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 if (mfct->mfchashtbl == NULL)
387 return (NULL);
388 LIST_FOREACH(rt, &mfct->mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
389 if (in_hosteq(rt->mfc_origin, *o) &&
390 in_hosteq(rt->mfc_mcastgrp, *g) &&
391 buf_ring_empty(rt->mfc_stall_ring))
392 break;
393 }
394
395 return (rt);
396 }
397
398 static __inline struct mfc *
mfc_alloc(void)399 mfc_alloc(void)
400 {
401 struct mfc *rt;
402 rt = malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT | M_ZERO);
403 if (rt == NULL)
404 return rt;
405
406 rt->mfc_stall_ring = buf_ring_alloc(MAX_UPQ, M_MRTABLE, M_NOWAIT,
407 NOLOCK);
408 if (rt->mfc_stall_ring == NULL) {
409 free(rt, M_MRTABLE);
410 return NULL;
411 }
412
413 return rt;
414 }
415
416 static struct mfctable *
somfctable(struct socket * so)417 somfctable(struct socket *so)
418 {
419 int fib;
420
421 fib = atomic_load_int(&so->so_fibnum);
422 KASSERT(fib >= 0 && fib < V_nmfctables,
423 ("%s: so_fibnum %d out of range", __func__, fib));
424 return (&V_mfctables[fib]);
425 }
426
427 /*
428 * Handle MRT setsockopt commands to modify the multicast forwarding tables.
429 */
430 static int
X_ip_mrouter_set(struct socket * so,struct sockopt * sopt)431 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
432 {
433 struct mfctable *mfct;
434 int error, optval;
435 vifi_t vifi;
436 struct vifctl vifc;
437 struct mfcctl2 mfc;
438 struct bw_upcall bw_upcall;
439 uint32_t i;
440
441 mfct = somfctable(so);
442 if (so != mfct->router && sopt->sopt_name != MRT_INIT)
443 return EPERM;
444
445 error = 0;
446 switch (sopt->sopt_name) {
447 case MRT_INIT:
448 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
449 if (error)
450 break;
451 error = ip_mrouter_init(so, optval);
452 break;
453 case MRT_DONE:
454 ip_mrouter_done(so);
455 break;
456 case MRT_ADD_VIF:
457 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
458 if (error)
459 break;
460 error = add_vif(mfct, so->so_fibnum, &vifc);
461 break;
462 case MRT_DEL_VIF:
463 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
464 if (error)
465 break;
466 error = del_vif(mfct, vifi);
467 break;
468 case MRT_ADD_MFC:
469 case MRT_DEL_MFC:
470 /*
471 * select data size depending on API version.
472 */
473 if (sopt->sopt_name == MRT_ADD_MFC &&
474 (mfct->api_config & MRT_API_FLAGS_ALL) != 0) {
475 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
476 sizeof(struct mfcctl2));
477 } else {
478 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
479 sizeof(struct mfcctl));
480 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
481 sizeof(mfc) - sizeof(struct mfcctl));
482 }
483 if (error)
484 break;
485 if (sopt->sopt_name == MRT_ADD_MFC)
486 error = add_mfc(mfct, &mfc);
487 else
488 error = del_mfc(mfct, &mfc);
489 break;
490
491 case MRT_ASSERT:
492 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
493 if (error)
494 break;
495 set_assert(mfct, optval);
496 break;
497
498 case MRT_API_CONFIG:
499 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
500 if (!error)
501 error = set_api_config(mfct, &i);
502 if (!error)
503 error = sooptcopyout(sopt, &i, sizeof i);
504 break;
505
506 case MRT_ADD_BW_UPCALL:
507 case MRT_DEL_BW_UPCALL:
508 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
509 sizeof bw_upcall);
510 if (error)
511 break;
512 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
513 error = add_bw_upcall(mfct, &bw_upcall);
514 else
515 error = del_bw_upcall(mfct, &bw_upcall);
516 break;
517
518 default:
519 error = EOPNOTSUPP;
520 break;
521 }
522 return error;
523 }
524
525 /*
526 * Handle MRT getsockopt commands
527 */
528 static int
X_ip_mrouter_get(struct socket * so,struct sockopt * sopt)529 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
530 {
531 struct mfctable *mfct;
532 int error;
533
534 mfct = somfctable(so);
535 switch (sopt->sopt_name) {
536 case MRT_VERSION:
537 error = sooptcopyout(sopt, &mrt_api_version,
538 sizeof mrt_api_version);
539 break;
540 case MRT_ASSERT:
541 error = sooptcopyout(sopt, &mfct->pim_assert_enabled,
542 sizeof(mfct->pim_assert_enabled));
543 break;
544 case MRT_API_SUPPORT:
545 error = sooptcopyout(sopt, &mrt_api_support,
546 sizeof mrt_api_support);
547 break;
548 case MRT_API_CONFIG:
549 error = sooptcopyout(sopt, &mfct->api_config,
550 sizeof(mfct->api_config));
551 break;
552 default:
553 error = EOPNOTSUPP;
554 break;
555 }
556 return error;
557 }
558
559 /*
560 * Handle ioctl commands to obtain information from the cache
561 */
562 static int
X_mrt_ioctl(u_long cmd,caddr_t data,int fibnum)563 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum)
564 {
565 struct mfctable *mfct;
566 int error;
567
568 error = priv_check(curthread, PRIV_NETINET_MROUTE);
569 if (error)
570 return (error);
571
572 mfct = &V_mfctables[fibnum];
573 switch (cmd) {
574 case SIOCGETVIFCNT:
575 error = get_vif_cnt(mfct, (struct sioc_vif_req *)data);
576 break;
577
578 case SIOCGETSGCNT:
579 error = get_sg_cnt(mfct, (struct sioc_sg_req *)data);
580 break;
581
582 default:
583 error = EINVAL;
584 break;
585 }
586 return error;
587 }
588
589 /*
590 * returns the packet, byte, rpf-failure count for the source group provided
591 */
592 static int
get_sg_cnt(struct mfctable * mfct,struct sioc_sg_req * req)593 get_sg_cnt(struct mfctable *mfct, struct sioc_sg_req *req)
594 {
595 struct mfc *rt;
596
597 MRW_RLOCK();
598 rt = mfc_find(mfct, &req->src, &req->grp);
599 if (rt == NULL) {
600 MRW_RUNLOCK();
601 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
602 return EADDRNOTAVAIL;
603 }
604 req->pktcnt = rt->mfc_pkt_cnt;
605 req->bytecnt = rt->mfc_byte_cnt;
606 req->wrong_if = rt->mfc_wrong_if;
607 MRW_RUNLOCK();
608 return 0;
609 }
610
611 /*
612 * returns the input and output packet and byte counts on the vif provided
613 */
614 static int
get_vif_cnt(struct mfctable * mfct,struct sioc_vif_req * req)615 get_vif_cnt(struct mfctable *mfct, struct sioc_vif_req *req)
616 {
617 struct vif *vif;
618 vifi_t vifi;
619
620 vifi = req->vifi;
621
622 MRW_RLOCK();
623 if (vifi >= mfct->numvifs) {
624 MRW_RUNLOCK();
625 return EINVAL;
626 }
627
628 vif = &mfct->viftable[vifi];
629 mtx_lock(&vif->v_mtx);
630 req->icount = vif->v_pkt_in;
631 req->ocount = vif->v_pkt_out;
632 req->ibytes = vif->v_bytes_in;
633 req->obytes = vif->v_bytes_out;
634 mtx_unlock(&vif->v_mtx);
635 MRW_RUNLOCK();
636
637 return 0;
638 }
639
640 /*
641 * Tear down multicast forwarder state associated with this ifnet.
642 * 1. Walk the vif list, matching vifs against this ifnet.
643 * 2. Walk the multicast forwarding cache (mfc) looking for
644 * inner matches with this vif's index.
645 * 3. Expire any matching multicast forwarding cache entries.
646 * 4. Free vif state. This should disable ALLMULTI on the interface.
647 */
648 static int
detach_ifnet(struct mfctable * mfct,struct ifnet * ifp)649 detach_ifnet(struct mfctable *mfct, struct ifnet *ifp)
650 {
651 struct ifnet *free_ptr, *multi_leave;
652 int count;
653
654 count = 0;
655 restart:
656 for (vifi_t vifi = 0; vifi < mfct->numvifs; vifi++) {
657 if (mfct->viftable[vifi].v_ifp != ifp)
658 continue;
659 for (u_long i = 0; i < mfchashsize; i++) {
660 struct mfc *rt, *nrt;
661
662 LIST_FOREACH_SAFE(rt, &mfct->mfchashtbl[i], mfc_hash,
663 nrt) {
664 if (rt->mfc_parent == vifi) {
665 expire_mfc(rt);
666 }
667 }
668 }
669 del_vif_locked(mfct, vifi, &multi_leave, &free_ptr);
670 if (free_ptr != NULL)
671 count++;
672 if (multi_leave) {
673 MRW_WUNLOCK();
674 if_allmulti(multi_leave, 0);
675 MRW_WLOCK();
676 goto restart;
677 }
678 }
679 return (count);
680 }
681
682 static void
if_detached_event(void * arg __unused,struct ifnet * ifp)683 if_detached_event(void *arg __unused, struct ifnet *ifp)
684 {
685 int count;
686
687 MRW_WLOCK();
688 if (!V_ip_mrouting_enabled) {
689 MRW_WUNLOCK();
690 return;
691 }
692
693 count = 0;
694 for (int i = 0; i < V_nmfctables; i++)
695 count += detach_ifnet(&V_mfctables[i], ifp);
696 MRW_WUNLOCK();
697
698 for (int i = 0; i < count; i++)
699 if_free(ifp);
700 }
701
702 static void
ip_mrouter_upcall_thread(void * arg,int pending __unused)703 ip_mrouter_upcall_thread(void *arg, int pending __unused)
704 {
705 CURVNET_SET((struct vnet *) arg);
706
707 MRW_WLOCK();
708 bw_upcalls_send_all();
709 MRW_WUNLOCK();
710
711 CURVNET_RESTORE();
712 }
713
714 /*
715 * Enable multicast forwarding.
