, struct sioc_vif_req *req)
613 {
614 struct vif *vif;
615 vifi_t vifi;
616
617 vifi = req->vifi;
618
619 MRW_RLOCK();
620 if (vifi >= mfct->numvifs) {
621 MRW_RUNLOCK();
622 return EINVAL;
623 }
624
625 vif = &mfct->viftable[vifi];
626 mtx_lock(&vif->v_mtx);
627 req->icount = vif->v_pkt_in;
628 req->ocount = vif->v_pkt_out;
629 req->ibytes = vif->v_bytes_in;
630 req->obytes = vif->v_bytes_out;
631 mtx_unlock(&vif->v_mtx);
632 MRW_RUNLOCK();
633
634 return 0;
635 }
636
637 /*
638 * Tear down multicast forwarder state associated with this ifnet.
639 * 1. Walk the vif list, matching vifs against this ifnet.
640 * 2. Walk the multicast forwarding cache (mfc) looking for
641 * inner matches with this vif's index.
642 * 3. Expire any matching multicast forwarding cache entries.
643 * 4. Free vif state. This should disable ALLMULTI on the interface.
644 */
645 static int
detach_ifnet(struct mfctable * mfct,struct ifnet * ifp)646 detach_ifnet(struct mfctable *mfct, struct ifnet *ifp)
647 {
648 struct ifnet *free_ptr, *multi_leave;
649 int count;
650
651 count = 0;
652 restart:
653 for (vifi_t vifi = 0; vifi < mfct->numvifs; vifi++) {
654 if (mfct->viftable[vifi].v_ifp != ifp)
655 continue;
656 for (u_long i = 0; i < mfchashsize; i++) {
657 struct mfc *rt, *nrt;
658
659 LIST_FOREACH_SAFE(rt, &mfct->mfchashtbl[i], mfc_hash,
660 nrt) {
661 if (rt->mfc_parent == vifi) {
662 expire_mfc(rt);
663 }
664 }
665 }
666 del_vif_locked(mfct, vifi, &multi_leave, &free_ptr);
667 if (free_ptr != NULL)
668 count++;
669 if (multi_leave) {
670 MRW_WUNLOCK();
671 if_allmulti(multi_leave, 0);
672 MRW_WLOCK();
673 goto restart;
674 }
675 }
676 return (count);
677 }
678
679 static void
if_detached_event(void * arg __unused,struct ifnet * ifp)680 if_detached_event(void *arg __unused, struct ifnet *ifp)
681 {
682 int count;
683
684 MRW_WLOCK();
685 if (!V_ip_mrouting_enabled) {
686 MRW_WUNLOCK();
687 return;
688 }
689
690 count = 0;
691 for (int i = 0; i < V_nmfctables; i++)
692 count += detach_ifnet(&V_mfctables[i], ifp);
693 MRW_WUNLOCK();
694
695 for (int i = 0; i < count; i++)
696 if_free(ifp);
697 }
698
699 static void
ip_mrouter_upcall_thread(void * arg,int pending __unused)700 ip_mrouter_upcall_thread(void *arg, int pending __unused)
701 {
702 CURVNET_SET((struct vnet *) arg);
703
704 MRW_WLOCK();
705 bw_upcalls_send_all();
706 MRW_WUNLOCK();
707
708 CURVNET_RESTORE();
709 }
710
711 /*
712 * Enable multicast forwarding.
713 */
714 static int
ip_mrouter_init(struct socket * so,int version)715 ip_mrouter_init(struct socket *so, int version)
716 {
717 struct mfctable *mfct;
718
719 CTR2(KTR_IPMF, "%s: so %p", __func__, so);
720
721 if (version != 1)
722 return ENOPROTOOPT;
723
724 mfct = somfctable(so);
725 MRW_TEARDOWN_WLOCK();
726 MRW_WLOCK();
727
728 if (ip_mrouter_unloading) {
729 MRW_WUNLOCK();
730 MRW_TEARDOWN_WUNLOCK();
731 return ENOPROTOOPT;
732 }
733
734 if (mfct->router != NULL) {
735 MRW_WUNLOCK();
736 MRW_TEARDOWN_WUNLOCK();
737 return EADDRINUSE;
738 }
739
740 mfct->mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
741 HASH_NOWAIT);
742 if (mfct->mfchashtbl == NULL) {
743 MRW_WUNLOCK();
744 MRW_TEARDOWN_WUNLOCK();
745 return (ENOMEM);
746 }
747
748 /* Create upcall ring */
749 mtx_init(&mfct->bw_upcalls_mtx, "mroute upcall buf_ring mtx", NULL,
750 MTX_DEF);
751 mfct->bw_upcalls = buf_ring_alloc(BW_UPCALLS_MAX, M_MRTABLE, M_NOWAIT,
752 &mfct->bw_upcalls_mtx);
753 if (mfct->bw_upcalls == NULL) {
754 MRW_WUNLOCK();
755 MRW_TEARDOWN_WUNLOCK();
756 return (ENOMEM);
757 }
758
759 TASK_INIT(&V_task, 0, ip_mrouter_upcall_thread, curvnet);
760 taskqueue_cancel(V_task_queue, &V_task, NULL);
761 taskqueue_unblock(V_task_queue);
762
763 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls_all,
764 curvnet);
765 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
766 curvnet);
767
768 mfct->router = so;
769 mfct->pim_assert_interval.tv_sec = 3;
770 V_ip_mrouting_enabled = true;
771 atomic_add_int(&ip_mrouter_cnt, 1);
772
773 MRW_WUNLOCK();
774 MRW_TEARDOWN_WUNLOCK();
775
776 CTR1(KTR_IPMF, "%s: done", __func__);
777
778 return 0;
779 }
780
781 /*
782 * Disable multicast forwarding.
783 */
784 static void
X_ip_mrouter_done(struct socket * so)785 X_ip_mrouter_done(struct socket *so)
786 {
787 struct mfctable *mfct;
788 struct ifnet **ifps;
789 int nifp;
790 u_long i;
791 vifi_t vifi;
792 struct bw_upcall *bu;
793
794 mfct = somfctable(so);
795
796 MRW_TEARDOWN_WLOCK();
797 if (so != mfct->router) {
798 MRW_TEARDOWN_WUNLOCK();
799 return;
800 }
801
802 /*
803 * Detach/disable hooks to the reset of the system.
804 */
805 mfct->router = NULL;
806 V_ip_mrouting_enabled = false;
807 atomic_subtract_int(&ip_mrouter_cnt, 1);
808 mfct->api_config = 0;
809
810 /*
811 * Wait for all epoch sections to complete to ensure the new value of
812 * V_ip_mrouting_enabled is visible to others.
813 */
814 NET_EPOCH_WAIT();
815
816 /* Stop and drain task queue */
817 taskqueue_block(V_task_queue);
818 while (taskqueue_cancel(V_task_queue, &V_task, NULL)) {
819 taskqueue_drain(V_task_queue, &V_task);
820 }
821
822 ifps = malloc(MAXVIFS * sizeof(*ifps), M_TEMP, M_WAITOK);
823
824 MRW_WLOCK();
825 taskqueue_cancel(V_task_queue, &V_task, NULL);
826
827 /* Destroy upcall ring */
828 while ((bu = buf_ring_dequeue_mc(mfct->bw_upcalls)) != NULL) {
829 free(bu, M_MRTABLE);
830 }
831 buf_ring_free(mfct->bw_upcalls, M_MRTABLE);
832 mtx_destroy(&mfct->bw_upcalls_mtx);
833
834 /*
835 * For each phyint in use, prepare to disable promiscuous reception
836 * of all IP multicasts. Defer the actual call until the lock is released;
837 * just record the list of interfaces while locked. Some interfaces use
838 * sx locks in their ioctl routines, which is not allowed while holding
839 * a non-sleepable lock.
840 */
841 KASSERT(mfct->numvifs <= MAXVIFS, ("More vifs than possible"));
842 for (vifi = 0, nifp = 0; vifi < mfct->numvifs; vifi++) {
843 struct vif *vif;
844
845 vif = &mfct->viftable[vifi];
846 if (!in_nullhost(vif->v_lcl_addr) &&
847 (vif->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)) == 0) {
848 ifps[nifp++] = vif->v_ifp;
849 }
850 }
851 bzero(mfct->viftable, sizeof(*mfct->viftable) * MAXVIFS);
852 mfct->numvifs = 0;
853 mfct->pim_assert_enabled = 0;
854
855 /*
856 * Free all multicast forwarding cache entries.
857 * Do not use hashdestroy(), as we must perform other cleanup.
858 */
859 for (i = 0; i < mfchashsize; i++) {
860 struct mfc *rt, *nrt;
861
862 LIST_FOREACH_SAFE(rt, &mfct->mfchashtbl[i], mfc_hash, nrt) {
863 expire_mfc(rt);
864 }
865 }
866 free(mfct->mfchashtbl, M_MRTABLE);
867 mfct->mfchashtbl = NULL;
868
869 bzero(mfct->nexpire, sizeof(mfct->nexpire[0]) * mfchashsize);
870 mfct->register_vif = VIFI_INVALID;
871
872 MRW_WUNLOCK();
873 MRW_TEARDOWN_WUNLOCK();
874
875 /*
876 * Now drop our claim on promiscuous multicast on the interfaces recorded
877 * above. This is safe to do now because ALLMULTI is reference counted.
