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/protosw.h>
89 #include <sys/signalvar.h>
90 #include <sys/socket.h>
91 #include <sys/socketvar.h>
92 #include <sys/sockio.h>
93 #include <sys/sx.h>
94 #include <sys/sysctl.h>
95 #include <sys/syslog.h>
96 #include <sys/systm.h>
97 #include <sys/taskqueue.h>
98 #include <sys/time.h>
99 #include <sys/counter.h>
100 #include <machine/atomic.h>
101
102 #include <net/if.h>
103 #include <net/if_var.h>
104 #include <net/if_private.h>
105 #include <net/if_types.h>
106 #include <net/netisr.h>
107 #include <net/route.h>
108 #include <net/vnet.h>
109
110 #include <netinet/in.h>
111 #include <netinet/igmp.h>
112 #include <netinet/in_systm.h>
113 #include <netinet/in_var.h>
114 #include <netinet/ip.h>
115 #include <netinet/ip_encap.h>
116 #include <netinet/ip_mroute.h>
117 #include <netinet/ip_var.h>
118 #include <netinet/ip_options.h>
119 #include <netinet/pim.h>
120 #include <netinet/pim_var.h>
121 #include <netinet/udp.h>
122
123 #include <machine/in_cksum.h>
124
125 #ifndef KTR_IPMF
126 #define KTR_IPMF KTR_INET
127 #endif
128
129 #define VIFI_INVALID ((vifi_t) -1)
130
131 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
132
133 /*
134 * Locking. We use two locks: one for the virtual interface table and
135 * one for the forwarding table. These locks may be nested in which case
136 * the VIF lock must always be taken first. Note that each lock is used
137 * to cover not only the specific data structure but also related data
138 * structures.
139 */
140
141 static struct sx __exclusive_cache_line mrouter_teardown;
142 #define MRW_TEARDOWN_WLOCK() sx_xlock(&mrouter_teardown)
143 #define MRW_TEARDOWN_WUNLOCK() sx_xunlock(&mrouter_teardown)
144 #define MRW_TEARDOWN_LOCK_INIT() \
145 sx_init(&mrouter_teardown, "IPv4 multicast forwarding teardown")
146 #define MRW_TEARDOWN_LOCK_DESTROY() sx_destroy(&mrouter_teardown)
147
148 static struct rwlock mrouter_lock;
149 #define MRW_RLOCK() rw_rlock(&mrouter_lock)
150 #define MRW_WLOCK() rw_wlock(&mrouter_lock)
151 #define MRW_RUNLOCK() rw_runlock(&mrouter_lock)
152 #define MRW_WUNLOCK() rw_wunlock(&mrouter_lock)
153 #define MRW_UNLOCK() rw_unlock(&mrouter_lock)
154 #define MRW_LOCK_ASSERT() rw_assert(&mrouter_lock, RA_LOCKED)
155 #define MRW_WLOCK_ASSERT() rw_assert(&mrouter_lock, RA_WLOCKED)
156 #define MRW_LOCK_TRY_UPGRADE() rw_try_upgrade(&mrouter_lock)
157 #define MRW_WOWNED() rw_wowned(&mrouter_lock)
158 #define MRW_LOCK_INIT() \
159 rw_init(&mrouter_lock, "IPv4 multicast forwarding")
160 #define MRW_LOCK_DESTROY() rw_destroy(&mrouter_lock)
161
162 static int ip_mrouter_cnt; /* # of vnets with active mrouters */
163 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
164
165 VNET_PCPUSTAT_DEFINE_STATIC(struct mrtstat, mrtstat);
166 VNET_PCPUSTAT_SYSINIT(mrtstat);
167 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
168 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
169 mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
170 "netinet/ip_mroute.h)");
171
172 VNET_DEFINE_STATIC(struct socket *, ip_mrouter);
173 #define V_ip_mrouter VNET(ip_mrouter)
174
175 VNET_DEFINE_STATIC(u_long, mfchash);
176 #define V_mfchash VNET(mfchash)
177 #define MFCHASH(a, g) \
178 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
179 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
180 #define MFCHASHSIZE 256
181
182 static u_long mfchashsize = MFCHASHSIZE; /* Hash size */
183 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, mfchashsize, CTLFLAG_RDTUN,
184 &mfchashsize, 0, "IPv4 Multicast Forwarding Table hash size");
185 VNET_DEFINE_STATIC(u_char *, nexpire); /* 0..mfchashsize-1 */
186 #define V_nexpire VNET(nexpire)
187 VNET_DEFINE_STATIC(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
188 #define V_mfchashtbl VNET(mfchashtbl)
189 VNET_DEFINE_STATIC(struct taskqueue *, task_queue);
190 #define V_task_queue VNET(task_queue)
191 VNET_DEFINE_STATIC(struct task, task);
192 #define V_task VNET(task)
193
194 VNET_DEFINE_STATIC(vifi_t, numvifs);
195 #define V_numvifs VNET(numvifs)
196 VNET_DEFINE_STATIC(struct vif *, viftable);
197 #define V_viftable VNET(viftable)
198
199 static eventhandler_tag if_detach_event_tag = NULL;
200
201 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
202 #define V_expire_upcalls_ch VNET(expire_upcalls_ch)
203
204 VNET_DEFINE_STATIC(struct mtx, buf_ring_mtx);
205 #define V_buf_ring_mtx VNET(buf_ring_mtx)
206
207 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
208 #define UPCALL_EXPIRE 6 /* number of timeouts */
209
210 /*
211 * Bandwidth meter variables and constants
212 */
213 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
214
215 /*
216 * Pending upcalls are stored in a ring which is flushed when
217 * full, or periodically
218 */
219 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
220 #define V_bw_upcalls_ch VNET(bw_upcalls_ch)
221 VNET_DEFINE_STATIC(struct buf_ring *, bw_upcalls_ring);
222 #define V_bw_upcalls_ring VNET(bw_upcalls_ring)
223 VNET_DEFINE_STATIC(struct mtx, bw_upcalls_ring_mtx);
224 #define V_bw_upcalls_ring_mtx VNET(bw_upcalls_ring_mtx)
225
226 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
227
228 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
229 VNET_PCPUSTAT_SYSINIT(pimstat);
230 VNET_PCPUSTAT_SYSUNINIT(pimstat);
231
232 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
233 "PIM");
234 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
235 pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
236
237 static u_long pim_squelch_wholepkt = 0;
238 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RWTUN,
239 &pim_squelch_wholepkt, 0,
240 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
241
242 static const struct encaptab *pim_encap_cookie;
243 static int pim_encapcheck(const struct mbuf *, int, int, void *);
244 static int pim_input(struct mbuf *, int, int, void *);
245
246 extern int in_mcast_loop;
247
248 static const struct encap_config ipv4_encap_cfg = {
249 .proto = IPPROTO_PIM,
250 .min_length = sizeof(struct ip) + PIM_MINLEN,
251 .exact_match = 8,
252 .check = pim_encapcheck,
253 .input = pim_input
254 };
255
256 /*
257 * Note: the PIM Register encapsulation adds the following in front of a
258 * data packet:
259 *
260 * struct pim_encap_hdr {
261 * struct ip ip;
262 * struct pim_encap_pimhdr pim;
263 * }
264 *
265 */
266
267 struct pim_encap_pimhdr {
268 struct pim pim;
269 uint32_t flags;
270 };
271 #define PIM_ENCAP_TTL 64
272
273 static struct ip pim_encap_iphdr = {
274 #if BYTE_ORDER == LITTLE_ENDIAN
275 sizeof(struct ip) >> 2,
276 IPVERSION,
277 #else
278 IPVERSION,
279 sizeof(struct ip) >> 2,
280 #endif
281 0, /* tos */
282 sizeof(struct ip), /* total length */
283 0, /* id */
284 0, /* frag offset */
285 PIM_ENCAP_TTL,
286 IPPROTO_PIM,
287 0, /* checksum */
288 };
289
290 static struct pim_encap_pimhdr pim_encap_pimhdr = {
291 {
292 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
293 0, /* reserved */
294 0, /* checksum */
295 },
296 0 /* flags */
297 };
298
299 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
300 #define V_reg_vif_num VNET(reg_vif_num)
301 VNET_DEFINE_STATIC(struct ifnet *, multicast_register_if);
302 #define V_multicast_register_if VNET(multicast_register_if)
303
304 /*
305 * Private variables.
306 */
307
308 static u_long X_ip_mcast_src(int);
309 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
310 struct ip_moptions *);
311 static void X_ip_mrouter_done(struct socket *);
312 static int X_ip_mrouter_get(struct socket *, struct sockopt *);
313 static int X_ip_mrouter_set(struct socket *, struct sockopt *);
314 static int X_legal_vif_num(int);
315 static int X_mrt_ioctl(u_long, caddr_t, int);
316
317 static int add_bw_upcall(struct bw_upcall *);
318 static int add_mfc(struct mfcctl2 *);
319 static int add_vif(struct vifctl *);
320 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
321 static void bw_meter_geq_receive_packet(struct bw_meter *, int,
322 struct timeval *);
323 static void bw_upcalls_send(void);
324 static int del_bw_upcall(struct bw_upcall *);
325 static int del_mfc(struct mfcctl2 *);
326 static int del_vif(vifi_t);
327 static int del_vif_locked(vifi_t, struct ifnet **, struct ifnet **);
328 static void expire_bw_upcalls_send(void *);
329 static void expire_mfc(struct mfc *);
330 static void expire_upcalls(void *);
331 static void free_bw_list(struct bw_meter *);
332 static int get_sg_cnt(struct sioc_sg_req *);
333 static int get_vif_cnt(struct sioc_vif_req *);
334 static void if_detached_event(void *, struct ifnet *);
335 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
336 static int ip_mrouter_init(struct socket *, int);
337 static __inline struct mfc *
338 mfc_find(struct in_addr *, struct in_addr *);
339 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
340 static struct mbuf *
341 pim_register_prepare(struct ip *, struct mbuf *);
342 static int pim_register_send(struct ip *, struct vif *,
343 struct mbuf *, struct mfc *);
344 static int pim_register_send_rp(struct ip *, struct vif *,
345 struct mbuf *, struct mfc *);
346 static int pim_register_send_upcall(struct ip *, struct vif *,
347 struct mbuf *, struct mfc *);
348 static void send_packet(struct vif *, struct mbuf *);
349 static int set_api_config(uint32_t *);
350 static int set_assert(int);
351 static int socket_send(struct socket *, struct mbuf *,
352 struct sockaddr_in *);
353
354 /*
355 * Kernel multicast forwarding API capabilities and setup.
356 * If more API capabilities are added to the kernel, they should be
357 * recorded in `mrt_api_support'.
358 */
359 #define MRT_API_VERSION 0x0305
360
361 static const int mrt_api_version = MRT_API_VERSION;
362 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
363 MRT_MFC_FLAGS_BORDER_VIF |
364 MRT_MFC_RP |
365 MRT_MFC_BW_UPCALL);
366 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
367 #define V_mrt_api_config VNET(mrt_api_config)
368 VNET_DEFINE_STATIC(int, pim_assert_enabled);
369 #define V_pim_assert_enabled VNET(pim_assert_enabled)
370 static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */
371
372 /*
373 * Find a route for a given origin IP address and multicast group address.
374 * Statistics must be updated by the caller.