716 */
717 static int
ip_mrouter_init(struct socket * so,int version)718 ip_mrouter_init(struct socket *so, int version)
719 {
720 struct mfctable *mfct;
721
722 CTR2(KTR_IPMF, "%s: so %p", __func__, so);
723
724 if (version != 1)
725 return ENOPROTOOPT;
726
727 mfct = somfctable(so);
728 MRW_TEARDOWN_WLOCK();
729 MRW_WLOCK();
730
731 if (ip_mrouter_unloading) {
732 MRW_WUNLOCK();
733 MRW_TEARDOWN_WUNLOCK();
734 return ENOPROTOOPT;
735 }
736
737 if (mfct->router != NULL) {
738 MRW_WUNLOCK();
739 MRW_TEARDOWN_WUNLOCK();
740 return EADDRINUSE;
741 }
742
743 mfct->mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
744 HASH_NOWAIT);
745 if (mfct->mfchashtbl == NULL) {
746 MRW_WUNLOCK();
747 MRW_TEARDOWN_WUNLOCK();
748 return (ENOMEM);
749 }
750
751 /* Create upcall ring */
752 mtx_init(&mfct->bw_upcalls_mtx, "mroute upcall buf_ring mtx", NULL,
753 MTX_DEF);
754 mfct->bw_upcalls = buf_ring_alloc(BW_UPCALLS_MAX, M_MRTABLE, M_NOWAIT,
755 &mfct->bw_upcalls_mtx);
756 if (mfct->bw_upcalls == NULL) {
757 MRW_WUNLOCK();
758 MRW_TEARDOWN_WUNLOCK();
759 return (ENOMEM);
760 }
761
762 TASK_INIT(&V_task, 0, ip_mrouter_upcall_thread, curvnet);
763 taskqueue_cancel(V_task_queue, &V_task, NULL);
764 taskqueue_unblock(V_task_queue);
765
766 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls_all,
767 curvnet);
768 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
769 curvnet);
770
771 mfct->router = so;
772 mfct->pim_assert_interval.tv_sec = 3;
773 V_ip_mrouting_enabled = true;
774 atomic_add_int(&ip_mrouter_cnt, 1);
775
776 MRW_WUNLOCK();
777 MRW_TEARDOWN_WUNLOCK();
778
779 CTR1(KTR_IPMF, "%s: done", __func__);
780
781 return 0;
782 }
783
784 /*
785 * Disable multicast forwarding.
786 */
787 static void
X_ip_mrouter_done(struct socket * so)788 X_ip_mrouter_done(struct socket *so)
789 {
790 struct mfctable *mfct;
791 struct ifnet **ifps;
792 int nifp;
793 u_long i;
794 vifi_t vifi;
795 struct bw_upcall *bu;
796
797 mfct = somfctable(so);
798
799 MRW_TEARDOWN_WLOCK();
800 if (so != mfct->router) {
801 MRW_TEARDOWN_WUNLOCK();
802 return;
803 }
804
805 /*
806 * Detach/disable hooks to the reset of the system.
807 */
808 mfct->router = NULL;
809 V_ip_mrouting_enabled = false;
810 atomic_subtract_int(&ip_mrouter_cnt, 1);
811 mfct->api_config = 0;
812
813 /*
814 * Wait for all epoch sections to complete to ensure the new value of
815 * V_ip_mrouting_enabled is visible to others.
816 */
817 NET_EPOCH_WAIT();
818
819 /* Stop and drain task queue */
820 taskqueue_block(V_task_queue);
821 while (taskqueue_cancel(V_task_queue, &V_task, NULL)) {
822 taskqueue_drain(V_task_queue, &V_task);
823 }
824
825 ifps = malloc(MAXVIFS * sizeof(*ifps), M_TEMP, M_WAITOK);
826
827 MRW_WLOCK();
828 taskqueue_cancel(V_task_queue, &V_task, NULL);
829
830 /* Destroy upcall ring */
831 while ((bu = buf_ring_dequeue_mc(mfct->bw_upcalls)) != NULL) {
832 free(bu, M_MRTABLE);
833 }
834 buf_ring_free(mfct->bw_upcalls, M_MRTABLE);
835 mtx_destroy(&mfct->bw_upcalls_mtx);
836
837 /*
838 * For each phyint in use, prepare to disable promiscuous reception
839 * of all IP multicasts. Defer the actual call until the lock is released;
840 * just record the list of interfaces while locked. Some interfaces use
841 * sx locks in their ioctl routines, which is not allowed while holding
842 * a non-sleepable lock.
843 */
844 KASSERT(mfct->numvifs <= MAXVIFS, ("More vifs than possible"));
845 for (vifi = 0, nifp = 0; vifi < mfct->numvifs; vifi++) {
846 struct vif *vif;
847
848 vif = &mfct->viftable[vifi];
849 if (!in_nullhost(vif->v_lcl_addr) &&
850 (vif->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)) == 0) {
851 ifps[nifp++] = vif->v_ifp;
852 }
853 }
854 bzero(mfct->viftable, sizeof(*mfct->viftable) * MAXVIFS);
855 mfct->numvifs = 0;
856 mfct->pim_assert_enabled = 0;
857
858 /*
859 * Free all multicast forwarding cache entries.
860 * Do not use hashdestroy(), as we must perform other cleanup.
861 */
862 for (i = 0; i < mfchashsize; i++) {
863 struct mfc *rt, *nrt;
864
865 LIST_FOREACH_SAFE(rt, &mfct->mfchashtbl[i], mfc_hash, nrt) {
866 expire_mfc(rt);
867 }
868 }
869 free(mfct->mfchashtbl, M_MRTABLE);
870 mfct->mfchashtbl = NULL;
871
872 bzero(mfct->nexpire, sizeof(mfct->nexpire[0]) * mfchashsize);
873 mfct->register_vif = VIFI_INVALID;
874
875 MRW_WUNLOCK();
876 MRW_TEARDOWN_WUNLOCK();
877
878 /*
879 * Now drop our claim on promiscuous multicast on the interfaces recorded
880 * above. This is safe to do now because ALLMULTI is reference counted.
881 */
882 for (vifi = 0; vifi < nifp; vifi++)
883 if_allmulti(ifps[vifi], 0);
884 free(ifps, M_TEMP);
885
886 CTR1(KTR_IPMF, "%s: done", __func__);
887 }
888
889 /*
890 * Set PIM assert processing global
891 */
892 static int
set_assert(struct mfctable * mfct,int i)893 set_assert(struct mfctable *mfct, int i)
894 {
895 if ((i != 1) && (i != 0))
896 return EINVAL;
897
898 mfct->pim_assert_enabled = i;
899
900 return 0;
901 }
902
903 /*
904 * Configure API capabilities
905 */
906 int
set_api_config(struct mfctable * mfct,uint32_t * apival)907 set_api_config(struct mfctable *mfct, uint32_t *apival)
908 {
909 u_long i;
910
911 /*
912 * We can set the API capabilities only if it is the first operation
913 * after MRT_INIT. I.e.:
914 * - there are no vifs installed
915 * - pim_assert is not enabled
916 * - the MFC table is empty
917 */
918 if (mfct->numvifs > 0) {
919 *apival = 0;
920 return EPERM;
921 }
922 if (mfct->pim_assert_enabled) {
923 *apival = 0;
924 return EPERM;
925 }
926
927 MRW_RLOCK();
928
929 for (i = 0; i < mfchashsize; i++) {
930 if (LIST_FIRST(&mfct->mfchashtbl[i]) != NULL) {
931 MRW_RUNLOCK();
932 *apival = 0;
933 return EPERM;
934 }
935 }
936
937 MRW_RUNLOCK();
938
939 mfct->api_config = *apival & mrt_api_support;
940 *apival = mfct->api_config;
941
942 return 0;
943 }
944
945 /*
946 * Add a vif to the vif table
947 */
948 static int
add_vif(struct mfctable * mfct,int fibnum,struct vifctl * vifcp)949 add_vif(struct mfctable *mfct, int fibnum, struct vifctl *vifcp)
950 {
951 struct vif *vifp;
952 struct sockaddr_in sin = {sizeof sin, AF_INET};
953 struct ifaddr *ifa;
954 struct ifnet *ifp;
955 int error;
956
957 if (vifcp->vifc_vifi >= MAXVIFS)
958 return EINVAL;
959
960 vifp = &mfct->viftable[vifcp->vifc_vifi];
961
962 /* rate limiting is no longer supported by this code */
963 if (vifcp->vifc_rate_limit != 0) {
964 log(LOG_ERR, "rate limiting is no longer supported\n");
965 return EINVAL;
966 }
967
968 if (in_nullhost(vifcp->vifc_lcl_addr))
969 return EADDRNOTAVAIL;
970
971 /* Find the interface with an address in AF_INET family */
972 if (vifcp->vifc_flags & VIFF_REGISTER) {
973 /*
974 * XXX: Because VIFF_REGISTER does not really need a valid
975 * local interface (e.g. it could be 127.0.0.2), we don't
976 * check its address.
977 */
978 ifp = NULL;
979 } else {
980 struct epoch_tracker et;
981
982 sin.sin_addr = vifcp->vifc_lcl_addr;
983 NET_EPOCH_ENTER(et);
984 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
985 if (ifa == NULL) {
986 NET_EPOCH_EXIT(et);
987 return EADDRNOTAVAIL;
988 }
989 ifp = ifa->ifa_ifp;
990 if (ifp->if_fib != fibnum) {
991 NET_EPOCH_EXIT(et);
992 return EADDRNOTAVAIL;
993 }
994 /* XXX FIXME we need to take a ref on ifp and cleanup properly! */
995 NET_EPOCH_EXIT(et);
996 }
997
998 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
999 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
1000 return EOPNOTSUPP;
1001 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
1002 ifp = mfct->register_if = if_alloc(IFT_LOOP);
1003 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
1004 if (mfct->register_vif == VIFI_INVALID) {
1005 if_initname(mfct->register_if, "register_vif", 0);
1006 mfct->register_vif = vifcp->vifc_vifi;
1007 }
1008 } else { /* Make sure the interface supports multicast */
1009 if ((ifp->if_flags & IFF_MULTICAST) == 0)
1010 return EOPNOTSUPP;
1011
1012 /* Enable promiscuous reception of all IP multicasts from the if */
1013 error = if_allmulti(ifp, 1);
1014 if (error)
1015 return error;
1016 }
1017
1018 MRW_WLOCK();
1019
1020 if (!in_nullhost(vifp->v_lcl_addr)) {
1021 if (ifp)
1022 mfct->register_if = NULL;
1023 MRW_WUNLOCK();
1024 if (ifp)
1025 if_free(ifp);
1026 return EADDRINUSE;
1027 }
1028
1029 vifp->v_flags = vifcp->vifc_flags;
1030 vifp->v_threshold = vifcp->vifc_threshold;
1031 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
1032 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
1033 vifp->v_ifp = ifp;
1034 /* initialize per vif pkt counters */
1035 vifp->v_pkt_in = 0;
1036 vifp->v_pkt_out = 0;
1037 vifp->v_bytes_in = 0;
1038 vifp->v_bytes_out = 0;
1039 sprintf(vifp->v_mtx_name, "BM[%d] mtx", vifcp->vifc_vifi);
1040 mtx_init(&vifp->v_mtx, vifp->v_mtx_name, NULL, MTX_DEF);
1041
1042 /* Adjust numvifs up if the vifi is higher than numvifs */
1043 if (mfct->numvifs <= vifcp->vifc_vifi)
1044 mfct->numvifs = vifcp->vifc_vifi + 1;
1045
1046 MRW_WUNLOCK();
1047
1048 CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
1049 (int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
1050 (int)vifcp->vifc_threshold);
1051
1052 return 0;
1053 }
1054
1055 /*
1056 * Delete a vif from the vif table
1057 */
1058 static int
del_vif_locked(struct mfctable * mfct,vifi_t vifi,struct ifnet ** ifp_multi_leave,struct ifnet ** ifp_free)1059 del_vif_locked(struct mfctable *mfct, vifi_t vifi,
1060 struct ifnet **ifp_multi_leave, struct ifnet **ifp_free)
1061 {
1062 struct vif *vifp;
1063
1064 *ifp_free = NULL;
1065 *ifp_multi_leave = NULL;
1066
1067 MRW_WLOCK_ASSERT();
1068
1069 if (vifi >= mfct->numvifs) {
1070 return EINVAL;
1071 }
1072 vifp = &mfct->viftable[vifi];
1073 if (in_nullhost(vifp->v_lcl_addr)) {
1074 return EADDRNOTAVAIL;
1075 }
1076
1077 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1078 *ifp_multi_leave = vifp->v_ifp;
1079
1080 if (vifp->v_flags & VIFF_REGISTER) {
1081 mfct->register_vif = VIFI_INVALID;
1082 if (vifp->v_ifp) {
1083 if (vifp->v_ifp == mfct->register_if)
1084 mfct->register_if = NULL;
1085 *ifp_free = vifp->v_ifp;
1086 }
1087 }
1088
1089 mtx_destroy(&vifp->v_mtx);
1090
1091 bzero((caddr_t)vifp, sizeof (*vifp));
1092
1093 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
1094
1095 /* Adjust numvifs down */
1096 for (vifi = mfct->numvifs; vifi > 0; vifi--)
1097 if (!in_nullhost(mfct->viftable[vifi - 1].v_lcl_addr))
1098 break;
1099 mfct->numvifs = vifi;
1100
1101 return 0;
1102 }
1103
1104 static int
del_vif(struct mfctable * mfct,vifi_t vifi)1105 del_vif(struct mfctable *mfct, vifi_t vifi)
1106 {
1107 int cc;
1108 struct ifnet *free_ptr, *multi_leave;
1109
1110 MRW_WLOCK();
1111 cc = del_vif_locked(mfct, vifi, &multi_leave, &free_ptr);
1112 MRW_WUNLOCK();
1113
1114 if (multi_leave)
1115 if_allmulti(multi_leave, 0);
1116 if (free_ptr) {
1117 if_free(free_ptr);
1118 }
1119
1120 return cc;
1121 }
1122
1123 /*
1124 * update an mfc entry without resetting counters and S,G addresses.