878 */
879 for (vifi = 0; vifi < nifp; vifi++)
880 if_allmulti(ifps[vifi], 0);
881 free(ifps, M_TEMP);
882
883 CTR1(KTR_IPMF, "%s: done", __func__);
884 }
885
886 /*
887 * Set PIM assert processing global
888 */
889 static int
set_assert(struct mfctable * mfct,int i)890 set_assert(struct mfctable *mfct, int i)
891 {
892 if ((i != 1) && (i != 0))
893 return EINVAL;
894
895 mfct->pim_assert_enabled = i;
896
897 return 0;
898 }
899
900 /*
901 * Configure API capabilities
902 */
903 int
set_api_config(struct mfctable * mfct,uint32_t * apival)904 set_api_config(struct mfctable *mfct, uint32_t *apival)
905 {
906 u_long i;
907
908 /*
909 * We can set the API capabilities only if it is the first operation
910 * after MRT_INIT. I.e.:
911 * - there are no vifs installed
912 * - pim_assert is not enabled
913 * - the MFC table is empty
914 */
915 if (mfct->numvifs > 0) {
916 *apival = 0;
917 return EPERM;
918 }
919 if (mfct->pim_assert_enabled) {
920 *apival = 0;
921 return EPERM;
922 }
923
924 MRW_RLOCK();
925
926 for (i = 0; i < mfchashsize; i++) {
927 if (LIST_FIRST(&mfct->mfchashtbl[i]) != NULL) {
928 MRW_RUNLOCK();
929 *apival = 0;
930 return EPERM;
931 }
932 }
933
934 MRW_RUNLOCK();
935
936 mfct->api_config = *apival & mrt_api_support;
937 *apival = mfct->api_config;
938
939 return 0;
940 }
941
942 /*
943 * Add a vif to the vif table
944 */
945 static int
add_vif(struct mfctable * mfct,int fibnum,struct vifctl * vifcp)946 add_vif(struct mfctable *mfct, int fibnum, struct vifctl *vifcp)
947 {
948 struct vif *vifp;
949 struct sockaddr_in sin = {sizeof sin, AF_INET};
950 struct ifaddr *ifa;
951 struct ifnet *ifp;
952 int error;
953
954 if (vifcp->vifc_vifi >= MAXVIFS)
955 return EINVAL;
956
957 vifp = &mfct->viftable[vifcp->vifc_vifi];
958
959 /* rate limiting is no longer supported by this code */
960 if (vifcp->vifc_rate_limit != 0) {
961 log(LOG_ERR, "rate limiting is no longer supported\n");
962 return EINVAL;
963 }
964
965 if (in_nullhost(vifcp->vifc_lcl_addr))
966 return EADDRNOTAVAIL;
967
968 /* Find the interface with an address in AF_INET family */
969 if (vifcp->vifc_flags & VIFF_REGISTER) {
970 /*
971 * XXX: Because VIFF_REGISTER does not really need a valid
972 * local interface (e.g. it could be 127.0.0.2), we don't
973 * check its address.
974 */
975 ifp = NULL;
976 } else {
977 struct epoch_tracker et;
978
979 sin.sin_addr = vifcp->vifc_lcl_addr;
980 NET_EPOCH_ENTER(et);
981 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
982 if (ifa == NULL) {
983 NET_EPOCH_EXIT(et);
984 return EADDRNOTAVAIL;
985 }
986 ifp = ifa->ifa_ifp;
987 if (ifp->if_fib != fibnum) {
988 NET_EPOCH_EXIT(et);
989 return EADDRNOTAVAIL;
990 }
991 /* XXX FIXME we need to take a ref on ifp and cleanup properly! */
992 NET_EPOCH_EXIT(et);
993 }
994
995 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
996 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
997 return EOPNOTSUPP;
998 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
999 ifp = mfct->register_if = if_alloc(IFT_LOOP);
1000 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
1001 if (mfct->register_vif == VIFI_INVALID) {
1002 if_initname(mfct->register_if, "register_vif", 0);
1003 mfct->register_vif = vifcp->vifc_vifi;
1004 }
1005 } else { /* Make sure the interface supports multicast */
1006 if ((ifp->if_flags & IFF_MULTICAST) == 0)
1007 return EOPNOTSUPP;
1008
1009 /* Enable promiscuous reception of all IP multicasts from the if */
1010 error = if_allmulti(ifp, 1);
1011 if (error)
1012 return error;
1013 }
1014
1015 MRW_WLOCK();
1016
1017 if (!in_nullhost(vifp->v_lcl_addr)) {
1018 if (ifp)
1019 mfct->register_if = NULL;
1020 MRW_WUNLOCK();
1021 if (ifp)
1022 if_free(ifp);
1023 return EADDRINUSE;
1024 }
1025
1026 vifp->v_flags = vifcp->vifc_flags;
1027 vifp->v_threshold = vifcp->vifc_threshold;
1028 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
1029 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
1030 vifp->v_ifp = ifp;
1031 /* initialize per vif pkt counters */
1032 vifp->v_pkt_in = 0;
1033 vifp->v_pkt_out = 0;
1034 vifp->v_bytes_in = 0;
1035 vifp->v_bytes_out = 0;
1036 sprintf(vifp->v_mtx_name, "BM[%d] mtx", vifcp->vifc_vifi);
1037 mtx_init(&vifp->v_mtx, vifp->v_mtx_name, NULL, MTX_DEF);
1038
1039 /* Adjust numvifs up if the vifi is higher than numvifs */
1040 if (mfct->numvifs <= vifcp->vifc_vifi)
1041 mfct->numvifs = vifcp->vifc_vifi + 1;
1042
1043 MRW_WUNLOCK();
1044
1045 CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
1046 (int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
1047 (int)vifcp->vifc_threshold);
1048
1049 return 0;
1050 }
1051
1052 /*
1053 * Delete a vif from the vif table
1054 */
1055 static int
del_vif_locked(struct mfctable * mfct,vifi_t vifi,struct ifnet ** ifp_multi_leave,struct ifnet ** ifp_free)1056 del_vif_locked(struct mfctable *mfct, vifi_t vifi,
1057 struct ifnet **ifp_multi_leave, struct ifnet **ifp_free)
1058 {
1059 struct vif *vifp;
1060
1061 *ifp_free = NULL;
1062 *ifp_multi_leave = NULL;
1063
1064 MRW_WLOCK_ASSERT();
1065
1066 if (vifi >= mfct->numvifs) {
1067 return EINVAL;
1068 }
1069 vifp = &mfct->viftable[vifi];
1070 if (in_nullhost(vifp->v_lcl_addr)) {
1071 return EADDRNOTAVAIL;
1072 }
1073
1074 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1075 *ifp_multi_leave = vifp->v_ifp;
1076
1077 if (vifp->v_flags & VIFF_REGISTER) {
1078 mfct->register_vif = VIFI_INVALID;
1079 if (vifp->v_ifp) {
1080 if (vifp->v_ifp == mfct->register_if)
1081 mfct->register_if = NULL;
1082 *ifp_free = vifp->v_ifp;
1083 }
1084 }
1085
1086 mtx_destroy(&vifp->v_mtx);
1087
1088 bzero((caddr_t)vifp, sizeof (*vifp));
1089
1090 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
1091
1092 /* Adjust numvifs down */
1093 for (vifi = mfct->numvifs; vifi > 0; vifi--)
1094 if (!in_nullhost(mfct->viftable[vifi - 1].v_lcl_addr))
1095 break;
1096 mfct->numvifs = vifi;
1097
1098 return 0;
1099 }
1100
1101 static int
del_vif(struct mfctable * mfct,vifi_t vifi)1102 del_vif(struct mfctable *mfct, vifi_t vifi)
1103 {
1104 int cc;
1105 struct ifnet *free_ptr, *multi_leave;
1106
1107 MRW_WLOCK();
1108 cc = del_vif_locked(mfct, vifi, &multi_leave, &free_ptr);
1109 MRW_WUNLOCK();
1110
1111 if (multi_leave)
1112 if_allmulti(multi_leave, 0);
1113 if (free_ptr) {
1114 if_free(free_ptr);
1115 }
1116
1117 return cc;
1118 }
1119
1120 /*
1121 * update an mfc entry without resetting counters and S,G addresses.
1122 */
1123 static void
update_mfc_params(struct mfctable * mfct,struct mfc * rt,struct mfcctl2 * mfccp)1124 update_mfc_params(struct mfctable *mfct, struct mfc *rt, struct mfcctl2 *mfccp)
1125 {
1126 int i;
1127
1128 rt->mfc_parent = mfccp->mfcc_parent;
1129 for (i = 0; i < mfct->numvifs; i++) {
1130 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1131 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mfct->api_config &
1132 MRT_MFC_FLAGS_ALL;
1133 }
1134 /* set the RP address */
1135 if (mfct->api_config & MRT_MFC_RP)
1136 rt->mfc_rp = mfccp->mfcc_rp;
1137 else
1138 rt->mfc_rp.s_addr = INADDR_ANY;
1139 }
1140
1141 /*
1142 * fully initialize an mfc entry from the parameter.