375 */
376 static __inline struct mfc *
mfc_find(struct in_addr * o,struct in_addr * g)377 mfc_find(struct in_addr *o, struct in_addr *g)
378 {
379 struct mfc *rt;
380
381 /*
382 * Might be called both RLOCK and WLOCK.
383 * Check if any, it's caller responsibility
384 * to choose correct option.
385 */
386 MRW_LOCK_ASSERT();
387
388 LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
389 if (in_hosteq(rt->mfc_origin, *o) &&
390 in_hosteq(rt->mfc_mcastgrp, *g) &&
391 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,
407 M_NOWAIT, &V_buf_ring_mtx);
408 if (rt->mfc_stall_ring == NULL) {
409 free(rt, M_MRTABLE);
410 return NULL;
411 }
412
413 return rt;
414 }
415
416 /*
417 * Handle MRT setsockopt commands to modify the multicast forwarding tables.
418 */
419 static int
X_ip_mrouter_set(struct socket * so,struct sockopt * sopt)420 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
421 {
422 int error, optval;
423 vifi_t vifi;
424 struct vifctl vifc;
425 struct mfcctl2 mfc;
426 struct bw_upcall bw_upcall;
427 uint32_t i;
428
429 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
430 return EPERM;
431
432 error = 0;
433 switch (sopt->sopt_name) {
434 case MRT_INIT:
435 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
436 if (error)
437 break;
438 error = ip_mrouter_init(so, optval);
439 break;
440 case MRT_DONE:
441 ip_mrouter_done(so);
442 break;
443 case MRT_ADD_VIF:
444 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
445 if (error)
446 break;
447 error = add_vif(&vifc);
448 break;
449 case MRT_DEL_VIF:
450 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
451 if (error)
452 break;
453 error = del_vif(vifi);
454 break;
455 case MRT_ADD_MFC:
456 case MRT_DEL_MFC:
457 /*
458 * select data size depending on API version.
459 */
460 if (sopt->sopt_name == MRT_ADD_MFC &&
461 V_mrt_api_config & MRT_API_FLAGS_ALL) {
462 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
463 sizeof(struct mfcctl2));
464 } else {
465 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
466 sizeof(struct mfcctl));
467 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
468 sizeof(mfc) - sizeof(struct mfcctl));
469 }
470 if (error)
471 break;
472 if (sopt->sopt_name == MRT_ADD_MFC)
473 error = add_mfc(&mfc);
474 else
475 error = del_mfc(&mfc);
476 break;
477
478 case MRT_ASSERT:
479 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
480 if (error)
481 break;
482 set_assert(optval);
483 break;
484
485 case MRT_API_CONFIG:
486 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
487 if (!error)
488 error = set_api_config(&i);
489 if (!error)
490 error = sooptcopyout(sopt, &i, sizeof i);
491 break;
492
493 case MRT_ADD_BW_UPCALL:
494 case MRT_DEL_BW_UPCALL:
495 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
496 sizeof bw_upcall);
497 if (error)
498 break;
499 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
500 error = add_bw_upcall(&bw_upcall);
501 else
502 error = del_bw_upcall(&bw_upcall);
503 break;
504
505 default:
506 error = EOPNOTSUPP;
507 break;
508 }
509 return error;
510 }
511
512 /*
513 * Handle MRT getsockopt commands
514 */
515 static int
X_ip_mrouter_get(struct socket * so,struct sockopt * sopt)516 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
517 {
518 int error;
519
520 switch (sopt->sopt_name) {
521 case MRT_VERSION:
522 error = sooptcopyout(sopt, &mrt_api_version,
523 sizeof mrt_api_version);
524 break;
525 case MRT_ASSERT:
526 error = sooptcopyout(sopt, &V_pim_assert_enabled,
527 sizeof V_pim_assert_enabled);
528 break;
529 case MRT_API_SUPPORT:
530 error = sooptcopyout(sopt, &mrt_api_support,
531 sizeof mrt_api_support);
532 break;
533 case MRT_API_CONFIG:
534 error = sooptcopyout(sopt, &V_mrt_api_config,
535 sizeof V_mrt_api_config);
536 break;
537 default:
538 error = EOPNOTSUPP;
539 break;
540 }
541 return error;
542 }
543
544 /*
545 * Handle ioctl commands to obtain information from the cache
546 */
547 static int
X_mrt_ioctl(u_long cmd,caddr_t data,int fibnum __unused)548 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
549 {
550 int error;
551
552 error = priv_check(curthread, PRIV_NETINET_MROUTE);
553 if (error)
554 return (error);
555 switch (cmd) {
556 case (SIOCGETVIFCNT):
557 error = get_vif_cnt((struct sioc_vif_req *)data);
558 break;
559
560 case (SIOCGETSGCNT):
561 error = get_sg_cnt((struct sioc_sg_req *)data);
562 break;
563
564 default:
565 error = EINVAL;
566 break;
567 }
568 return error;
569 }
570
571 /*
572 * returns the packet, byte, rpf-failure count for the source group provided
573 */
574 static int
get_sg_cnt(struct sioc_sg_req * req)575 get_sg_cnt(struct sioc_sg_req *req)
576 {
577 struct mfc *rt;
578
579 MRW_RLOCK();
580 rt = mfc_find(&req->src, &req->grp);
581 if (rt == NULL) {
582 MRW_RUNLOCK();
583 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
584 return EADDRNOTAVAIL;
585 }
586 req->pktcnt = rt->mfc_pkt_cnt;
587 req->bytecnt = rt->mfc_byte_cnt;
588 req->wrong_if = rt->mfc_wrong_if;
589 MRW_RUNLOCK();
590 return 0;
591 }
592
593 /*
594 * returns the input and output packet and byte counts on the vif provided
595 */
596 static int
get_vif_cnt(struct sioc_vif_req * req)597 get_vif_cnt(struct sioc_vif_req *req)
598 {
599 struct vif *vif;
600 vifi_t vifi;
601
602 vifi = req->vifi;
603
604 MRW_RLOCK();
605 if (vifi >= V_numvifs) {
606 MRW_RUNLOCK();
607 return EINVAL;
608 }
609
610 vif = &V_viftable[vifi];
611 mtx_lock(&vif->v_mtx);
612 req->icount = vif->v_pkt_in;
613 req->ocount = vif->v_pkt_out;
614 req->ibytes = vif->v_bytes_in;
615 req->obytes = vif->v_bytes_out;
616 mtx_unlock(&vif->v_mtx);
617 MRW_RUNLOCK();
618
619 return 0;
620 }
621
622 static void
if_detached_event(void * arg __unused,struct ifnet * ifp)623 if_detached_event(void *arg __unused, struct ifnet *ifp)
624 {
625 vifi_t vifi;
626 u_long i, vifi_cnt = 0;
627 struct ifnet *free_ptr, *multi_leave;
628
629 MRW_WLOCK();
630 if (!V_ip_mrouting_enabled) {
631 MRW_WUNLOCK();
632 return;
633 }
634
635 /*
636 * Tear down multicast forwarder state associated with this ifnet.
637 * 1. Walk the vif list, matching vifs against this ifnet.
638 * 2. Walk the multicast forwarding cache (mfc) looking for
639 * inner matches with this vif's index.
640 * 3. Expire any matching multicast forwarding cache entries.
641 * 4. Free vif state. This should disable ALLMULTI on the interface.
642 */
643 restart:
644 for (vifi = 0; vifi < V_numvifs; vifi++) {
645 if (V_viftable[vifi].v_ifp != ifp)
646 continue;
647 for (i = 0; i < mfchashsize; i++) {
648 struct mfc *rt, *nrt;
649
650 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
651 if (rt->mfc_parent == vifi) {
652 expire_mfc(rt);
653 }
654 }
655 }
656 del_vif_locked(vifi, &multi_leave, &free_ptr);
657 if (free_ptr != NULL)
658 vifi_cnt++;
659 if (multi_leave) {
660 MRW_WUNLOCK();
661 if_allmulti(multi_leave, 0);
662 MRW_WLOCK();
663 goto restart;
664 }
665 }
666
667 MRW_WUNLOCK();
668
669 /*
670 * Free IFP. We don't have to use free_ptr here as it is the same
671 * that ifp. Perform free as many times as required in case
672 * refcount is greater than 1.
673 */
674 for (i = 0; i < vifi_cnt; i++)
675 if_free(ifp);
676 }
677
678 static void
ip_mrouter_upcall_thread(void * arg,int pending __unused)679 ip_mrouter_upcall_thread(void *arg, int pending __unused)
680 {
681 CURVNET_SET((struct vnet *) arg);
682
683 MRW_WLOCK();
684 bw_upcalls_send();
685 MRW_WUNLOCK();
686
687 CURVNET_RESTORE();
688 }
689
690 /*
691 * Enable multicast forwarding.
692 */
693 static int
ip_mrouter_init(struct socket * so,int version)694 ip_mrouter_init(struct socket *so, int version)
695 {
696
697 CTR2(KTR_IPMF, "%s: so %p", __func__, so);
698
699 if (version != 1)
700 return ENOPROTOOPT;
701
702 MRW_TEARDOWN_WLOCK();
703 MRW_WLOCK();
704
705 if (ip_mrouter_unloading) {
706 MRW_WUNLOCK();
707 MRW_TEARDOWN_WUNLOCK();
708 return ENOPROTOOPT;
709 }
710
711 if (V_ip_mrouter != NULL) {
712 MRW_WUNLOCK();
713 MRW_TEARDOWN_WUNLOCK();
714 return EADDRINUSE;
715 }
716
717 V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
718 HASH_NOWAIT);
719 if (V_mfchashtbl == NULL) {
720 MRW_WUNLOCK();
721 MRW_TEARDOWN_WUNLOCK();
722 return (ENOMEM);
723 }
724
725 /* Create upcall ring */
726 mtx_init(&V_bw_upcalls_ring_mtx, "mroute upcall buf_ring mtx", NULL, MTX_DEF);
727 V_bw_upcalls_ring = buf_ring_alloc(BW_UPCALLS_MAX, M_MRTABLE,
728 M_NOWAIT, &V_bw_upcalls_ring_mtx);
729 if (!V_bw_upcalls_ring) {
730 MRW_WUNLOCK();
731 MRW_TEARDOWN_WUNLOCK();
732 return (ENOMEM);
733 }
734
735 TASK_INIT(&V_task, 0, ip_mrouter_upcall_thread, curvnet);
736 taskqueue_cancel(V_task_queue, &V_task, NULL);
737 taskqueue_unblock(V_task_queue);
738
739 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
740 curvnet);
741 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
742 curvnet);
743
744 V_ip_mrouter = so;
745 V_ip_mrouting_enabled = true;
746 atomic_add_int(&ip_mrouter_cnt, 1);
747
748 /* This is a mutex required by buf_ring init, but not used internally */
749 mtx_init(&V_buf_ring_mtx, "mroute buf_ring mtx", NULL, MTX_DEF);
750
751 MRW_WUNLOCK();
752 MRW_TEARDOWN_WUNLOCK();
753
754 CTR1(KTR_IPMF, "%s: done", __func__);
755
756 return 0;
757 }
758
759 /*
760 * Disable multicast forwarding.
761 */
762 static void
X_ip_mrouter_done(struct socket * so)763 X_ip_mrouter_done(struct socket *so)
764 {
765 struct ifnet **ifps;
766 int nifp;
767 u_long i;
768 vifi_t vifi;
769 struct bw_upcall *bu;
770
771 MRW_TEARDOWN_WLOCK();
772 if (so != V_ip_mrouter) {
773 MRW_TEARDOWN_WUNLOCK();
774 return;
775 }
776
777 /*
778 * Detach/disable hooks to the reset of the system.