1125 */
1126 static void
update_mfc_params(struct mfctable * mfct,struct mfc * rt,struct mfcctl2 * mfccp)1127 update_mfc_params(struct mfctable *mfct, struct mfc *rt, struct mfcctl2 *mfccp)
1128 {
1129 int i;
1130
1131 rt->mfc_parent = mfccp->mfcc_parent;
1132 for (i = 0; i < mfct->numvifs; i++) {
1133 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1134 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mfct->api_config &
1135 MRT_MFC_FLAGS_ALL;
1136 }
1137 /* set the RP address */
1138 if (mfct->api_config & MRT_MFC_RP)
1139 rt->mfc_rp = mfccp->mfcc_rp;
1140 else
1141 rt->mfc_rp.s_addr = INADDR_ANY;
1142 }
1143
1144 /*
1145 * fully initialize an mfc entry from the parameter.
1146 */
1147 static void
init_mfc_params(struct mfctable * mfct,struct mfc * rt,struct mfcctl2 * mfccp)1148 init_mfc_params(struct mfctable *mfct, struct mfc *rt, struct mfcctl2 *mfccp)
1149 {
1150 rt->mfc_origin = mfccp->mfcc_origin;
1151 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1152
1153 update_mfc_params(mfct, rt, mfccp);
1154
1155 /* initialize pkt counters per src-grp */
1156 rt->mfc_pkt_cnt = 0;
1157 rt->mfc_byte_cnt = 0;
1158 rt->mfc_wrong_if = 0;
1159 timevalclear(&rt->mfc_last_assert);
1160 }
1161
1162 static void
expire_mfc(struct mfc * rt)1163 expire_mfc(struct mfc *rt)
1164 {
1165 struct rtdetq *rte;
1166
1167 MRW_WLOCK_ASSERT();
1168
1169 free_bw_list(rt->mfc_bw_meter_leq);
1170 free_bw_list(rt->mfc_bw_meter_geq);
1171
1172 while (!buf_ring_empty(rt->mfc_stall_ring)) {
1173 rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1174 if (rte) {
1175 m_freem(rte->m);
1176 free(rte, M_MRTABLE);
1177 }
1178 }
1179 buf_ring_free(rt->mfc_stall_ring, M_MRTABLE);
1180
1181 LIST_REMOVE(rt, mfc_hash);
1182 free(rt, M_MRTABLE);
1183 }
1184
1185 /*
1186 * Add an mfc entry
1187 */
1188 static int
add_mfc(struct mfctable * mfct,struct mfcctl2 * mfccp)1189 add_mfc(struct mfctable *mfct, struct mfcctl2 *mfccp)
1190 {
1191 struct mfc *rt;
1192 struct rtdetq *rte;
1193 u_long hash = 0;
1194 u_short nstl;
1195 struct epoch_tracker et;
1196
1197 MRW_WLOCK();
1198 rt = mfc_find(mfct, &mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1199
1200 /* If an entry already exists, just update the fields */
1201 if (rt) {
1202 CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1203 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1204 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1205 mfccp->mfcc_parent);
1206 update_mfc_params(mfct, rt, mfccp);
1207 MRW_WUNLOCK();
1208 return (0);
1209 }
1210
1211 /*
1212 * Find the entry for which the upcall was made and update
1213 */
1214 nstl = 0;
1215 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1216 NET_EPOCH_ENTER(et);
1217 LIST_FOREACH(rt, &mfct->mfchashtbl[hash], mfc_hash) {
1218 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1219 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1220 !buf_ring_empty(rt->mfc_stall_ring)) {
1221 CTR5(KTR_IPMF,
1222 "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1223 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1224 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1225 mfccp->mfcc_parent,
1226 rt->mfc_stall_ring);
1227 if (nstl++)
1228 CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1229
1230 init_mfc_params(mfct, rt, mfccp);
1231 rt->mfc_expire = 0; /* Don't clean this guy up */
1232 mfct->nexpire[hash]--;
1233
1234 /* Free queued packets, but attempt to forward them first. */
1235 while (!buf_ring_empty(rt->mfc_stall_ring)) {
1236 rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1237 if (rte->ifp != NULL)
1238 ip_mdq(mfct, rte->m, rte->ifp, rt, -1);
1239 m_freem(rte->m);
1240 free(rte, M_MRTABLE);
1241 }
1242 }
1243 }
1244 NET_EPOCH_EXIT(et);
1245
1246 /*
1247 * It is possible that an entry is being inserted without an upcall
1248 */
1249 if (nstl == 0) {
1250 CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1251 LIST_FOREACH(rt, &mfct->mfchashtbl[hash], mfc_hash) {
1252 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1253 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1254 init_mfc_params(mfct, rt, mfccp);
1255 if (rt->mfc_expire)
1256 mfct->nexpire[hash]--;
1257 rt->mfc_expire = 0;
1258 break; /* XXX */
1259 }
1260 }
1261
1262 if (rt == NULL) { /* no upcall, so make a new entry */
1263 rt = mfc_alloc();
1264 if (rt == NULL) {
1265 MRW_WUNLOCK();
1266 return (ENOBUFS);
1267 }
1268
1269 init_mfc_params(mfct, rt, mfccp);
1270
1271 rt->mfc_expire = 0;
1272 rt->mfc_bw_meter_leq = NULL;
1273 rt->mfc_bw_meter_geq = NULL;
1274
1275 /* insert new entry at head of hash chain */
1276 LIST_INSERT_HEAD(&mfct->mfchashtbl[hash], rt, mfc_hash);
1277 }
1278 }
1279
1280 MRW_WUNLOCK();
1281
1282 return (0);
1283 }
1284
1285 /*
1286 * Delete an mfc entry
1287 */
1288 static int
del_mfc(struct mfctable * mfct,struct mfcctl2 * mfccp)1289 del_mfc(struct mfctable *mfct, struct mfcctl2 *mfccp)
1290 {
1291 struct in_addr origin;
1292 struct in_addr mcastgrp;
1293 struct mfc *rt;
1294
1295 origin = mfccp->mfcc_origin;
1296 mcastgrp = mfccp->mfcc_mcastgrp;
1297
1298 CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1299 ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1300
1301 MRW_WLOCK();
1302
1303 LIST_FOREACH(rt, &mfct->mfchashtbl[MFCHASH(origin, mcastgrp)],
1304 mfc_hash) {
1305 if (in_hosteq(rt->mfc_origin, origin) &&
1306 in_hosteq(rt->mfc_mcastgrp, mcastgrp))
1307 break;
1308 }
1309 if (rt == NULL) {
1310 MRW_WUNLOCK();
1311 return EADDRNOTAVAIL;
1312 }
1313
1314 expire_mfc(rt);
1315
1316 MRW_WUNLOCK();
1317
1318 return (0);
1319 }
1320
1321 /*
1322 * Send a message to the routing daemon on the multicast routing socket.
1323 */
1324 static int
socket_send(struct socket * s,struct mbuf * mm,struct sockaddr_in * src)1325 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1326 {
1327 if (s) {
1328 SOCKBUF_LOCK(&s->so_rcv);
1329 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1330 NULL) != 0) {
1331 sorwakeup_locked(s);
1332 return 0;
1333 }
1334 soroverflow_locked(s);
1335 }
1336 m_freem(mm);
1337 return -1;
1338 }
1339
1340 /*
1341 * IP multicast forwarding function. This function assumes that the packet
1342 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1343 * pointed to by "ifp", and the packet is to be relayed to other networks
1344 * that have members of the packet's destination IP multicast group.
1345 *
1346 * The packet is returned unscathed to the caller, unless it is
1347 * erroneous, in which case a non-zero return value tells the caller to
1348 * discard it.
1349 */
1350
1351 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1352
1353 static int
X_ip_mforward(struct ip * ip,struct ifnet * ifp,struct mbuf * m,struct ip_moptions * imo)1354 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1355 struct ip_moptions *imo)
1356 {
1357 struct mfc *rt;
1358 struct mfctable *mfct;
1359 int error;
1360 vifi_t vifi;
1361 struct mbuf *mb0;
1362 struct rtdetq *rte;
1363 u_long hash;
1364 int hlen;
1365
1366 M_ASSERTMAPPED(m);
1367
1368 CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1369 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1370
1371 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1372 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1373 /*
1374 * Packet arrived via a physical interface or
1375 * an encapsulated tunnel or a register_vif.
1376 */
1377 } else {
1378 /*
1379 * Packet arrived through a source-route tunnel.
1380 * Source-route tunnels are no longer supported.
1381 */
1382 return (1);
1383 }
1384
1385 mfct = &V_mfctables[M_GETFIB(m)];
1386
1387 /*
1388 * BEGIN: MCAST ROUTING HOT PATH
1389 */
1390 MRW_RLOCK();
1391 if (__predict_false(mfct->router == NULL)) {
1392 MRW_RUNLOCK();
1393 return (EADDRNOTAVAIL);
1394 }
1395
1396 if (imo && ((vifi = imo->imo_multicast_vif) < mfct->numvifs)) {
1397 if (ip->ip_ttl < MAXTTL)
1398 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1399 error = ip_mdq(mfct, m, ifp, NULL, vifi);
1400 MRW_RUNLOCK();
1401 return error;
1402 }
1403
1404 /*
1405 * Don't forward a packet with time-to-live of zero or one,
1406 * or a packet destined to a local-only group.
