Annotation of sys/netinet/ip_mroute.c, Revision 1.1.1.1
1.1 nbrk 1: /* $OpenBSD: ip_mroute.c,v 1.48 2007/05/22 09:51:13 michele Exp $ */
2: /* $NetBSD: ip_mroute.c,v 1.85 2004/04/26 01:31:57 matt Exp $ */
3:
4: /*
5: * Copyright (c) 1989 Stephen Deering
6: * Copyright (c) 1992, 1993
7: * The Regents of the University of California. All rights reserved.
8: *
9: * This code is derived from software contributed to Berkeley by
10: * Stephen Deering of Stanford University.
11: *
12: * Redistribution and use in source and binary forms, with or without
13: * modification, are permitted provided that the following conditions
14: * are met:
15: * 1. Redistributions of source code must retain the above copyright
16: * notice, this list of conditions and the following disclaimer.
17: * 2. Redistributions in binary form must reproduce the above copyright
18: * notice, this list of conditions and the following disclaimer in the
19: * documentation and/or other materials provided with the distribution.
20: * 3. Neither the name of the University nor the names of its contributors
21: * may be used to endorse or promote products derived from this software
22: * without specific prior written permission.
23: *
24: * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27: * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34: * SUCH DAMAGE.
35: *
36: * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
37: */
38:
39: /*
40: * IP multicast forwarding procedures
41: *
42: * Written by David Waitzman, BBN Labs, August 1988.
43: * Modified by Steve Deering, Stanford, February 1989.
44: * Modified by Mark J. Steiglitz, Stanford, May, 1991
45: * Modified by Van Jacobson, LBL, January 1993
46: * Modified by Ajit Thyagarajan, PARC, August 1993
47: * Modified by Bill Fenner, PARC, April 1994
48: * Modified by Charles M. Hannum, NetBSD, May 1995.
49: * Modified by Ahmed Helmy, SGI, June 1996
50: * Modified by George Edmond Eddy (Rusty), ISI, February 1998
51: * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
52: * Modified by Hitoshi Asaeda, WIDE, August 2000
53: * Modified by Pavlin Radoslavov, ICSI, October 2002
54: *
55: * MROUTING Revision: 1.2
56: * and PIM-SMv2 and PIM-DM support, advanced API support,
57: * bandwidth metering and signaling
58: */
59:
60: #ifdef PIM
61: #define _PIM_VT 1
62: #endif
63:
64: #include <sys/param.h>
65: #include <sys/systm.h>
66: #include <sys/mbuf.h>
67: #include <sys/socket.h>
68: #include <sys/socketvar.h>
69: #include <sys/protosw.h>
70: #include <sys/errno.h>
71: #include <sys/time.h>
72: #include <sys/kernel.h>
73: #include <sys/ioctl.h>
74: #include <sys/syslog.h>
75: #include <sys/timeout.h>
76:
77: #include <net/if.h>
78: #include <net/route.h>
79: #include <net/raw_cb.h>
80:
81: #include <netinet/in.h>
82: #include <netinet/in_var.h>
83: #include <netinet/in_systm.h>
84: #include <netinet/ip.h>
85: #include <netinet/ip_var.h>
86: #include <netinet/in_pcb.h>
87: #include <netinet/udp.h>
88: #include <netinet/igmp.h>
89: #include <netinet/igmp_var.h>
90: #include <netinet/ip_mroute.h>
91: #ifdef PIM
92: #include <netinet/pim.h>
93: #include <netinet/pim_var.h>
94: #endif
95:
96: #include <sys/stdarg.h>
97:
98: #define IP_MULTICASTOPTS 0
99: #define M_PULLUP(m, len) \
100: do { \
101: if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \
102: (m) = m_pullup((m), (len)); \
103: } while (/*CONSTCOND*/ 0)
104:
105: /*
106: * Globals. All but ip_mrouter and ip_mrtproto could be static,
107: * except for netstat or debugging purposes.