779 */
780 V_ip_mrouter = NULL;
781 V_ip_mrouting_enabled = false;
782 atomic_subtract_int(&ip_mrouter_cnt, 1);
783 V_mrt_api_config = 0;
784
785 /*
786 * Wait for all epoch sections to complete to ensure the new value of
787 * V_ip_mrouting_enabled is visible to others.
788 */
789 NET_EPOCH_WAIT();
790
791 /* Stop and drain task queue */
792 taskqueue_block(V_task_queue);
793 while (taskqueue_cancel(V_task_queue, &V_task, NULL)) {
794 taskqueue_drain(V_task_queue, &V_task);
795 }
796
797 ifps = malloc(MAXVIFS * sizeof(*ifps), M_TEMP, M_WAITOK);
798
799 MRW_WLOCK();
800 taskqueue_cancel(V_task_queue, &V_task, NULL);
801
802 /* Destroy upcall ring */
803 while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
804 free(bu, M_MRTABLE);
805 }
806 buf_ring_free(V_bw_upcalls_ring, M_MRTABLE);
807 mtx_destroy(&V_bw_upcalls_ring_mtx);
808
809 /*
810 * For each phyint in use, prepare to disable promiscuous reception
811 * of all IP multicasts. Defer the actual call until the lock is released;
812 * just record the list of interfaces while locked. Some interfaces use
813 * sx locks in their ioctl routines, which is not allowed while holding
814 * a non-sleepable lock.
815 */
816 KASSERT(V_numvifs <= MAXVIFS, ("More vifs than possible"));
817 for (vifi = 0, nifp = 0; vifi < V_numvifs; vifi++) {
818 if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
819 !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
820 ifps[nifp++] = V_viftable[vifi].v_ifp;
821 }
822 }
823 bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
824 V_numvifs = 0;
825 V_pim_assert_enabled = 0;
826
827 callout_stop(&V_expire_upcalls_ch);
828 callout_stop(&V_bw_upcalls_ch);
829
830 /*
831 * Free all multicast forwarding cache entries.
832 * Do not use hashdestroy(), as we must perform other cleanup.
833 */
834 for (i = 0; i < mfchashsize; i++) {
835 struct mfc *rt, *nrt;
836
837 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
838 expire_mfc(rt);
839 }
840 }
841 free(V_mfchashtbl, M_MRTABLE);
842 V_mfchashtbl = NULL;
843
844 bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
845
846 V_reg_vif_num = VIFI_INVALID;
847
848 mtx_destroy(&V_buf_ring_mtx);
849
850 MRW_WUNLOCK();
851 MRW_TEARDOWN_WUNLOCK();
852
853 /*
854 * Now drop our claim on promiscuous multicast on the interfaces recorded
855 * above. This is safe to do now because ALLMULTI is reference counted.
856 */
857 for (vifi = 0; vifi < nifp; vifi++)
858 if_allmulti(ifps[vifi], 0);
859 free(ifps, M_TEMP);
860
861 CTR1(KTR_IPMF, "%s: done", __func__);
862 }
863
864 /*
865 * Set PIM assert processing global
866 */
867 static int
set_assert(int i)868 set_assert(int i)
869 {
870 if ((i != 1) && (i != 0))
871 return EINVAL;
872
873 V_pim_assert_enabled = i;
874
875 return 0;
876 }
877
878 /*
879 * Configure API capabilities
880 */
881 int
set_api_config(uint32_t * apival)882 set_api_config(uint32_t *apival)
883 {
884 u_long i;
885
886 /*
887 * We can set the API capabilities only if it is the first operation
888 * after MRT_INIT. I.e.:
889 * - there are no vifs installed
890 * - pim_assert is not enabled
891 * - the MFC table is empty
892 */
893 if (V_numvifs > 0) {
894 *apival = 0;
895 return EPERM;
896 }
897 if (V_pim_assert_enabled) {
898 *apival = 0;
899 return EPERM;
900 }
901
902 MRW_RLOCK();
903
904 for (i = 0; i < mfchashsize; i++) {
905 if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
906 MRW_RUNLOCK();
907 *apival = 0;
908 return EPERM;
909 }
910 }
911
912 MRW_RUNLOCK();
913
914 V_mrt_api_config = *apival & mrt_api_support;
915 *apival = V_mrt_api_config;
916
917 return 0;
918 }
919
920 /*
921 * Add a vif to the vif table
922 */
923 static int
add_vif(struct vifctl * vifcp)924 add_vif(struct vifctl *vifcp)
925 {
926 struct vif *vifp = V_viftable + vifcp->vifc_vifi;
927 struct sockaddr_in sin = {sizeof sin, AF_INET};
928 struct ifaddr *ifa;
929 struct ifnet *ifp;
930 int error;
931
932 if (vifcp->vifc_vifi >= MAXVIFS)
933 return EINVAL;
934 /* rate limiting is no longer supported by this code */
935 if (vifcp->vifc_rate_limit != 0) {
936 log(LOG_ERR, "rate limiting is no longer supported\n");
937 return EINVAL;
938 }
939
940 if (in_nullhost(vifcp->vifc_lcl_addr))
941 return EADDRNOTAVAIL;
942
943 /* Find the interface with an address in AF_INET family */
944 if (vifcp->vifc_flags & VIFF_REGISTER) {
945 /*
946 * XXX: Because VIFF_REGISTER does not really need a valid
947 * local interface (e.g. it could be 127.0.0.2), we don't
948 * check its address.
949 */
950 ifp = NULL;
951 } else {
952 struct epoch_tracker et;
953
954 sin.sin_addr = vifcp->vifc_lcl_addr;
955 NET_EPOCH_ENTER(et);
956 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
957 if (ifa == NULL) {
958 NET_EPOCH_EXIT(et);
959 return EADDRNOTAVAIL;
960 }
961 ifp = ifa->ifa_ifp;
962 /* XXX FIXME we need to take a ref on ifp and cleanup properly! */
963 NET_EPOCH_EXIT(et);
964 }
965
966 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
967 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
968 return EOPNOTSUPP;
969 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
970 ifp = V_multicast_register_if = if_alloc(IFT_LOOP);
971 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
972 if (V_reg_vif_num == VIFI_INVALID) {
973 if_initname(V_multicast_register_if, "register_vif", 0);
974 V_reg_vif_num = vifcp->vifc_vifi;
975 }
976 } else { /* Make sure the interface supports multicast */
977 if ((ifp->if_flags & IFF_MULTICAST) == 0)
978 return EOPNOTSUPP;
979
980 /* Enable promiscuous reception of all IP multicasts from the if */
981 error = if_allmulti(ifp, 1);
982 if (error)
983 return error;
984 }
985
986 MRW_WLOCK();
987
988 if (!in_nullhost(vifp->v_lcl_addr)) {
989 if (ifp)
990 V_multicast_register_if = NULL;
991 MRW_WUNLOCK();
992 if (ifp)
993 if_free(ifp);
994 return EADDRINUSE;
995 }
996
997 vifp->v_flags = vifcp->vifc_flags;
998 vifp->v_threshold = vifcp->vifc_threshold;
999 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
1000 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
1001 vifp->v_ifp = ifp;
1002 /* initialize per vif pkt counters */
1003 vifp->v_pkt_in = 0;
1004 vifp->v_pkt_out = 0;
1005 vifp->v_bytes_in = 0;
1006 vifp->v_bytes_out = 0;
1007 sprintf(vifp->v_mtx_name, "BM[%d] mtx", vifcp->vifc_vifi);
1008 mtx_init(&vifp->v_mtx, vifp->v_mtx_name, NULL, MTX_DEF);
1009
1010 /* Adjust numvifs up if the vifi is higher than numvifs */
1011 if (V_numvifs <= vifcp->vifc_vifi)
1012 V_numvifs = vifcp->vifc_vifi + 1;
1013
1014 MRW_WUNLOCK();
1015
1016 CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
1017 (int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
1018 (int)vifcp->vifc_threshold);
1019
1020 return 0;
1021 }
1022
1023 /*
1024 * Delete a vif from the vif table
1025 */
1026 static int
del_vif_locked(vifi_t vifi,struct ifnet ** ifp_multi_leave,struct ifnet ** ifp_free)1027 del_vif_locked(vifi_t vifi, struct ifnet **ifp_multi_leave, struct ifnet **ifp_free)
1028 {
1029 struct vif *vifp;
1030
1031 *ifp_free = NULL;
1032 *ifp_multi_leave = NULL;
1033
1034 MRW_WLOCK_ASSERT();
1035
1036 if (vifi >= V_numvifs) {
1037 return EINVAL;
1038 }
1039 vifp = &V_viftable[vifi];
1040 if (in_nullhost(vifp->v_lcl_addr)) {
1041 return EADDRNOTAVAIL;
1042 }
1043
1044 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1045 *ifp_multi_leave = vifp->v_ifp;
1046
1047 if (vifp->v_flags & VIFF_REGISTER) {
1048 V_reg_vif_num = VIFI_INVALID;
1049 if (vifp->v_ifp) {
1050 if (vifp->v_ifp == V_multicast_register_if)
1051 V_multicast_register_if = NULL;
1052 *ifp_free = vifp->v_ifp;
1053 }
1054 }
1055
1056 mtx_destroy(&vifp->v_mtx);
1057
1058 bzero((caddr_t)vifp, sizeof (*vifp));
1059
1060 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
1061
1062 /* Adjust numvifs down */
1063 for (vifi = V_numvifs; vifi > 0; vifi--)
1064 if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
1065 break;
1066 V_numvifs = vifi;
1067
1068 return 0;
1069 }
1070
1071 static int
del_vif(vifi_t vifi)1072 del_vif(vifi_t vifi)
1073 {
1074 int cc;
1075 struct ifnet *free_ptr, *multi_leave;
1076
1077 MRW_WLOCK();
1078 cc = del_vif_locked(vifi, &multi_leave, &free_ptr);
1079 MRW_WUNLOCK();
1080
1081 if (multi_leave)
1082 if_allmulti(multi_leave, 0);
1083 if (free_ptr) {
1084 if_free(free_ptr);
1085 }
1086
1087 return cc;
1088 }
1089
1090 /*
1091 * update an mfc entry without resetting counters and S,G addresses.
1092 */
1093 static void
update_mfc_params(struct mfc * rt,struct mfcctl2 * mfccp)1094 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1095 {
1096 int i;
1097
1098 rt->mfc_parent = mfccp->mfcc_parent;
1099 for (i = 0; i < V_numvifs; i++) {
1100 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1101 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1102 MRT_MFC_FLAGS_ALL;
1103 }
1104 /* set the RP address */
1105 if (V_mrt_api_config & MRT_MFC_RP)
1106 rt->mfc_rp = mfccp->mfcc_rp;
1107 else
1108 rt->mfc_rp.s_addr = INADDR_ANY;
1109 }
1110
1111 /*
1112 * fully initialize an mfc entry from the parameter.
