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(u_long, mfchash);
173 #define V_mfchash VNET(mfchash)
174 #define MFCHASH(a, g) \
175 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
176 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
177 #define MFCHASHSIZE 256
178
179 static u_long mfchashsize = MFCHASHSIZE; /* Hash size */
180 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, mfchashsize, CTLFLAG_RDTUN,
181 &mfchashsize, 0, "IPv4 Multicast Forwarding Table hash size");
182 VNET_DEFINE_STATIC(u_char *, nexpire); /* 0..mfchashsize-1 */
183 #define V_nexpire VNET(nexpire)
184 VNET_DEFINE_STATIC(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
185 #define V_mfchashtbl VNET(mfchashtbl)
186 VNET_DEFINE_STATIC(struct taskqueue *, task_queue);
187 #define V_task_queue VNET(task_queue)
188 VNET_DEFINE_STATIC(struct task, task);
189 #define V_task VNET(task)
190
191 VNET_DEFINE_STATIC(vifi_t, numvifs);
192 #define V_numvifs VNET(numvifs)
193 VNET_DEFINE_STATIC(struct vif *, viftable);
194 #define V_viftable VNET(viftable)
195
196 static eventhandler_tag if_detach_event_tag = NULL;
197
198 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
199 #define V_expire_upcalls_ch VNET(expire_upcalls_ch)
200
201 VNET_DEFINE_STATIC(struct mtx, buf_ring_mtx);
202 #define V_buf_ring_mtx VNET(buf_ring_mtx)
203
204 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
205 #define UPCALL_EXPIRE 6 /* number of timeouts */
206
207 /*
208 * Bandwidth meter variables and constants
209 */
210 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
211
212 /*
213 * Pending upcalls are stored in a ring which is flushed when
214 * full, or periodically
215 */
216 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
217 #define V_bw_upcalls_ch VNET(bw_upcalls_ch)
218 VNET_DEFINE_STATIC(struct buf_ring *, bw_upcalls_ring);
219 #define V_bw_upcalls_ring VNET(bw_upcalls_ring)
220 VNET_DEFINE_STATIC(struct mtx, bw_upcalls_ring_mtx);
221 #define V_bw_upcalls_ring_mtx VNET(bw_upcalls_ring_mtx)
222
223 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
224
225 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
226 VNET_PCPUSTAT_SYSINIT(pimstat);
227 VNET_PCPUSTAT_SYSUNINIT(pimstat);
228
229 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
230 "PIM");
231 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
232 pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
233
234 static u_long pim_squelch_wholepkt = 0;
235 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RWTUN,
236 &pim_squelch_wholepkt, 0,
237 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
238
239 static const struct encaptab *pim_encap_cookie;
240 static int pim_encapcheck(const struct mbuf *, int, int, void *);
241 static int pim_input(struct mbuf *, int, int, void *);
242
243 extern int in_mcast_loop;
244
245 static const struct encap_config ipv4_encap_cfg = {
246 .proto = IPPROTO_PIM,
247 .min_length = sizeof(struct ip) + PIM_MINLEN,
248 .exact_match = 8,
249 .check = pim_encapcheck,
250 .input = pim_input
251 };
252
253 /*
254 * Note: the PIM Register encapsulation adds the following in front of a
255 * data packet:
256 *
257 * struct pim_encap_hdr {
258 * struct ip ip;
259 * struct pim_encap_pimhdr pim;
260 * }
261 *
262 */
263
264 struct pim_encap_pimhdr {
265 struct pim pim;
266 uint32_t flags;
267 };
268 #define PIM_ENCAP_TTL 64
269
270 static struct ip pim_encap_iphdr = {
271 #if BYTE_ORDER == LITTLE_ENDIAN
272 sizeof(struct ip) >> 2,
273 IPVERSION,
274 #else
275 IPVERSION,
276 sizeof(struct ip) >> 2,
277 #endif
278 0, /* tos */
279 sizeof(struct ip), /* total length */
280 0, /* id */
281 0, /* frag offset */
282 PIM_ENCAP_TTL,
283 IPPROTO_PIM,
284 0, /* checksum */
285 };
286
287 static struct pim_encap_pimhdr pim_encap_pimhdr = {
288 {
289 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
290 0, /* reserved */
291 0, /* checksum */
292 },
293 0 /* flags */
294 };
295
296 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
297 #define V_reg_vif_num VNET(reg_vif_num)
298 VNET_DEFINE_STATIC(struct ifnet *, multicast_register_if);
299 #define V_multicast_register_if VNET(multicast_register_if)
300
301 /*
302 * Private variables.
303 */
304
305 static u_long X_ip_mcast_src(int);
306 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
307 struct ip_moptions *);
308 static int X_ip_mrouter_done(void);
309 static int X_ip_mrouter_get(struct socket *, struct sockopt *);
310 static int X_ip_mrouter_set(struct socket *, struct sockopt *);
311 static int X_legal_vif_num(int);
312 static int X_mrt_ioctl(u_long, caddr_t, int);
313
314 static int add_bw_upcall(struct bw_upcall *);
315 static int add_mfc(struct mfcctl2 *);
316 static int add_vif(struct vifctl *);
317 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
318 static void bw_meter_geq_receive_packet(struct bw_meter *, int,
319 struct timeval *);
320 static void bw_upcalls_send(void);
321 static int del_bw_upcall(struct bw_upcall *);
322 static int del_mfc(struct mfcctl2 *);
323 static int del_vif(vifi_t);
324 static int del_vif_locked(vifi_t, struct ifnet **, struct ifnet **);
325 static void expire_bw_upcalls_send(void *);
326 static void expire_mfc(struct mfc *);
327 static void expire_upcalls(void *);
328 static void free_bw_list(struct bw_meter *);
329 static int get_sg_cnt(struct sioc_sg_req *);
330 static int get_vif_cnt(struct sioc_vif_req *);
331 static void if_detached_event(void *, struct ifnet *);
332 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
333 static int ip_mrouter_init(struct socket *, int);
334 static __inline struct mfc *
335 mfc_find(struct in_addr *, struct in_addr *);
336 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
337 static struct mbuf *
338 pim_register_prepare(struct ip *, struct mbuf *);
339 static int pim_register_send(struct ip *, struct vif *,
340 struct mbuf *, struct mfc *);
341 static int pim_register_send_rp(struct ip *, struct vif *,
342 struct mbuf *, struct mfc *);
343 static int pim_register_send_upcall(struct ip *, struct vif *,
344 struct mbuf *, struct mfc *);
345 static void send_packet(struct vif *, struct mbuf *);
346 static int set_api_config(uint32_t *);
347 static int set_assert(int);
348 static int socket_send(struct socket *, struct mbuf *,
349 struct sockaddr_in *);
350
351 /*
352 * Kernel multicast forwarding API capabilities and setup.
353 * If more API capabilities are added to the kernel, they should be
354 * recorded in `mrt_api_support'.
355 */
356 #define MRT_API_VERSION 0x0305
357
358 static const int mrt_api_version = MRT_API_VERSION;
359 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
360 MRT_MFC_FLAGS_BORDER_VIF |
361 MRT_MFC_RP |
362 MRT_MFC_BW_UPCALL);
363 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
364 #define V_mrt_api_config VNET(mrt_api_config)
365 VNET_DEFINE_STATIC(int, pim_assert_enabled);
366 #define V_pim_assert_enabled VNET(pim_assert_enabled)
367 static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */
368
369 /*
370 * Find a route for a given origin IP address and multicast group address.
371 * Statistics must be updated by the caller.
372 */
373 static __inline struct mfc *
mfc_find(struct in_addr * o,struct in_addr * g)374 mfc_find(struct in_addr *o, struct in_addr *g)
375 {
376 struct mfc *rt;
377
378 /*
379 * Might be called both RLOCK and WLOCK.
380 * Check if any, it's caller responsibility
381 * to choose correct option.