1407 */
1408 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1409 MRW_RUNLOCK();
1410 return 0;
1411 }
1412
1413 mfc_find_retry:
1414 /*
1415 * Determine forwarding vifs from the forwarding cache table
1416 */
1417 MRTSTAT_INC(mrts_mfc_lookups);
1418 rt = mfc_find(mfct, &ip->ip_src, &ip->ip_dst);
1419
1420 /* Entry exists, so forward if necessary */
1421 if (rt != NULL) {
1422 error = ip_mdq(mfct, m, ifp, rt, -1);
1423 /* Generic unlock here as we might release R or W lock */
1424 MRW_UNLOCK();
1425 return error;
1426 }
1427
1428 /*
1429 * END: MCAST ROUTING HOT PATH
1430 */
1431
1432 /* Further processing must be done with WLOCK taken */
1433 if ((MRW_WOWNED() == 0) && (MRW_LOCK_TRY_UPGRADE() == 0)) {
1434 MRW_RUNLOCK();
1435 MRW_WLOCK();
1436 goto mfc_find_retry;
1437 }
1438
1439 /*
1440 * If we don't have a route for packet's origin,
1441 * Make a copy of the packet & send message to routing daemon
1442 */
1443 hlen = ip->ip_hl << 2;
1444
1445 MRTSTAT_INC(mrts_mfc_misses);
1446 MRTSTAT_INC(mrts_no_route);
1447 CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1448 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1449
1450 /*
1451 * Allocate mbufs early so that we don't do extra work if we are
1452 * just going to fail anyway. Make sure to pullup the header so
1453 * that other people can't step on it.
1454 */
1455 rte = malloc((sizeof *rte), M_MRTABLE, M_NOWAIT|M_ZERO);
1456 if (rte == NULL) {
1457 MRW_WUNLOCK();
1458 return ENOBUFS;
1459 }
1460
1461 mb0 = m_copypacket(m, M_NOWAIT);
1462 if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1463 mb0 = m_pullup(mb0, hlen);
1464 if (mb0 == NULL) {
1465 free(rte, M_MRTABLE);
1466 MRW_WUNLOCK();
1467 return ENOBUFS;
1468 }
1469
1470 /* is there an upcall waiting for this flow ? */
1471 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1472 LIST_FOREACH(rt, &mfct->mfchashtbl[hash], mfc_hash) {
1473 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1474 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1475 !buf_ring_empty(rt->mfc_stall_ring))
1476 break;
1477 }
1478
1479 if (rt == NULL) {
1480 int i;
1481 struct igmpmsg *im;
1482 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1483 struct mbuf *mm;
1484
1485 /*
1486 * Locate the vifi for the incoming interface for this packet.
1487 * If none found, drop packet.
1488 */
1489 for (vifi = 0; vifi < mfct->numvifs &&
1490 mfct->viftable[vifi].v_ifp != ifp; vifi++)
1491 ;
1492 if (vifi >= mfct->numvifs) /* vif not found, drop packet */
1493 goto non_fatal;
1494
1495 /* no upcall, so make a new entry */
1496 rt = mfc_alloc();
1497 if (rt == NULL)
1498 goto fail;
1499
1500 /* Make a copy of the header to send to the user level process */
1501 mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1502 if (mm == NULL)
1503 goto fail1;
1504
1505 /*
1506 * Send message to routing daemon to install
1507 * a route into the kernel table
1508 */
1509
1510 im = mtod(mm, struct igmpmsg*);
1511 im->im_msgtype = IGMPMSG_NOCACHE;
1512 im->im_mbz = 0;
1513 im->im_vif = vifi;
1514
1515 MRTSTAT_INC(mrts_upcalls);
1516
1517 k_igmpsrc.sin_addr = ip->ip_src;
1518 if (socket_send(mfct->router, mm, &k_igmpsrc) < 0) {
1519 CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1520 MRTSTAT_INC(mrts_upq_sockfull);
1521 fail1: free(rt, M_MRTABLE);
1522 fail: free(rte, M_MRTABLE);
1523 m_freem(mb0);
1524 MRW_WUNLOCK();
1525 return ENOBUFS;
1526 }
1527
1528 /* insert new entry at head of hash chain */
1529 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1530 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1531 rt->mfc_expire = UPCALL_EXPIRE;
1532 mfct->nexpire[hash]++;
1533 for (i = 0; i < mfct->numvifs; i++) {
1534 rt->mfc_ttls[i] = 0;
1535 rt->mfc_flags[i] = 0;
1536 }
1537 rt->mfc_parent = -1;
1538
1539 /* clear the RP address */
1540 rt->mfc_rp.s_addr = INADDR_ANY;
1541 rt->mfc_bw_meter_leq = NULL;
1542 rt->mfc_bw_meter_geq = NULL;
1543
1544 /* initialize pkt counters per src-grp */
1545 rt->mfc_pkt_cnt = 0;
1546 rt->mfc_byte_cnt = 0;
1547 rt->mfc_wrong_if = 0;
1548 timevalclear(&rt->mfc_last_assert);
1549
1550 buf_ring_enqueue(rt->mfc_stall_ring, rte);
1551
1552 /* Add RT to hashtable as it didn't exist before */
1553 LIST_INSERT_HEAD(&mfct->mfchashtbl[hash], rt, mfc_hash);
1554 } else {
1555 /* determine if queue has overflowed */
1556 if (buf_ring_full(rt->mfc_stall_ring)) {
1557 MRTSTAT_INC(mrts_upq_ovflw);
1558 non_fatal: free(rte, M_MRTABLE);
1559 m_freem(mb0);
1560 MRW_WUNLOCK();
1561 return (0);
1562 }
1563
1564 buf_ring_enqueue(rt->mfc_stall_ring, rte);
1565 }
1566
1567 rte->m = mb0;
1568 rte->ifp = ifp;
1569
1570 MRW_WUNLOCK();
1571
1572 return 0;
1573 }
1574
1575 static void
expire_upcalls(struct mfctable * mfct)1576 expire_upcalls(struct mfctable *mfct)
1577 {
1578 for (u_long i = 0; i < mfchashsize; i++) {
1579 struct mfc *rt, *nrt;
1580
1581 if (mfct->nexpire[i] == 0)
1582 continue;
1583
1584 LIST_FOREACH_SAFE(rt, &mfct->mfchashtbl[i], mfc_hash, nrt) {
1585 if (buf_ring_empty(rt->mfc_stall_ring))
1586 continue;
1587
1588 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1589 continue;
1590
1591 MRTSTAT_INC(mrts_cache_cleanups);
1592 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1593 (u_long)ntohl(rt->mfc_origin.s_addr),
1594 (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1595
1596 expire_mfc(rt);
1597 }
1598 }
1599 }
1600
1601 /*
1602 * Clean up the cache entry if upcall is not serviced
1603 */
1604 static void
expire_upcalls_all(void * arg)1605 expire_upcalls_all(void *arg)
1606 {
1607 CURVNET_SET((struct vnet *)arg);
1608
1609 MRW_LOCK_ASSERT();
1610
1611 for (int i = 0; i < V_nmfctables; i++)
1612 expire_upcalls(&V_mfctables[i]);
1613
1614 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls_all,
1615 curvnet);
1616
1617 CURVNET_RESTORE();
1618 }
1619
1620 /*
1621 * Packet forwarding routine once entry in the cache is made
1622 */
1623 static int
ip_mdq(struct mfctable * mfct,struct mbuf * m,struct ifnet * ifp,struct mfc * rt,vifi_t xmt_vif)1624 ip_mdq(struct mfctable *mfct, struct mbuf *m, struct ifnet *ifp, struct mfc *rt,
1625 vifi_t xmt_vif)
1626 {
1627 struct ip *ip = mtod(m, struct ip *);
1628 struct vif *vif;
1629 vifi_t vifi;
1630 int plen = ntohs(ip->ip_len);
1631
1632 M_ASSERTMAPPED(m);
1633 MRW_LOCK_ASSERT();
1634 NET_EPOCH_ASSERT();
1635
1636 /*
1637 * If xmt_vif is not -1, send on only the requested vif.
1638 *
1639 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1640 */
1641 if (xmt_vif < mfct->numvifs) {
1642 if (mfct->viftable[xmt_vif].v_flags & VIFF_REGISTER)
1643 pim_register_send(mfct, ip, &mfct->viftable[xmt_vif], m,
1644 rt);
1645 else
1646 phyint_send(ip, &mfct->viftable[xmt_vif], m);
1647 return 1;
1648 }
1649
1650 /*
1651 * Don't forward if it didn't arrive from the parent vif for its origin.
1652 */
1653 vifi = rt->mfc_parent;
1654 vif = &mfct->viftable[vifi];
1655 if (vifi >= mfct->numvifs || vif->v_ifp != ifp) {
1656 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1657 __func__, ifp, (int)vifi, vif->v_ifp);
1658 MRTSTAT_INC(mrts_wrong_if);
1659 ++rt->mfc_wrong_if;
1660 /*
1661 * If we are doing PIM assert processing, send a message
1662 * to the routing daemon.
1663 *
1664 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1665 * can complete the SPT switch, regardless of the type
1666 * of the iif (broadcast media, GRE tunnel, etc).
1667 */
1668 if (mfct->pim_assert_enabled && (vifi < mfct->numvifs) &&
1669 vif->v_ifp != NULL) {
1670 if (ifp == mfct->register_if)
1671 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1672
1673 /* Get vifi for the incoming packet */
1674 for (vifi = 0; vifi < mfct->numvifs &&
1675 mfct->viftable[vifi].v_ifp != ifp; vifi++)
1676 ;
1677 if (vifi >= mfct->numvifs)
1678 return 0; /* The iif is not found: ignore the packet. */
1679
1680 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1681 return 0; /* WRONGVIF disabled: ignore the packet */
1682
1683 if (ratecheck(&rt->mfc_last_assert,
1684 &mfct->pim_assert_interval)) {
1685 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1686 struct igmpmsg *im;
1687 int hlen = ip->ip_hl << 2;
1688 struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1689
1690 if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1691 mm = m_pullup(mm, hlen);
1692 if (mm == NULL)
1693 return ENOBUFS;
1694
1695 im = mtod(mm, struct igmpmsg *);
1696 im->im_msgtype = IGMPMSG_WRONGVIF;
1697 im->im_mbz = 0;
1698 im->im_vif = vifi;
1699
1700 MRTSTAT_INC(mrts_upcalls);
1701
1702 k_igmpsrc.sin_addr = im->im_src;
1703 if (socket_send(mfct->router, mm,
1704 &k_igmpsrc) < 0) {
1705 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1706 MRTSTAT_INC(mrts_upq_sockfull);
1707 return ENOBUFS;
1708 }
1709 }
1710 }
1711 return 0;
1712 }
1713
1714 /* If I sourced this packet, it counts as output, else it was input. */
1715 mtx_lock(&vif->v_mtx);
1716 if (in_hosteq(ip->ip_src, vif->v_lcl_addr)) {
1717 vif->v_pkt_out++;
1718 vif->v_bytes_out += plen;
1719 } else {
1720 vif->v_pkt_in++;
1721 vif->v_bytes_in += plen;
1722 }
1723 mtx_unlock(&vif->v_mtx);
1724
1725 rt->mfc_pkt_cnt++;
1726 rt->mfc_byte_cnt += plen;
1727
1728 /*
1729 * For each vif, decide if a copy of the packet should be forwarded.