108: */
109: struct socket *ip_mrouter = NULL;
110: int ip_mrtproto = IGMP_DVMRP; /* for netstat only */
111:
112: #define NO_RTE_FOUND 0x1
113: #define RTE_FOUND 0x2
114:
115: #define MFCHASH(a, g) \
116: ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
117: ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash)
118: LIST_HEAD(mfchashhdr, mfc) *mfchashtbl;
119: u_long mfchash;
120:
121: u_char nexpire[MFCTBLSIZ];
122: struct vif viftable[MAXVIFS];
123: struct mrtstat mrtstat;
124: u_int mrtdebug = 0; /* debug level */
125: #define DEBUG_MFC 0x02
126: #define DEBUG_FORWARD 0x04
127: #define DEBUG_EXPIRE 0x08
128: #define DEBUG_XMIT 0x10
129: #define DEBUG_PIM 0x20
130:
131: #define VIFI_INVALID ((vifi_t) -1)
132:
133: u_int tbfdebug = 0; /* tbf debug level */
134: #ifdef RSVP_ISI
135: u_int rsvpdebug = 0; /* rsvp debug level */
136: extern struct socket *ip_rsvpd;
137: extern int rsvp_on;
138: #endif /* RSVP_ISI */
139:
140: #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
141: #define UPCALL_EXPIRE 6 /* number of timeouts */
142: struct timeout expire_upcalls_ch;
143:
144: /*
145: * Define the token bucket filter structures
146: */
147:
148: #define TBF_REPROCESS (hz / 100) /* 100x / second */
149:
150: static int get_sg_cnt(struct sioc_sg_req *);
151: static int get_vif_cnt(struct sioc_vif_req *);
152: static int ip_mrouter_init(struct socket *, struct mbuf *);
153: static int get_version(struct mbuf *);
154: static int set_assert(struct mbuf *);
155: static int get_assert(struct mbuf *);
156: static int add_vif(struct mbuf *);
157: static int del_vif(struct mbuf *);
158: static void update_mfc_params(struct mfc *, struct mfcctl2 *);
159: static void init_mfc_params(struct mfc *, struct mfcctl2 *);
160: static void expire_mfc(struct mfc *);
161: static int add_mfc(struct mbuf *);
162: #ifdef UPCALL_TIMING
163: static void collate(struct timeval *);
164: #endif
165: static int del_mfc(struct mbuf *);
166: static int set_api_config(struct mbuf *); /* chose API capabilities */
167: static int get_api_support(struct mbuf *);
168: static int get_api_config(struct mbuf *);
169: static int socket_send(struct socket *, struct mbuf *,
170: struct sockaddr_in *);
171: static void expire_upcalls(void *);
172: #ifdef RSVP_ISI
173: static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
174: #else
175: static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *);
176: #endif
177: static void phyint_send(struct ip *, struct vif *, struct mbuf *);
178: static void encap_send(struct ip *, struct vif *, struct mbuf *);
179: static void tbf_control(struct vif *, struct mbuf *, struct ip *,
180: u_int32_t);
181: static void tbf_queue(struct vif *, struct mbuf *);
182: static void tbf_process_q(struct vif *);
183: static void tbf_reprocess_q(void *);
184: static int tbf_dq_sel(struct vif *, struct ip *);
185: static void tbf_send_packet(struct vif *, struct mbuf *);
186: static void tbf_update_tokens(struct vif *);
187: static int priority(struct vif *, struct ip *);
188:
189: /*
190: * Bandwidth monitoring
191: */
192: static void free_bw_list(struct bw_meter *);
193: static int add_bw_upcall(struct mbuf *);
194: static int del_bw_upcall(struct mbuf *);
195: static void bw_meter_receive_packet(struct bw_meter *, int , struct timeval *);
196: static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
197: static void bw_upcalls_send(void);
198: static void schedule_bw_meter(struct bw_meter *, struct timeval *);
199: static void unschedule_bw_meter(struct bw_meter *);
200: static void bw_meter_process(void);
201: static void expire_bw_upcalls_send(void *);
202: static void expire_bw_meter_process(void *);
203:
204: #ifdef PIM
205: static int pim_register_send(struct ip *, struct vif *,
206: struct mbuf *, struct mfc *);
207: static int pim_register_send_rp(struct ip *, struct vif *,
208: struct mbuf *, struct mfc *);
209: static int pim_register_send_upcall(struct ip *, struct vif *,
210: struct mbuf *, struct mfc *);
211: static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
212: #endif
213:
214: /*
215: * 'Interfaces' associated with decapsulator (so we can tell
216: * packets that went through it from ones that get reflected
217: * by a broken gateway). These interfaces are never linked into
218: * the system ifnet list & no routes point to them. I.e., packets
219: * can't be sent this way. They only exist as a placeholder for
220: * multicast source verification.