1113 */
1114 static void
init_mfc_params(struct mfc * rt,struct mfcctl2 * mfccp)1115 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1116 {
1117 rt->mfc_origin = mfccp->mfcc_origin;
1118 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1119
1120 update_mfc_params(rt, mfccp);
1121
1122 /* initialize pkt counters per src-grp */
1123 rt->mfc_pkt_cnt = 0;
1124 rt->mfc_byte_cnt = 0;
1125 rt->mfc_wrong_if = 0;
1126 timevalclear(&rt->mfc_last_assert);
1127 }
1128
1129 static void
expire_mfc(struct mfc * rt)1130 expire_mfc(struct mfc *rt)
1131 {
1132 struct rtdetq *rte;
1133
1134 MRW_WLOCK_ASSERT();
1135
1136 free_bw_list(rt->mfc_bw_meter_leq);
1137 free_bw_list(rt->mfc_bw_meter_geq);
1138
1139 while (!buf_ring_empty(rt->mfc_stall_ring)) {
1140 rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1141 if (rte) {
1142 m_freem(rte->m);
1143 free(rte, M_MRTABLE);
1144 }
1145 }
1146 buf_ring_free(rt->mfc_stall_ring, M_MRTABLE);
1147
1148 LIST_REMOVE(rt, mfc_hash);
1149 free(rt, M_MRTABLE);
1150 }
1151
1152 /*
1153 * Add an mfc entry
1154 */
1155 static int
add_mfc(struct mfcctl2 * mfccp)1156 add_mfc(struct mfcctl2 *mfccp)
1157 {
1158 struct mfc *rt;
1159 struct rtdetq *rte;
1160 u_long hash = 0;
1161 u_short nstl;
1162 struct epoch_tracker et;
1163
1164 MRW_WLOCK();
1165 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1166
1167 /* If an entry already exists, just update the fields */
1168 if (rt) {
1169 CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1170 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1171 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1172 mfccp->mfcc_parent);
1173 update_mfc_params(rt, mfccp);
1174 MRW_WUNLOCK();
1175 return (0);
1176 }
1177
1178 /*
1179 * Find the entry for which the upcall was made and update
1180 */
1181 nstl = 0;
1182 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1183 NET_EPOCH_ENTER(et);
1184 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1185 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1186 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1187 !buf_ring_empty(rt->mfc_stall_ring)) {
1188 CTR5(KTR_IPMF,
1189 "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1190 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1191 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1192 mfccp->mfcc_parent,
1193 rt->mfc_stall_ring);
1194 if (nstl++)
1195 CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1196
1197 init_mfc_params(rt, mfccp);
1198 rt->mfc_expire = 0; /* Don't clean this guy up */
1199 V_nexpire[hash]--;
1200
1201 /* Free queued packets, but attempt to forward them first. */
1202 while (!buf_ring_empty(rt->mfc_stall_ring)) {
1203 rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1204 if (rte->ifp != NULL)
1205 ip_mdq(rte->m, rte->ifp, rt, -1);
1206 m_freem(rte->m);
1207 free(rte, M_MRTABLE);
1208 }
1209 }
1210 }
1211 NET_EPOCH_EXIT(et);
1212
1213 /*
1214 * It is possible that an entry is being inserted without an upcall
1215 */
1216 if (nstl == 0) {
1217 CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1218 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1219 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1220 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1221 init_mfc_params(rt, mfccp);
1222 if (rt->mfc_expire)
1223 V_nexpire[hash]--;
1224 rt->mfc_expire = 0;
1225 break; /* XXX */
1226 }
1227 }
1228
1229 if (rt == NULL) { /* no upcall, so make a new entry */
1230 rt = mfc_alloc();
1231 if (rt == NULL) {
1232 MRW_WUNLOCK();
1233 return (ENOBUFS);
1234 }
1235
1236 init_mfc_params(rt, mfccp);
1237
1238 rt->mfc_expire = 0;
1239 rt->mfc_bw_meter_leq = NULL;
1240 rt->mfc_bw_meter_geq = NULL;
1241
1242 /* insert new entry at head of hash chain */
1243 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1244 }
1245 }
1246
1247 MRW_WUNLOCK();
1248
1249 return (0);
1250 }
1251
1252 /*
1253 * Delete an mfc entry
1254 */
1255 static int
del_mfc(struct mfcctl2 * mfccp)1256 del_mfc(struct mfcctl2 *mfccp)
1257 {
1258 struct in_addr origin;
1259 struct in_addr mcastgrp;
1260 struct mfc *rt;
1261
1262 origin = mfccp->mfcc_origin;
1263 mcastgrp = mfccp->mfcc_mcastgrp;
1264
1265 CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1266 ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1267
1268 MRW_WLOCK();
1269
1270 LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(origin, mcastgrp)], mfc_hash) {
1271 if (in_hosteq(rt->mfc_origin, origin) &&
1272 in_hosteq(rt->mfc_mcastgrp, mcastgrp))
1273 break;
1274 }
1275 if (rt == NULL) {
1276 MRW_WUNLOCK();
1277 return EADDRNOTAVAIL;
1278 }
1279
1280 expire_mfc(rt);
1281
1282 MRW_WUNLOCK();
1283
1284 return (0);
1285 }
1286
1287 /*
1288 * Send a message to the routing daemon on the multicast routing socket.
1289 */
1290 static int
socket_send(struct socket * s,struct mbuf * mm,struct sockaddr_in * src)1291 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1292 {
1293 if (s) {
1294 SOCKBUF_LOCK(&s->so_rcv);
1295 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1296 NULL) != 0) {
1297 sorwakeup_locked(s);
1298 return 0;
1299 }
1300 soroverflow_locked(s);
1301 }
1302 m_freem(mm);
1303 return -1;
1304 }
1305
1306 /*
1307 * IP multicast forwarding function. This function assumes that the packet
1308 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1309 * pointed to by "ifp", and the packet is to be relayed to other networks
1310 * that have members of the packet's destination IP multicast group.
1311 *
1312 * The packet is returned unscathed to the caller, unless it is
1313 * erroneous, in which case a non-zero return value tells the caller to
1314 * discard it.
1315 */
1316
1317 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1318
1319 static int
X_ip_mforward(struct ip * ip,struct ifnet * ifp,struct mbuf * m,struct ip_moptions * imo)1320 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1321 struct ip_moptions *imo)
1322 {
1323 struct mfc *rt;
1324 int error;
1325 vifi_t vifi;
1326 struct mbuf *mb0;
1327 struct rtdetq *rte;
1328 u_long hash;
1329 int hlen;
1330
1331 M_ASSERTMAPPED(m);
1332
1333 CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1334 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1335
1336 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1337 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1338 /*
1339 * Packet arrived via a physical interface or
1340 * an encapsulated tunnel or a register_vif.
1341 */
1342 } else {
1343 /*
1344 * Packet arrived through a source-route tunnel.
1345 * Source-route tunnels are no longer supported.
1346 */
1347 return (1);
1348 }
1349
1350 /*
1351 * BEGIN: MCAST ROUTING HOT PATH
1352 */
1353 MRW_RLOCK();
1354 if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1355 if (ip->ip_ttl < MAXTTL)
1356 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1357 error = ip_mdq(m, ifp, NULL, vifi);
1358 MRW_RUNLOCK();
1359 return error;
1360 }
1361
1362 /*
1363 * Don't forward a packet with time-to-live of zero or one,
1364 * or a packet destined to a local-only group.
1365 */
1366 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1367 MRW_RUNLOCK();
1368 return 0;
1369 }
1370
1371 mfc_find_retry:
1372 /*
1373 * Determine forwarding vifs from the forwarding cache table
1374 */
1375 MRTSTAT_INC(mrts_mfc_lookups);
1376 rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1377
1378 /* Entry exists, so forward if necessary */
1379 if (rt != NULL) {
1380 error = ip_mdq(m, ifp, rt, -1);
1381 /* Generic unlock here as we might release R or W lock */
1382 MRW_UNLOCK();
1383 return error;
1384 }
1385
1386 /*
1387 * END: MCAST ROUTING HOT PATH
1388 */
1389
1390 /* Further processing must be done with WLOCK taken */
1391 if ((MRW_WOWNED() == 0) && (MRW_LOCK_TRY_UPGRADE() == 0)) {
1392 MRW_RUNLOCK();
1393 MRW_WLOCK();
1394 goto mfc_find_retry;
1395 }
1396
1397 /*
1398 * If we don't have a route for packet's origin,
1399 * Make a copy of the packet & send message to routing daemon
1400 */
1401 hlen = ip->ip_hl << 2;
1402
1403 MRTSTAT_INC(mrts_mfc_misses);
1404 MRTSTAT_INC(mrts_no_route);
1405 CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1406 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1407
1408 /*
1409 * Allocate mbufs early so that we don't do extra work if we are
1410 * just going to fail anyway. Make sure to pullup the header so
1411 * that other people can't step on it.
1412 */
1413 rte = malloc((sizeof *rte), M_MRTABLE, M_NOWAIT|M_ZERO);
1414 if (rte == NULL) {
1415 MRW_WUNLOCK();
1416 return ENOBUFS;
1417 }
1418
1419 mb0 = m_copypacket(m, M_NOWAIT);
1420 if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1421 mb0 = m_pullup(mb0, hlen);
1422 if (mb0 == NULL) {
1423 free(rte, M_MRTABLE);
1424 MRW_WUNLOCK();
1425 return ENOBUFS;
1426 }
1427
1428 /* is there an upcall waiting for this flow ? */
1429 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1430 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash)
1431 {
1432 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1433 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1434 !buf_ring_empty(rt->mfc_stall_ring))
1435 break;
1436 }
1437
1438 if (rt == NULL) {
1439 int i;
1440 struct igmpmsg *im;
1441 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1442 struct mbuf *mm;
1443
1444 /*
1445 * Locate the vifi for the incoming interface for this packet.
1446 * If none found, drop packet.