1730 * Forward if:
1731 * - the ttl exceeds the vif's threshold
1732 * - there are group members downstream on interface
1733 */
1734 for (vifi = 0; vifi < mfct->numvifs; vifi++)
1735 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1736 vif = &mfct->viftable[vifi];
1737 vif->v_pkt_out++;
1738 vif->v_bytes_out += plen;
1739 if (vif->v_flags & VIFF_REGISTER)
1740 pim_register_send(mfct, ip, vif, m, rt);
1741 else
1742 phyint_send(ip, vif, m);
1743 }
1744
1745 /*
1746 * Perform upcall-related bw measuring.
1747 */
1748 if ((rt->mfc_bw_meter_geq != NULL) || (rt->mfc_bw_meter_leq != NULL)) {
1749 struct bw_meter *x;
1750 struct timeval now;
1751
1752 microtime(&now);
1753 /* Process meters for Greater-or-EQual case */
1754 for (x = rt->mfc_bw_meter_geq; x != NULL; x = x->bm_mfc_next)
1755 bw_meter_geq_receive_packet(x, plen, &now);
1756
1757 /* Process meters for Lower-or-EQual case */
1758 for (x = rt->mfc_bw_meter_leq; x != NULL; x = x->bm_mfc_next) {
1759 /*
1760 * Record that a packet is received.
1761 * A lock has to be taken as callout context
1762 * (expire_bw_meter_leq) might modify these fields
1763 * as well
1764 */
1765 mtx_lock(&x->bm_mtx);
1766 x->bm_measured.b_packets++;
1767 x->bm_measured.b_bytes += plen;
1768 mtx_unlock(&x->bm_mtx);
1769 }
1770 }
1771
1772 return 0;
1773 }
1774
1775 /*
1776 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1777 */
1778 static int
X_legal_vif_num(int fibnum,int vif)1779 X_legal_vif_num(int fibnum, int vif)
1780 {
1781 struct mfctable *mfct;
1782 int ret;
1783
1784 ret = 0;
1785 if (vif < 0)
1786 return (ret);
1787
1788 mfct = &V_mfctables[fibnum];
1789 MRW_RLOCK();
1790 if (vif < mfct->numvifs)
1791 ret = 1;
1792 MRW_RUNLOCK();
1793
1794 return (ret);
1795 }
1796
1797 /*
1798 * Return the local address used by this vif
1799 */
1800 static u_long
X_ip_mcast_src(int fibnum,int vifi)1801 X_ip_mcast_src(int fibnum, int vifi)
1802 {
1803 struct mfctable *mfct;
1804 in_addr_t addr;
1805
1806 addr = INADDR_ANY;
1807 if (vifi < 0)
1808 return (addr);
1809
1810 mfct = &V_mfctables[fibnum];
1811 MRW_RLOCK();
1812 if (vifi < mfct->numvifs)
1813 addr = mfct->viftable[vifi].v_lcl_addr.s_addr;
1814 MRW_RUNLOCK();
1815
1816 return (addr);
1817 }
1818
1819 static void
phyint_send(struct ip * ip,struct vif * vifp,struct mbuf * m)1820 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1821 {
1822 struct mbuf *mb_copy;
1823 int hlen = ip->ip_hl << 2;
1824
1825 MRW_LOCK_ASSERT();
1826 M_ASSERTMAPPED(m);
1827
1828 /*
1829 * Make a new reference to the packet; make sure that
1830 * the IP header is actually copied, not just referenced,
1831 * so that ip_output() only scribbles on the copy.
1832 */
1833 mb_copy = m_copypacket(m, M_NOWAIT);
1834 if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1835 mb_copy = m_pullup(mb_copy, hlen);
1836 if (mb_copy == NULL)
1837 return;
1838
1839 send_packet(vifp, mb_copy);
1840 }
1841
1842 static void
send_packet(struct vif * vifp,struct mbuf * m)1843 send_packet(struct vif *vifp, struct mbuf *m)
1844 {
1845 struct ip_moptions imo;
1846 int error __unused;
1847
1848 MRW_LOCK_ASSERT();
1849 NET_EPOCH_ASSERT();
1850
1851 imo.imo_multicast_ifp = vifp->v_ifp;
1852 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1853 imo.imo_multicast_loop = !!in_mcast_loop;
1854 imo.imo_multicast_vif = -1;
1855 STAILQ_INIT(&imo.imo_head);
1856
1857 /*
1858 * Re-entrancy should not be a problem here, because
1859 * the packets that we send out and are looped back at us
1860 * should get rejected because they appear to come from
1861 * the loopback interface, thus preventing looping.
1862 */
1863 error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1864 }
1865
1866 /*
1867 * Stubs for old RSVP socket shim implementation.
1868 */
1869
1870 static int
X_ip_rsvp_vif(struct socket * so __unused,struct sockopt * sopt __unused)1871 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1872 {
1873
1874 return (EOPNOTSUPP);
1875 }
1876
1877 static void
X_ip_rsvp_force_done(struct socket * so __unused)1878 X_ip_rsvp_force_done(struct socket *so __unused)
1879 {
1880
1881 }
1882
1883 static int
X_rsvp_input(struct mbuf ** mp,int * offp,int proto)1884 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1885 {
1886 struct mbuf *m;
1887
1888 m = *mp;
1889 *mp = NULL;
1890 if (!V_rsvp_on)
1891 m_freem(m);
1892 return (IPPROTO_DONE);
1893 }
1894
1895 /*
1896 * Code for bandwidth monitors
1897 */
1898
1899 /*
1900 * Define common interface for timeval-related methods
1901 */
1902 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1903 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1904 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1905
1906 static uint32_t
compute_bw_meter_flags(struct bw_upcall * req)1907 compute_bw_meter_flags(struct bw_upcall *req)
1908 {
1909 uint32_t flags = 0;
1910
1911 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1912 flags |= BW_METER_UNIT_PACKETS;
1913 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1914 flags |= BW_METER_UNIT_BYTES;
1915 if (req->bu_flags & BW_UPCALL_GEQ)
1916 flags |= BW_METER_GEQ;
1917 if (req->bu_flags & BW_UPCALL_LEQ)
1918 flags |= BW_METER_LEQ;
1919
1920 return flags;
1921 }
1922
1923 static void
expire_bw_meter_leq(void * arg)1924 expire_bw_meter_leq(void *arg)
1925 {
1926 struct bw_meter *x = arg;
1927 struct timeval now;
1928
1929 CURVNET_SET((struct vnet *)x->arg);
1930
1931 MRW_LOCK_ASSERT();
1932
1933 microtime(&now);
1934
1935 /*
1936 * Test if we should deliver an upcall
1937 */
1938 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1939 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1940 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1941 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1942 /* Prepare an upcall for delivery */
1943 bw_meter_prepare_upcall(x, &now);
1944 }
1945
1946 /* Send all upcalls that are pending delivery */
1947 taskqueue_enqueue(V_task_queue, &V_task);
1948
1949 /* Reset counters */
1950 x->bm_start_time = now;
1951 /*
1952 * The lock has to be taken as ip_forward context
1953 * might modify these fields as well
1954 */
1955 mtx_lock(&x->bm_mtx);
1956 x->bm_measured.b_bytes = 0;
1957 x->bm_measured.b_packets = 0;
1958 mtx_unlock(&x->bm_mtx);
1959
1960 callout_schedule(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time));
1961
1962 CURVNET_RESTORE();
1963 }
1964
1965 /*
1966 * Add a bw_meter entry
1967 */
1968 static int
add_bw_upcall(struct mfctable * mfct,struct bw_upcall * req)1969 add_bw_upcall(struct mfctable *mfct, struct bw_upcall *req)
1970 {
1971 struct mfc *mfc;
1972 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1973 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1974 struct timeval now;
1975 struct bw_meter *x, **bwm_ptr;
1976 uint32_t flags;
1977
1978 if (!(mfct->api_config & MRT_MFC_BW_UPCALL))
1979 return EOPNOTSUPP;
1980
1981 /* Test if the flags are valid */
1982 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1983 return EINVAL;
1984 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1985 return EINVAL;
1986 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1987 return EINVAL;
1988
1989 /* Test if the threshold time interval is valid */
1990 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1991 return EINVAL;
1992
1993 flags = compute_bw_meter_flags(req);
1994
1995 /*
1996 * Find if we have already same bw_meter entry
1997 */
1998 MRW_WLOCK();
1999 mfc = mfc_find(mfct, &req->bu_src, &req->bu_dst);
2000 if (mfc == NULL) {
2001 MRW_WUNLOCK();
2002 return EADDRNOTAVAIL;
2003 }
2004
2005 /* Choose an appropriate bw_meter list */
2006 if (req->bu_flags & BW_UPCALL_GEQ)
2007 bwm_ptr = &mfc->mfc_bw_meter_geq;
2008 else
2009 bwm_ptr = &mfc->mfc_bw_meter_leq;
2010
2011 for (x = *bwm_ptr; x != NULL; x = x->bm_mfc_next) {
2012 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2013 &req->bu_threshold.b_time, ==))
2014 && (x->bm_threshold.b_packets
2015 == req->bu_threshold.b_packets)
2016 && (x->bm_threshold.b_bytes
2017 == req->bu_threshold.b_bytes)
2018 && (x->bm_flags & BW_METER_USER_FLAGS)
2019 == flags) {
2020 MRW_WUNLOCK();
2021 return 0; /* XXX Already installed */
2022 }
2023 }
2024
2025 /* Allocate the new bw_meter entry */
2026 x = malloc(sizeof(*x), M_BWMETER, M_ZERO | M_NOWAIT);
2027 if (x == NULL) {
2028 MRW_WUNLOCK();
2029 return ENOBUFS;
2030 }
2031
2032 /* Set the new bw_meter entry */
2033 x->bm_threshold.b_time = req->bu_threshold.b_time;
2034 microtime(&now);
2035 x->bm_start_time = now;
2036 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2037 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2038 x->bm_measured.b_packets = 0;
2039 x->bm_measured.b_bytes = 0;
2040 x->bm_flags = flags;
2041 x->bm_time_next = NULL;
2042 x->bm_mfc = mfc;
2043 x->bm_mfctable = mfct;
2044 x->arg = curvnet;
2045 sprintf(x->bm_mtx_name, "BM mtx %p", x);
2046 mtx_init(&x->bm_mtx, x->bm_mtx_name, NULL, MTX_DEF);
2047
2048 /* For LEQ case create periodic callout */
2049 if (req->bu_flags & BW_UPCALL_LEQ) {
2050 callout_init_rw(&x->bm_meter_callout, &mrouter_lock, CALLOUT_SHAREDLOCK);
2051 callout_reset(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time),
2052 expire_bw_meter_leq, x);
2053 }
2054
2055 /* Add the new bw_meter entry to the front of entries for this MFC */
2056 x->bm_mfc_next = *bwm_ptr;
2057 *bwm_ptr = x;
2058
2059 MRW_WUNLOCK();
2060
2061 return 0;
2062 }
2063
2064 static void
free_bw_list(struct bw_meter * list)2065 free_bw_list(struct bw_meter *list)
2066 {
2067 while (list != NULL) {
2068 struct bw_meter *x = list;
2069
2070 /* MRW_WLOCK must be held here */
2071 if (x->bm_flags & BW_METER_LEQ) {
2072 callout_drain(&x->bm_meter_callout);
2073 mtx_destroy(&x->bm_mtx);
2074 }
2075
2076 list = list->bm_mfc_next;
2077 free(x, M_BWMETER);