1143 */
1144 static void
init_mfc_params(struct mfctable * mfct,struct mfc * rt,struct mfcctl2 * mfccp)1145 init_mfc_params(struct mfctable *mfct, struct mfc *rt, struct mfcctl2 *mfccp)
1146 {
1147 rt->mfc_origin = mfccp->mfcc_origin;
1148 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1149
1150 update_mfc_params(mfct, rt, mfccp);
1151
1152 /* initialize pkt counters per src-grp */
1153 rt->mfc_pkt_cnt = 0;
1154 rt->mfc_byte_cnt = 0;
1155 rt->mfc_wrong_if = 0;
1156 timevalclear(&rt->mfc_last_assert);
1157 }
1158
1159 static void
expire_mfc(struct mfc * rt)1160 expire_mfc(struct mfc *rt)
1161 {
1162 struct rtdetq *rte;
1163
1164 MRW_WLOCK_ASSERT();
1165
1166 free_bw_list(rt->mfc_bw_meter_leq);
1167 free_bw_list(rt->mfc_bw_meter_geq);
1168
1169 while (!buf_ring_empty(rt->mfc_stall_ring)) {
1170 rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1171 if (rte) {
1172 m_freem(rte->m);
1173 free(rte, M_MRTABLE);
1174 }
1175 }
1176 buf_ring_free(rt->mfc_stall_ring, M_MRTABLE);
1177
1178 LIST_REMOVE(rt, mfc_hash);
1179 free(rt, M_MRTABLE);
1180 }
1181
1182 /*
1183 * Add an mfc entry
1184 */
1185 static int
add_mfc(struct mfctable * mfct,struct mfcctl2 * mfccp)1186 add_mfc(struct mfctable *mfct, struct mfcctl2 *mfccp)
1187 {
1188 struct mfc *rt;
1189 struct rtdetq *rte;
1190 u_long hash = 0;
1191 u_short nstl;
1192 struct epoch_tracker et;
1193
1194 MRW_WLOCK();
1195 rt = mfc_find(mfct, &mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1196
1197 /* If an entry already exists, just update the fields */
1198 if (rt) {
1199 CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1200 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1201 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1202 mfccp->mfcc_parent);
1203 update_mfc_params(mfct, rt, mfccp);
1204 MRW_WUNLOCK();
1205 return (0);
1206 }
1207
1208 /*
1209 * Find the entry for which the upcall was made and update
1210 */
1211 nstl = 0;
1212 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1213 NET_EPOCH_ENTER(et);
1214 LIST_FOREACH(rt, &mfct->mfchashtbl[hash], mfc_hash) {
1215 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1216 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1217 !buf_ring_empty(rt->mfc_stall_ring)) {
1218 CTR5(KTR_IPMF,
1219 "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1220 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1221 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1222 mfccp->mfcc_parent,
1223 rt->mfc_stall_ring);
1224 if (nstl++)
1225 CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1226
1227 init_mfc_params(mfct, rt, mfccp);
1228 rt->mfc_expire = 0; /* Don't clean this guy up */
1229 mfct->nexpire[hash]--;
1230
1231 /* Free queued packets, but attempt to forward them first. */
1232 while (!buf_ring_empty(rt->mfc_stall_ring)) {
1233 rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1234 if (rte->ifp != NULL)
1235 ip_mdq(mfct, rte->m, rte->ifp, rt, -1);
1236 m_freem(rte->m);
1237 free(rte, M_MRTABLE);
1238 }
1239 }
1240 }
1241 NET_EPOCH_EXIT(et);
1242
1243 /*
1244 * It is possible that an entry is being inserted without an upcall
1245 */
1246 if (nstl == 0) {
1247 CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1248 LIST_FOREACH(rt, &mfct->mfchashtbl[hash], mfc_hash) {
1249 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1250 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1251 init_mfc_params(mfct, rt, mfccp);
1252 if (rt->mfc_expire)
1253 mfct->nexpire[hash]--;
1254 rt->mfc_expire = 0;
1255 break; /* XXX */
1256 }
1257 }
1258
1259 if (rt == NULL) { /* no upcall, so make a new entry */
1260 rt = mfc_alloc();
1261 if (rt == NULL) {
1262 MRW_WUNLOCK();
1263 return (ENOBUFS);
1264 }
1265
1266 init_mfc_params(mfct, rt, mfccp);
1267
1268 rt->mfc_expire = 0;
1269 rt->mfc_bw_meter_leq = NULL;
1270 rt->mfc_bw_meter_geq = NULL;
1271
1272 /* insert new entry at head of hash chain */
1273 LIST_INSERT_HEAD(&mfct->mfchashtbl[hash], rt, mfc_hash);
1274 }
1275 }
1276
1277 MRW_WUNLOCK();
1278
1279 return (0);
1280 }
1281
1282 /*
1283 * Delete an mfc entry
1284 */
1285 static int
del_mfc(struct mfctable * mfct,struct mfcctl2 * mfccp)1286 del_mfc(struct mfctable *mfct, struct mfcctl2 *mfccp)
1287 {
1288 struct in_addr origin;
1289 struct in_addr mcastgrp;
1290 struct mfc *rt;
1291
1292 origin = mfccp->mfcc_origin;
1293 mcastgrp = mfccp->mfcc_mcastgrp;
1294
1295 CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1296 ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1297
1298 MRW_WLOCK();
1299
1300 LIST_FOREACH(rt, &mfct->mfchashtbl[MFCHASH(origin, mcastgrp)],
1301 mfc_hash) {
1302 if (in_hosteq(rt->mfc_origin, origin) &&
1303 in_hosteq(rt->mfc_mcastgrp, mcastgrp))
1304 break;
1305 }
1306 if (rt == NULL) {
1307 MRW_WUNLOCK();
1308 return EADDRNOTAVAIL;
1309 }
1310
1311 expire_mfc(rt);
1312
1313 MRW_WUNLOCK();
1314
1315 return (0);
1316 }
1317
1318 /*
1319 * Send a message to the routing daemon on the multicast routing socket.
1320 */
1321 static int
socket_send(struct socket * s,struct mbuf * mm,struct sockaddr_in * src)1322 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1323 {
1324 if (s) {
1325 SOCKBUF_LOCK(&s->so_rcv);
1326 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1327 NULL) != 0) {
1328 sorwakeup_locked(s);
1329 return 0;
1330 }
1331 soroverflow_locked(s);
1332 }
1333 m_freem(mm);
1334 return -1;
1335 }
1336
1337 /*
1338 * IP multicast forwarding function. This function assumes that the packet
1339 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1340 * pointed to by "ifp", and the packet is to be relayed to other networks
1341 * that have members of the packet's destination IP multicast group.
1342 *
1343 * The packet is returned unscathed to the caller, unless it is
1344 * erroneous, in which case a non-zero return value tells the caller to
1345 * discard it.
1346 */
1347
1348 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1349
1350 static int
X_ip_mforward(struct ip * ip,struct ifnet * ifp,struct mbuf * m,struct ip_moptions * imo)1351 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1352 struct ip_moptions *imo)
1353 {
1354 struct mfc *rt;
1355 struct mfctable *mfct;
1356 int error;
1357 vifi_t vifi;
1358 struct mbuf *mb0;
1359 struct rtdetq *rte;
1360 u_long hash;
1361 int hlen;
1362
1363 M_ASSERTMAPPED(m);
1364
1365 CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1366 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1367
1368 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1369 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1370 /*
1371 * Packet arrived via a physical interface or
1372 * an encapsulated tunnel or a register_vif.
1373 */
1374 } else {
1375 /*
1376 * Packet arrived through a source-route tunnel.
1377 * Source-route tunnels are no longer supported.
1378 */
1379 return (1);
1380 }
1381
1382 mfct = &V_mfctables[M_GETFIB(m)];
1383
1384 /*
1385 * BEGIN: MCAST ROUTING HOT PATH
1386 */
1387 MRW_RLOCK();
1388 if (__predict_false(mfct->router == NULL)) {
1389 MRW_RUNLOCK();
1390 return (EADDRNOTAVAIL);
1391 }
1392
1393 if (imo && ((vifi = imo->imo_multicast_vif) < mfct->numvifs)) {
1394 if (ip->ip_ttl < MAXTTL)
1395 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1396 error = ip_mdq(mfct, m, ifp, NULL, vifi);
1397 MRW_RUNLOCK();
1398 return error;
1399 }
1400
1401 /*
1402 * Don't forward a packet with time-to-live of zero or one,
1403 * or a packet destined to a local-only group.
1404 */
1405 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1406 MRW_RUNLOCK();
1407 return 0;
1408 }
1409
1410 mfc_find_retry:
1411 /*
1412 * Determine forwarding vifs from the forwarding cache table
1413 */
1414 MRTSTAT_INC(mrts_mfc_lookups);
1415 rt = mfc_find(mfct, &ip->ip_src, &ip->ip_dst);
1416
1417 /* Entry exists, so forward if necessary */
1418 if (rt != NULL) {
1419 error = ip_mdq(mfct, m, ifp, rt, -1);
1420 /* Generic unlock here as we might release R or W lock */
1421 MRW_UNLOCK();
1422 return error;
1423 }
1424
1425 /*
1426 * END: MCAST ROUTING HOT PATH
1427 */
1428
1429 /* Further processing must be done with WLOCK taken */
1430 if ((MRW_WOWNED() == 0) && (MRW_LOCK_TRY_UPGRADE() == 0)) {
1431 MRW_RUNLOCK();
1432 MRW_WLOCK();
1433 goto mfc_find_retry;
1434 }
1435
1436 /*
1437 * If we don't have a route for packet's origin,
1438 * Make a copy of the packet & send message to routing daemon
1439 */
1440 hlen = ip->ip_hl << 2;
1441
1442 MRTSTAT_INC(mrts_mfc_misses);
1443 MRTSTAT_INC(mrts_no_route);
1444 CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1445 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1446
1447 /*
1448 * Allocate mbufs early so that we don't do extra work if we are
1449 * just going to fail anyway. Make sure to pullup the header so
1450 * that other people can't step on it.
1451 */
1452 rte = malloc((sizeof *rte), M_MRTABLE, M_NOWAIT|M_ZERO);
1453 if (rte == NULL) {
1454 MRW_WUNLOCK();
1455 return ENOBUFS;
1456 }
1457
1458 mb0 = m_copypacket(m, M_NOWAIT);
1459 if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1460 mb0 = m_pullup(mb0, hlen);
1461 if (mb0 == NULL) {
1462 free(rte, M_MRTABLE);
1463 MRW_WUNLOCK();
1464 return ENOBUFS;
1465 }
1466
1467 /* is there an upcall waiting for this flow ? */
1468 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1469 LIST_FOREACH(rt, &mfct->mfchashtbl[hash], mfc_hash) {
1470 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1471 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1472 !buf_ring_empty(rt->mfc_stall_ring))
1473 break;
1474 }
1475
1476 if (rt == NULL) {
1477 int i;
1478 struct igmpmsg *im;
1479 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1480 struct mbuf *mm;
1481
1482 /*
1483 * Locate the vifi for the incoming interface for this packet.
1484 * If none found, drop packet.