382 */
383 MRW_LOCK_ASSERT();
384
385 LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
386 if (in_hosteq(rt->mfc_origin, *o) &&
387 in_hosteq(rt->mfc_mcastgrp, *g) &&
388 buf_ring_empty(rt->mfc_stall_ring))
389 break;
390 }
391
392 return (rt);
393 }
394
395 static __inline struct mfc *
mfc_alloc(void)396 mfc_alloc(void)
397 {
398 struct mfc *rt;
399 rt = malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT | M_ZERO);
400 if (rt == NULL)
401 return rt;
402
403 rt->mfc_stall_ring = buf_ring_alloc(MAX_UPQ, M_MRTABLE,
404 M_NOWAIT, &V_buf_ring_mtx);
405 if (rt->mfc_stall_ring == NULL) {
406 free(rt, M_MRTABLE);
407 return NULL;
408 }
409
410 return rt;
411 }
412
413 /*
414 * Handle MRT setsockopt commands to modify the multicast forwarding tables.
415 */
416 static int
X_ip_mrouter_set(struct socket * so,struct sockopt * sopt)417 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
418 {
419 int error, optval;
420 vifi_t vifi;
421 struct vifctl vifc;
422 struct mfcctl2 mfc;
423 struct bw_upcall bw_upcall;
424 uint32_t i;
425
426 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
427 return EPERM;
428
429 error = 0;
430 switch (sopt->sopt_name) {
431 case MRT_INIT:
432 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
433 if (error)
434 break;
435 error = ip_mrouter_init(so, optval);
436 break;
437 case MRT_DONE:
438 error = ip_mrouter_done();
439 break;
440 case MRT_ADD_VIF:
441 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
442 if (error)
443 break;
444 error = add_vif(&vifc);
445 break;
446 case MRT_DEL_VIF:
447 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
448 if (error)
449 break;
450 error = del_vif(vifi);
451 break;
452 case MRT_ADD_MFC:
453 case MRT_DEL_MFC:
454 /*
455 * select data size depending on API version.
456 */
457 if (sopt->sopt_name == MRT_ADD_MFC &&
458 V_mrt_api_config & MRT_API_FLAGS_ALL) {
459 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
460 sizeof(struct mfcctl2));
461 } else {
462 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
463 sizeof(struct mfcctl));
464 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
465 sizeof(mfc) - sizeof(struct mfcctl));
466 }
467 if (error)
468 break;
469 if (sopt->sopt_name == MRT_ADD_MFC)
470 error = add_mfc(&mfc);
471 else
472 error = del_mfc(&mfc);
473 break;
474
475 case MRT_ASSERT:
476 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
477 if (error)
478 break;
479 set_assert(optval);
480 break;
481
482 case MRT_API_CONFIG:
483 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
484 if (!error)
485 error = set_api_config(&i);
486 if (!error)
487 error = sooptcopyout(sopt, &i, sizeof i);
488 break;
489
490 case MRT_ADD_BW_UPCALL:
491 case MRT_DEL_BW_UPCALL:
492 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
493 sizeof bw_upcall);
494 if (error)
495 break;
496 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
497 error = add_bw_upcall(&bw_upcall);
498 else
499 error = del_bw_upcall(&bw_upcall);
500 break;
501
502 default:
503 error = EOPNOTSUPP;
504 break;
505 }
506 return error;
507 }
508
509 /*
510 * Handle MRT getsockopt commands
511 */
512 static int
X_ip_mrouter_get(struct socket * so,struct sockopt * sopt)513 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
514 {
515 int error;
516
517 switch (sopt->sopt_name) {
518 case MRT_VERSION:
519 error = sooptcopyout(sopt, &mrt_api_version,
520 sizeof mrt_api_version);
521 break;
522 case MRT_ASSERT:
523 error = sooptcopyout(sopt, &V_pim_assert_enabled,
524 sizeof V_pim_assert_enabled);
525 break;
526 case MRT_API_SUPPORT:
527 error = sooptcopyout(sopt, &mrt_api_support,
528 sizeof mrt_api_support);
529 break;
530 case MRT_API_CONFIG:
531 error = sooptcopyout(sopt, &V_mrt_api_config,
532 sizeof V_mrt_api_config);
533 break;
534 default:
535 error = EOPNOTSUPP;
536 break;
537 }
538 return error;
539 }
540
541 /*
542 * Handle ioctl commands to obtain information from the cache
543 */
544 static int
X_mrt_ioctl(u_long cmd,caddr_t data,int fibnum __unused)545 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
546 {
547 int error;
548
549 /*
550 * Currently the only function calling this ioctl routine is rtioctl_fib().
551 * Typically, only root can create the raw socket in order to execute
552 * this ioctl method, however the request might be coming from a prison
553 */
554 error = priv_check(curthread, PRIV_NETINET_MROUTE);
555 if (error)
556 return (error);
557 switch (cmd) {
558 case (SIOCGETVIFCNT):
559 error = get_vif_cnt((struct sioc_vif_req *)data);
560 break;
561
562 case (SIOCGETSGCNT):
563 error = get_sg_cnt((struct sioc_sg_req *)data);
564 break;
565
566 default:
567 error = EINVAL;
568 break;
569 }
570 return error;
571 }
572
573 /*
574 * returns the packet, byte, rpf-failure count for the source group provided
575 */
576 static int
get_sg_cnt(struct sioc_sg_req * req)577 get_sg_cnt(struct sioc_sg_req *req)
578 {
579 struct mfc *rt;
580
581 MRW_RLOCK();
582 rt = mfc_find(&req->src, &req->grp);
583 if (rt == NULL) {
584 MRW_RUNLOCK();
585 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
586 return EADDRNOTAVAIL;
587 }
588 req->pktcnt = rt->mfc_pkt_cnt;
589 req->bytecnt = rt->mfc_byte_cnt;
590 req->wrong_if = rt->mfc_wrong_if;
591 MRW_RUNLOCK();
592 return 0;
593 }
594
595 /*
596 * returns the input and output packet and byte counts on the vif provided
597 */
598 static int
get_vif_cnt(struct sioc_vif_req * req)599 get_vif_cnt(struct sioc_vif_req *req)
600 {
601 vifi_t vifi = req->vifi;
602
603 MRW_RLOCK();
604 if (vifi >= V_numvifs) {
605 MRW_RUNLOCK();
606 return EINVAL;
607 }
608
609 mtx_lock_spin(&V_viftable[vifi].v_spin);
610 req->icount = V_viftable[vifi].v_pkt_in;
611 req->ocount = V_viftable[vifi].v_pkt_out;
612 req->ibytes = V_viftable[vifi].v_bytes_in;
613 req->obytes = V_viftable[vifi].v_bytes_out;
614 mtx_unlock_spin(&V_viftable[vifi].v_spin);
615 MRW_RUNLOCK();
616
617 return 0;
618 }
619
620 static void
if_detached_event(void * arg __unused,struct ifnet * ifp)621 if_detached_event(void *arg __unused, struct ifnet *ifp)
622 {
623 vifi_t vifi;
624 u_long i, vifi_cnt = 0;
625 struct ifnet *free_ptr, *multi_leave;
626
627 MRW_WLOCK();
628
629 if (V_ip_mrouter == NULL) {
630 MRW_WUNLOCK();
631 return;
632 }
633
634 /*
635 * Tear down multicast forwarder state associated with this ifnet.
636 * 1. Walk the vif list, matching vifs against this ifnet.
637 * 2. Walk the multicast forwarding cache (mfc) looking for
638 * inner matches with this vif's index.
639 * 3. Expire any matching multicast forwarding cache entries.
640 * 4. Free vif state. This should disable ALLMULTI on the interface.
641 */
642 restart:
643 for (vifi = 0; vifi < V_numvifs; vifi++) {
644 if (V_viftable[vifi].v_ifp != ifp)
645 continue;
646 for (i = 0; i < mfchashsize; i++) {
647 struct mfc *rt, *nrt;
648
649 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
650 if (rt->mfc_parent == vifi) {
651 expire_mfc(rt);
652 }
653 }
654 }
655 del_vif_locked(vifi, &multi_leave, &free_ptr);
656 if (free_ptr != NULL)
657 vifi_cnt++;
658 if (multi_leave) {
659 MRW_WUNLOCK();
660 if_allmulti(multi_leave, 0);
661 MRW_WLOCK();
662 goto restart;
663 }
664 }
665
666 MRW_WUNLOCK();
667
668 /*
669 * Free IFP. We don't have to use free_ptr here as it is the same
670 * that ifp. Perform free as many times as required in case
671 * refcount is greater than 1.