2078 }
2079 }
2080
2081 /*
2082 * Delete one or multiple bw_meter entries
2083 */
2084 static int
del_bw_upcall(struct mfctable * mfct,struct bw_upcall * req)2085 del_bw_upcall(struct mfctable *mfct, struct bw_upcall *req)
2086 {
2087 struct mfc *mfc;
2088 struct bw_meter *x, **bwm_ptr;
2089
2090 if (!(mfct->api_config & MRT_MFC_BW_UPCALL))
2091 return EOPNOTSUPP;
2092
2093 MRW_WLOCK();
2094
2095 /* Find the corresponding MFC entry */
2096 mfc = mfc_find(mfct, &req->bu_src, &req->bu_dst);
2097 if (mfc == NULL) {
2098 MRW_WUNLOCK();
2099 return EADDRNOTAVAIL;
2100 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2101 /*
2102 * Delete all bw_meter entries for this mfc
2103 */
2104 struct bw_meter *list;
2105
2106 /* Free LEQ list */
2107 list = mfc->mfc_bw_meter_leq;
2108 mfc->mfc_bw_meter_leq = NULL;
2109 free_bw_list(list);
2110
2111 /* Free GEQ list */
2112 list = mfc->mfc_bw_meter_geq;
2113 mfc->mfc_bw_meter_geq = NULL;
2114 free_bw_list(list);
2115 MRW_WUNLOCK();
2116 return 0;
2117 } else { /* Delete a single bw_meter entry */
2118 struct bw_meter *prev;
2119 uint32_t flags = 0;
2120
2121 flags = compute_bw_meter_flags(req);
2122
2123 /* Choose an appropriate bw_meter list */
2124 if (req->bu_flags & BW_UPCALL_GEQ)
2125 bwm_ptr = &mfc->mfc_bw_meter_geq;
2126 else
2127 bwm_ptr = &mfc->mfc_bw_meter_leq;
2128
2129 /* Find the bw_meter entry to delete */
2130 for (prev = NULL, x = *bwm_ptr; x != NULL;
2131 prev = x, x = x->bm_mfc_next) {
2132 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, &req->bu_threshold.b_time, ==)) &&
2133 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2134 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2135 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2136 break;
2137 }
2138 if (x != NULL) { /* Delete entry from the list for this MFC */
2139 if (prev != NULL)
2140 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
2141 else
2142 *bwm_ptr = x->bm_mfc_next;/* new head of list */
2143
2144 if (req->bu_flags & BW_UPCALL_LEQ)
2145 callout_stop(&x->bm_meter_callout);
2146
2147 MRW_WUNLOCK();
2148 /* Free the bw_meter entry */
2149 free(x, M_BWMETER);
2150 return 0;
2151 } else {
2152 MRW_WUNLOCK();
2153 return EINVAL;
2154 }
2155 }
2156 __assert_unreachable();
2157 }
2158
2159 /*
2160 * Perform bandwidth measurement processing that may result in an upcall
2161 */
2162 static void
bw_meter_geq_receive_packet(struct bw_meter * x,int plen,struct timeval * nowp)2163 bw_meter_geq_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2164 {
2165 struct timeval delta;
2166
2167 MRW_LOCK_ASSERT();
2168
2169 delta = *nowp;
2170 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2171
2172 /*
2173 * Processing for ">=" type of bw_meter entry.
2174 * bm_mtx does not have to be hold here as in GEQ
2175 * case this is the only context accessing bm_measured.
2176 */
2177 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2178 /* Reset the bw_meter entry */
2179 x->bm_start_time = *nowp;
2180 x->bm_measured.b_packets = 0;
2181 x->bm_measured.b_bytes = 0;
2182 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2183 }
2184
2185 /* Record that a packet is received */
2186 x->bm_measured.b_packets++;
2187 x->bm_measured.b_bytes += plen;
2188
2189 /*
2190 * Test if we should deliver an upcall
2191 */
2192 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2193 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2194 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2195 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2196 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2197 /* Prepare an upcall for delivery */
2198 bw_meter_prepare_upcall(x, nowp);
2199 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2200 }
2201 }
2202 }
2203
2204 /*
2205 * Prepare a bandwidth-related upcall
2206 */
2207 static void
bw_meter_prepare_upcall(struct bw_meter * x,struct timeval * nowp)2208 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2209 {
2210 struct timeval delta;
2211 struct bw_upcall *u;
2212
2213 MRW_LOCK_ASSERT();
2214
2215 /*
2216 * Compute the measured time interval
2217 */
2218 delta = *nowp;
2219 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2220
2221 /*
2222 * Set the bw_upcall entry
2223 */
2224 u = malloc(sizeof(struct bw_upcall), M_MRTABLE, M_NOWAIT | M_ZERO);
2225 if (!u) {
2226 log(LOG_WARNING, "bw_meter_prepare_upcall: cannot allocate entry\n");
2227 return;
2228 }
2229 u->bu_src = x->bm_mfc->mfc_origin;
2230 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2231 u->bu_threshold.b_time = x->bm_threshold.b_time;
2232 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2233 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2234 u->bu_measured.b_time = delta;
2235 u->bu_measured.b_packets = x->bm_measured.b_packets;
2236 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2237 u->bu_flags = 0;
2238 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2239 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2240 if (x->bm_flags & BW_METER_UNIT_BYTES)
2241 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2242 if (x->bm_flags & BW_METER_GEQ)
2243 u->bu_flags |= BW_UPCALL_GEQ;
2244 if (x->bm_flags & BW_METER_LEQ)
2245 u->bu_flags |= BW_UPCALL_LEQ;
2246
2247 if (buf_ring_enqueue(x->bm_mfctable->bw_upcalls, u))
2248 log(LOG_WARNING, "bw_meter_prepare_upcall: cannot enqueue upcall\n");
2249 if (buf_ring_count(x->bm_mfctable->bw_upcalls) > (BW_UPCALLS_MAX / 2)) {
2250 taskqueue_enqueue(V_task_queue, &V_task);
2251 }
2252 }
2253 /*
2254 * Send the pending bandwidth-related upcalls
2255 */
2256 static void
bw_upcalls_send(struct mfctable * mfct)2257 bw_upcalls_send(struct mfctable *mfct)
2258 {
2259 struct mbuf *m;
2260 int len = 0;
2261 struct bw_upcall *bu;
2262 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2263 static struct igmpmsg igmpmsg = {
2264 0, /* unused1 */
2265 0, /* unused2 */
2266 IGMPMSG_BW_UPCALL,/* im_msgtype */
2267 0, /* im_mbz */
2268 0, /* im_vif */
2269 0, /* unused3 */
2270 { 0 }, /* im_src */
2271 { 0 } /* im_dst */
2272 };
2273
2274 MRW_LOCK_ASSERT();
2275
2276 if (buf_ring_empty(mfct->bw_upcalls))
2277 return;
2278
2279 /*
2280 * Allocate a new mbuf, initialize it with the header and
2281 * the payload for the pending calls.
2282 */
2283 m = m_gethdr(M_NOWAIT, MT_DATA);
2284 if (m == NULL) {
2285 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2286 return;
2287 }
2288
2289 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2290 len += sizeof(struct igmpmsg);
2291 while ((bu = buf_ring_dequeue_mc(mfct->bw_upcalls)) != NULL) {
2292 m_copyback(m, len, sizeof(struct bw_upcall), (caddr_t)bu);
2293 len += sizeof(struct bw_upcall);
2294 free(bu, M_MRTABLE);
2295 }
2296
2297 /*
2298 * Send the upcalls
2299 * XXX do we need to set the address in k_igmpsrc ?
2300 */
2301 MRTSTAT_INC(mrts_upcalls);
2302 if (socket_send(mfct->router, m, &k_igmpsrc) < 0) {
2303 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2304 MRTSTAT_INC(mrts_upq_sockfull);
2305 }
2306 }
2307
2308 static void
bw_upcalls_send_all(void)2309 bw_upcalls_send_all(void)
2310 {
2311 for (int i = 0; i < V_nmfctables; i++) {
2312 struct mfctable *mfct;
2313
2314 mfct = &V_mfctables[i];
2315 if (mfct->router != NULL)
2316 bw_upcalls_send(mfct);
2317 }
2318 }
2319
2320 /*
2321 * A periodic function for sending all upcalls that are pending delivery
2322 */
2323 static void
expire_bw_upcalls_send(void * arg)2324 expire_bw_upcalls_send(void *arg)
2325 {
2326 CURVNET_SET((struct vnet *) arg);
2327
2328 /* This callout is run with MRW_RLOCK taken */
2329
2330 bw_upcalls_send_all();
2331
2332 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2333 curvnet);
2334 CURVNET_RESTORE();
2335 }
2336
2337 /*
2338 * End of bandwidth monitoring code
2339 */
2340
2341 /*
2342 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2343 *
2344 */
2345 static int
pim_register_send(struct mfctable * mfct,struct ip * ip,struct vif * vifp,struct mbuf * m,struct mfc * rt)2346 pim_register_send(struct mfctable *mfct, struct ip *ip, struct vif *vifp,
2347 struct mbuf *m, struct mfc *rt)
2348 {
2349 struct mbuf *mb_copy, *mm;
2350
2351 /*
2352 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2353 * rendezvous point was unspecified, and we were told not to.
2354 */
2355 if (pim_squelch_wholepkt != 0 && (mfct->api_config & MRT_MFC_RP) &&
2356 in_nullhost(rt->mfc_rp))
2357 return 0;
2358
2359 mb_copy = pim_register_prepare(ip, m);
2360 if (mb_copy == NULL)
2361 return ENOBUFS;
2362
2363 /*
2364 * Send all the fragments. Note that the mbuf for each fragment
2365 * is freed by the sending machinery.
2366 */
2367 for (mm = mb_copy; mm; mm = mb_copy) {
2368 mb_copy = mm->m_nextpkt;
2369 mm->m_nextpkt = 0;
2370 mm = m_pullup(mm, sizeof(struct ip));
2371 if (mm != NULL) {
2372 ip = mtod(mm, struct ip *);
2373 if ((mfct->api_config & MRT_MFC_RP) &&
2374 !in_nullhost(rt->mfc_rp)) {
2375 pim_register_send_rp(mfct, ip, vifp, mm, rt);
2376 } else {
2377 pim_register_send_upcall(mfct, ip, vifp, mm,
2378 rt);
2379 }
2380 }
2381 }
2382
2383 return 0;
2384 }
2385
2386 /*
2387 * Return a copy of the data packet that is ready for PIM Register
2388 * encapsulation.
2389 * XXX: Note that in the returned copy the IP header is a valid one.
2390 */
2391 static struct mbuf *
pim_register_prepare(struct ip * ip,struct mbuf * m)2392 pim_register_prepare(struct ip *ip, struct mbuf *m)
2393 {
2394 struct mbuf *mb_copy = NULL;
2395 int mtu;
2396
2397 /* Take care of delayed checksums */
2398 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2399 in_delayed_cksum(m);
2400 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2401 }
2402
2403 /*
2404 * Copy the old packet & pullup its IP header into the
2405 * new mbuf so we can modify it.