1485 */
1486 for (vifi = 0; vifi < mfct->numvifs &&
1487 mfct->viftable[vifi].v_ifp != ifp; vifi++)
1488 ;
1489 if (vifi >= mfct->numvifs) /* vif not found, drop packet */
1490 goto non_fatal;
1491
1492 /* no upcall, so make a new entry */
1493 rt = mfc_alloc();
1494 if (rt == NULL)
1495 goto fail;
1496
1497 /* Make a copy of the header to send to the user level process */
1498 mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1499 if (mm == NULL)
1500 goto fail1;
1501
1502 /*
1503 * Send message to routing daemon to install
1504 * a route into the kernel table
1505 */
1506
1507 im = mtod(mm, struct igmpmsg*);
1508 im->im_msgtype = IGMPMSG_NOCACHE;
1509 im->im_mbz = 0;
1510 im->im_vif = vifi;
1511
1512 MRTSTAT_INC(mrts_upcalls);
1513
1514 k_igmpsrc.sin_addr = ip->ip_src;
1515 if (socket_send(mfct->router, mm, &k_igmpsrc) < 0) {
1516 CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1517 MRTSTAT_INC(mrts_upq_sockfull);
1518 fail1: free(rt, M_MRTABLE);
1519 fail: free(rte, M_MRTABLE);
1520 m_freem(mb0);
1521 MRW_WUNLOCK();
1522 return ENOBUFS;
1523 }
1524
1525 /* insert new entry at head of hash chain */
1526 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1527 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1528 rt->mfc_expire = UPCALL_EXPIRE;
1529 mfct->nexpire[hash]++;
1530 for (i = 0; i < mfct->numvifs; i++) {
1531 rt->mfc_ttls[i] = 0;
1532 rt->mfc_flags[i] = 0;
1533 }
1534 rt->mfc_parent = -1;
1535
1536 /* clear the RP address */
1537 rt->mfc_rp.s_addr = INADDR_ANY;
1538 rt->mfc_bw_meter_leq = NULL;
1539 rt->mfc_bw_meter_geq = NULL;
1540
1541 /* initialize pkt counters per src-grp */
1542 rt->mfc_pkt_cnt = 0;
1543 rt->mfc_byte_cnt = 0;
1544 rt->mfc_wrong_if = 0;
1545 timevalclear(&rt->mfc_last_assert);
1546
1547 buf_ring_enqueue(rt->mfc_stall_ring, rte);
1548
1549 /* Add RT to hashtable as it didn't exist before */
1550 LIST_INSERT_HEAD(&mfct->mfchashtbl[hash], rt, mfc_hash);
1551 } else {
1552 /* determine if queue has overflowed */
1553 if (buf_ring_full(rt->mfc_stall_ring)) {
1554 MRTSTAT_INC(mrts_upq_ovflw);
1555 non_fatal: free(rte, M_MRTABLE);
1556 m_freem(mb0);
1557 MRW_WUNLOCK();
1558 return (0);
1559 }
1560
1561 buf_ring_enqueue(rt->mfc_stall_ring, rte);
1562 }
1563
1564 rte->m = mb0;
1565 rte->ifp = ifp;
1566
1567 MRW_WUNLOCK();
1568
1569 return 0;
1570 }
1571
1572 static void
expire_upcalls(struct mfctable * mfct)1573 expire_upcalls(struct mfctable *mfct)
1574 {
1575 for (u_long i = 0; i < mfchashsize; i++) {
1576 struct mfc *rt, *nrt;
1577
1578 if (mfct->nexpire[i] == 0)
1579 continue;
1580
1581 LIST_FOREACH_SAFE(rt, &mfct->mfchashtbl[i], mfc_hash, nrt) {
1582 if (buf_ring_empty(rt->mfc_stall_ring))
1583 continue;
1584
1585 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1586 continue;
1587
1588 MRTSTAT_INC(mrts_cache_cleanups);
1589 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1590 (u_long)ntohl(rt->mfc_origin.s_addr),
1591 (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1592
1593 expire_mfc(rt);
1594 }
1595 }
1596 }
1597
1598 /*
1599 * Clean up the cache entry if upcall is not serviced
1600 */
1601 static void
expire_upcalls_all(void * arg)1602 expire_upcalls_all(void *arg)
1603 {
1604 CURVNET_SET((struct vnet *)arg);
1605
1606 MRW_LOCK_ASSERT();
1607
1608 for (int i = 0; i < V_nmfctables; i++)
1609 expire_upcalls(&V_mfctables[i]);
1610
1611 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls_all,
1612 curvnet);
1613
1614 CURVNET_RESTORE();
1615 }
1616
1617 /*
1618 * Packet forwarding routine once entry in the cache is made
1619 */
1620 static int
ip_mdq(struct mfctable * mfct,struct mbuf * m,struct ifnet * ifp,struct mfc * rt,vifi_t xmt_vif)1621 ip_mdq(struct mfctable *mfct, struct mbuf *m, struct ifnet *ifp, struct mfc *rt,
1622 vifi_t xmt_vif)
1623 {
1624 struct ip *ip = mtod(m, struct ip *);
1625 struct vif *vif;
1626 vifi_t vifi;
1627 int plen = ntohs(ip->ip_len);
1628
1629 M_ASSERTMAPPED(m);
1630 MRW_LOCK_ASSERT();
1631 NET_EPOCH_ASSERT();
1632
1633 /*
1634 * If xmt_vif is not -1, send on only the requested vif.
1635 *
1636 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1637 */
1638 if (xmt_vif < mfct->numvifs) {
1639 if (mfct->viftable[xmt_vif].v_flags & VIFF_REGISTER)
1640 pim_register_send(mfct, ip, &mfct->viftable[xmt_vif], m,
1641 rt);
1642 else
1643 phyint_send(ip, &mfct->viftable[xmt_vif], m);
1644 return 1;
1645 }
1646
1647 /*
1648 * Don't forward if it didn't arrive from the parent vif for its origin.
1649 */
1650 vifi = rt->mfc_parent;
1651 vif = &mfct->viftable[vifi];
1652 if (vifi >= mfct->numvifs || vif->v_ifp != ifp) {
1653 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1654 __func__, ifp, (int)vifi, vif->v_ifp);
1655 MRTSTAT_INC(mrts_wrong_if);
1656 ++rt->mfc_wrong_if;
1657 /*
1658 * If we are doing PIM assert processing, send a message
1659 * to the routing daemon.
1660 *
1661 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1662 * can complete the SPT switch, regardless of the type
1663 * of the iif (broadcast media, GRE tunnel, etc).
1664 */
1665 if (mfct->pim_assert_enabled && (vifi < mfct->numvifs) &&
1666 vif->v_ifp != NULL) {
1667 if (ifp == mfct->register_if)
1668 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1669
1670 /* Get vifi for the incoming packet */
1671 for (vifi = 0; vifi < mfct->numvifs &&
1672 mfct->viftable[vifi].v_ifp != ifp; vifi++)
1673 ;
1674 if (vifi >= mfct->numvifs)
1675 return 0; /* The iif is not found: ignore the packet. */
1676
1677 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1678 return 0; /* WRONGVIF disabled: ignore the packet */
1679
1680 if (ratecheck(&rt->mfc_last_assert,
1681 &mfct->pim_assert_interval)) {
1682 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1683 struct igmpmsg *im;
1684 int hlen = ip->ip_hl << 2;
1685 struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1686
1687 if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1688 mm = m_pullup(mm, hlen);
1689 if (mm == NULL)
1690 return ENOBUFS;
1691
1692 im = mtod(mm, struct igmpmsg *);
1693 im->im_msgtype = IGMPMSG_WRONGVIF;
1694 im->im_mbz = 0;
1695 im->im_vif = vifi;
1696
1697 MRTSTAT_INC(mrts_upcalls);
1698
1699 k_igmpsrc.sin_addr = im->im_src;
1700 if (socket_send(mfct->router, mm,
1701 &k_igmpsrc) < 0) {
1702 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1703 MRTSTAT_INC(mrts_upq_sockfull);
1704 return ENOBUFS;
1705 }
1706 }
1707 }
1708 return 0;
1709 }
1710
1711 /* If I sourced this packet, it counts as output, else it was input. */
1712 mtx_lock(&vif->v_mtx);
1713 if (in_hosteq(ip->ip_src, vif->v_lcl_addr)) {
1714 vif->v_pkt_out++;
1715 vif->v_bytes_out += plen;
1716 } else {
1717 vif->v_pkt_in++;
1718 vif->v_bytes_in += plen;
1719 }
1720 mtx_unlock(&vif->v_mtx);
1721
1722 rt->mfc_pkt_cnt++;
1723 rt->mfc_byte_cnt += plen;
1724
1725 /*
1726 * For each vif, decide if a copy of the packet should be forwarded.
1727 * Forward if:
1728 * - the ttl exceeds the vif's threshold
1729 * - there are group members downstream on interface
1730 */
1731 for (vifi = 0; vifi < mfct->numvifs; vifi++)
1732 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1733 vif = &mfct->viftable[vifi];
1734 vif->v_pkt_out++;
1735 vif->v_bytes_out += plen;
1736 if (vif->v_flags & VIFF_REGISTER)
1737 pim_register_send(mfct, ip, vif, m, rt);
1738 else
1739 phyint_send(ip, vif, m);
1740 }
1741
1742 /*
1743 * Perform upcall-related bw measuring.
1744 */
1745 if ((rt->mfc_bw_meter_geq != NULL) || (rt->mfc_bw_meter_leq != NULL)) {
1746 struct bw_meter *x;
1747 struct timeval now;
1748
1749 microtime(&now);
1750 /* Process meters for Greater-or-EQual case */
1751 for (x = rt->mfc_bw_meter_geq; x != NULL; x = x->bm_mfc_next)
1752 bw_meter_geq_receive_packet(x, plen, &now);
1753
1754 /* Process meters for Lower-or-EQual case */
1755 for (x = rt->mfc_bw_meter_leq; x != NULL; x = x->bm_mfc_next) {
1756 /*
1757 * Record that a packet is received.
1758 * A lock has to be taken as callout context
1759 * (expire_bw_meter_leq) might modify these fields
1760 * as well
1761 */
1762 mtx_lock(&x->bm_mtx);
1763 x->bm_measured.b_packets++;
1764 x->bm_measured.b_bytes += plen;
1765 mtx_unlock(&x->bm_mtx);
1766 }
1767 }
1768
1769 return 0;
1770 }
1771
1772 /*
1773 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1774 */
1775 static int
X_legal_vif_num(int fibnum,int vif)1776 X_legal_vif_num(int fibnum, int vif)
1777 {
1778 struct mfctable *mfct;
1779 int ret;
1780
1781 ret = 0;
1782 if (vif < 0)
1783 return (ret);
1784
1785 mfct = &V_mfctables[fibnum];
1786 MRW_RLOCK();
1787 if (vif < mfct->numvifs)
1788 ret = 1;
1789 MRW_RUNLOCK();
1790
1791 return (ret);
1792 }
1793
1794 /*
1795 * Return the local address used by this vif
1796 */
1797 static u_long
X_ip_mcast_src(int fibnum,int vifi)1798 X_ip_mcast_src(int fibnum, int vifi)
1799 {
1800 struct mfctable *mfct;
1801 in_addr_t addr;
1802
1803 addr = INADDR_ANY;
1804 if (vifi < 0)
1805 return (addr);
1806
1807 mfct = &V_mfctables[fibnum];
1808 MRW_RLOCK();
1809 if (vifi < mfct->numvifs)
1810 addr = mfct->viftable[vifi].v_lcl_addr.s_addr;
1811 MRW_RUNLOCK();
1812
1813 return (addr);
1814 }
1815
1816 static void
phyint_send(struct ip * ip,struct vif * vifp,struct mbuf * m)1817 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1818 {
1819 struct mbuf *mb_copy;
1820 int hlen = ip->ip_hl << 2;
1821
1822 MRW_LOCK_ASSERT();
1823 M_ASSERTMAPPED(m);
1824
1825 /*
1826 * Make a new reference to the packet; make sure that
1827 * the IP header is actually copied, not just referenced,
1828 * so that ip_output() only scribbles on the copy.