221: */
222: #if 0
223: struct ifnet multicast_decap_if[MAXVIFS];
224: #endif
225:
226: #define ENCAP_TTL 64
227: #define ENCAP_PROTO IPPROTO_IPIP /* 4 */
228:
229: /* prototype IP hdr for encapsulated packets */
230: struct ip multicast_encap_iphdr = {
231: #if BYTE_ORDER == LITTLE_ENDIAN
232: sizeof(struct ip) >> 2, IPVERSION,
233: #else
234: IPVERSION, sizeof(struct ip) >> 2,
235: #endif
236: 0, /* tos */
237: sizeof(struct ip), /* total length */
238: 0, /* id */
239: 0, /* frag offset */
240: ENCAP_TTL, ENCAP_PROTO,
241: 0, /* checksum */
242: };
243:
244: /*
245: * Bandwidth meter variables and constants
246: */
247:
248: /*
249: * Pending timeouts are stored in a hash table, the key being the
250: * expiration time. Periodically, the entries are analysed and processed.
251: */
252: #define BW_METER_BUCKETS 1024
253: static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
254: struct timeout bw_meter_ch;
255: #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
256:
257: /*
258: * Pending upcalls are stored in a vector which is flushed when
259: * full, or periodically
260: */
261: static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX];
262: static u_int bw_upcalls_n; /* # of pending upcalls */
263: struct timeout bw_upcalls_ch;
264: #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
265:
266: #ifdef PIM
267: struct pimstat pimstat;
268:
269: /*
270: * Note: the PIM Register encapsulation adds the following in front of a
271: * data packet:
272: *
273: * struct pim_encap_hdr {
274: * struct ip ip;
275: * struct pim_encap_pimhdr pim;
276: * }
277: *
278: */
279:
280: struct pim_encap_pimhdr {
281: struct pim pim;
282: uint32_t flags;
283: };
284:
285: static struct ip pim_encap_iphdr = {
286: #if BYTE_ORDER == LITTLE_ENDIAN
287: sizeof(struct ip) >> 2,
288: IPVERSION,
289: #else
290: IPVERSION,
291: sizeof(struct ip) >> 2,
292: #endif
293: 0, /* tos */
294: sizeof(struct ip), /* total length */
295: 0, /* id */
296: 0, /* frag offset */
297: ENCAP_TTL,
298: IPPROTO_PIM,
299: 0, /* checksum */
300: };
301:
302: static struct pim_encap_pimhdr pim_encap_pimhdr = {
303: {
304: PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
305: 0, /* reserved */
306: 0, /* checksum */
307: },
308: 0 /* flags */
309: };
310:
311: static struct ifnet multicast_register_if;
312: static vifi_t reg_vif_num = VIFI_INVALID;
313: #endif /* PIM */
314:
315:
316: /*
317: * Private variables.
318: */
319: static vifi_t numvifs = 0;
320: static int have_encap_tunnel = 0;
321:
322: /*
323: * whether or not special PIM assert processing is enabled.
324: */
325: static int pim_assert;
326: /*
327: * Rate limit for assert notification messages, in usec
328: */
329: #define ASSERT_MSG_TIME 3000000
330:
331: /*
332: * Kernel multicast routing API capabilities and setup.
333: * If more API capabilities are added to the kernel, they should be
334: * recorded in `mrt_api_support'.
335: */
336: static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
337: MRT_MFC_FLAGS_BORDER_VIF |
338: MRT_MFC_RP |
339: MRT_MFC_BW_UPCALL);
340: static u_int32_t mrt_api_config = 0;
341:
342: /*
343: * Find a route for a given origin IP address and Multicast group address
344: * Type of service parameter to be added in the future!!!
345: * Statistics are updated by the caller if needed
346: * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
347: */
348: static struct mfc *
349: mfc_find(struct in_addr *o, struct in_addr *g)
350: {
351: struct mfc *rt;
352:
353: LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
354: if (in_hosteq(rt->mfc_origin, *o) &&
355: in_hosteq(rt->mfc_mcastgrp, *g) &&
356: (rt->mfc_stall == NULL))
357: break;
358: }
359:
360: return (rt);
361: }
362:
363: /*
364: * Macros to compute elapsed time efficiently
365: * Borrowed from Van Jacobson's scheduling code
366: */
367: #define TV_DELTA(a, b, delta) do { \
368: int xxs; \
369: delta = (a).tv_usec - (b).tv_usec; \
370: xxs = (a).tv_sec - (b).tv_sec; \
371: switch (xxs) { \
372: case 2: \
373: delta += 1000000; \
374: /* FALLTHROUGH */ \
375: case 1: \
376: delta += 1000000; \
377: /* FALLTHROUGH */ \
378: case 0: \
379: break; \
380: default: \
381: delta += (1000000 * xxs); \
382: break; \
383: } \
384: } while (/*CONSTCOND*/ 0)
385:
386: #ifdef UPCALL_TIMING
387: u_int32_t upcall_data[51];
388: #endif /* UPCALL_TIMING */
389:
390: /*
391: * Handle MRT setsockopt commands to modify the multicast routing tables.