672 */
673 for (i = 0; i < vifi_cnt; i++)
674 if_free(ifp);
675 }
676
677 static void
ip_mrouter_upcall_thread(void * arg,int pending __unused)678 ip_mrouter_upcall_thread(void *arg, int pending __unused)
679 {
680 CURVNET_SET((struct vnet *) arg);
681
682 MRW_WLOCK();
683 bw_upcalls_send();
684 MRW_WUNLOCK();
685
686 CURVNET_RESTORE();
687 }
688
689 /*
690 * Enable multicast forwarding.
691 */
692 static int
ip_mrouter_init(struct socket * so,int version)693 ip_mrouter_init(struct socket *so, int version)
694 {
695
696 CTR2(KTR_IPMF, "%s: so %p", __func__, so);
697
698 if (version != 1)
699 return ENOPROTOOPT;
700
701 MRW_TEARDOWN_WLOCK();
702 MRW_WLOCK();
703
704 if (ip_mrouter_unloading) {
705 MRW_WUNLOCK();
706 MRW_TEARDOWN_WUNLOCK();
707 return ENOPROTOOPT;
708 }
709
710 if (V_ip_mrouter != NULL) {
711 MRW_WUNLOCK();
712 MRW_TEARDOWN_WUNLOCK();
713 return EADDRINUSE;
714 }
715
716 V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
717 HASH_NOWAIT);
718 if (V_mfchashtbl == NULL) {
719 MRW_WUNLOCK();
720 MRW_TEARDOWN_WUNLOCK();
721 return (ENOMEM);
722 }
723
724 /* Create upcall ring */
725 mtx_init(&V_bw_upcalls_ring_mtx, "mroute upcall buf_ring mtx", NULL, MTX_DEF);
726 V_bw_upcalls_ring = buf_ring_alloc(BW_UPCALLS_MAX, M_MRTABLE,
727 M_NOWAIT, &V_bw_upcalls_ring_mtx);
728 if (!V_bw_upcalls_ring) {
729 MRW_WUNLOCK();
730 MRW_TEARDOWN_WUNLOCK();
731 return (ENOMEM);
732 }
733
734 TASK_INIT(&V_task, 0, ip_mrouter_upcall_thread, curvnet);
735 taskqueue_cancel(V_task_queue, &V_task, NULL);
736 taskqueue_unblock(V_task_queue);
737
738 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
739 curvnet);
740 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
741 curvnet);
742
743 V_ip_mrouter = so;
744 atomic_add_int(&ip_mrouter_cnt, 1);
745
746 /* This is a mutex required by buf_ring init, but not used internally */
747 mtx_init(&V_buf_ring_mtx, "mroute buf_ring mtx", NULL, MTX_DEF);
748
749 MRW_WUNLOCK();
750 MRW_TEARDOWN_WUNLOCK();
751
752 CTR1(KTR_IPMF, "%s: done", __func__);
753
754 return 0;
755 }
756
757 /*
758 * Disable multicast forwarding.
759 */
760 static int
X_ip_mrouter_done(void)761 X_ip_mrouter_done(void)
762 {
763 struct ifnet **ifps;
764 int nifp;
765 u_long i;
766 vifi_t vifi;
767 struct bw_upcall *bu;
768
769 MRW_TEARDOWN_WLOCK();
770
771 if (V_ip_mrouter == NULL) {
772 MRW_TEARDOWN_WUNLOCK();
773 return (EINVAL);
774 }
775
776 /*
777 * Detach/disable hooks to the reset of the system.
778 */
779 V_ip_mrouter = NULL;
780 atomic_subtract_int(&ip_mrouter_cnt, 1);
781 V_mrt_api_config = 0;
782
783 /*
784 * Wait for all epoch sections to complete to ensure
785 * V_ip_mrouter = NULL is visible to others.
786 */
787 NET_EPOCH_WAIT();
788
789 /* Stop and drain task queue */
790 taskqueue_block(V_task_queue);
791 while (taskqueue_cancel(V_task_queue, &V_task, NULL)) {
792 taskqueue_drain(V_task_queue, &V_task);
793 }
794
795 ifps = malloc(MAXVIFS * sizeof(*ifps), M_TEMP, M_WAITOK);
796
797 MRW_WLOCK();
798 taskqueue_cancel(V_task_queue, &V_task, NULL);
799
800 /* Destroy upcall ring */
801 while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
802 free(bu, M_MRTABLE);
803 }
804 buf_ring_free(V_bw_upcalls_ring, M_MRTABLE);
805 mtx_destroy(&V_bw_upcalls_ring_mtx);
806
807 /*
808 * For each phyint in use, prepare to disable promiscuous reception
809 * of all IP multicasts. Defer the actual call until the lock is released;
810 * just record the list of interfaces while locked. Some interfaces use
811 * sx locks in their ioctl routines, which is not allowed while holding
812 * a non-sleepable lock.
813 */
814 KASSERT(V_numvifs <= MAXVIFS, ("More vifs than possible"));
815 for (vifi = 0, nifp = 0; vifi < V_numvifs; vifi++) {
816 if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
817 !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
818 ifps[nifp++] = V_viftable[vifi].v_ifp;
819 }
820 }
821 bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
822 V_numvifs = 0;
823 V_pim_assert_enabled = 0;
824
825 callout_stop(&V_expire_upcalls_ch);
826 callout_stop(&V_bw_upcalls_ch);
827
828 /*
829 * Free all multicast forwarding cache entries.
830 * Do not use hashdestroy(), as we must perform other cleanup.
831 */
832 for (i = 0; i < mfchashsize; i++) {
833 struct mfc *rt, *nrt;
834
835 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
836 expire_mfc(rt);
837 }
838 }
839 free(V_mfchashtbl, M_MRTABLE);
840 V_mfchashtbl = NULL;
841
842 bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
843
844 V_reg_vif_num = VIFI_INVALID;
845
846 mtx_destroy(&V_buf_ring_mtx);
847
848 MRW_WUNLOCK();
849 MRW_TEARDOWN_WUNLOCK();
850
851 /*
852 * Now drop our claim on promiscuous multicast on the interfaces recorded
853 * above. This is safe to do now because ALLMULTI is reference counted.
854 */
855 for (vifi = 0; vifi < nifp; vifi++)
856 if_allmulti(ifps[vifi], 0);
857 free(ifps, M_TEMP);
858
859 CTR1(KTR_IPMF, "%s: done", __func__);
860
861 return 0;
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_spin_name, "BM[%d] spin", vifcp->vifc_vifi);
1008 mtx_init(&vifp->v_spin, vifp->v_spin_name, NULL, MTX_SPIN);
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_spin);
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 vifi_t vifi;
1582 int plen = ntohs(ip->ip_len);
1583
1584 M_ASSERTMAPPED(m);
1585 MRW_LOCK_ASSERT();
1586 NET_EPOCH_ASSERT();
1587
1588 /*
1589 * If xmt_vif is not -1, send on only the requested vif.
1590 *
1591 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1592 */
1593 if (xmt_vif < V_numvifs) {
1594 if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1595 pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1596 else
1597 phyint_send(ip, V_viftable + xmt_vif, m);
1598 return 1;
1599 }
1600
1601 /*
1602 * Don't forward if it didn't arrive from the parent vif for its origin.
1603 */
1604 vifi = rt->mfc_parent;
1605 if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1606 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1607 __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1608 MRTSTAT_INC(mrts_wrong_if);
1609 ++rt->mfc_wrong_if;
1610 /*
1611 * If we are doing PIM assert processing, send a message
1612 * to the routing daemon.
1613 *
1614 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1615 * can complete the SPT switch, regardless of the type
1616 * of the iif (broadcast media, GRE tunnel, etc).