1447 */
1448 for (vifi = 0; vifi < V_numvifs &&
1449 V_viftable[vifi].v_ifp != ifp; vifi++)
1450 ;
1451 if (vifi >= V_numvifs) /* vif not found, drop packet */
1452 goto non_fatal;
1453
1454 /* no upcall, so make a new entry */
1455 rt = mfc_alloc();
1456 if (rt == NULL)
1457 goto fail;
1458
1459 /* Make a copy of the header to send to the user level process */
1460 mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1461 if (mm == NULL)
1462 goto fail1;
1463
1464 /*
1465 * Send message to routing daemon to install
1466 * a route into the kernel table
1467 */
1468
1469 im = mtod(mm, struct igmpmsg*);
1470 im->im_msgtype = IGMPMSG_NOCACHE;
1471 im->im_mbz = 0;
1472 im->im_vif = vifi;
1473
1474 MRTSTAT_INC(mrts_upcalls);
1475
1476 k_igmpsrc.sin_addr = ip->ip_src;
1477 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1478 CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1479 MRTSTAT_INC(mrts_upq_sockfull);
1480 fail1: free(rt, M_MRTABLE);
1481 fail: free(rte, M_MRTABLE);
1482 m_freem(mb0);
1483 MRW_WUNLOCK();
1484 return ENOBUFS;
1485 }
1486
1487 /* insert new entry at head of hash chain */
1488 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1489 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1490 rt->mfc_expire = UPCALL_EXPIRE;
1491 V_nexpire[hash]++;
1492 for (i = 0; i < V_numvifs; i++) {
1493 rt->mfc_ttls[i] = 0;
1494 rt->mfc_flags[i] = 0;
1495 }
1496 rt->mfc_parent = -1;
1497
1498 /* clear the RP address */
1499 rt->mfc_rp.s_addr = INADDR_ANY;
1500 rt->mfc_bw_meter_leq = NULL;
1501 rt->mfc_bw_meter_geq = NULL;
1502
1503 /* initialize pkt counters per src-grp */
1504 rt->mfc_pkt_cnt = 0;
1505 rt->mfc_byte_cnt = 0;
1506 rt->mfc_wrong_if = 0;
1507 timevalclear(&rt->mfc_last_assert);
1508
1509 buf_ring_enqueue(rt->mfc_stall_ring, rte);
1510
1511 /* Add RT to hashtable as it didn't exist before */
1512 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1513 } else {
1514 /* determine if queue has overflowed */
1515 if (buf_ring_full(rt->mfc_stall_ring)) {
1516 MRTSTAT_INC(mrts_upq_ovflw);
1517 non_fatal: free(rte, M_MRTABLE);
1518 m_freem(mb0);
1519 MRW_WUNLOCK();
1520 return (0);
1521 }
1522
1523 buf_ring_enqueue(rt->mfc_stall_ring, rte);
1524 }
1525
1526 rte->m = mb0;
1527 rte->ifp = ifp;
1528
1529 MRW_WUNLOCK();
1530
1531 return 0;
1532 }
1533
1534 /*
1535 * Clean up the cache entry if upcall is not serviced
1536 */
1537 static void
expire_upcalls(void * arg)1538 expire_upcalls(void *arg)
1539 {
1540 u_long i;
1541
1542 CURVNET_SET((struct vnet *) arg);
1543
1544 /*This callout is always run with MRW_WLOCK taken. */
1545
1546 for (i = 0; i < mfchashsize; i++) {
1547 struct mfc *rt, *nrt;
1548
1549 if (V_nexpire[i] == 0)
1550 continue;
1551
1552 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1553 if (buf_ring_empty(rt->mfc_stall_ring))
1554 continue;
1555
1556 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1557 continue;
1558
1559 MRTSTAT_INC(mrts_cache_cleanups);
1560 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1561 (u_long)ntohl(rt->mfc_origin.s_addr),
1562 (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1563
1564 expire_mfc(rt);
1565 }
1566 }
1567
1568 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1569 curvnet);
1570
1571 CURVNET_RESTORE();
1572 }
1573
1574 /*
1575 * Packet forwarding routine once entry in the cache is made
1576 */
1577 static int
ip_mdq(struct mbuf * m,struct ifnet * ifp,struct mfc * rt,vifi_t xmt_vif)1578 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1579 {
1580 struct ip *ip = mtod(m, struct ip *);
1581 struct vif *vif;
1582 vifi_t vifi;
1583 int plen = ntohs(ip->ip_len);
1584
1585 M_ASSERTMAPPED(m);
1586 MRW_LOCK_ASSERT();
1587 NET_EPOCH_ASSERT();
1588
1589 /*
1590 * If xmt_vif is not -1, send on only the requested vif.
1591 *
1592 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1593 */
1594 if (xmt_vif < V_numvifs) {
1595 if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1596 pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1597 else
1598 phyint_send(ip, V_viftable + xmt_vif, m);
1599 return 1;
1600 }
1601
1602 /*
1603 * Don't forward if it didn't arrive from the parent vif for its origin.
1604 */
1605 vifi = rt->mfc_parent;
1606 vif = &V_viftable[vifi];
1607 if (vifi >= V_numvifs || vif->v_ifp != ifp) {
1608 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1609 __func__, ifp, (int)vifi, vif->v_ifp);
1610 MRTSTAT_INC(mrts_wrong_if);
1611 ++rt->mfc_wrong_if;
1612 /*
1613 * If we are doing PIM assert processing, send a message
1614 * to the routing daemon.
1615 *
1616 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1617 * can complete the SPT switch, regardless of the type
1618 * of the iif (broadcast media, GRE tunnel, etc).
1619 */
1620 if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1621 vif->v_ifp != NULL) {
1622 if (ifp == V_multicast_register_if)
1623 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1624
1625 /* Get vifi for the incoming packet */
1626 for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp; vifi++)
1627 ;
1628 if (vifi >= V_numvifs)
1629 return 0; /* The iif is not found: ignore the packet. */
1630
1631 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1632 return 0; /* WRONGVIF disabled: ignore the packet */
1633
1634 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1635 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1636 struct igmpmsg *im;
1637 int hlen = ip->ip_hl << 2;
1638 struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1639
1640 if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1641 mm = m_pullup(mm, hlen);
1642 if (mm == NULL)
1643 return ENOBUFS;
1644
1645 im = mtod(mm, struct igmpmsg *);
1646 im->im_msgtype = IGMPMSG_WRONGVIF;
1647 im->im_mbz = 0;
1648 im->im_vif = vifi;
1649
1650 MRTSTAT_INC(mrts_upcalls);
1651
1652 k_igmpsrc.sin_addr = im->im_src;
1653 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1654 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1655 MRTSTAT_INC(mrts_upq_sockfull);
1656 return ENOBUFS;
1657 }
1658 }
1659 }
1660 return 0;
1661 }
1662
1663 /* If I sourced this packet, it counts as output, else it was input. */
1664 mtx_lock(&vif->v_mtx);
1665 if (in_hosteq(ip->ip_src, vif->v_lcl_addr)) {
1666 vif->v_pkt_out++;
1667 vif->v_bytes_out += plen;
1668 } else {
1669 vif->v_pkt_in++;
1670 vif->v_bytes_in += plen;
1671 }
1672 mtx_unlock(&vif->v_mtx);
1673
1674 rt->mfc_pkt_cnt++;
1675 rt->mfc_byte_cnt += plen;
1676
1677 /*
1678 * For each vif, decide if a copy of the packet should be forwarded.
1679 * Forward if:
1680 * - the ttl exceeds the vif's threshold
1681 * - there are group members downstream on interface
1682 */
1683 for (vifi = 0; vifi < V_numvifs; vifi++)
1684 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1685 vif = &V_viftable[vifi];
1686 vif->v_pkt_out++;
1687 vif->v_bytes_out += plen;
1688 if (vif->v_flags & VIFF_REGISTER)
1689 pim_register_send(ip, vif, m, rt);
1690 else
1691 phyint_send(ip, vif, m);
1692 }
1693
1694 /*
1695 * Perform upcall-related bw measuring.
1696 */
1697 if ((rt->mfc_bw_meter_geq != NULL) || (rt->mfc_bw_meter_leq != NULL)) {
1698 struct bw_meter *x;
1699 struct timeval now;
1700
1701 microtime(&now);
1702 /* Process meters for Greater-or-EQual case */
1703 for (x = rt->mfc_bw_meter_geq; x != NULL; x = x->bm_mfc_next)
1704 bw_meter_geq_receive_packet(x, plen, &now);
1705
1706 /* Process meters for Lower-or-EQual case */
1707 for (x = rt->mfc_bw_meter_leq; x != NULL; x = x->bm_mfc_next) {
1708 /*
1709 * Record that a packet is received.
1710 * A lock has to be taken as callout context
1711 * (expire_bw_meter_leq) might modify these fields
1712 * as well
1713 */
1714 mtx_lock(&x->bm_mtx);
1715 x->bm_measured.b_packets++;
1716 x->bm_measured.b_bytes += plen;
1717 mtx_unlock(&x->bm_mtx);
1718 }
1719 }
1720
1721 return 0;
1722 }
1723
1724 /*
1725 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1726 */
1727 static int
X_legal_vif_num(int vif)1728 X_legal_vif_num(int vif)
1729 {
1730 int ret;
1731
1732 ret = 0;
1733 if (vif < 0)
1734 return (ret);
1735
1736 MRW_RLOCK();
1737 if (vif < V_numvifs)
1738 ret = 1;
1739 MRW_RUNLOCK();
1740
1741 return (ret);
1742 }
1743
1744 /*
1745 * Return the local address used by this vif
1746 */
1747 static u_long
X_ip_mcast_src(int vifi)1748 X_ip_mcast_src(int vifi)
1749 {
1750 in_addr_t addr;
1751
1752 addr = INADDR_ANY;
1753 if (vifi < 0)
1754 return (addr);
1755
1756 MRW_RLOCK();
1757 if (vifi < V_numvifs)
1758 addr = V_viftable[vifi].v_lcl_addr.s_addr;
1759 MRW_RUNLOCK();
1760
1761 return (addr);
1762 }
1763
1764 static void
phyint_send(struct ip * ip,struct vif * vifp,struct mbuf * m)1765 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1766 {
1767 struct mbuf *mb_copy;
1768 int hlen = ip->ip_hl << 2;
1769
1770 MRW_LOCK_ASSERT();
1771 M_ASSERTMAPPED(m);
1772
1773 /*
1774 * Make a new reference to the packet; make sure that
1775 * the IP header is actually copied, not just referenced,
1776 * so that ip_output() only scribbles on the copy.
1777 */
1778 mb_copy = m_copypacket(m, M_NOWAIT);
1779 if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1780 mb_copy = m_pullup(mb_copy, hlen);
1781 if (mb_copy == NULL)
1782 return;
1783
1784 send_packet(vifp, mb_copy);
1785 }
1786
1787 static void
send_packet(struct vif * vifp,struct mbuf * m)1788 send_packet(struct vif *vifp, struct mbuf *m)
1789 {
1790 struct ip_moptions imo;
1791 int error __unused;
1792
1793 MRW_LOCK_ASSERT();
1794 NET_EPOCH_ASSERT();
1795
1796 imo.imo_multicast_ifp = vifp->v_ifp;
1797 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1798 imo.imo_multicast_loop = !!in_mcast_loop;
1799 imo.imo_multicast_vif = -1;
1800 STAILQ_INIT(&imo.imo_head);
1801
1802 /*
1803 * Re-entrancy should not be a problem here, because
1804 * the packets that we send out and are looped back at us
1805 * should get rejected because they appear to come from
1806 * the loopback interface, thus preventing looping.
1807 */
1808 error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1809 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1810 (ptrdiff_t)(vifp - V_viftable), error);
1811 }
1812
1813 /*
1814 * Stubs for old RSVP socket shim implementation.