392: */
393: int
394: ip_mrouter_set(struct socket *so, int optname, struct mbuf **m)
395: {
396: int error;
397:
398: if (optname != MRT_INIT && so != ip_mrouter)
399: error = ENOPROTOOPT;
400: else
401: switch (optname) {
402: case MRT_INIT:
403: error = ip_mrouter_init(so, *m);
404: break;
405: case MRT_DONE:
406: error = ip_mrouter_done();
407: break;
408: case MRT_ADD_VIF:
409: error = add_vif(*m);
410: break;
411: case MRT_DEL_VIF:
412: error = del_vif(*m);
413: break;
414: case MRT_ADD_MFC:
415: error = add_mfc(*m);
416: break;
417: case MRT_DEL_MFC:
418: error = del_mfc(*m);
419: break;
420: case MRT_ASSERT:
421: error = set_assert(*m);
422: break;
423: case MRT_API_CONFIG:
424: error = set_api_config(*m);
425: break;
426: case MRT_ADD_BW_UPCALL:
427: error = add_bw_upcall(*m);
428: break;
429: case MRT_DEL_BW_UPCALL:
430: error = del_bw_upcall(*m);
431: break;
432: default:
433: error = ENOPROTOOPT;
434: break;
435: }
436:
437: if (*m)
438: m_free(*m);
439: return (error);
440: }
441:
442: /*
443: * Handle MRT getsockopt commands
444: */
445: int
446: ip_mrouter_get(struct socket *so, int optname, struct mbuf **m)
447: {
448: int error;
449:
450: if (so != ip_mrouter)
451: error = ENOPROTOOPT;
452: else {
453: *m = m_get(M_WAIT, MT_SOOPTS);
454:
455: switch (optname) {
456: case MRT_VERSION:
457: error = get_version(*m);
458: break;
459: case MRT_ASSERT:
460: error = get_assert(*m);
461: break;
462: case MRT_API_SUPPORT:
463: error = get_api_support(*m);
464: break;
465: case MRT_API_CONFIG:
466: error = get_api_config(*m);
467: break;
468: default:
469: error = ENOPROTOOPT;
470: break;
471: }
472:
473: if (error)
474: m_free(*m);
475: }
476:
477: return (error);
478: }
479:
480: /*
481: * Handle ioctl commands to obtain information from the cache
482: */
483: int
484: mrt_ioctl(struct socket *so, u_long cmd, caddr_t data)
485: {
486: int error;
487:
488: if (so != ip_mrouter)
489: error = EINVAL;
490: else
491: switch (cmd) {
492: case SIOCGETVIFCNT:
493: error = get_vif_cnt((struct sioc_vif_req *)data);
494: break;
495: case SIOCGETSGCNT:
496: error = get_sg_cnt((struct sioc_sg_req *)data);
497: break;
498: default:
499: error = EINVAL;
500: break;
501: }
502:
503: return (error);
504: }
505:
506: /*
507: * returns the packet, byte, rpf-failure count for the source group provided
508: */
509: static int
510: get_sg_cnt(struct sioc_sg_req *req)
511: {
512: int s;
513: struct mfc *rt;
514:
515: s = splsoftnet();
516: rt = mfc_find(&req->src, &req->grp);
517: if (rt == NULL) {
518: splx(s);
519: req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
520: return (EADDRNOTAVAIL);
521: }
522: req->pktcnt = rt->mfc_pkt_cnt;
523: req->bytecnt = rt->mfc_byte_cnt;
524: req->wrong_if = rt->mfc_wrong_if;
525: splx(s);
526:
527: return (0);
528: }
529:
530: /*
531: * returns the input and output packet and byte counts on the vif provided
532: */
533: static int
534: get_vif_cnt(struct sioc_vif_req *req)
535: {
536: vifi_t vifi = req->vifi;
537:
538: if (vifi >= numvifs)
539: return (EINVAL);
540:
541: req->icount = viftable[vifi].v_pkt_in;
542: req->ocount = viftable[vifi].v_pkt_out;
543: req->ibytes = viftable[vifi].v_bytes_in;
544: req->obytes = viftable[vifi].v_bytes_out;
545:
546: return (0);
547: }
548:
549: /*
550: * Enable multicast routing
551: */
552: static int
553: ip_mrouter_init(struct socket *so, struct mbuf *m)
554: {
555: int *v;
556:
557: if (mrtdebug)
558: log(LOG_DEBUG,