1617 */
1618 if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1619 V_viftable[vifi].v_ifp) {
1620 if (ifp == V_multicast_register_if)
1621 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1622
1623 /* Get vifi for the incoming packet */
1624 for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp; vifi++)
1625 ;
1626 if (vifi >= V_numvifs)
1627 return 0; /* The iif is not found: ignore the packet. */
1628
1629 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1630 return 0; /* WRONGVIF disabled: ignore the packet */
1631
1632 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1633 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1634 struct igmpmsg *im;
1635 int hlen = ip->ip_hl << 2;
1636 struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1637
1638 if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1639 mm = m_pullup(mm, hlen);
1640 if (mm == NULL)
1641 return ENOBUFS;
1642
1643 im = mtod(mm, struct igmpmsg *);
1644 im->im_msgtype = IGMPMSG_WRONGVIF;
1645 im->im_mbz = 0;
1646 im->im_vif = vifi;
1647
1648 MRTSTAT_INC(mrts_upcalls);
1649
1650 k_igmpsrc.sin_addr = im->im_src;
1651 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1652 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1653 MRTSTAT_INC(mrts_upq_sockfull);
1654 return ENOBUFS;
1655 }
1656 }
1657 }
1658 return 0;
1659 }
1660
1661 /* If I sourced this packet, it counts as output, else it was input. */
1662 mtx_lock_spin(&V_viftable[vifi].v_spin);
1663 if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1664 V_viftable[vifi].v_pkt_out++;
1665 V_viftable[vifi].v_bytes_out += plen;
1666 } else {
1667 V_viftable[vifi].v_pkt_in++;
1668 V_viftable[vifi].v_bytes_in += plen;
1669 }
1670 mtx_unlock_spin(&V_viftable[vifi].v_spin);
1671
1672 rt->mfc_pkt_cnt++;
1673 rt->mfc_byte_cnt += plen;
1674
1675 /*
1676 * For each vif, decide if a copy of the packet should be forwarded.
1677 * Forward if:
1678 * - the ttl exceeds the vif's threshold
1679 * - there are group members downstream on interface
1680 */
1681 for (vifi = 0; vifi < V_numvifs; vifi++)
1682 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1683 V_viftable[vifi].v_pkt_out++;
1684 V_viftable[vifi].v_bytes_out += plen;
1685 if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1686 pim_register_send(ip, V_viftable + vifi, m, rt);
1687 else
1688 phyint_send(ip, V_viftable + vifi, m);
1689 }
1690
1691 /*
1692 * Perform upcall-related bw measuring.
1693 */
1694 if ((rt->mfc_bw_meter_geq != NULL) || (rt->mfc_bw_meter_leq != NULL)) {
1695 struct bw_meter *x;
1696 struct timeval now;
1697
1698 microtime(&now);
1699 /* Process meters for Greater-or-EQual case */
1700 for (x = rt->mfc_bw_meter_geq; x != NULL; x = x->bm_mfc_next)
1701 bw_meter_geq_receive_packet(x, plen, &now);
1702
1703 /* Process meters for Lower-or-EQual case */
1704 for (x = rt->mfc_bw_meter_leq; x != NULL; x = x->bm_mfc_next) {
1705 /*
1706 * Record that a packet is received.
1707 * Spin lock has to be taken as callout context
1708 * (expire_bw_meter_leq) might modify these fields
1709 * as well
1710 */
1711 mtx_lock_spin(&x->bm_spin);
1712 x->bm_measured.b_packets++;
1713 x->bm_measured.b_bytes += plen;
1714 mtx_unlock_spin(&x->bm_spin);
1715 }
1716 }
1717
1718 return 0;
1719 }
1720
1721 /*
1722 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1723 */
1724 static int
X_legal_vif_num(int vif)1725 X_legal_vif_num(int vif)
1726 {
1727 int ret;
1728
1729 ret = 0;
1730 if (vif < 0)
1731 return (ret);
1732
1733 MRW_RLOCK();
1734 if (vif < V_numvifs)
1735 ret = 1;
1736 MRW_RUNLOCK();
1737
1738 return (ret);
1739 }
1740
1741 /*
1742 * Return the local address used by this vif
1743 */
1744 static u_long
X_ip_mcast_src(int vifi)1745 X_ip_mcast_src(int vifi)
1746 {
1747 in_addr_t addr;
1748
1749 addr = INADDR_ANY;
1750 if (vifi < 0)
1751 return (addr);
1752
1753 MRW_RLOCK();
1754 if (vifi < V_numvifs)
1755 addr = V_viftable[vifi].v_lcl_addr.s_addr;
1756 MRW_RUNLOCK();
1757
1758 return (addr);
1759 }
1760
1761 static void
phyint_send(struct ip * ip,struct vif * vifp,struct mbuf * m)1762 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1763 {
1764 struct mbuf *mb_copy;
1765 int hlen = ip->ip_hl << 2;
1766
1767 MRW_LOCK_ASSERT();
1768 M_ASSERTMAPPED(m);
1769
1770 /*
1771 * Make a new reference to the packet; make sure that
1772 * the IP header is actually copied, not just referenced,
1773 * so that ip_output() only scribbles on the copy.
1774 */
1775 mb_copy = m_copypacket(m, M_NOWAIT);
1776 if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1777 mb_copy = m_pullup(mb_copy, hlen);
1778 if (mb_copy == NULL)
1779 return;
1780
1781 send_packet(vifp, mb_copy);
1782 }
1783
1784 static void
send_packet(struct vif * vifp,struct mbuf * m)1785 send_packet(struct vif *vifp, struct mbuf *m)
1786 {
1787 struct ip_moptions imo;
1788 int error __unused;
1789
1790 MRW_LOCK_ASSERT();
1791 NET_EPOCH_ASSERT();
1792
1793 imo.imo_multicast_ifp = vifp->v_ifp;
1794 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1795 imo.imo_multicast_loop = !!in_mcast_loop;
1796 imo.imo_multicast_vif = -1;
1797 STAILQ_INIT(&imo.imo_head);
1798
1799 /*
1800 * Re-entrancy should not be a problem here, because
1801 * the packets that we send out and are looped back at us
1802 * should get rejected because they appear to come from
1803 * the loopback interface, thus preventing looping.
1804 */
1805 error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1806 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1807 (ptrdiff_t)(vifp - V_viftable), error);
1808 }
1809
1810 /*
1811 * Stubs for old RSVP socket shim implementation.
1812 */
1813
1814 static int
X_ip_rsvp_vif(struct socket * so __unused,struct sockopt * sopt __unused)1815 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1816 {
1817
1818 return (EOPNOTSUPP);
1819 }
1820
1821 static void
X_ip_rsvp_force_done(struct socket * so __unused)1822 X_ip_rsvp_force_done(struct socket *so __unused)
1823 {
1824
1825 }
1826
1827 static int
X_rsvp_input(struct mbuf ** mp,int * offp,int proto)1828 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1829 {
1830 struct mbuf *m;
1831
1832 m = *mp;
1833 *mp = NULL;
1834 if (!V_rsvp_on)
1835 m_freem(m);
1836 return (IPPROTO_DONE);
1837 }
1838
1839 /*
1840 * Code for bandwidth monitors
1841 */
1842
1843 /*
1844 * Define common interface for timeval-related methods
1845 */
1846 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1847 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1848 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1849
1850 static uint32_t
compute_bw_meter_flags(struct bw_upcall * req)1851 compute_bw_meter_flags(struct bw_upcall *req)
1852 {
1853 uint32_t flags = 0;
1854
1855 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1856 flags |= BW_METER_UNIT_PACKETS;
1857 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1858 flags |= BW_METER_UNIT_BYTES;
1859 if (req->bu_flags & BW_UPCALL_GEQ)
1860 flags |= BW_METER_GEQ;
1861 if (req->bu_flags & BW_UPCALL_LEQ)
1862 flags |= BW_METER_LEQ;
1863
1864 return flags;
1865 }
1866
1867 static void
expire_bw_meter_leq(void * arg)1868 expire_bw_meter_leq(void *arg)
1869 {
1870 struct bw_meter *x = arg;
1871 struct timeval now;
1872 /*