1815 */
1816
1817 static int
X_ip_rsvp_vif(struct socket * so __unused,struct sockopt * sopt __unused)1818 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1819 {
1820
1821 return (EOPNOTSUPP);
1822 }
1823
1824 static void
X_ip_rsvp_force_done(struct socket * so __unused)1825 X_ip_rsvp_force_done(struct socket *so __unused)
1826 {
1827
1828 }
1829
1830 static int
X_rsvp_input(struct mbuf ** mp,int * offp,int proto)1831 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1832 {
1833 struct mbuf *m;
1834
1835 m = *mp;
1836 *mp = NULL;
1837 if (!V_rsvp_on)
1838 m_freem(m);
1839 return (IPPROTO_DONE);
1840 }
1841
1842 /*
1843 * Code for bandwidth monitors
1844 */
1845
1846 /*
1847 * Define common interface for timeval-related methods
1848 */
1849 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1850 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1851 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1852
1853 static uint32_t
compute_bw_meter_flags(struct bw_upcall * req)1854 compute_bw_meter_flags(struct bw_upcall *req)
1855 {
1856 uint32_t flags = 0;
1857
1858 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1859 flags |= BW_METER_UNIT_PACKETS;
1860 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1861 flags |= BW_METER_UNIT_BYTES;
1862 if (req->bu_flags & BW_UPCALL_GEQ)
1863 flags |= BW_METER_GEQ;
1864 if (req->bu_flags & BW_UPCALL_LEQ)
1865 flags |= BW_METER_LEQ;
1866
1867 return flags;
1868 }
1869
1870 static void
expire_bw_meter_leq(void * arg)1871 expire_bw_meter_leq(void *arg)
1872 {
1873 struct bw_meter *x = arg;
1874 struct timeval now;
1875 /*
1876 * INFO:
1877 * callout is always executed with MRW_WLOCK taken
1878 */
1879
1880 CURVNET_SET((struct vnet *)x->arg);
1881
1882 microtime(&now);
1883
1884 /*
1885 * Test if we should deliver an upcall
1886 */
1887 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1888 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1889 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1890 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1891 /* Prepare an upcall for delivery */
1892 bw_meter_prepare_upcall(x, &now);
1893 }
1894
1895 /* Send all upcalls that are pending delivery */
1896 taskqueue_enqueue(V_task_queue, &V_task);
1897
1898 /* Reset counters */
1899 x->bm_start_time = now;
1900 /*
1901 * The lock has to be taken as ip_forward context
1902 * might modify these fields as well
1903 */
1904 mtx_lock(&x->bm_mtx);
1905 x->bm_measured.b_bytes = 0;
1906 x->bm_measured.b_packets = 0;
1907 mtx_unlock(&x->bm_mtx);
1908
1909 callout_schedule(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time));
1910
1911 CURVNET_RESTORE();
1912 }
1913
1914 /*
1915 * Add a bw_meter entry
1916 */
1917 static int
add_bw_upcall(struct bw_upcall * req)1918 add_bw_upcall(struct bw_upcall *req)
1919 {
1920 struct mfc *mfc;
1921 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1922 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1923 struct timeval now;
1924 struct bw_meter *x, **bwm_ptr;
1925 uint32_t flags;
1926
1927 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1928 return EOPNOTSUPP;
1929
1930 /* Test if the flags are valid */
1931 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1932 return EINVAL;
1933 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1934 return EINVAL;
1935 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1936 return EINVAL;
1937
1938 /* Test if the threshold time interval is valid */
1939 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1940 return EINVAL;
1941
1942 flags = compute_bw_meter_flags(req);
1943
1944 /*
1945 * Find if we have already same bw_meter entry
1946 */
1947 MRW_WLOCK();
1948 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1949 if (mfc == NULL) {
1950 MRW_WUNLOCK();
1951 return EADDRNOTAVAIL;
1952 }
1953
1954 /* Choose an appropriate bw_meter list */
1955 if (req->bu_flags & BW_UPCALL_GEQ)
1956 bwm_ptr = &mfc->mfc_bw_meter_geq;
1957 else
1958 bwm_ptr = &mfc->mfc_bw_meter_leq;
1959
1960 for (x = *bwm_ptr; x != NULL; x = x->bm_mfc_next) {
1961 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1962 &req->bu_threshold.b_time, ==))
1963 && (x->bm_threshold.b_packets
1964 == req->bu_threshold.b_packets)
1965 && (x->bm_threshold.b_bytes
1966 == req->bu_threshold.b_bytes)
1967 && (x->bm_flags & BW_METER_USER_FLAGS)
1968 == flags) {
1969 MRW_WUNLOCK();
1970 return 0; /* XXX Already installed */
1971 }
1972 }
1973
1974 /* Allocate the new bw_meter entry */
1975 x = malloc(sizeof(*x), M_BWMETER, M_ZERO | M_NOWAIT);
1976 if (x == NULL) {
1977 MRW_WUNLOCK();
1978 return ENOBUFS;
1979 }
1980
1981 /* Set the new bw_meter entry */
1982 x->bm_threshold.b_time = req->bu_threshold.b_time;
1983 microtime(&now);
1984 x->bm_start_time = now;
1985 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1986 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1987 x->bm_measured.b_packets = 0;
1988 x->bm_measured.b_bytes = 0;
1989 x->bm_flags = flags;
1990 x->bm_time_next = NULL;
1991 x->bm_mfc = mfc;
1992 x->arg = curvnet;
1993 sprintf(x->bm_mtx_name, "BM mtx %p", x);
1994 mtx_init(&x->bm_mtx, x->bm_mtx_name, NULL, MTX_DEF);
1995
1996 /* For LEQ case create periodic callout */
1997 if (req->bu_flags & BW_UPCALL_LEQ) {
1998 callout_init_rw(&x->bm_meter_callout, &mrouter_lock, CALLOUT_SHAREDLOCK);
1999 callout_reset(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time),
2000 expire_bw_meter_leq, x);
2001 }
2002
2003 /* Add the new bw_meter entry to the front of entries for this MFC */
2004 x->bm_mfc_next = *bwm_ptr;
2005 *bwm_ptr = x;
2006
2007 MRW_WUNLOCK();
2008
2009 return 0;
2010 }
2011
2012 static void
free_bw_list(struct bw_meter * list)2013 free_bw_list(struct bw_meter *list)
2014 {
2015 while (list != NULL) {
2016 struct bw_meter *x = list;
2017
2018 /* MRW_WLOCK must be held here */
2019 if (x->bm_flags & BW_METER_LEQ) {
2020 callout_drain(&x->bm_meter_callout);
2021 mtx_destroy(&x->bm_mtx);
2022 }
2023
2024 list = list->bm_mfc_next;
2025 free(x, M_BWMETER);
2026 }
2027 }
2028
2029 /*
2030 * Delete one or multiple bw_meter entries
2031 */
2032 static int
del_bw_upcall(struct bw_upcall * req)2033 del_bw_upcall(struct bw_upcall *req)
2034 {
2035 struct mfc *mfc;
2036 struct bw_meter *x, **bwm_ptr;
2037
2038 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
2039 return EOPNOTSUPP;
2040
2041 MRW_WLOCK();
2042
2043 /* Find the corresponding MFC entry */
2044 mfc = mfc_find(&req->bu_src, &req->bu_dst);
2045 if (mfc == NULL) {
2046 MRW_WUNLOCK();
2047 return EADDRNOTAVAIL;
2048 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2049 /*
2050 * Delete all bw_meter entries for this mfc
2051 */
2052 struct bw_meter *list;
2053
2054 /* Free LEQ list */
2055 list = mfc->mfc_bw_meter_leq;
2056 mfc->mfc_bw_meter_leq = NULL;
2057 free_bw_list(list);
2058
2059 /* Free GEQ list */
2060 list = mfc->mfc_bw_meter_geq;
2061 mfc->mfc_bw_meter_geq = NULL;
2062 free_bw_list(list);
2063 MRW_WUNLOCK();
2064 return 0;
2065 } else { /* Delete a single bw_meter entry */
2066 struct bw_meter *prev;
2067 uint32_t flags = 0;
2068
2069 flags = compute_bw_meter_flags(req);
2070
2071 /* Choose an appropriate bw_meter list */
2072 if (req->bu_flags & BW_UPCALL_GEQ)
2073 bwm_ptr = &mfc->mfc_bw_meter_geq;
2074 else
2075 bwm_ptr = &mfc->mfc_bw_meter_leq;
2076
2077 /* Find the bw_meter entry to delete */
2078 for (prev = NULL, x = *bwm_ptr; x != NULL;
2079 prev = x, x = x->bm_mfc_next) {
2080 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, &req->bu_threshold.b_time, ==)) &&
2081 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2082 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2083 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2084 break;
2085 }
2086 if (x != NULL) { /* Delete entry from the list for this MFC */
2087 if (prev != NULL)
2088 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
2089 else
2090 *bwm_ptr = x->bm_mfc_next;/* new head of list */
2091
2092 if (req->bu_flags & BW_UPCALL_LEQ)
2093 callout_stop(&x->bm_meter_callout);
2094
2095 MRW_WUNLOCK();
2096 /* Free the bw_meter entry */
2097 free(x, M_BWMETER);
2098 return 0;
2099 } else {
2100 MRW_WUNLOCK();
2101 return EINVAL;
2102 }
2103 }
2104 __assert_unreachable();
2105 }
2106
2107 /*
2108 * Perform bandwidth measurement processing that may result in an upcall
2109 */
2110 static void
bw_meter_geq_receive_packet(struct bw_meter * x,int plen,struct timeval * nowp)2111 bw_meter_geq_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2112 {
2113 struct timeval delta;
2114
2115 MRW_LOCK_ASSERT();
2116
2117 delta = *nowp;
2118 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2119
2120 /*
2121 * Processing for ">=" type of bw_meter entry.
2122 * bm_mtx does not have to be hold here as in GEQ
2123 * case this is the only context accessing bm_measured.
2124 */
2125 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2126 /* Reset the bw_meter entry */
2127 x->bm_start_time = *nowp;
2128 x->bm_measured.b_packets = 0;
2129 x->bm_measured.b_bytes = 0;
2130 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2131 }
2132
2133 /* Record that a packet is received */
2134 x->bm_measured.b_packets++;
2135 x->bm_measured.b_bytes += plen;
2136
2137 /*
2138 * Test if we should deliver an upcall
2139 */
2140 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2141 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2142 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2143 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2144 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2145 /* Prepare an upcall for delivery */
2146 bw_meter_prepare_upcall(x, nowp);
2147 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2148 }
2149 }
2150 }
2151
2152 /*
2153 * Prepare a bandwidth-related upcall
2154 */
2155 static void
bw_meter_prepare_upcall(struct bw_meter * x,struct timeval * nowp)2156 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2157 {
2158 struct timeval delta;
2159 struct bw_upcall *u;
2160
2161 MRW_LOCK_ASSERT();
2162
2163 /*
2164 * Compute the measured time interval
2165 */
2166 delta = *nowp;
2167 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2168
2169 /*
2170 * Set the bw_upcall entry
2171 */
2172 u = malloc(sizeof(struct bw_upcall), M_MRTABLE, M_NOWAIT | M_ZERO);
2173 if (!u) {
2174 log(LOG_WARNING, "bw_meter_prepare_upcall: cannot allocate entry\n");
2175 return;
2176 }
2177 u->bu_src = x->bm_mfc->mfc_origin;
2178 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2179 u->bu_threshold.b_time = x->bm_threshold.b_time;
2180 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2181 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2182 u->bu_measured.b_time = delta;
2183 u->bu_measured.b_packets = x->bm_measured.b_packets;
2184 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2185 u->bu_flags = 0;
2186 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2187 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2188 if (x->bm_flags & BW_METER_UNIT_BYTES)
2189 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2190 if (x->bm_flags & BW_METER_GEQ)
2191 u->bu_flags |= BW_UPCALL_GEQ;
2192 if (x->bm_flags & BW_METER_LEQ)
2193 u->bu_flags |= BW_UPCALL_LEQ;
2194
2195 if (buf_ring_enqueue(V_bw_upcalls_ring, u))
2196 log(LOG_WARNING, "bw_meter_prepare_upcall: cannot enqueue upcall\n");
2197 if (buf_ring_count(V_bw_upcalls_ring) > (BW_UPCALLS_MAX / 2)) {
2198 taskqueue_enqueue(V_task_queue, &V_task);
2199 }
2200 }
2201 /*
2202 * Send the pending bandwidth-related upcalls
2203 */
2204 static void
bw_upcalls_send(void)2205 bw_upcalls_send(void)
2206 {
2207 struct mbuf *m;
2208 int len = 0;
2209 struct bw_upcall *bu;
2210 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2211 static struct igmpmsg igmpmsg = {
2212 0, /* unused1 */
2213 0, /* unused2 */
2214 IGMPMSG_BW_UPCALL,/* im_msgtype */
2215 0, /* im_mbz */
2216 0, /* im_vif */
2217 0, /* unused3 */
2218 { 0 }, /* im_src */
2219 { 0 } /* im_dst */
2220 };
2221
2222 MRW_LOCK_ASSERT();
2223
2224 if (buf_ring_empty(V_bw_upcalls_ring))
2225 return;
2226
2227 /*
2228 * Allocate a new mbuf, initialize it with the header and
2229 * the payload for the pending calls.