1829 */
1830 mb_copy = m_copypacket(m, M_NOWAIT);
1831 if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1832 mb_copy = m_pullup(mb_copy, hlen);
1833 if (mb_copy == NULL)
1834 return;
1835
1836 send_packet(vifp, mb_copy);
1837 }
1838
1839 static void
send_packet(struct vif * vifp,struct mbuf * m)1840 send_packet(struct vif *vifp, struct mbuf *m)
1841 {
1842 struct ip_moptions imo;
1843 int error __unused;
1844
1845 MRW_LOCK_ASSERT();
1846 NET_EPOCH_ASSERT();
1847
1848 imo.imo_multicast_ifp = vifp->v_ifp;
1849 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1850 imo.imo_multicast_loop = !!in_mcast_loop;
1851 imo.imo_multicast_vif = -1;
1852 STAILQ_INIT(&imo.imo_head);
1853
1854 /*
1855 * Re-entrancy should not be a problem here, because
1856 * the packets that we send out and are looped back at us
1857 * should get rejected because they appear to come from
1858 * the loopback interface, thus preventing looping.
1859 */
1860 error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1861 }
1862
1863 /*
1864 * Stubs for old RSVP socket shim implementation.
1865 */
1866
1867 static int
X_ip_rsvp_vif(struct socket * so __unused,struct sockopt * sopt __unused)1868 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1869 {
1870
1871 return (EOPNOTSUPP);
1872 }
1873
1874 static void
X_ip_rsvp_force_done(struct socket * so __unused)1875 X_ip_rsvp_force_done(struct socket *so __unused)
1876 {
1877
1878 }
1879
1880 static int
X_rsvp_input(struct mbuf ** mp,int * offp,int proto)1881 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1882 {
1883 struct mbuf *m;
1884
1885 m = *mp;
1886 *mp = NULL;
1887 if (!V_rsvp_on)
1888 m_freem(m);
1889 return (IPPROTO_DONE);
1890 }
1891
1892 /*
1893 * Code for bandwidth monitors
1894 */
1895
1896 /*
1897 * Define common interface for timeval-related methods
1898 */
1899 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1900 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1901 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1902
1903 static uint32_t
compute_bw_meter_flags(struct bw_upcall * req)1904 compute_bw_meter_flags(struct bw_upcall *req)
1905 {
1906 uint32_t flags = 0;
1907
1908 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1909 flags |= BW_METER_UNIT_PACKETS;
1910 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1911 flags |= BW_METER_UNIT_BYTES;
1912 if (req->bu_flags & BW_UPCALL_GEQ)
1913 flags |= BW_METER_GEQ;
1914 if (req->bu_flags & BW_UPCALL_LEQ)
1915 flags |= BW_METER_LEQ;
1916
1917 return flags;
1918 }
1919
1920 static void
expire_bw_meter_leq(void * arg)1921 expire_bw_meter_leq(void *arg)
1922 {
1923 struct bw_meter *x = arg;
1924 struct timeval now;
1925
1926 CURVNET_SET((struct vnet *)x->arg);
1927
1928 MRW_LOCK_ASSERT();
1929
1930 microtime(&now);
1931
1932 /*
1933 * Test if we should deliver an upcall
1934 */
1935 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1936 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1937 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1938 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1939 /* Prepare an upcall for delivery */
1940 bw_meter_prepare_upcall(x, &now);
1941 }
1942
1943 /* Send all upcalls that are pending delivery */
1944 taskqueue_enqueue(V_task_queue, &V_task);
1945
1946 /* Reset counters */
1947 x->bm_start_time = now;
1948 /*
1949 * The lock has to be taken as ip_forward context
1950 * might modify these fields as well
1951 */
1952 mtx_lock(&x->bm_mtx);
1953 x->bm_measured.b_bytes = 0;
1954 x->bm_measured.b_packets = 0;
1955 mtx_unlock(&x->bm_mtx);
1956
1957 callout_schedule(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time));
1958
1959 CURVNET_RESTORE();
1960 }
1961
1962 /*
1963 * Add a bw_meter entry
1964 */
1965 static int
add_bw_upcall(struct mfctable * mfct,struct bw_upcall * req)1966 add_bw_upcall(struct mfctable *mfct, struct bw_upcall *req)
1967 {
1968 struct mfc *mfc;
1969 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1970 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1971 struct timeval now;
1972 struct bw_meter *x, **bwm_ptr;
1973 uint32_t flags;
1974
1975 if (!(mfct->api_config & MRT_MFC_BW_UPCALL))
1976 return EOPNOTSUPP;
1977
1978 /* Test if the flags are valid */
1979 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1980 return EINVAL;
1981 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1982 return EINVAL;
1983 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1984 return EINVAL;
1985
1986 /* Test if the threshold time interval is valid */
1987 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1988 return EINVAL;
1989
1990 flags = compute_bw_meter_flags(req);
1991
1992 /*
1993 * Find if we have already same bw_meter entry
1994 */
1995 MRW_WLOCK();
1996 mfc = mfc_find(mfct, &req->bu_src, &req->bu_dst);
1997 if (mfc == NULL) {
1998 MRW_WUNLOCK();
1999 return EADDRNOTAVAIL;
2000 }
2001
2002 /* Choose an appropriate bw_meter list */
2003 if (req->bu_flags & BW_UPCALL_GEQ)
2004 bwm_ptr = &mfc->mfc_bw_meter_geq;
2005 else
2006 bwm_ptr = &mfc->mfc_bw_meter_leq;
2007
2008 for (x = *bwm_ptr; x != NULL; x = x->bm_mfc_next) {
2009 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2010 &req->bu_threshold.b_time, ==))
2011 && (x->bm_threshold.b_packets
2012 == req->bu_threshold.b_packets)
2013 && (x->bm_threshold.b_bytes
2014 == req->bu_threshold.b_bytes)
2015 && (x->bm_flags & BW_METER_USER_FLAGS)
2016 == flags) {
2017 MRW_WUNLOCK();
2018 return 0; /* XXX Already installed */
2019 }
2020 }
2021
2022 /* Allocate the new bw_meter entry */
2023 x = malloc(sizeof(*x), M_BWMETER, M_ZERO | M_NOWAIT);
2024 if (x == NULL) {
2025 MRW_WUNLOCK();
2026 return ENOBUFS;
2027 }
2028
2029 /* Set the new bw_meter entry */
2030 x->bm_threshold.b_time = req->bu_threshold.b_time;
2031 microtime(&now);
2032 x->bm_start_time = now;
2033 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2034 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2035 x->bm_measured.b_packets = 0;
2036 x->bm_measured.b_bytes = 0;
2037 x->bm_flags = flags;
2038 x->bm_time_next = NULL;
2039 x->bm_mfc = mfc;
2040 x->bm_mfctable = mfct;
2041 x->arg = curvnet;
2042 sprintf(x->bm_mtx_name, "BM mtx %p", x);
2043 mtx_init(&x->bm_mtx, x->bm_mtx_name, NULL, MTX_DEF);
2044
2045 /* For LEQ case create periodic callout */
2046 if (req->bu_flags & BW_UPCALL_LEQ) {
2047 callout_init_rw(&x->bm_meter_callout, &mrouter_lock, CALLOUT_SHAREDLOCK);
2048 callout_reset(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time),
2049 expire_bw_meter_leq, x);
2050 }
2051
2052 /* Add the new bw_meter entry to the front of entries for this MFC */
2053 x->bm_mfc_next = *bwm_ptr;
2054 *bwm_ptr = x;
2055
2056 MRW_WUNLOCK();
2057
2058 return 0;
2059 }
2060
2061 static void
free_bw_list(struct bw_meter * list)2062 free_bw_list(struct bw_meter *list)
2063 {
2064 while (list != NULL) {
2065 struct bw_meter *x = list;
2066
2067 /* MRW_WLOCK must be held here */
2068 if (x->bm_flags & BW_METER_LEQ) {
2069 callout_drain(&x->bm_meter_callout);
2070 mtx_destroy(&x->bm_mtx);
2071 }
2072
2073 list = list->bm_mfc_next;
2074 free(x, M_BWMETER);
2075 }
2076 }
2077
2078 /*
2079 * Delete one or multiple bw_meter entries
2080 */
2081 static int
del_bw_upcall(struct mfctable * mfct,struct bw_upcall * req)2082 del_bw_upcall(struct mfctable *mfct, struct bw_upcall *req)
2083 {
2084 struct mfc *mfc;
2085 struct bw_meter *x, **bwm_ptr;
2086
2087 if (!(mfct->api_config & MRT_MFC_BW_UPCALL))
2088 return EOPNOTSUPP;
2089
2090 MRW_WLOCK();
2091
2092 /* Find the corresponding MFC entry */
2093 mfc = mfc_find(mfct, &req->bu_src, &req->bu_dst);
2094 if (mfc == NULL) {
2095 MRW_WUNLOCK();
2096 return EADDRNOTAVAIL;
2097 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2098 /*
2099 * Delete all bw_meter entries for this mfc
2100 */
2101 struct bw_meter *list;
2102
2103 /* Free LEQ list */
2104 list = mfc->mfc_bw_meter_leq;
2105 mfc->mfc_bw_meter_leq = NULL;
2106 free_bw_list(list);
2107
2108 /* Free GEQ list */
2109 list = mfc->mfc_bw_meter_geq;
2110 mfc->mfc_bw_meter_geq = NULL;
2111 free_bw_list(list);
2112 MRW_WUNLOCK();
2113 return 0;
2114 } else { /* Delete a single bw_meter entry */
2115 struct bw_meter *prev;
2116 uint32_t flags = 0;
2117
2118 flags = compute_bw_meter_flags(req);
2119
2120 /* Choose an appropriate bw_meter list */
2121 if (req->bu_flags & BW_UPCALL_GEQ)
2122 bwm_ptr = &mfc->mfc_bw_meter_geq;
2123 else
2124 bwm_ptr = &mfc->mfc_bw_meter_leq;
2125
2126 /* Find the bw_meter entry to delete */
2127 for (prev = NULL, x = *bwm_ptr; x != NULL;
2128 prev = x, x = x->bm_mfc_next) {
2129 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, &req->bu_threshold.b_time, ==)) &&
2130 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2131 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2132 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2133 break;
2134 }
2135 if (x != NULL) { /* Delete entry from the list for this MFC */
2136 if (prev != NULL)
2137 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
2138 else
2139 *bwm_ptr = x->bm_mfc_next;/* new head of list */
2140
2141 if (req->bu_flags & BW_UPCALL_LEQ)
2142 callout_stop(&x->bm_meter_callout);
2143
2144 MRW_WUNLOCK();
2145 /* Free the bw_meter entry */
2146 free(x, M_BWMETER);
2147 return 0;
2148 } else {
2149 MRW_WUNLOCK();
2150 return EINVAL;
2151 }
2152 }
2153 __assert_unreachable();
2154 }
2155
2156 /*
2157 * Perform bandwidth measurement processing that may result in an upcall
2158 */
2159 static void
bw_meter_geq_receive_packet(struct bw_meter * x,int plen,struct timeval * nowp)2160 bw_meter_geq_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2161 {
2162 struct timeval delta;
2163
2164 MRW_LOCK_ASSERT();
2165
2166 delta = *nowp;
2167 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2168
2169 /*
2170 * Processing for ">=" type of bw_meter entry.