1873 * INFO:
1874 * callout is always executed with MRW_WLOCK taken
1875 */
1876
1877 CURVNET_SET((struct vnet *)x->arg);
1878
1879 microtime(&now);
1880
1881 /*
1882 * Test if we should deliver an upcall
1883 */
1884 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1885 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1886 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1887 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1888 /* Prepare an upcall for delivery */
1889 bw_meter_prepare_upcall(x, &now);
1890 }
1891
1892 /* Send all upcalls that are pending delivery */
1893 taskqueue_enqueue(V_task_queue, &V_task);
1894
1895 /* Reset counters */
1896 x->bm_start_time = now;
1897 /* Spin lock has to be taken as ip_forward context
1898 * might modify these fields as well
1899 */
1900 mtx_lock_spin(&x->bm_spin);
1901 x->bm_measured.b_bytes = 0;
1902 x->bm_measured.b_packets = 0;
1903 mtx_unlock_spin(&x->bm_spin);
1904
1905 callout_schedule(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time));
1906
1907 CURVNET_RESTORE();
1908 }
1909
1910 /*
1911 * Add a bw_meter entry
1912 */
1913 static int
add_bw_upcall(struct bw_upcall * req)1914 add_bw_upcall(struct bw_upcall *req)
1915 {
1916 struct mfc *mfc;
1917 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1918 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1919 struct timeval now;
1920 struct bw_meter *x, **bwm_ptr;
1921 uint32_t flags;
1922
1923 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1924 return EOPNOTSUPP;
1925
1926 /* Test if the flags are valid */
1927 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1928 return EINVAL;
1929 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1930 return EINVAL;
1931 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1932 return EINVAL;
1933
1934 /* Test if the threshold time interval is valid */
1935 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1936 return EINVAL;
1937
1938 flags = compute_bw_meter_flags(req);
1939
1940 /*
1941 * Find if we have already same bw_meter entry
1942 */
1943 MRW_WLOCK();
1944 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1945 if (mfc == NULL) {
1946 MRW_WUNLOCK();
1947 return EADDRNOTAVAIL;
1948 }
1949
1950 /* Choose an appropriate bw_meter list */
1951 if (req->bu_flags & BW_UPCALL_GEQ)
1952 bwm_ptr = &mfc->mfc_bw_meter_geq;
1953 else
1954 bwm_ptr = &mfc->mfc_bw_meter_leq;
1955
1956 for (x = *bwm_ptr; x != NULL; x = x->bm_mfc_next) {
1957 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1958 &req->bu_threshold.b_time, ==))
1959 && (x->bm_threshold.b_packets
1960 == req->bu_threshold.b_packets)
1961 && (x->bm_threshold.b_bytes
1962 == req->bu_threshold.b_bytes)
1963 && (x->bm_flags & BW_METER_USER_FLAGS)
1964 == flags) {
1965 MRW_WUNLOCK();
1966 return 0; /* XXX Already installed */
1967 }
1968 }
1969
1970 /* Allocate the new bw_meter entry */
1971 x = malloc(sizeof(*x), M_BWMETER, M_ZERO | M_NOWAIT);
1972 if (x == NULL) {
1973 MRW_WUNLOCK();
1974 return ENOBUFS;
1975 }
1976
1977 /* Set the new bw_meter entry */
1978 x->bm_threshold.b_time = req->bu_threshold.b_time;
1979 microtime(&now);
1980 x->bm_start_time = now;
1981 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1982 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1983 x->bm_measured.b_packets = 0;
1984 x->bm_measured.b_bytes = 0;
1985 x->bm_flags = flags;
1986 x->bm_time_next = NULL;
1987 x->bm_mfc = mfc;
1988 x->arg = curvnet;
1989 sprintf(x->bm_spin_name, "BM spin %p", x);
1990 mtx_init(&x->bm_spin, x->bm_spin_name, NULL, MTX_SPIN);
1991
1992 /* For LEQ case create periodic callout */
1993 if (req->bu_flags & BW_UPCALL_LEQ) {
1994 callout_init_rw(&x->bm_meter_callout, &mrouter_lock, CALLOUT_SHAREDLOCK);
1995 callout_reset(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time),
1996 expire_bw_meter_leq, x);
1997 }
1998
1999 /* Add the new bw_meter entry to the front of entries for this MFC */
2000 x->bm_mfc_next = *bwm_ptr;
2001 *bwm_ptr = x;
2002
2003 MRW_WUNLOCK();
2004
2005 return 0;
2006 }
2007
2008 static void
free_bw_list(struct bw_meter * list)2009 free_bw_list(struct bw_meter *list)
2010 {
2011 while (list != NULL) {
2012 struct bw_meter *x = list;
2013
2014 /* MRW_WLOCK must be held here */
2015 if (x->bm_flags & BW_METER_LEQ) {
2016 callout_drain(&x->bm_meter_callout);
2017 mtx_destroy(&x->bm_spin);
2018 }
2019
2020 list = list->bm_mfc_next;
2021 free(x, M_BWMETER);
2022 }
2023 }
2024
2025 /*
2026 * Delete one or multiple bw_meter entries
2027 */
2028 static int
del_bw_upcall(struct bw_upcall * req)2029 del_bw_upcall(struct bw_upcall *req)
2030 {
2031 struct mfc *mfc;
2032 struct bw_meter *x, **bwm_ptr;
2033
2034 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
2035 return EOPNOTSUPP;
2036
2037 MRW_WLOCK();
2038
2039 /* Find the corresponding MFC entry */
2040 mfc = mfc_find(&req->bu_src, &req->bu_dst);
2041 if (mfc == NULL) {
2042 MRW_WUNLOCK();
2043 return EADDRNOTAVAIL;
2044 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2045 /*
2046 * Delete all bw_meter entries for this mfc
2047 */
2048 struct bw_meter *list;
2049
2050 /* Free LEQ list */
2051 list = mfc->mfc_bw_meter_leq;
2052 mfc->mfc_bw_meter_leq = NULL;
2053 free_bw_list(list);
2054
2055 /* Free GEQ list */
2056 list = mfc->mfc_bw_meter_geq;
2057 mfc->mfc_bw_meter_geq = NULL;
2058 free_bw_list(list);
2059 MRW_WUNLOCK();
2060 return 0;
2061 } else { /* Delete a single bw_meter entry */
2062 struct bw_meter *prev;
2063 uint32_t flags = 0;
2064
2065 flags = compute_bw_meter_flags(req);
2066
2067 /* Choose an appropriate bw_meter list */
2068 if (req->bu_flags & BW_UPCALL_GEQ)
2069 bwm_ptr = &mfc->mfc_bw_meter_geq;
2070 else
2071 bwm_ptr = &mfc->mfc_bw_meter_leq;
2072
2073 /* Find the bw_meter entry to delete */
2074 for (prev = NULL, x = *bwm_ptr; x != NULL;
2075 prev = x, x = x->bm_mfc_next) {
2076 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, &req->bu_threshold.b_time, ==)) &&
2077 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2078 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2079 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2080 break;
2081 }
2082 if (x != NULL) { /* Delete entry from the list for this MFC */
2083 if (prev != NULL)
2084 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
2085 else
2086 *bwm_ptr = x->bm_mfc_next;/* new head of list */
2087
2088 if (req->bu_flags & BW_UPCALL_LEQ)
2089 callout_stop(&x->bm_meter_callout);
2090
2091 MRW_WUNLOCK();
2092 /* Free the bw_meter entry */
2093 free(x, M_BWMETER);
2094 return 0;
2095 } else {
2096 MRW_WUNLOCK();
2097 return EINVAL;
2098 }
2099 }
2100 __assert_unreachable();
2101 }
2102
2103 /*
2104 * Perform bandwidth measurement processing that may result in an upcall
2105 */
2106 static void
bw_meter_geq_receive_packet(struct bw_meter * x,int plen,struct timeval * nowp)2107 bw_meter_geq_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2108 {
2109 struct timeval delta;
2110
2111 MRW_LOCK_ASSERT();
2112
2113 delta = *nowp;
2114 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2115
2116 /*
2117 * Processing for ">=" type of bw_meter entry.
2118 * bm_spin does not have to be hold here as in GEQ
2119 * case this is the only context accessing bm_measured.