2230 */
2231 m = m_gethdr(M_NOWAIT, MT_DATA);
2232 if (m == NULL) {
2233 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2234 return;
2235 }
2236
2237 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2238 len += sizeof(struct igmpmsg);
2239 while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
2240 m_copyback(m, len, sizeof(struct bw_upcall), (caddr_t)bu);
2241 len += sizeof(struct bw_upcall);
2242 free(bu, M_MRTABLE);
2243 }
2244
2245 /*
2246 * Send the upcalls
2247 * XXX do we need to set the address in k_igmpsrc ?
2248 */
2249 MRTSTAT_INC(mrts_upcalls);
2250 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2251 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2252 MRTSTAT_INC(mrts_upq_sockfull);
2253 }
2254 }
2255
2256 /*
2257 * A periodic function for sending all upcalls that are pending delivery
2258 */
2259 static void
expire_bw_upcalls_send(void * arg)2260 expire_bw_upcalls_send(void *arg)
2261 {
2262 CURVNET_SET((struct vnet *) arg);
2263
2264 /* This callout is run with MRW_RLOCK taken */
2265
2266 bw_upcalls_send();
2267
2268 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2269 curvnet);
2270 CURVNET_RESTORE();
2271 }
2272
2273 /*
2274 * End of bandwidth monitoring code
2275 */
2276
2277 /*
2278 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2279 *
2280 */
2281 static int
pim_register_send(struct ip * ip,struct vif * vifp,struct mbuf * m,struct mfc * rt)2282 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2283 struct mfc *rt)
2284 {
2285 struct mbuf *mb_copy, *mm;
2286
2287 /*
2288 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2289 * rendezvous point was unspecified, and we were told not to.
2290 */
2291 if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2292 in_nullhost(rt->mfc_rp))
2293 return 0;
2294
2295 mb_copy = pim_register_prepare(ip, m);
2296 if (mb_copy == NULL)
2297 return ENOBUFS;
2298
2299 /*
2300 * Send all the fragments. Note that the mbuf for each fragment
2301 * is freed by the sending machinery.
2302 */
2303 for (mm = mb_copy; mm; mm = mb_copy) {
2304 mb_copy = mm->m_nextpkt;
2305 mm->m_nextpkt = 0;
2306 mm = m_pullup(mm, sizeof(struct ip));
2307 if (mm != NULL) {
2308 ip = mtod(mm, struct ip *);
2309 if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2310 pim_register_send_rp(ip, vifp, mm, rt);
2311 } else {
2312 pim_register_send_upcall(ip, vifp, mm, rt);
2313 }
2314 }
2315 }
2316
2317 return 0;
2318 }
2319
2320 /*
2321 * Return a copy of the data packet that is ready for PIM Register
2322 * encapsulation.
2323 * XXX: Note that in the returned copy the IP header is a valid one.
2324 */
2325 static struct mbuf *
pim_register_prepare(struct ip * ip,struct mbuf * m)2326 pim_register_prepare(struct ip *ip, struct mbuf *m)
2327 {
2328 struct mbuf *mb_copy = NULL;
2329 int mtu;
2330
2331 /* Take care of delayed checksums */
2332 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2333 in_delayed_cksum(m);
2334 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2335 }
2336
2337 /*
2338 * Copy the old packet & pullup its IP header into the
2339 * new mbuf so we can modify it.
2340 */
2341 mb_copy = m_copypacket(m, M_NOWAIT);
2342 if (mb_copy == NULL)
2343 return NULL;
2344 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2345 if (mb_copy == NULL)
2346 return NULL;
2347
2348 /* take care of the TTL */
2349 ip = mtod(mb_copy, struct ip *);
2350 --ip->ip_ttl;
2351
2352 /* Compute the MTU after the PIM Register encapsulation */
2353 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2354
2355 if (ntohs(ip->ip_len) <= mtu) {
2356 /* Turn the IP header into a valid one */
2357 ip->ip_sum = 0;
2358 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2359 } else {
2360 /* Fragment the packet */
2361 mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2362 if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2363 m_freem(mb_copy);
2364 return NULL;
2365 }
2366 }
2367 return mb_copy;
2368 }
2369
2370 /*
2371 * Send an upcall with the data packet to the user-level process.
2372 */
2373 static int
pim_register_send_upcall(struct ip * ip,struct vif * vifp,struct mbuf * mb_copy,struct mfc * rt)2374 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2375 struct mbuf *mb_copy, struct mfc *rt)
2376 {
2377 struct mbuf *mb_first;
2378 int len = ntohs(ip->ip_len);
2379 struct igmpmsg *im;
2380 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2381
2382 MRW_LOCK_ASSERT();
2383
2384 /*
2385 * Add a new mbuf with an upcall header
2386 */
2387 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2388 if (mb_first == NULL) {
2389 m_freem(mb_copy);
2390 return ENOBUFS;
2391 }
2392 mb_first->m_data += max_linkhdr;
2393 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2394 mb_first->m_len = sizeof(struct igmpmsg);
2395 mb_first->m_next = mb_copy;
2396
2397 /* Send message to routing daemon */
2398 im = mtod(mb_first, struct igmpmsg *);
2399 im->im_msgtype = IGMPMSG_WHOLEPKT;
2400 im->im_mbz = 0;
2401 im->im_vif = vifp - V_viftable;
2402 im->im_src = ip->ip_src;
2403 im->im_dst = ip->ip_dst;
2404
2405 k_igmpsrc.sin_addr = ip->ip_src;
2406
2407 MRTSTAT_INC(mrts_upcalls);
2408
2409 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2410 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2411 MRTSTAT_INC(mrts_upq_sockfull);
2412 return ENOBUFS;
2413 }
2414
2415 /* Keep statistics */
2416 PIMSTAT_INC(pims_snd_registers_msgs);
2417 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2418
2419 return 0;
2420 }
2421
2422 /*
2423 * Encapsulate the data packet in PIM Register message and send it to the RP.
2424 */
2425 static int
pim_register_send_rp(struct ip * ip,struct vif * vifp,struct mbuf * mb_copy,struct mfc * rt)2426 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2427 struct mfc *rt)
2428 {
2429 struct mbuf *mb_first;
2430 struct ip *ip_outer;
2431 struct pim_encap_pimhdr *pimhdr;
2432 int len = ntohs(ip->ip_len);
2433 vifi_t vifi = rt->mfc_parent;
2434
2435 MRW_LOCK_ASSERT();
2436
2437 if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2438 m_freem(mb_copy);
2439 return EADDRNOTAVAIL; /* The iif vif is invalid */
2440 }
2441
2442 /*
2443 * Add a new mbuf with the encapsulating header
2444 */
2445 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2446 if (mb_first == NULL) {
2447 m_freem(mb_copy);
2448 return ENOBUFS;
2449 }
2450 mb_first->m_data += max_linkhdr;
2451 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2452 mb_first->m_next = mb_copy;
2453
2454 mb_first->m_pkthdr.len = len + mb_first->m_len;
2455
2456 /*
2457 * Fill in the encapsulating IP and PIM header
2458 */
2459 ip_outer = mtod(mb_first, struct ip *);
2460 *ip_outer = pim_encap_iphdr;
2461 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2462 sizeof(pim_encap_pimhdr));
2463 ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2464 ip_outer->ip_dst = rt->mfc_rp;
2465 /*
2466 * Copy the inner header TOS to the outer header, and take care of the
2467 * IP_DF bit.
2468 */
2469 ip_outer->ip_tos = ip->ip_tos;
2470 if (ip->ip_off & htons(IP_DF))
2471 ip_outer->ip_off |= htons(IP_DF);
2472 ip_fillid(ip_outer, V_ip_random_id);
2473 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2474 + sizeof(pim_encap_iphdr));
2475 *pimhdr = pim_encap_pimhdr;
2476 /* If the iif crosses a border, set the Border-bit */
2477 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2478 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2479
2480 mb_first->m_data += sizeof(pim_encap_iphdr);
2481 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2482 mb_first->m_data -= sizeof(pim_encap_iphdr);
2483
2484 send_packet(vifp, mb_first);
2485
2486 /* Keep statistics */
2487 PIMSTAT_INC(pims_snd_registers_msgs);
2488 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2489
2490 return 0;
2491 }
2492
2493 /*
2494 * pim_encapcheck() is called by the encap4_input() path at runtime to
2495 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2496 * into the kernel.
2497 */
2498 static int
pim_encapcheck(const struct mbuf * m __unused,int off __unused,int proto __unused,void * arg __unused)2499 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2500 int proto __unused, void *arg __unused)
2501 {
2502
2503 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2504 return (8); /* claim the datagram. */
2505 }
2506
2507 /*
2508 * PIM-SMv2 and PIM-DM messages processing.
2509 * Receives and verifies the PIM control messages, and passes them
2510 * up to the listening socket, using rip_input().
2511 * The only message with special processing is the PIM_REGISTER message
2512 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2513 * is passed to if_simloop().
2514 */
2515 static int
pim_input(struct mbuf * m,int off,int proto,void * arg __unused)2516 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2517 {
2518 struct ip *ip = mtod(m, struct ip *);
2519 struct pim *pim;
2520 int iphlen = off;
2521 int minlen;
2522 int datalen = ntohs(ip->ip_len) - iphlen;
2523 int ip_tos;
2524
2525 /* Keep statistics */
2526 PIMSTAT_INC(pims_rcv_total_msgs);
2527 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2528
2529 /*
2530 * Validate lengths
2531 */
2532 if (datalen < PIM_MINLEN) {
2533 PIMSTAT_INC(pims_rcv_tooshort);
2534 CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2535 __func__, datalen, ntohl(ip->ip_src.s_addr));
2536 m_freem(m);
2537 return (IPPROTO_DONE);
2538 }
2539
2540 /*
2541 * If the packet is at least as big as a REGISTER, go agead
2542 * and grab the PIM REGISTER header size, to avoid another
2543 * possible m_pullup() later.
2544 *
2545 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2546 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2547 */
2548 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2549 /*
2550 * Get the IP and PIM headers in contiguous memory, and
2551 * possibly the PIM REGISTER header.
2552 */
2553 if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2554 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2555 return (IPPROTO_DONE);
2556 }
2557
2558 /* m_pullup() may have given us a new mbuf so reset ip. */
2559 ip = mtod(m, struct ip *);
2560 ip_tos = ip->ip_tos;
2561
2562 /* adjust mbuf to point to the PIM header */
2563 m->m_data += iphlen;
2564 m->m_len -= iphlen;
2565 pim = mtod(m, struct pim *);
2566
2567 /*
2568 * Validate checksum. If PIM REGISTER, exclude the data packet.