2171 * bm_mtx does not have to be hold here as in GEQ
2172 * case this is the only context accessing bm_measured.
2173 */
2174 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2175 /* Reset the bw_meter entry */
2176 x->bm_start_time = *nowp;
2177 x->bm_measured.b_packets = 0;
2178 x->bm_measured.b_bytes = 0;
2179 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2180 }
2181
2182 /* Record that a packet is received */
2183 x->bm_measured.b_packets++;
2184 x->bm_measured.b_bytes += plen;
2185
2186 /*
2187 * Test if we should deliver an upcall
2188 */
2189 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2190 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2191 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2192 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2193 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2194 /* Prepare an upcall for delivery */
2195 bw_meter_prepare_upcall(x, nowp);
2196 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2197 }
2198 }
2199 }
2200
2201 /*
2202 * Prepare a bandwidth-related upcall
2203 */
2204 static void
bw_meter_prepare_upcall(struct bw_meter * x,struct timeval * nowp)2205 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2206 {
2207 struct timeval delta;
2208 struct bw_upcall *u;
2209
2210 MRW_LOCK_ASSERT();
2211
2212 /*
2213 * Compute the measured time interval
2214 */
2215 delta = *nowp;
2216 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2217
2218 /*
2219 * Set the bw_upcall entry
2220 */
2221 u = malloc(sizeof(struct bw_upcall), M_MRTABLE, M_NOWAIT | M_ZERO);
2222 if (!u) {
2223 log(LOG_WARNING, "bw_meter_prepare_upcall: cannot allocate entry\n");
2224 return;
2225 }
2226 u->bu_src = x->bm_mfc->mfc_origin;
2227 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2228 u->bu_threshold.b_time = x->bm_threshold.b_time;
2229 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2230 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2231 u->bu_measured.b_time = delta;
2232 u->bu_measured.b_packets = x->bm_measured.b_packets;
2233 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2234 u->bu_flags = 0;
2235 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2236 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2237 if (x->bm_flags & BW_METER_UNIT_BYTES)
2238 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2239 if (x->bm_flags & BW_METER_GEQ)
2240 u->bu_flags |= BW_UPCALL_GEQ;
2241 if (x->bm_flags & BW_METER_LEQ)
2242 u->bu_flags |= BW_UPCALL_LEQ;
2243
2244 if (buf_ring_enqueue(x->bm_mfctable->bw_upcalls, u))
2245 log(LOG_WARNING, "bw_meter_prepare_upcall: cannot enqueue upcall\n");
2246 if (buf_ring_count(x->bm_mfctable->bw_upcalls) > (BW_UPCALLS_MAX / 2)) {
2247 taskqueue_enqueue(V_task_queue, &V_task);
2248 }
2249 }
2250 /*
2251 * Send the pending bandwidth-related upcalls
2252 */
2253 static void
bw_upcalls_send(struct mfctable * mfct)2254 bw_upcalls_send(struct mfctable *mfct)
2255 {
2256 struct mbuf *m;
2257 int len = 0;
2258 struct bw_upcall *bu;
2259 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2260 static struct igmpmsg igmpmsg = {
2261 0, /* unused1 */
2262 0, /* unused2 */
2263 IGMPMSG_BW_UPCALL,/* im_msgtype */
2264 0, /* im_mbz */
2265 0, /* im_vif */
2266 0, /* unused3 */
2267 { 0 }, /* im_src */
2268 { 0 } /* im_dst */
2269 };
2270
2271 MRW_LOCK_ASSERT();
2272
2273 if (buf_ring_empty(mfct->bw_upcalls))
2274 return;
2275
2276 /*
2277 * Allocate a new mbuf, initialize it with the header and
2278 * the payload for the pending calls.
2279 */
2280 m = m_gethdr(M_NOWAIT, MT_DATA);
2281 if (m == NULL) {
2282 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2283 return;
2284 }
2285
2286 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2287 len += sizeof(struct igmpmsg);
2288 while ((bu = buf_ring_dequeue_mc(mfct->bw_upcalls)) != NULL) {
2289 m_copyback(m, len, sizeof(struct bw_upcall), (caddr_t)bu);
2290 len += sizeof(struct bw_upcall);
2291 free(bu, M_MRTABLE);
2292 }
2293
2294 /*
2295 * Send the upcalls
2296 * XXX do we need to set the address in k_igmpsrc ?
2297 */
2298 MRTSTAT_INC(mrts_upcalls);
2299 if (socket_send(mfct->router, m, &k_igmpsrc) < 0) {
2300 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2301 MRTSTAT_INC(mrts_upq_sockfull);
2302 }
2303 }
2304
2305 static void
bw_upcalls_send_all(void)2306 bw_upcalls_send_all(void)
2307 {
2308 for (int i = 0; i < V_nmfctables; i++) {
2309 struct mfctable *mfct;
2310
2311 mfct = &V_mfctables[i];
2312 if (mfct->router != NULL)
2313 bw_upcalls_send(mfct);
2314 }
2315 }
2316
2317 /*
2318 * A periodic function for sending all upcalls that are pending delivery
2319 */
2320 static void
expire_bw_upcalls_send(void * arg)2321 expire_bw_upcalls_send(void *arg)
2322 {
2323 CURVNET_SET((struct vnet *) arg);
2324
2325 /* This callout is run with MRW_RLOCK taken */
2326
2327 bw_upcalls_send_all();
2328
2329 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2330 curvnet);
2331 CURVNET_RESTORE();
2332 }
2333
2334 /*
2335 * End of bandwidth monitoring code
2336 */
2337
2338 /*
2339 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2340 *
2341 */
2342 static int
pim_register_send(struct mfctable * mfct,struct ip * ip,struct vif * vifp,struct mbuf * m,struct mfc * rt)2343 pim_register_send(struct mfctable *mfct, struct ip *ip, struct vif *vifp,
2344 struct mbuf *m, struct mfc *rt)
2345 {
2346 struct mbuf *mb_copy, *mm;
2347
2348 /*
2349 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2350 * rendezvous point was unspecified, and we were told not to.
2351 */
2352 if (pim_squelch_wholepkt != 0 && (mfct->api_config & MRT_MFC_RP) &&
2353 in_nullhost(rt->mfc_rp))
2354 return 0;
2355
2356 mb_copy = pim_register_prepare(ip, m);
2357 if (mb_copy == NULL)
2358 return ENOBUFS;
2359
2360 /*
2361 * Send all the fragments. Note that the mbuf for each fragment
2362 * is freed by the sending machinery.
2363 */
2364 for (mm = mb_copy; mm; mm = mb_copy) {
2365 mb_copy = mm->m_nextpkt;
2366 mm->m_nextpkt = 0;
2367 mm = m_pullup(mm, sizeof(struct ip));
2368 if (mm != NULL) {
2369 ip = mtod(mm, struct ip *);
2370 if ((mfct->api_config & MRT_MFC_RP) &&
2371 !in_nullhost(rt->mfc_rp)) {
2372 pim_register_send_rp(mfct, ip, vifp, mm, rt);
2373 } else {
2374 pim_register_send_upcall(mfct, ip, vifp, mm,
2375 rt);
2376 }
2377 }
2378 }
2379
2380 return 0;
2381 }
2382
2383 /*
2384 * Return a copy of the data packet that is ready for PIM Register
2385 * encapsulation.
2386 * XXX: Note that in the returned copy the IP header is a valid one.
2387 */
2388 static struct mbuf *
pim_register_prepare(struct ip * ip,struct mbuf * m)2389 pim_register_prepare(struct ip *ip, struct mbuf *m)
2390 {
2391 struct mbuf *mb_copy = NULL;
2392 int mtu;
2393
2394 /* Take care of delayed checksums */
2395 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2396 in_delayed_cksum(m);
2397 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2398 }
2399
2400 /*
2401 * Copy the old packet & pullup its IP header into the
2402 * new mbuf so we can modify it.
2403 */
2404 mb_copy = m_copypacket(m, M_NOWAIT);
2405 if (mb_copy == NULL)
2406 return NULL;
2407 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2408 if (mb_copy == NULL)
2409 return NULL;
2410
2411 /* take care of the TTL */
2412 ip = mtod(mb_copy, struct ip *);
2413 --ip->ip_ttl;
2414
2415 /* Compute the MTU after the PIM Register encapsulation */
2416 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2417
2418 if (ntohs(ip->ip_len) <= mtu) {
2419 /* Turn the IP header into a valid one */
2420 ip->ip_sum = 0;
2421 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2422 } else {
2423 /* Fragment the packet */
2424 mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2425 if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2426 m_freem(mb_copy);
2427 return NULL;
2428 }
2429 }
2430 return mb_copy;
2431 }
2432
2433 /*
2434 * Send an upcall with the data packet to the user-level process.