2120 */
2121 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2122 /* Reset the bw_meter entry */
2123 x->bm_start_time = *nowp;
2124 x->bm_measured.b_packets = 0;
2125 x->bm_measured.b_bytes = 0;
2126 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2127 }
2128
2129 /* Record that a packet is received */
2130 x->bm_measured.b_packets++;
2131 x->bm_measured.b_bytes += plen;
2132
2133 /*
2134 * Test if we should deliver an upcall
2135 */
2136 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2137 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2138 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2139 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2140 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2141 /* Prepare an upcall for delivery */
2142 bw_meter_prepare_upcall(x, nowp);
2143 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2144 }
2145 }
2146 }
2147
2148 /*
2149 * Prepare a bandwidth-related upcall
2150 */
2151 static void
bw_meter_prepare_upcall(struct bw_meter * x,struct timeval * nowp)2152 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2153 {
2154 struct timeval delta;
2155 struct bw_upcall *u;
2156
2157 MRW_LOCK_ASSERT();
2158
2159 /*
2160 * Compute the measured time interval
2161 */
2162 delta = *nowp;
2163 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2164
2165 /*
2166 * Set the bw_upcall entry
2167 */
2168 u = malloc(sizeof(struct bw_upcall), M_MRTABLE, M_NOWAIT | M_ZERO);
2169 if (!u) {
2170 log(LOG_WARNING, "bw_meter_prepare_upcall: cannot allocate entry\n");
2171 return;
2172 }
2173 u->bu_src = x->bm_mfc->mfc_origin;
2174 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2175 u->bu_threshold.b_time = x->bm_threshold.b_time;
2176 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2177 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2178 u->bu_measured.b_time = delta;
2179 u->bu_measured.b_packets = x->bm_measured.b_packets;
2180 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2181 u->bu_flags = 0;
2182 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2183 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2184 if (x->bm_flags & BW_METER_UNIT_BYTES)
2185 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2186 if (x->bm_flags & BW_METER_GEQ)
2187 u->bu_flags |= BW_UPCALL_GEQ;
2188 if (x->bm_flags & BW_METER_LEQ)
2189 u->bu_flags |= BW_UPCALL_LEQ;
2190
2191 if (buf_ring_enqueue(V_bw_upcalls_ring, u))
2192 log(LOG_WARNING, "bw_meter_prepare_upcall: cannot enqueue upcall\n");
2193 if (buf_ring_count(V_bw_upcalls_ring) > (BW_UPCALLS_MAX / 2)) {
2194 taskqueue_enqueue(V_task_queue, &V_task);
2195 }
2196 }
2197 /*
2198 * Send the pending bandwidth-related upcalls
2199 */
2200 static void
bw_upcalls_send(void)2201 bw_upcalls_send(void)
2202 {
2203 struct mbuf *m;
2204 int len = 0;
2205 struct bw_upcall *bu;
2206 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2207 static struct igmpmsg igmpmsg = {
2208 0, /* unused1 */
2209 0, /* unused2 */
2210 IGMPMSG_BW_UPCALL,/* im_msgtype */
2211 0, /* im_mbz */
2212 0, /* im_vif */
2213 0, /* unused3 */
2214 { 0 }, /* im_src */
2215 { 0 } /* im_dst */
2216 };
2217
2218 MRW_LOCK_ASSERT();
2219
2220 if (buf_ring_empty(V_bw_upcalls_ring))
2221 return;
2222
2223 /*
2224 * Allocate a new mbuf, initialize it with the header and
2225 * the payload for the pending calls.
2226 */
2227 m = m_gethdr(M_NOWAIT, MT_DATA);
2228 if (m == NULL) {
2229 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2230 return;
2231 }
2232
2233 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2234 len += sizeof(struct igmpmsg);
2235 while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
2236 m_copyback(m, len, sizeof(struct bw_upcall), (caddr_t)bu);
2237 len += sizeof(struct bw_upcall);
2238 free(bu, M_MRTABLE);
2239 }
2240
2241 /*
2242 * Send the upcalls
2243 * XXX do we need to set the address in k_igmpsrc ?
2244 */
2245 MRTSTAT_INC(mrts_upcalls);
2246 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2247 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2248 MRTSTAT_INC(mrts_upq_sockfull);
2249 }
2250 }
2251
2252 /*
2253 * A periodic function for sending all upcalls that are pending delivery
2254 */
2255 static void
expire_bw_upcalls_send(void * arg)2256 expire_bw_upcalls_send(void *arg)
2257 {
2258 CURVNET_SET((struct vnet *) arg);
2259
2260 /* This callout is run with MRW_RLOCK taken */
2261
2262 bw_upcalls_send();
2263
2264 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2265 curvnet);
2266 CURVNET_RESTORE();
2267 }
2268
2269 /*
2270 * End of bandwidth monitoring code
2271 */
2272
2273 /*
2274 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2275 *
2276 */
2277 static int
pim_register_send(struct ip * ip,struct vif * vifp,struct mbuf * m,struct mfc * rt)2278 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2279 struct mfc *rt)
2280 {
2281 struct mbuf *mb_copy, *mm;
2282
2283 /*
2284 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2285 * rendezvous point was unspecified, and we were told not to.
2286 */
2287 if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2288 in_nullhost(rt->mfc_rp))
2289 return 0;
2290
2291 mb_copy = pim_register_prepare(ip, m);
2292 if (mb_copy == NULL)
2293 return ENOBUFS;
2294
2295 /*
2296 * Send all the fragments. Note that the mbuf for each fragment
2297 * is freed by the sending machinery.
2298 */
2299 for (mm = mb_copy; mm; mm = mb_copy) {
2300 mb_copy = mm->m_nextpkt;
2301 mm->m_nextpkt = 0;
2302 mm = m_pullup(mm, sizeof(struct ip));
2303 if (mm != NULL) {
2304 ip = mtod(mm, struct ip *);
2305 if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2306 pim_register_send_rp(ip, vifp, mm, rt);
2307 } else {
2308 pim_register_send_upcall(ip, vifp, mm, rt);
2309 }
2310 }
2311 }
2312
2313 return 0;
2314 }
2315
2316 /*
2317 * Return a copy of the data packet that is ready for PIM Register
2318 * encapsulation.
2319 * XXX: Note that in the returned copy the IP header is a valid one.
2320 */
2321 static struct mbuf *
pim_register_prepare(struct ip * ip,struct mbuf * m)2322 pim_register_prepare(struct ip *ip, struct mbuf *m)
2323 {
2324 struct mbuf *mb_copy = NULL;
2325 int mtu;
2326
2327 /* Take care of delayed checksums */
2328 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2329 in_delayed_cksum(m);
2330 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2331 }
2332
2333 /*
2334 * Copy the old packet & pullup its IP header into the
2335 * new mbuf so we can modify it.
2336 */
2337 mb_copy = m_copypacket(m, M_NOWAIT);
2338 if (mb_copy == NULL)
2339 return NULL;
2340 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2341 if (mb_copy == NULL)
2342 return NULL;
2343
2344 /* take care of the TTL */
2345 ip = mtod(mb_copy, struct ip *);
2346 --ip->ip_ttl;
2347
2348 /* Compute the MTU after the PIM Register encapsulation */
2349 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2350
2351 if (ntohs(ip->ip_len) <= mtu) {
2352 /* Turn the IP header into a valid one */
2353 ip->ip_sum = 0;
2354 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2355 } else {
2356 /* Fragment the packet */
2357 mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2358 if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2359 m_freem(mb_copy);
2360 return NULL;
2361 }
2362 }
2363 return mb_copy;
2364 }
2365
2366 /*
2367 * Send an upcall with the data packet to the user-level process.
2368 */
2369 static int
pim_register_send_upcall(struct ip * ip,struct vif * vifp,struct mbuf * mb_copy,struct mfc * rt)2370 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2371 struct mbuf *mb_copy, struct mfc *rt)
2372 {
2373 struct mbuf *mb_first;
2374 int len = ntohs(ip->ip_len);
2375 struct igmpmsg *im;
2376 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2377
2378 MRW_LOCK_ASSERT();
2379
2380 /*
2381 * Add a new mbuf with an upcall header
2382 */
2383 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2384 if (mb_first == NULL) {
2385 m_freem(mb_copy);
2386 return ENOBUFS;
2387 }
2388 mb_first->m_data += max_linkhdr;
2389 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2390 mb_first->m_len = sizeof(struct igmpmsg);
2391 mb_first->m_next = mb_copy;
2392
2393 /* Send message to routing daemon */
2394 im = mtod(mb_first, struct igmpmsg *);
2395 im->im_msgtype = IGMPMSG_WHOLEPKT;
2396 im->im_mbz = 0;
2397 im->im_vif = vifp - V_viftable;
2398 im->im_src = ip->ip_src;
2399 im->im_dst = ip->ip_dst;
2400
2401 k_igmpsrc.sin_addr = ip->ip_src;
2402
2403 MRTSTAT_INC(mrts_upcalls);
2404
2405 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2406 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2407 MRTSTAT_INC(mrts_upq_sockfull);
2408 return ENOBUFS;
2409 }
2410
2411 /* Keep statistics */
2412 PIMSTAT_INC(pims_snd_registers_msgs);
2413 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2414
2415 return 0;
2416 }
2417
2418 /*
2419 * Encapsulate the data packet in PIM Register message and send it to the RP.