2406 */
2407 mb_copy = m_copypacket(m, M_NOWAIT);
2408 if (mb_copy == NULL)
2409 return NULL;
2410 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2411 if (mb_copy == NULL)
2412 return NULL;
2413
2414 /* take care of the TTL */
2415 ip = mtod(mb_copy, struct ip *);
2416 --ip->ip_ttl;
2417
2418 /* Compute the MTU after the PIM Register encapsulation */
2419 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2420
2421 if (ntohs(ip->ip_len) <= mtu) {
2422 /* Turn the IP header into a valid one */
2423 ip->ip_sum = 0;
2424 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2425 } else {
2426 /* Fragment the packet */
2427 mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2428 if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2429 m_freem(mb_copy);
2430 return NULL;
2431 }
2432 }
2433 return mb_copy;
2434 }
2435
2436 /*
2437 * Send an upcall with the data packet to the user-level process.
2438 */
2439 static int
pim_register_send_upcall(struct mfctable * mfct,struct ip * ip,struct vif * vifp,struct mbuf * mb_copy,struct mfc * rt)2440 pim_register_send_upcall(struct mfctable *mfct, struct ip *ip, struct vif *vifp,
2441 struct mbuf *mb_copy, struct mfc *rt)
2442 {
2443 struct mbuf *mb_first;
2444 int len = ntohs(ip->ip_len);
2445 struct igmpmsg *im;
2446 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2447
2448 MRW_LOCK_ASSERT();
2449
2450 /*
2451 * Add a new mbuf with an upcall header
2452 */
2453 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2454 if (mb_first == NULL) {
2455 m_freem(mb_copy);
2456 return ENOBUFS;
2457 }
2458 mb_first->m_data += max_linkhdr;
2459 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2460 mb_first->m_len = sizeof(struct igmpmsg);
2461 mb_first->m_next = mb_copy;
2462
2463 /* Send message to routing daemon */
2464 im = mtod(mb_first, struct igmpmsg *);
2465 memset(im, 0, sizeof(*im));
2466 im->im_msgtype = IGMPMSG_WHOLEPKT;
2467 im->im_mbz = 0;
2468 im->im_vif = vifp - mfct->viftable;
2469 im->im_src = ip->ip_src;
2470 im->im_dst = ip->ip_dst;
2471
2472 k_igmpsrc.sin_addr = ip->ip_src;
2473
2474 MRTSTAT_INC(mrts_upcalls);
2475
2476 if (socket_send(mfct->router, mb_first, &k_igmpsrc) < 0) {
2477 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2478 MRTSTAT_INC(mrts_upq_sockfull);
2479 return ENOBUFS;
2480 }
2481
2482 /* Keep statistics */
2483 PIMSTAT_INC(pims_snd_registers_msgs);
2484 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2485
2486 return 0;
2487 }
2488
2489 /*
2490 * Encapsulate the data packet in PIM Register message and send it to the RP.
2491 */
2492 static int
pim_register_send_rp(struct mfctable * mfct,struct ip * ip,struct vif * vifp,struct mbuf * mb_copy,struct mfc * rt)2493 pim_register_send_rp(struct mfctable *mfct, struct ip *ip, struct vif *vifp,
2494 struct mbuf *mb_copy, struct mfc *rt)
2495 {
2496 struct mbuf *mb_first;
2497 struct ip *ip_outer;
2498 struct pim_encap_pimhdr *pimhdr;
2499 int len = ntohs(ip->ip_len);
2500 vifi_t vifi = rt->mfc_parent;
2501
2502 MRW_LOCK_ASSERT();
2503
2504 if (vifi >= mfct->numvifs ||
2505 in_nullhost(mfct->viftable[vifi].v_lcl_addr)) {
2506 m_freem(mb_copy);
2507 return EADDRNOTAVAIL; /* The iif vif is invalid */
2508 }
2509
2510 /*
2511 * Add a new mbuf with the encapsulating header
2512 */
2513 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2514 if (mb_first == NULL) {
2515 m_freem(mb_copy);
2516 return ENOBUFS;
2517 }
2518 mb_first->m_data += max_linkhdr;
2519 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2520 mb_first->m_next = mb_copy;
2521
2522 mb_first->m_pkthdr.len = len + mb_first->m_len;
2523
2524 /*
2525 * Fill in the encapsulating IP and PIM header
2526 */
2527 ip_outer = mtod(mb_first, struct ip *);
2528 *ip_outer = pim_encap_iphdr;
2529 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2530 sizeof(pim_encap_pimhdr));
2531 ip_outer->ip_src = mfct->viftable[vifi].v_lcl_addr;
2532 ip_outer->ip_dst = rt->mfc_rp;
2533 /*
2534 * Copy the inner header TOS to the outer header, and take care of the
2535 * IP_DF bit.
2536 */
2537 ip_outer->ip_tos = ip->ip_tos;
2538 if (ip->ip_off & htons(IP_DF))
2539 ip_outer->ip_off |= htons(IP_DF);
2540 ip_fillid(ip_outer, V_ip_random_id);
2541 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2542 + sizeof(pim_encap_iphdr));
2543 *pimhdr = pim_encap_pimhdr;
2544 /* If the iif crosses a border, set the Border-bit */
2545 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mfct->api_config)
2546 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2547
2548 mb_first->m_data += sizeof(pim_encap_iphdr);
2549 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2550 mb_first->m_data -= sizeof(pim_encap_iphdr);
2551
2552 send_packet(vifp, mb_first);
2553
2554 /* Keep statistics */
2555 PIMSTAT_INC(pims_snd_registers_msgs);
2556 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2557
2558 return 0;
2559 }
2560
2561 /*
2562 * pim_encapcheck() is called by the encap4_input() path at runtime to
2563 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2564 * into the kernel.
2565 */
2566 static int
pim_encapcheck(const struct mbuf * m __unused,int off __unused,int proto __unused,void * arg __unused)2567 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2568 int proto __unused, void *arg __unused)
2569 {
2570
2571 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2572 return (8); /* claim the datagram. */
2573 }
2574
2575 /*
2576 * PIM-SMv2 and PIM-DM messages processing.
2577 * Receives and verifies the PIM control messages, and passes them
2578 * up to the listening socket, using rip_input().
2579 * The only message with special processing is the PIM_REGISTER message
2580 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2581 * is passed to if_simloop().
2582 */
2583 static int
pim_input(struct mbuf * m,int off,int proto,void * arg __unused)2584 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2585 {
2586 struct mfctable *mfct;
2587 struct ip *ip = mtod(m, struct ip *);
2588 struct pim *pim;
2589 int iphlen = off;
2590 int minlen;
2591 int datalen = ntohs(ip->ip_len) - iphlen;
2592 int ip_tos;
2593
2594 mfct = &V_mfctables[M_GETFIB(m)];
2595
2596 /* Keep statistics */
2597 PIMSTAT_INC(pims_rcv_total_msgs);
2598 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2599
2600 /*
2601 * Validate lengths
2602 */
2603 if (datalen < PIM_MINLEN) {
2604 PIMSTAT_INC(pims_rcv_tooshort);
2605 CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2606 __func__, datalen, ntohl(ip->ip_src.s_addr));
2607 m_freem(m);
2608 return (IPPROTO_DONE);
2609 }
2610
2611 /*
2612 * If the packet is at least as big as a REGISTER, go agead
2613 * and grab the PIM REGISTER header size, to avoid another
2614 * possible m_pullup() later.
2615 *
2616 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2617 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2618 */
2619 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2620 /*
2621 * Get the IP and PIM headers in contiguous memory, and
2622 * possibly the PIM REGISTER header.
2623 */
2624 if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2625 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2626 return (IPPROTO_DONE);
2627 }
2628
2629 /* m_pullup() may have given us a new mbuf so reset ip. */
2630 ip = mtod(m, struct ip *);
2631 ip_tos = ip->ip_tos;
2632
2633 /* adjust mbuf to point to the PIM header */
2634 m->m_data += iphlen;
2635 m->m_len -= iphlen;
2636 pim = mtod(m, struct pim *);
2637
2638 /*
2639 * Validate checksum. If PIM REGISTER, exclude the data packet.
2640 *
2641 * XXX: some older PIMv2 implementations don't make this distinction,
2642 * so for compatibility reason perform the checksum over part of the
2643 * message, and if error, then over the whole message.
2644 */
2645 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2646 /* do nothing, checksum okay */
2647 } else if (in_cksum(m, datalen)) {
2648 PIMSTAT_INC(pims_rcv_badsum);
2649 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2650 m_freem(m);
2651 return (IPPROTO_DONE);
2652 }
2653
2654 /* PIM version check */
2655 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2656 PIMSTAT_INC(pims_rcv_badversion);
2657 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2658 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2659 m_freem(m);
2660 return (IPPROTO_DONE);
2661 }
2662
2663 /* restore mbuf back to the outer IP */
2664 m->m_data -= iphlen;
2665 m->m_len += iphlen;
2666
2667 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2668 /*
2669 * Since this is a REGISTER, we'll make a copy of the register
2670 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2671 * routing daemon.
2672 */
2673 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2674 struct mbuf *mcp;
2675 struct ip *encap_ip;
2676 u_int32_t *reghdr;
2677 struct ifnet *vifp;
2678
2679 MRW_RLOCK();
2680 if (mfct->register_vif >= mfct->numvifs ||
2681 mfct->register_vif == VIFI_INVALID) {
2682 MRW_RUNLOCK();
2683 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2684 (int)mfct->register_vif);
2685 m_freem(m);
2686 return (IPPROTO_DONE);
2687 }
2688 /* XXX need refcnt? */
2689 vifp = mfct->viftable[mfct->register_vif].v_ifp;
2690 MRW_RUNLOCK();
2691
2692 /*
2693 * Validate length
2694 */
2695 if (datalen < PIM_REG_MINLEN) {
2696 PIMSTAT_INC(pims_rcv_tooshort);
2697 PIMSTAT_INC(pims_rcv_badregisters);
2698 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2699 m_freem(m);
2700 return (IPPROTO_DONE);
2701 }
2702
2703 reghdr = (u_int32_t *)(pim + 1);
2704 encap_ip = (struct ip *)(reghdr + 1);
2705
2706 CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2707 __func__, ntohl(encap_ip->ip_src.s_addr),
2708 ntohs(encap_ip->ip_len));
2709
2710 /* verify the version number of the inner packet */
2711 if (encap_ip->ip_v != IPVERSION) {
2712 PIMSTAT_INC(pims_rcv_badregisters);
2713 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2714 m_freem(m);
2715 return (IPPROTO_DONE);
2716 }
2717
2718 /* verify the inner packet is destined to a mcast group */
2719 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2720 PIMSTAT_INC(pims_rcv_badregisters);
2721 CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2722 ntohl(encap_ip->ip_dst.s_addr));
2723 m_freem(m);
2724 return (IPPROTO_DONE);
2725 }
2726
2727 /* If a NULL_REGISTER, pass it to the daemon */
2728 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2729 goto pim_input_to_daemon;
2730
2731 /*
2732 * Copy the TOS from the outer IP header to the inner IP header.