2435 */
2436 static int
pim_register_send_upcall(struct mfctable * mfct,struct ip * ip,struct vif * vifp,struct mbuf * mb_copy,struct mfc * rt)2437 pim_register_send_upcall(struct mfctable *mfct, struct ip *ip, struct vif *vifp,
2438 struct mbuf *mb_copy, struct mfc *rt)
2439 {
2440 struct mbuf *mb_first;
2441 int len = ntohs(ip->ip_len);
2442 struct igmpmsg *im;
2443 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2444
2445 MRW_LOCK_ASSERT();
2446
2447 /*
2448 * Add a new mbuf with an upcall header
2449 */
2450 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2451 if (mb_first == NULL) {
2452 m_freem(mb_copy);
2453 return ENOBUFS;
2454 }
2455 mb_first->m_data += max_linkhdr;
2456 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2457 mb_first->m_len = sizeof(struct igmpmsg);
2458 mb_first->m_next = mb_copy;
2459
2460 /* Send message to routing daemon */
2461 im = mtod(mb_first, struct igmpmsg *);
2462 memset(im, 0, sizeof(*im));
2463 im->im_msgtype = IGMPMSG_WHOLEPKT;
2464 im->im_mbz = 0;
2465 im->im_vif = vifp - mfct->viftable;
2466 im->im_src = ip->ip_src;
2467 im->im_dst = ip->ip_dst;
2468
2469 k_igmpsrc.sin_addr = ip->ip_src;
2470
2471 MRTSTAT_INC(mrts_upcalls);
2472
2473 if (socket_send(mfct->router, mb_first, &k_igmpsrc) < 0) {
2474 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2475 MRTSTAT_INC(mrts_upq_sockfull);
2476 return ENOBUFS;
2477 }
2478
2479 /* Keep statistics */
2480 PIMSTAT_INC(pims_snd_registers_msgs);
2481 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2482
2483 return 0;
2484 }
2485
2486 /*
2487 * Encapsulate the data packet in PIM Register message and send it to the RP.
2488 */
2489 static int
pim_register_send_rp(struct mfctable * mfct,struct ip * ip,struct vif * vifp,struct mbuf * mb_copy,struct mfc * rt)2490 pim_register_send_rp(struct mfctable *mfct, struct ip *ip, struct vif *vifp,
2491 struct mbuf *mb_copy, struct mfc *rt)
2492 {
2493 struct mbuf *mb_first;
2494 struct ip *ip_outer;
2495 struct pim_encap_pimhdr *pimhdr;
2496 int len = ntohs(ip->ip_len);
2497 vifi_t vifi = rt->mfc_parent;
2498
2499 MRW_LOCK_ASSERT();
2500
2501 if (vifi >= mfct->numvifs ||
2502 in_nullhost(mfct->viftable[vifi].v_lcl_addr)) {
2503 m_freem(mb_copy);
2504 return EADDRNOTAVAIL; /* The iif vif is invalid */
2505 }
2506
2507 /*
2508 * Add a new mbuf with the encapsulating header
2509 */
2510 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2511 if (mb_first == NULL) {
2512 m_freem(mb_copy);
2513 return ENOBUFS;
2514 }
2515 mb_first->m_data += max_linkhdr;
2516 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2517 mb_first->m_next = mb_copy;
2518
2519 mb_first->m_pkthdr.len = len + mb_first->m_len;
2520
2521 /*
2522 * Fill in the encapsulating IP and PIM header
2523 */
2524 ip_outer = mtod(mb_first, struct ip *);
2525 *ip_outer = pim_encap_iphdr;
2526 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2527 sizeof(pim_encap_pimhdr));
2528 ip_outer->ip_src = mfct->viftable[vifi].v_lcl_addr;
2529 ip_outer->ip_dst = rt->mfc_rp;
2530 /*
2531 * Copy the inner header TOS to the outer header, and take care of the
2532 * IP_DF bit.
2533 */
2534 ip_outer->ip_tos = ip->ip_tos;
2535 if (ip->ip_off & htons(IP_DF))
2536 ip_outer->ip_off |= htons(IP_DF);
2537 ip_fillid(ip_outer, V_ip_random_id);
2538 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2539 + sizeof(pim_encap_iphdr));
2540 *pimhdr = pim_encap_pimhdr;
2541 /* If the iif crosses a border, set the Border-bit */
2542 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mfct->api_config)
2543 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2544
2545 mb_first->m_data += sizeof(pim_encap_iphdr);
2546 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2547 mb_first->m_data -= sizeof(pim_encap_iphdr);
2548
2549 send_packet(vifp, mb_first);
2550
2551 /* Keep statistics */
2552 PIMSTAT_INC(pims_snd_registers_msgs);
2553 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2554
2555 return 0;
2556 }
2557
2558 /*
2559 * pim_encapcheck() is called by the encap4_input() path at runtime to
2560 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2561 * into the kernel.
2562 */
2563 static int
pim_encapcheck(const struct mbuf * m __unused,int off __unused,int proto __unused,void * arg __unused)2564 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2565 int proto __unused, void *arg __unused)
2566 {
2567
2568 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2569 return (8); /* claim the datagram. */
2570 }
2571
2572 /*
2573 * PIM-SMv2 and PIM-DM messages processing.
2574 * Receives and verifies the PIM control messages, and passes them
2575 * up to the listening socket, using rip_input().
2576 * The only message with special processing is the PIM_REGISTER message
2577 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2578 * is passed to if_simloop().
2579 */
2580 static int
pim_input(struct mbuf * m,int off,int proto,void * arg __unused)2581 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2582 {
2583 struct mfctable *mfct;
2584 struct ip *ip = mtod(m, struct ip *);
2585 struct pim *pim;
2586 int iphlen = off;
2587 int minlen;
2588 int datalen = ntohs(ip->ip_len) - iphlen;
2589 int ip_tos;
2590
2591 mfct = &V_mfctables[M_GETFIB(m)];
2592
2593 /* Keep statistics */
2594 PIMSTAT_INC(pims_rcv_total_msgs);
2595 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2596
2597 /*
2598 * Validate lengths
2599 */
2600 if (datalen < PIM_MINLEN) {
2601 PIMSTAT_INC(pims_rcv_tooshort);
2602 CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2603 __func__, datalen, ntohl(ip->ip_src.s_addr));
2604 m_freem(m);
2605 return (IPPROTO_DONE);
2606 }
2607
2608 /*
2609 * If the packet is at least as big as a REGISTER, go agead
2610 * and grab the PIM REGISTER header size, to avoid another
2611 * possible m_pullup() later.
2612 *
2613 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2614 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2615 */
2616 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2617 /*
2618 * Get the IP and PIM headers in contiguous memory, and
2619 * possibly the PIM REGISTER header.
2620 */
2621 if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2622 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2623 return (IPPROTO_DONE);
2624 }
2625
2626 /* m_pullup() may have given us a new mbuf so reset ip. */
2627 ip = mtod(m, struct ip *);
2628 ip_tos = ip->ip_tos;
2629
2630 /* adjust mbuf to point to the PIM header */
2631 m->m_data += iphlen;
2632 m->m_len -= iphlen;
2633 pim = mtod(m, struct pim *);
2634
2635 /*
2636 * Validate checksum. If PIM REGISTER, exclude the data packet.
2637 *
2638 * XXX: some older PIMv2 implementations don't make this distinction,
2639 * so for compatibility reason perform the checksum over part of the
2640 * message, and if error, then over the whole message.
2641 */
2642 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2643 /* do nothing, checksum okay */
2644 } else if (in_cksum(m, datalen)) {
2645 PIMSTAT_INC(pims_rcv_badsum);
2646 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2647 m_freem(m);
2648 return (IPPROTO_DONE);
2649 }
2650
2651 /* PIM version check */
2652 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2653 PIMSTAT_INC(pims_rcv_badversion);
2654 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2655 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2656 m_freem(m);
2657 return (IPPROTO_DONE);
2658 }
2659
2660 /* restore mbuf back to the outer IP */
2661 m->m_data -= iphlen;
2662 m->m_len += iphlen;
2663
2664 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2665 /*
2666 * Since this is a REGISTER, we'll make a copy of the register
2667 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2668 * routing daemon.
2669 */
2670 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2671 struct mbuf *mcp;
2672 struct ip *encap_ip;
2673 u_int32_t *reghdr;
2674 struct ifnet *vifp;
2675
2676 MRW_RLOCK();
2677 if (mfct->register_vif >= mfct->numvifs ||
2678 mfct->register_vif == VIFI_INVALID) {
2679 MRW_RUNLOCK();
2680 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2681 (int)mfct->register_vif);
2682 m_freem(m);
2683 return (IPPROTO_DONE);
2684 }
2685 /* XXX need refcnt? */
2686 vifp = mfct->viftable[mfct->register_vif].v_ifp;
2687 MRW_RUNLOCK();
2688
2689 /*
2690 * Validate length
2691 */
2692 if (datalen < PIM_REG_MINLEN) {
2693 PIMSTAT_INC(pims_rcv_tooshort);
2694 PIMSTAT_INC(pims_rcv_badregisters);
2695 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2696 m_freem(m);
2697 return (IPPROTO_DONE);
2698 }
2699
2700 reghdr = (u_int32_t *)(pim + 1);
2701 encap_ip = (struct ip *)(reghdr + 1);
2702
2703 CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2704 __func__, ntohl(encap_ip->ip_src.s_addr),
2705 ntohs(encap_ip->ip_len));
2706
2707 /* verify the version number of the inner packet */
2708 if (encap_ip->ip_v != IPVERSION) {
2709 PIMSTAT_INC(pims_rcv_badregisters);
2710 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2711 m_freem(m);
2712 return (IPPROTO_DONE);
2713 }
2714
2715 /* verify the inner packet is destined to a mcast group */
2716 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2717 PIMSTAT_INC(pims_rcv_badregisters);
2718 CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2719 ntohl(encap_ip->ip_dst.s_addr));
2720 m_freem(m);
2721 return (IPPROTO_DONE);
2722 }
2723
2724 /* If a NULL_REGISTER, pass it to the daemon */
2725 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2726 goto pim_input_to_daemon;
2727
2728 /*
2729 * Copy the TOS from the outer IP header to the inner IP header.
2730 */
2731 if (encap_ip->ip_tos != ip_tos) {
2732 /* Outer TOS -> inner TOS */
2733 encap_ip->ip_tos = ip_tos;
2734 /* Recompute the inner header checksum. Sigh... */
2735
2736 /* adjust mbuf to point to the inner IP header */
2737 m->m_data += (iphlen + PIM_MINLEN);
2738 m->m_len -= (iphlen + PIM_MINLEN);
2739
2740 encap_ip->ip_sum = 0;
2741 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2742
2743 /* restore mbuf to point back to the outer IP header */
2744 m->m_data -= (iphlen + PIM_MINLEN);
2745 m->m_len += (iphlen + PIM_MINLEN);
2746 }
2747
2748 /*
2749 * Decapsulate the inner IP packet and loopback to forward it
2750 * as a normal multicast packet. Also, make a copy of the
2751 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2752 * to pass to the daemon later, so it can take the appropriate
2753 * actions (e.g., send back PIM_REGISTER_STOP).