2420 */
2421 static int
pim_register_send_rp(struct ip * ip,struct vif * vifp,struct mbuf * mb_copy,struct mfc * rt)2422 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2423 struct mfc *rt)
2424 {
2425 struct mbuf *mb_first;
2426 struct ip *ip_outer;
2427 struct pim_encap_pimhdr *pimhdr;
2428 int len = ntohs(ip->ip_len);
2429 vifi_t vifi = rt->mfc_parent;
2430
2431 MRW_LOCK_ASSERT();
2432
2433 if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2434 m_freem(mb_copy);
2435 return EADDRNOTAVAIL; /* The iif vif is invalid */
2436 }
2437
2438 /*
2439 * Add a new mbuf with the encapsulating header
2440 */
2441 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2442 if (mb_first == NULL) {
2443 m_freem(mb_copy);
2444 return ENOBUFS;
2445 }
2446 mb_first->m_data += max_linkhdr;
2447 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2448 mb_first->m_next = mb_copy;
2449
2450 mb_first->m_pkthdr.len = len + mb_first->m_len;
2451
2452 /*
2453 * Fill in the encapsulating IP and PIM header
2454 */
2455 ip_outer = mtod(mb_first, struct ip *);
2456 *ip_outer = pim_encap_iphdr;
2457 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2458 sizeof(pim_encap_pimhdr));
2459 ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2460 ip_outer->ip_dst = rt->mfc_rp;
2461 /*
2462 * Copy the inner header TOS to the outer header, and take care of the
2463 * IP_DF bit.
2464 */
2465 ip_outer->ip_tos = ip->ip_tos;
2466 if (ip->ip_off & htons(IP_DF))
2467 ip_outer->ip_off |= htons(IP_DF);
2468 ip_fillid(ip_outer, V_ip_random_id);
2469 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2470 + sizeof(pim_encap_iphdr));
2471 *pimhdr = pim_encap_pimhdr;
2472 /* If the iif crosses a border, set the Border-bit */
2473 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2474 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2475
2476 mb_first->m_data += sizeof(pim_encap_iphdr);
2477 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2478 mb_first->m_data -= sizeof(pim_encap_iphdr);
2479
2480 send_packet(vifp, mb_first);
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 * pim_encapcheck() is called by the encap4_input() path at runtime to
2491 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2492 * into the kernel.
2493 */
2494 static int
pim_encapcheck(const struct mbuf * m __unused,int off __unused,int proto __unused,void * arg __unused)2495 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2496 int proto __unused, void *arg __unused)
2497 {
2498
2499 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2500 return (8); /* claim the datagram. */
2501 }
2502
2503 /*
2504 * PIM-SMv2 and PIM-DM messages processing.
2505 * Receives and verifies the PIM control messages, and passes them
2506 * up to the listening socket, using rip_input().
2507 * The only message with special processing is the PIM_REGISTER message
2508 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2509 * is passed to if_simloop().
2510 */
2511 static int
pim_input(struct mbuf * m,int off,int proto,void * arg __unused)2512 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2513 {
2514 struct ip *ip = mtod(m, struct ip *);
2515 struct pim *pim;
2516 int iphlen = off;
2517 int minlen;
2518 int datalen = ntohs(ip->ip_len) - iphlen;
2519 int ip_tos;
2520
2521 /* Keep statistics */
2522 PIMSTAT_INC(pims_rcv_total_msgs);
2523 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2524
2525 /*
2526 * Validate lengths
2527 */
2528 if (datalen < PIM_MINLEN) {
2529 PIMSTAT_INC(pims_rcv_tooshort);
2530 CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2531 __func__, datalen, ntohl(ip->ip_src.s_addr));
2532 m_freem(m);
2533 return (IPPROTO_DONE);
2534 }
2535
2536 /*
2537 * If the packet is at least as big as a REGISTER, go agead
2538 * and grab the PIM REGISTER header size, to avoid another
2539 * possible m_pullup() later.
2540 *
2541 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2542 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2543 */
2544 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2545 /*
2546 * Get the IP and PIM headers in contiguous memory, and
2547 * possibly the PIM REGISTER header.
2548 */
2549 if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2550 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2551 return (IPPROTO_DONE);
2552 }
2553
2554 /* m_pullup() may have given us a new mbuf so reset ip. */
2555 ip = mtod(m, struct ip *);
2556 ip_tos = ip->ip_tos;
2557
2558 /* adjust mbuf to point to the PIM header */
2559 m->m_data += iphlen;
2560 m->m_len -= iphlen;
2561 pim = mtod(m, struct pim *);
2562
2563 /*
2564 * Validate checksum. If PIM REGISTER, exclude the data packet.
2565 *
2566 * XXX: some older PIMv2 implementations don't make this distinction,
2567 * so for compatibility reason perform the checksum over part of the
2568 * message, and if error, then over the whole message.
2569 */
2570 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2571 /* do nothing, checksum okay */
2572 } else if (in_cksum(m, datalen)) {
2573 PIMSTAT_INC(pims_rcv_badsum);
2574 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2575 m_freem(m);
2576 return (IPPROTO_DONE);
2577 }
2578
2579 /* PIM version check */
2580 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2581 PIMSTAT_INC(pims_rcv_badversion);
2582 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2583 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2584 m_freem(m);
2585 return (IPPROTO_DONE);
2586 }
2587
2588 /* restore mbuf back to the outer IP */
2589 m->m_data -= iphlen;
2590 m->m_len += iphlen;
2591
2592 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2593 /*
2594 * Since this is a REGISTER, we'll make a copy of the register
2595 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2596 * routing daemon.
2597 */
2598 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2599 struct mbuf *mcp;
2600 struct ip *encap_ip;
2601 u_int32_t *reghdr;
2602 struct ifnet *vifp;
2603
2604 MRW_RLOCK();
2605 if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2606 MRW_RUNLOCK();
2607 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2608 (int)V_reg_vif_num);
2609 m_freem(m);
2610 return (IPPROTO_DONE);
2611 }
2612 /* XXX need refcnt? */
2613 vifp = V_viftable[V_reg_vif_num].v_ifp;
2614 MRW_RUNLOCK();
2615
2616 /*
2617 * Validate length
2618 */
2619 if (datalen < PIM_REG_MINLEN) {
2620 PIMSTAT_INC(pims_rcv_tooshort);
2621 PIMSTAT_INC(pims_rcv_badregisters);
2622 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2623 m_freem(m);
2624 return (IPPROTO_DONE);
2625 }
2626
2627 reghdr = (u_int32_t *)(pim + 1);
2628 encap_ip = (struct ip *)(reghdr + 1);
2629
2630 CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2631 __func__, ntohl(encap_ip->ip_src.s_addr),
2632 ntohs(encap_ip->ip_len));
2633
2634 /* verify the version number of the inner packet */
2635 if (encap_ip->ip_v != IPVERSION) {
2636 PIMSTAT_INC(pims_rcv_badregisters);
2637 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2638 m_freem(m);
2639 return (IPPROTO_DONE);
2640 }
2641
2642 /* verify the inner packet is destined to a mcast group */
2643 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2644 PIMSTAT_INC(pims_rcv_badregisters);
2645 CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2646 ntohl(encap_ip->ip_dst.s_addr));
2647 m_freem(m);
2648 return (IPPROTO_DONE);
2649 }
2650
2651 /* If a NULL_REGISTER, pass it to the daemon */
2652 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2653 goto pim_input_to_daemon;
2654
2655 /*
2656 * Copy the TOS from the outer IP header to the inner IP header.
2657 */
2658 if (encap_ip->ip_tos != ip_tos) {
2659 /* Outer TOS -> inner TOS */
2660 encap_ip->ip_tos = ip_tos;
2661 /* Recompute the inner header checksum. Sigh... */
2662
2663 /* adjust mbuf to point to the inner IP header */
2664 m->m_data += (iphlen + PIM_MINLEN);
2665 m->m_len -= (iphlen + PIM_MINLEN);
2666
2667 encap_ip->ip_sum = 0;
2668 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2669
2670 /* restore mbuf to point back to the outer IP header */
2671 m->m_data -= (iphlen + PIM_MINLEN);
2672 m->m_len += (iphlen + PIM_MINLEN);
2673 }
2674
2675 /*
2676 * Decapsulate the inner IP packet and loopback to forward it
2677 * as a normal multicast packet. Also, make a copy of the
2678 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2679 * to pass to the daemon later, so it can take the appropriate
2680 * actions (e.g., send back PIM_REGISTER_STOP).