2569 *
2570 * XXX: some older PIMv2 implementations don't make this distinction,
2571 * so for compatibility reason perform the checksum over part of the
2572 * message, and if error, then over the whole message.
2573 */
2574 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2575 /* do nothing, checksum okay */
2576 } else if (in_cksum(m, datalen)) {
2577 PIMSTAT_INC(pims_rcv_badsum);
2578 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2579 m_freem(m);
2580 return (IPPROTO_DONE);
2581 }
2582
2583 /* PIM version check */
2584 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2585 PIMSTAT_INC(pims_rcv_badversion);
2586 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2587 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2588 m_freem(m);
2589 return (IPPROTO_DONE);
2590 }
2591
2592 /* restore mbuf back to the outer IP */
2593 m->m_data -= iphlen;
2594 m->m_len += iphlen;
2595
2596 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2597 /*
2598 * Since this is a REGISTER, we'll make a copy of the register
2599 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2600 * routing daemon.
2601 */
2602 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2603 struct mbuf *mcp;
2604 struct ip *encap_ip;
2605 u_int32_t *reghdr;
2606 struct ifnet *vifp;
2607
2608 MRW_RLOCK();
2609 if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2610 MRW_RUNLOCK();
2611 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2612 (int)V_reg_vif_num);
2613 m_freem(m);
2614 return (IPPROTO_DONE);
2615 }
2616 /* XXX need refcnt? */
2617 vifp = V_viftable[V_reg_vif_num].v_ifp;
2618 MRW_RUNLOCK();
2619
2620 /*
2621 * Validate length
2622 */
2623 if (datalen < PIM_REG_MINLEN) {
2624 PIMSTAT_INC(pims_rcv_tooshort);
2625 PIMSTAT_INC(pims_rcv_badregisters);
2626 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2627 m_freem(m);
2628 return (IPPROTO_DONE);
2629 }
2630
2631 reghdr = (u_int32_t *)(pim + 1);
2632 encap_ip = (struct ip *)(reghdr + 1);
2633
2634 CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2635 __func__, ntohl(encap_ip->ip_src.s_addr),
2636 ntohs(encap_ip->ip_len));
2637
2638 /* verify the version number of the inner packet */
2639 if (encap_ip->ip_v != IPVERSION) {
2640 PIMSTAT_INC(pims_rcv_badregisters);
2641 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2642 m_freem(m);
2643 return (IPPROTO_DONE);
2644 }
2645
2646 /* verify the inner packet is destined to a mcast group */
2647 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2648 PIMSTAT_INC(pims_rcv_badregisters);
2649 CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2650 ntohl(encap_ip->ip_dst.s_addr));
2651 m_freem(m);
2652 return (IPPROTO_DONE);
2653 }
2654
2655 /* If a NULL_REGISTER, pass it to the daemon */
2656 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2657 goto pim_input_to_daemon;
2658
2659 /*
2660 * Copy the TOS from the outer IP header to the inner IP header.
2661 */
2662 if (encap_ip->ip_tos != ip_tos) {
2663 /* Outer TOS -> inner TOS */
2664 encap_ip->ip_tos = ip_tos;
2665 /* Recompute the inner header checksum. Sigh... */
2666
2667 /* adjust mbuf to point to the inner IP header */
2668 m->m_data += (iphlen + PIM_MINLEN);
2669 m->m_len -= (iphlen + PIM_MINLEN);
2670
2671 encap_ip->ip_sum = 0;
2672 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2673
2674 /* restore mbuf to point back to the outer IP header */
2675 m->m_data -= (iphlen + PIM_MINLEN);
2676 m->m_len += (iphlen + PIM_MINLEN);
2677 }
2678
2679 /*
2680 * Decapsulate the inner IP packet and loopback to forward it
2681 * as a normal multicast packet. Also, make a copy of the
2682 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2683 * to pass to the daemon later, so it can take the appropriate
2684 * actions (e.g., send back PIM_REGISTER_STOP).
2685 * XXX: here m->m_data points to the outer IP header.
2686 */
2687 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2688 if (mcp == NULL) {
2689 CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2690 m_freem(m);
2691 return (IPPROTO_DONE);
2692 }
2693
2694 /* Keep statistics */
2695 /* XXX: registers_bytes include only the encap. mcast pkt */
2696 PIMSTAT_INC(pims_rcv_registers_msgs);
2697 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2698
2699 /*
2700 * forward the inner ip packet; point m_data at the inner ip.
2701 */
2702 m_adj(m, iphlen + PIM_MINLEN);
2703
2704 CTR4(KTR_IPMF,
2705 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2706 __func__,
2707 (u_long)ntohl(encap_ip->ip_src.s_addr),
2708 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2709 (int)V_reg_vif_num);
2710
2711 /* NB: vifp was collected above; can it change on us? */
2712 if_simloop(vifp, m, dst.sin_family, 0);
2713
2714 /* prepare the register head to send to the mrouting daemon */
2715 m = mcp;
2716 }
2717
2718 pim_input_to_daemon:
2719 /*
2720 * Pass the PIM message up to the daemon; if it is a Register message,
2721 * pass the 'head' only up to the daemon. This includes the
2722 * outer IP header, PIM header, PIM-Register header and the
2723 * inner IP header.
2724 * XXX: the outer IP header pkt size of a Register is not adjust to
2725 * reflect the fact that the inner multicast data is truncated.
2726 */
2727 return (rip_input(&m, &off, proto));
2728 }
2729
2730 static int
sysctl_mfctable(SYSCTL_HANDLER_ARGS)2731 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2732 {
2733 struct mfc *rt;
2734 int error, i;
2735
2736 if (req->newptr)
2737 return (EPERM);
2738 if (V_mfchashtbl == NULL) /* XXX unlocked */
2739 return (0);
2740 error = sysctl_wire_old_buffer(req, 0);
2741 if (error)
2742 return (error);
2743
2744 MRW_RLOCK();
2745 if (V_mfchashtbl == NULL)
2746 goto out_locked;
2747
2748 for (i = 0; i < mfchashsize; i++) {
2749 LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2750 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2751 if (error)
2752 goto out_locked;
2753 }
2754 }
2755 out_locked:
2756 MRW_RUNLOCK();
2757 return (error);
2758 }
2759
2760 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2761 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2762 "IPv4 Multicast Forwarding Table "
2763 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2764
2765 static int
sysctl_viflist(SYSCTL_HANDLER_ARGS)2766 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2767 {
2768 int error, i;
2769
2770 if (req->newptr)
2771 return (EPERM);
2772 if (V_viftable == NULL) /* XXX unlocked */
2773 return (0);
2774 error = sysctl_wire_old_buffer(req, MROUTE_VIF_SYSCTL_LEN * MAXVIFS);
2775 if (error)
2776 return (error);
2777
2778 MRW_RLOCK();
2779 /* Copy out user-visible portion of vif entry. */
2780 for (i = 0; i < MAXVIFS; i++) {
2781 error = SYSCTL_OUT(req, &V_viftable[i], MROUTE_VIF_SYSCTL_LEN);
2782 if (error)
2783 break;
2784 }
2785 MRW_RUNLOCK();
2786 return (error);
2787 }
2788
2789 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2790 CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2791 sysctl_viflist, "S,vif[MAXVIFS]",
2792 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2793
2794 static void
vnet_mroute_init(const void * unused __unused)2795 vnet_mroute_init(const void *unused __unused)
2796 {
2797
2798 V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2799
2800 V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2801 M_MRTABLE, M_WAITOK|M_ZERO);
2802
2803 callout_init_rw(&V_expire_upcalls_ch, &mrouter_lock, 0);
2804 callout_init_rw(&V_bw_upcalls_ch, &mrouter_lock, 0);
2805
2806 /* Prepare taskqueue */
2807 V_task_queue = taskqueue_create_fast("ip_mroute_tskq", M_NOWAIT,
2808 taskqueue_thread_enqueue, &V_task_queue);
2809 taskqueue_start_threads(&V_task_queue, 1, PI_NET, "ip_mroute_tskq task");
2810 }
2811
2812 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2813 NULL);
2814
2815 static void
vnet_mroute_uninit(const void * unused __unused)2816 vnet_mroute_uninit(const void *unused __unused)
2817 {
2818
2819 /* Taskqueue should be cancelled and drained before freeing */
2820 taskqueue_free(V_task_queue);
2821
2822 free(V_viftable, M_MRTABLE);
2823 free(V_nexpire, M_MRTABLE);
2824 V_nexpire = NULL;
2825 }
2826
2827 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2828 vnet_mroute_uninit, NULL);
2829
2830 static int
ip_mroute_modevent(module_t mod,int type,void * unused)2831 ip_mroute_modevent(module_t mod, int type, void *unused)
2832 {
2833
2834 switch (type) {
2835 case MOD_LOAD:
2836 MRW_TEARDOWN_LOCK_INIT();
2837 MRW_LOCK_INIT();
2838
2839 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2840 if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2841
2842 if (!powerof2(mfchashsize)) {
2843 printf("WARNING: %s not a power of 2; using default\n",
2844 "net.inet.ip.mfchashsize");
2845 mfchashsize = MFCHASHSIZE;
2846 }
2847
2848 pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2849
2850 ip_mcast_src = X_ip_mcast_src;
2851 ip_mforward = X_ip_mforward;
2852 ip_mrouter_done = X_ip_mrouter_done;
2853 ip_mrouter_get = X_ip_mrouter_get;
2854 ip_mrouter_set = X_ip_mrouter_set;
2855
2856 ip_rsvp_force_done = X_ip_rsvp_force_done;
2857 ip_rsvp_vif = X_ip_rsvp_vif;
2858
2859 legal_vif_num = X_legal_vif_num;
2860 mrt_ioctl = X_mrt_ioctl;
2861 rsvp_input_p = X_rsvp_input;
2862 break;
2863
2864 case MOD_UNLOAD:
2865 /*
2866 * Typically module unload happens after the user-level
2867 * process has shutdown the kernel services (the check
2868 * below insures someone can't just yank the module out
2869 * from under a running process). But if the module is
2870 * just loaded and then unloaded w/o starting up a user
2871 * process we still need to cleanup.
2872 */
2873 MRW_WLOCK();
2874 if (ip_mrouter_cnt != 0) {
2875 MRW_WUNLOCK();
2876 return (EBUSY);
2877 }
2878 ip_mrouter_unloading = 1;
2879 MRW_WUNLOCK();
2880
2881 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2882
2883 if (pim_encap_cookie) {
2884 ip_encap_detach(pim_encap_cookie);
2885 pim_encap_cookie = NULL;
2886 }
2887
2888 ip_mcast_src = NULL;
2889 ip_mforward = NULL;
2890 ip_mrouter_done = NULL;
2891 ip_mrouter_get = NULL;
2892 ip_mrouter_set = NULL;
2893
2894 ip_rsvp_force_done = NULL;
2895 ip_rsvp_vif = NULL;
2896
2897 legal_vif_num = NULL;
2898 mrt_ioctl = NULL;
2899 rsvp_input_p = NULL;
2900
2901 MRW_LOCK_DESTROY();
2902 MRW_TEARDOWN_LOCK_DESTROY();
2903 break;
2904
2905 default:
2906 return EOPNOTSUPP;
2907 }
2908 return 0;
2909 }
2910
2911 static moduledata_t ip_mroutemod = {
2912 "ip_mroute",
2913 ip_mroute_modevent,
2914 0
2915 };
2916
2917 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2918