2754 * XXX: here m->m_data points to the outer IP header.
2755 */
2756 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2757 if (mcp == NULL) {
2758 CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2759 m_freem(m);
2760 return (IPPROTO_DONE);
2761 }
2762
2763 /* Keep statistics */
2764 /* XXX: registers_bytes include only the encap. mcast pkt */
2765 PIMSTAT_INC(pims_rcv_registers_msgs);
2766 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2767
2768 /*
2769 * forward the inner ip packet; point m_data at the inner ip.
2770 */
2771 m_adj(m, iphlen + PIM_MINLEN);
2772
2773 CTR4(KTR_IPMF,
2774 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2775 __func__,
2776 (u_long)ntohl(encap_ip->ip_src.s_addr),
2777 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2778 (int)mfct->register_vif);
2779
2780 /* NB: vifp was collected above; can it change on us? */
2781 if_simloop(vifp, m, dst.sin_family, 0);
2782
2783 /* prepare the register head to send to the mrouting daemon */
2784 m = mcp;
2785 }
2786
2787 pim_input_to_daemon:
2788 /*
2789 * Pass the PIM message up to the daemon; if it is a Register message,
2790 * pass the 'head' only up to the daemon. This includes the
2791 * outer IP header, PIM header, PIM-Register header and the
2792 * inner IP header.
2793 * XXX: the outer IP header pkt size of a Register is not adjust to
2794 * reflect the fact that the inner multicast data is truncated.
2795 */
2796 return (rip_input(&m, &off, proto));
2797 }
2798
2799 static int
sysctl_mfctable(SYSCTL_HANDLER_ARGS)2800 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2801 {
2802 struct mfctable *mfct;
2803 struct mfc *rt;
2804 int error, i;
2805
2806 if (req->newptr)
2807 return (EPERM);
2808 mfct = &V_mfctables[curthread->td_proc->p_fibnum];
2809 if (mfct->mfchashtbl == NULL) /* XXX unlocked */
2810 return (0);
2811 error = sysctl_wire_old_buffer(req, 0);
2812 if (error)
2813 return (error);
2814
2815 MRW_RLOCK();
2816 if (mfct->mfchashtbl == NULL)
2817 goto out_locked;
2818
2819 for (i = 0; i < mfchashsize; i++) {
2820 LIST_FOREACH(rt, &mfct->mfchashtbl[i], mfc_hash) {
2821 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2822 if (error)
2823 goto out_locked;
2824 }
2825 }
2826 out_locked:
2827 MRW_RUNLOCK();
2828 return (error);
2829 }
2830 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2831 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2832 "IPv4 Multicast Forwarding Table "
2833 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2834
2835 static int
sysctl_viflist(SYSCTL_HANDLER_ARGS)2836 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2837 {
2838 struct mfctable *mfct;
2839 int error, i;
2840
2841 if (req->newptr)
2842 return (EPERM);
2843 error = sysctl_wire_old_buffer(req, MROUTE_VIF_SYSCTL_LEN * MAXVIFS);
2844 if (error)
2845 return (error);
2846
2847 mfct = &V_mfctables[curthread->td_proc->p_fibnum];
2848 MRW_RLOCK();
2849 /* Copy out user-visible portion of vif entry. */
2850 for (i = 0; i < MAXVIFS; i++) {
2851 error = SYSCTL_OUT(req, &mfct->viftable[i],
2852 MROUTE_VIF_SYSCTL_LEN);
2853 if (error)
2854 break;
2855 }
2856 MRW_RUNLOCK();
2857 return (error);
2858 }
2859 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2860 CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2861 sysctl_viflist, "S,vif[MAXVIFS]",
2862 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2863
2864 static void
ip_mroute_rtnumfibs_change(void * arg __unused,uint32_t ntables)2865 ip_mroute_rtnumfibs_change(void *arg __unused, uint32_t ntables)
2866 {
2867 struct mfctable *mfctables, *omfctables;
2868
2869 KASSERT(ntables >= V_nmfctables,
2870 ("%s: ntables %u nmfctables %u", __func__, ntables, V_nmfctables));
2871
2872 mfctables = mallocarray(ntables, sizeof(*mfctables), M_MRTABLE,
2873 M_WAITOK | M_ZERO);
2874 omfctables = V_mfctables;
2875
2876 for (int i = V_nmfctables; i < ntables; i++) {
2877 struct mfctable *mfct;
2878
2879 mfct = &mfctables[i];
2880 mfct->nexpire = malloc(mfchashsize, M_MRTABLE,
2881 M_WAITOK | M_ZERO);
2882 mfct->register_vif = VIFI_INVALID;
2883 }
2884
2885 MRW_TEARDOWN_WLOCK();
2886 MRW_WLOCK();
2887 for (int i = 0; i < V_nmfctables; i++)
2888 memcpy(&mfctables[i], &omfctables[i], sizeof(*mfctables));
2889 atomic_store_rel_ptr((uintptr_t *)&V_mfctables, (uintptr_t)mfctables);
2890 MRW_WUNLOCK();
2891 MRW_TEARDOWN_WUNLOCK();
2892
2893 NET_EPOCH_WAIT();
2894
2895 V_nmfctables = ntables;
2896 free(omfctables, M_MRTABLE);
2897 }
2898
2899 static void
vnet_mroute_init(const void * unused __unused)2900 vnet_mroute_init(const void *unused __unused)
2901 {
2902 ip_mroute_rtnumfibs_change(NULL, V_rt_numfibs);
2903
2904 callout_init_rw(&V_expire_upcalls_ch, &mrouter_lock, 0);
2905 callout_init_rw(&V_bw_upcalls_ch, &mrouter_lock, 0);
2906
2907 /* Prepare taskqueue */
2908 V_task_queue = taskqueue_create_fast("ip_mroute_tskq", M_NOWAIT,
2909 taskqueue_thread_enqueue, &V_task_queue);
2910 taskqueue_start_threads(&V_task_queue, 1, PI_NET, "ip_mroute_tskq task");
2911 }
2912 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2913 NULL);
2914
2915 static void
vnet_mroute_uninit(const void * unused __unused)2916 vnet_mroute_uninit(const void *unused __unused)
2917 {
2918 /* Taskqueue should be cancelled and drained before freeing */
2919 taskqueue_free(V_task_queue);
2920
2921 for (int i = 0; i < V_rt_numfibs; i++) {
2922 struct mfctable *mfct;
2923
2924 mfct = &V_mfctables[i];
2925 free(mfct->nexpire, M_MRTABLE);
2926 }
2927 free(V_mfctables, M_MRTABLE);
2928
2929 callout_drain(&V_expire_upcalls_ch);
2930 callout_drain(&V_bw_upcalls_ch);
2931 }
2932 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2933 vnet_mroute_uninit, NULL);
2934
2935 static int
ip_mroute_modevent(module_t mod,int type,void * unused)2936 ip_mroute_modevent(module_t mod, int type, void *unused)
2937 {
2938
2939 switch (type) {
2940 case MOD_LOAD:
2941 MRW_TEARDOWN_LOCK_INIT();
2942 MRW_LOCK_INIT();
2943
2944 if_detach_event_tag = EVENTHANDLER_REGISTER(
2945 ifnet_departure_event, if_detached_event, NULL,
2946 EVENTHANDLER_PRI_ANY);
2947 rtnumfibs_change_tag = EVENTHANDLER_REGISTER(
2948 rtnumfibs_change, ip_mroute_rtnumfibs_change,
2949 NULL, EVENTHANDLER_PRI_ANY);
2950
2951 if (!powerof2(mfchashsize)) {
2952 printf("WARNING: %s not a power of 2; using default\n",
2953 "net.inet.ip.mfchashsize");
2954 mfchashsize = MFCHASHSIZE;
2955 }
2956
2957 pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2958
2959 ip_mcast_src = X_ip_mcast_src;
2960 ip_mforward = X_ip_mforward;
2961 ip_mrouter_done = X_ip_mrouter_done;
2962 ip_mrouter_get = X_ip_mrouter_get;
2963 ip_mrouter_set = X_ip_mrouter_set;
2964
2965 ip_rsvp_force_done = X_ip_rsvp_force_done;
2966 ip_rsvp_vif = X_ip_rsvp_vif;
2967
2968 legal_vif_num = X_legal_vif_num;
2969 mrt_ioctl = X_mrt_ioctl;
2970 rsvp_input_p = X_rsvp_input;
2971 break;
2972
2973 case MOD_UNLOAD:
2974 /*
2975 * Typically module unload happens after the user-level
2976 * process has shutdown the kernel services (the check
2977 * below insures someone can't just yank the module out
2978 * from under a running process). But if the module is
2979 * just loaded and then unloaded w/o starting up a user
2980 * process we still need to cleanup.
2981 */
2982 MRW_WLOCK();
2983 if (ip_mrouter_cnt != 0) {
2984 MRW_WUNLOCK();
2985 return (EBUSY);
2986 }
2987 ip_mrouter_unloading = 1;
2988 MRW_WUNLOCK();
2989
2990 EVENTHANDLER_DEREGISTER(rtnumfibs_change,
2991 rtnumfibs_change_tag);
2992 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
2993 if_detach_event_tag);
2994
2995 if (pim_encap_cookie) {
2996 ip_encap_detach(pim_encap_cookie);
2997 pim_encap_cookie = NULL;
2998 }
2999
3000 ip_mcast_src = NULL;
3001 ip_mforward = NULL;
3002 ip_mrouter_done = NULL;
3003 ip_mrouter_get = NULL;
3004 ip_mrouter_set = NULL;
3005
3006 ip_rsvp_force_done = NULL;
3007 ip_rsvp_vif = NULL;
3008
3009 legal_vif_num = NULL;
3010 mrt_ioctl = NULL;
3011 rsvp_input_p = NULL;
3012
3013 MRW_LOCK_DESTROY();
3014 MRW_TEARDOWN_LOCK_DESTROY();
3015 break;
3016
3017 default:
3018 return EOPNOTSUPP;
3019 }
3020 return 0;
3021 }
3022
3023 static moduledata_t ip_mroutemod = {
3024 "ip_mroute",
3025 ip_mroute_modevent,
3026 0
3027 };
3028
3029 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
3030