2681 * XXX: here m->m_data points to the outer IP header.
2682 */
2683 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2684 if (mcp == NULL) {
2685 CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2686 m_freem(m);
2687 return (IPPROTO_DONE);
2688 }
2689
2690 /* Keep statistics */
2691 /* XXX: registers_bytes include only the encap. mcast pkt */
2692 PIMSTAT_INC(pims_rcv_registers_msgs);
2693 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2694
2695 /*
2696 * forward the inner ip packet; point m_data at the inner ip.
2697 */
2698 m_adj(m, iphlen + PIM_MINLEN);
2699
2700 CTR4(KTR_IPMF,
2701 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2702 __func__,
2703 (u_long)ntohl(encap_ip->ip_src.s_addr),
2704 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2705 (int)V_reg_vif_num);
2706
2707 /* NB: vifp was collected above; can it change on us? */
2708 if_simloop(vifp, m, dst.sin_family, 0);
2709
2710 /* prepare the register head to send to the mrouting daemon */
2711 m = mcp;
2712 }
2713
2714 pim_input_to_daemon:
2715 /*
2716 * Pass the PIM message up to the daemon; if it is a Register message,
2717 * pass the 'head' only up to the daemon. This includes the
2718 * outer IP header, PIM header, PIM-Register header and the
2719 * inner IP header.
2720 * XXX: the outer IP header pkt size of a Register is not adjust to
2721 * reflect the fact that the inner multicast data is truncated.
2722 */
2723 return (rip_input(&m, &off, proto));
2724 }
2725
2726 static int
sysctl_mfctable(SYSCTL_HANDLER_ARGS)2727 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2728 {
2729 struct mfc *rt;
2730 int error, i;
2731
2732 if (req->newptr)
2733 return (EPERM);
2734 if (V_mfchashtbl == NULL) /* XXX unlocked */
2735 return (0);
2736 error = sysctl_wire_old_buffer(req, 0);
2737 if (error)
2738 return (error);
2739
2740 MRW_RLOCK();
2741 if (V_mfchashtbl == NULL)
2742 goto out_locked;
2743
2744 for (i = 0; i < mfchashsize; i++) {
2745 LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2746 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2747 if (error)
2748 goto out_locked;
2749 }
2750 }
2751 out_locked:
2752 MRW_RUNLOCK();
2753 return (error);
2754 }
2755
2756 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2757 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2758 "IPv4 Multicast Forwarding Table "
2759 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2760
2761 static int
sysctl_viflist(SYSCTL_HANDLER_ARGS)2762 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2763 {
2764 int error, i;
2765
2766 if (req->newptr)
2767 return (EPERM);
2768 if (V_viftable == NULL) /* XXX unlocked */
2769 return (0);
2770 error = sysctl_wire_old_buffer(req, MROUTE_VIF_SYSCTL_LEN * MAXVIFS);
2771 if (error)
2772 return (error);
2773
2774 MRW_RLOCK();
2775 /* Copy out user-visible portion of vif entry. */
2776 for (i = 0; i < MAXVIFS; i++) {
2777 error = SYSCTL_OUT(req, &V_viftable[i], MROUTE_VIF_SYSCTL_LEN);
2778 if (error)
2779 break;
2780 }
2781 MRW_RUNLOCK();
2782 return (error);
2783 }
2784
2785 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2786 CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2787 sysctl_viflist, "S,vif[MAXVIFS]",
2788 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2789
2790 static void
vnet_mroute_init(const void * unused __unused)2791 vnet_mroute_init(const void *unused __unused)
2792 {
2793
2794 V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2795
2796 V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2797 M_MRTABLE, M_WAITOK|M_ZERO);
2798
2799 callout_init_rw(&V_expire_upcalls_ch, &mrouter_lock, 0);
2800 callout_init_rw(&V_bw_upcalls_ch, &mrouter_lock, 0);
2801
2802 /* Prepare taskqueue */
2803 V_task_queue = taskqueue_create_fast("ip_mroute_tskq", M_NOWAIT,
2804 taskqueue_thread_enqueue, &V_task_queue);
2805 taskqueue_start_threads(&V_task_queue, 1, PI_NET, "ip_mroute_tskq task");
2806 }
2807
2808 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2809 NULL);
2810
2811 static void
vnet_mroute_uninit(const void * unused __unused)2812 vnet_mroute_uninit(const void *unused __unused)
2813 {
2814
2815 /* Taskqueue should be cancelled and drained before freeing */
2816 taskqueue_free(V_task_queue);
2817
2818 free(V_viftable, M_MRTABLE);
2819 free(V_nexpire, M_MRTABLE);
2820 V_nexpire = NULL;
2821 }
2822
2823 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2824 vnet_mroute_uninit, NULL);
2825
2826 static int
ip_mroute_modevent(module_t mod,int type,void * unused)2827 ip_mroute_modevent(module_t mod, int type, void *unused)
2828 {
2829
2830 switch (type) {
2831 case MOD_LOAD:
2832 MRW_TEARDOWN_LOCK_INIT();
2833 MRW_LOCK_INIT();
2834
2835 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2836 if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2837 if (if_detach_event_tag == NULL) {
2838 printf("ip_mroute: unable to register "
2839 "ifnet_departure_event handler\n");
2840 MRW_LOCK_DESTROY();
2841 return (EINVAL);
2842 }
2843
2844 if (!powerof2(mfchashsize)) {
2845 printf("WARNING: %s not a power of 2; using default\n",
2846 "net.inet.ip.mfchashsize");
2847 mfchashsize = MFCHASHSIZE;
2848 }
2849
2850 pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2851
2852 ip_mcast_src = X_ip_mcast_src;
2853 ip_mforward = X_ip_mforward;
2854 ip_mrouter_done = X_ip_mrouter_done;
2855 ip_mrouter_get = X_ip_mrouter_get;
2856 ip_mrouter_set = X_ip_mrouter_set;
2857
2858 ip_rsvp_force_done = X_ip_rsvp_force_done;
2859 ip_rsvp_vif = X_ip_rsvp_vif;
2860
2861 legal_vif_num = X_legal_vif_num;
2862 mrt_ioctl = X_mrt_ioctl;
2863 rsvp_input_p = X_rsvp_input;
2864 break;
2865
2866 case MOD_UNLOAD:
2867 /*
2868 * Typically module unload happens after the user-level
2869 * process has shutdown the kernel services (the check
2870 * below insures someone can't just yank the module out
2871 * from under a running process). But if the module is
2872 * just loaded and then unloaded w/o starting up a user
2873 * process we still need to cleanup.
2874 */
2875 MRW_WLOCK();
2876 if (ip_mrouter_cnt != 0) {
2877 MRW_WUNLOCK();
2878 return (EINVAL);
2879 }
2880 ip_mrouter_unloading = 1;
2881 MRW_WUNLOCK();
2882
2883 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2884
2885 if (pim_encap_cookie) {
2886 ip_encap_detach(pim_encap_cookie);
2887 pim_encap_cookie = NULL;
2888 }
2889
2890 ip_mcast_src = NULL;
2891 ip_mforward = NULL;
2892 ip_mrouter_done = NULL;
2893 ip_mrouter_get = NULL;
2894 ip_mrouter_set = NULL;
2895
2896 ip_rsvp_force_done = NULL;
2897 ip_rsvp_vif = NULL;
2898
2899 legal_vif_num = NULL;
2900 mrt_ioctl = NULL;
2901 rsvp_input_p = NULL;
2902
2903 MRW_LOCK_DESTROY();
2904 MRW_TEARDOWN_LOCK_DESTROY();
2905 break;
2906
2907 default:
2908 return EOPNOTSUPP;
2909 }
2910 return 0;
2911 }
2912
2913 static moduledata_t ip_mroutemod = {
2914 "ip_mroute",
2915 ip_mroute_modevent,
2916 0
2917 };
2918
2919 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2920