2733 */
2734 if (encap_ip->ip_tos != ip_tos) {
2735 /* Outer TOS -> inner TOS */
2736 encap_ip->ip_tos = ip_tos;
2737 /* Recompute the inner header checksum. Sigh... */
2738
2739 /* adjust mbuf to point to the inner IP header */
2740 m->m_data += (iphlen + PIM_MINLEN);
2741 m->m_len -= (iphlen + PIM_MINLEN);
2742
2743 encap_ip->ip_sum = 0;
2744 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2745
2746 /* restore mbuf to point back to the outer IP header */
2747 m->m_data -= (iphlen + PIM_MINLEN);
2748 m->m_len += (iphlen + PIM_MINLEN);
2749 }
2750
2751 /*
2752 * Decapsulate the inner IP packet and loopback to forward it
2753 * as a normal multicast packet. Also, make a copy of the
2754 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2755 * to pass to the daemon later, so it can take the appropriate
2756 * actions (e.g., send back PIM_REGISTER_STOP).
2757 * XXX: here m->m_data points to the outer IP header.
2758 */
2759 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2760 if (mcp == NULL) {
2761 CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2762 m_freem(m);
2763 return (IPPROTO_DONE);
2764 }
2765
2766 /* Keep statistics */
2767 /* XXX: registers_bytes include only the encap. mcast pkt */
2768 PIMSTAT_INC(pims_rcv_registers_msgs);
2769 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2770
2771 /*
2772 * forward the inner ip packet; point m_data at the inner ip.
2773 */
2774 m_adj(m, iphlen + PIM_MINLEN);
2775
2776 CTR4(KTR_IPMF,
2777 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2778 __func__,
2779 (u_long)ntohl(encap_ip->ip_src.s_addr),
2780 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2781 (int)mfct->register_vif);
2782
2783 /* NB: vifp was collected above; can it change on us? */
2784 if_simloop(vifp, m, dst.sin_family, 0);
2785
2786 /* prepare the register head to send to the mrouting daemon */
2787 m = mcp;
2788 }
2789
2790 pim_input_to_daemon:
2791 /*
2792 * Pass the PIM message up to the daemon; if it is a Register message,
2793 * pass the 'head' only up to the daemon. This includes the
2794 * outer IP header, PIM header, PIM-Register header and the
2795 * inner IP header.
2796 * XXX: the outer IP header pkt size of a Register is not adjust to
2797 * reflect the fact that the inner multicast data is truncated.
2798 */
2799 return (rip_input(&m, &off, proto));
2800 }
2801
2802 static int
sysctl_mfctable(SYSCTL_HANDLER_ARGS)2803 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2804 {
2805 struct mfctable *mfct;
2806 struct mfc *rt;
2807 int error, i;
2808
2809 if (req->newptr)
2810 return (EPERM);
2811 mfct = &V_mfctables[curthread->td_proc->p_fibnum];
2812 if (mfct->mfchashtbl == NULL) /* XXX unlocked */
2813 return (0);
2814 error = sysctl_wire_old_buffer(req, 0);
2815 if (error)
2816 return (error);
2817
2818 MRW_RLOCK();
2819 if (mfct->mfchashtbl == NULL)
2820 goto out_locked;
2821
2822 for (i = 0; i < mfchashsize; i++) {
2823 LIST_FOREACH(rt, &mfct->mfchashtbl[i], mfc_hash) {
2824 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2825 if (error)
2826 goto out_locked;
2827 }
2828 }
2829 out_locked:
2830 MRW_RUNLOCK();
2831 return (error);
2832 }
2833 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2834 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2835 "IPv4 Multicast Forwarding Table "
2836 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2837
2838 static int
sysctl_viflist(SYSCTL_HANDLER_ARGS)2839 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2840 {
2841 struct mfctable *mfct;
2842 int error, i;
2843
2844 if (req->newptr)
2845 return (EPERM);
2846 error = sysctl_wire_old_buffer(req, MROUTE_VIF_SYSCTL_LEN * MAXVIFS);
2847 if (error)
2848 return (error);
2849
2850 mfct = &V_mfctables[curthread->td_proc->p_fibnum];
2851 MRW_RLOCK();
2852 /* Copy out user-visible portion of vif entry. */
2853 for (i = 0; i < MAXVIFS; i++) {
2854 error = SYSCTL_OUT(req, &mfct->viftable[i],
2855 MROUTE_VIF_SYSCTL_LEN);
2856 if (error)
2857 break;
2858 }
2859 MRW_RUNLOCK();
2860 return (error);
2861 }
2862 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2863 CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2864 sysctl_viflist, "S,vif[MAXVIFS]",
2865 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2866
2867 static void
ip_mroute_rtnumfibs_change(void * arg __unused,uint32_t ntables)2868 ip_mroute_rtnumfibs_change(void *arg __unused, uint32_t ntables)
2869 {
2870 struct mfctable *mfctables, *omfctables;
2871
2872 KASSERT(ntables >= V_nmfctables,
2873 ("%s: ntables %u nmfctables %u", __func__, ntables, V_nmfctables));
2874
2875 mfctables = mallocarray(ntables, sizeof(*mfctables), M_MRTABLE,
2876 M_WAITOK | M_ZERO);
2877 omfctables = V_mfctables;
2878
2879 for (int i = V_nmfctables; i < ntables; i++) {
2880 struct mfctable *mfct;
2881
2882 mfct = &mfctables[i];
2883 mfct->nexpire = malloc(mfchashsize, M_MRTABLE,
2884 M_WAITOK | M_ZERO);
2885 mfct->register_vif = VIFI_INVALID;
2886 }
2887
2888 MRW_TEARDOWN_WLOCK();
2889 MRW_WLOCK();
2890 for (int i = 0; i < V_nmfctables; i++)
2891 memcpy(&mfctables[i], &omfctables[i], sizeof(*mfctables));
2892 atomic_store_rel_ptr((uintptr_t *)&V_mfctables, (uintptr_t)mfctables);
2893 MRW_WUNLOCK();
2894 MRW_TEARDOWN_WUNLOCK();
2895
2896 NET_EPOCH_WAIT();
2897
2898 V_nmfctables = ntables;
2899 free(omfctables, M_MRTABLE);
2900 }
2901
2902 static void
vnet_mroute_init(const void * unused __unused)2903 vnet_mroute_init(const void *unused __unused)
2904 {
2905 ip_mroute_rtnumfibs_change(NULL, V_rt_numfibs);
2906
2907 callout_init_rw(&V_expire_upcalls_ch, &mrouter_lock, 0);
2908 callout_init_rw(&V_bw_upcalls_ch, &mrouter_lock, 0);
2909
2910 /* Prepare taskqueue */
2911 V_task_queue = taskqueue_create_fast("ip_mroute_tskq", M_NOWAIT,
2912 taskqueue_thread_enqueue, &V_task_queue);
2913 taskqueue_start_threads(&V_task_queue, 1, PI_NET, "ip_mroute_tskq task");
2914 }
2915 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2916 NULL);
2917
2918 static void
vnet_mroute_uninit(const void * unused __unused)2919 vnet_mroute_uninit(const void *unused __unused)
2920 {
2921 /* Taskqueue should be cancelled and drained before freeing */
2922 taskqueue_free(V_task_queue);
2923
2924 for (int i = 0; i < V_rt_numfibs; i++) {
2925 struct mfctable *mfct;
2926
2927 mfct = &V_mfctables[i];
2928 free(mfct->nexpire, M_MRTABLE);
2929 }
2930 free(V_mfctables, M_MRTABLE);
2931
2932 callout_drain(&V_expire_upcalls_ch);
2933 callout_drain(&V_bw_upcalls_ch);
2934 }
2935 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2936 vnet_mroute_uninit, NULL);
2937
2938 static int
ip_mroute_modevent(module_t mod,int type,void * unused)2939 ip_mroute_modevent(module_t mod, int type, void *unused)
2940 {
2941
2942 switch (type) {
2943 case MOD_LOAD:
2944 MRW_TEARDOWN_LOCK_INIT();
2945 MRW_LOCK_INIT();
2946
2947 if_detach_event_tag = EVENTHANDLER_REGISTER(
2948 ifnet_departure_event, if_detached_event, NULL,
2949 EVENTHANDLER_PRI_ANY);
2950 rtnumfibs_change_tag = EVENTHANDLER_REGISTER(
2951 rtnumfibs_change, ip_mroute_rtnumfibs_change,
2952 NULL, EVENTHANDLER_PRI_ANY);
2953
2954 if (!powerof2(mfchashsize)) {
2955 printf("WARNING: %s not a power of 2; using default\n",
2956 "net.inet.ip.mfchashsize");
2957 mfchashsize = MFCHASHSIZE;
2958 }
2959
2960 pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2961
2962 ip_mcast_src = X_ip_mcast_src;
2963 ip_mforward = X_ip_mforward;
2964 ip_mrouter_done = X_ip_mrouter_done;
2965 ip_mrouter_get = X_ip_mrouter_get;
2966 ip_mrouter_set = X_ip_mrouter_set;
2967
2968 ip_rsvp_force_done = X_ip_rsvp_force_done;
2969 ip_rsvp_vif = X_ip_rsvp_vif;
2970
2971 legal_vif_num = X_legal_vif_num;
2972 mrt_ioctl = X_mrt_ioctl;
2973 rsvp_input_p = X_rsvp_input;
2974 break;
2975
2976 case MOD_UNLOAD:
2977 /*
2978 * Typically module unload happens after the user-level
2979 * process has shutdown the kernel services (the check
2980 * below insures someone can't just yank the module out
2981 * from under a running process). But if the module is
2982 * just loaded and then unloaded w/o starting up a user
2983 * process we still need to cleanup.
2984 */
2985 MRW_WLOCK();
2986 if (ip_mrouter_cnt != 0) {
2987 MRW_WUNLOCK();
2988 return (EBUSY);
2989 }
2990 ip_mrouter_unloading = 1;
2991 MRW_WUNLOCK();
2992
2993 EVENTHANDLER_DEREGISTER(rtnumfibs_change,
2994 rtnumfibs_change_tag);
2995 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
2996 if_detach_event_tag);
2997
2998 if (pim_encap_cookie) {
2999 ip_encap_detach(pim_encap_cookie);
3000 pim_encap_cookie = NULL;
3001 }
3002
3003 ip_mcast_src = NULL;
3004 ip_mforward = NULL;
3005 ip_mrouter_done = NULL;
3006 ip_mrouter_get = NULL;
3007 ip_mrouter_set = NULL;
3008
3009 ip_rsvp_force_done = NULL;
3010 ip_rsvp_vif = NULL;
3011
3012 legal_vif_num = NULL;
3013 mrt_ioctl = NULL;
3014 rsvp_input_p = NULL;
3015
3016 MRW_LOCK_DESTROY();
3017 MRW_TEARDOWN_LOCK_DESTROY();
3018 break;
3019
3020 default:
3021 return EOPNOTSUPP;
3022 }
3023 return 0;
3024 }
3025
3026 static moduledata_t ip_mroutemod = {
3027 "ip_mroute",
3028 ip_mroute_modevent,
3029 0
3030 };
3031
3032 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
3033