559: "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
560: so->so_type, so->so_proto->pr_protocol);
561:
562: if (so->so_type != SOCK_RAW ||
563: so->so_proto->pr_protocol != IPPROTO_IGMP)
564: return (EOPNOTSUPP);
565:
566: if (m == NULL || m->m_len < sizeof(int))
567: return (EINVAL);
568:
569: v = mtod(m, int *);
570: if (*v != 1)
571: return (EINVAL);
572:
573: if (ip_mrouter != NULL)
574: return (EADDRINUSE);
575:
576: ip_mrouter = so;
577:
578: mfchashtbl = hashinit(MFCTBLSIZ, M_MRTABLE, M_WAITOK, &mfchash);
579: bzero((caddr_t)nexpire, sizeof(nexpire));
580:
581: pim_assert = 0;
582:
583: timeout_set(&expire_upcalls_ch, expire_upcalls, NULL);
584: timeout_add(&expire_upcalls_ch, EXPIRE_TIMEOUT);
585:
586: timeout_set(&bw_upcalls_ch, expire_bw_upcalls_send, NULL);
587: timeout_add(&bw_upcalls_ch, BW_UPCALLS_PERIOD);
588:
589: timeout_set(&bw_meter_ch, expire_bw_meter_process, NULL);
590: timeout_add(&bw_meter_ch, BW_METER_PERIOD);
591:
592: if (mrtdebug)
593: log(LOG_DEBUG, "ip_mrouter_init\n");
594:
595: return (0);
596: }
597:
598: /*
599: * Disable multicast routing
600: */
601: int
602: ip_mrouter_done()
603: {
604: vifi_t vifi;
605: struct vif *vifp;
606: int i;
607: int s;
608:
609: s = splsoftnet();
610:
611: /* Clear out all the vifs currently in use. */
612: for (vifi = 0; vifi < numvifs; vifi++) {
613: vifp = &viftable[vifi];
614: if (!in_nullhost(vifp->v_lcl_addr))
615: reset_vif(vifp);
616: }
617:
618: numvifs = 0;
619: pim_assert = 0;
620: mrt_api_config = 0;
621:
622: timeout_del(&expire_upcalls_ch);
623: timeout_del(&bw_upcalls_ch);
624: timeout_del(&bw_meter_ch);
625:
626: /*
627: * Free all multicast forwarding cache entries.
628: */
629: for (i = 0; i < MFCTBLSIZ; i++) {
630: struct mfc *rt, *nrt;
631:
632: for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
633: nrt = LIST_NEXT(rt, mfc_hash);
634:
635: expire_mfc(rt);
636: }
637: }
638:
639: bzero((caddr_t)nexpire, sizeof(nexpire));
640: free(mfchashtbl, M_MRTABLE);
641: mfchashtbl = NULL;
642:
643: bw_upcalls_n = 0;
644: bzero(bw_meter_timers, sizeof(bw_meter_timers));
645:
646: /* Reset de-encapsulation cache. */
647: have_encap_tunnel = 0;
648:
649: ip_mrouter = NULL;
650:
651: splx(s);
652:
653: if (mrtdebug)
654: log(LOG_DEBUG, "ip_mrouter_done\n");
655:
656: return (0);
657: }
658:
659: void
660: ip_mrouter_detach(struct ifnet *ifp)
661: {
662: int vifi, i;
663: struct vif *vifp;
664: struct mfc *rt;
665: struct rtdetq *rte;
666:
667: /* XXX not sure about side effect to userland routing daemon */
668: for (vifi = 0; vifi < numvifs; vifi++) {
669: vifp = &viftable[vifi];
670: if (vifp->v_ifp == ifp)
671: reset_vif(vifp);
672: }
673: for (i = 0; i < MFCTBLSIZ; i++) {
674: if (nexpire[i] == 0)
675: continue;
676: LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
677: for (rte = rt->mfc_stall; rte; rte = rte->next) {
678: if (rte->ifp == ifp)
679: rte->ifp = NULL;
680: }
681: }
682: }
683: }
684:
685: static int
686: get_version(struct mbuf *m)
687: {
688: int *v = mtod(m, int *);
689:
690: *v = 0x0305; /* XXX !!!! */
691: m->m_len = sizeof(int);
692: return (0);
693: }
694:
695: /*
696: * Set PIM assert processing global
697: */
698: static int
699: set_assert(struct mbuf *m)
700: {
701: int *i;
702:
703: if (m == NULL || m->m_len < sizeof(int))
704: return (EINVAL);
705:
706: i = mtod(m, int *);
707: pim_assert = !!*i;
708: return (0);
709: }
710:
711: /*
712: * Get PIM assert processing global
713: */
714: static int
715: get_assert(struct mbuf *m)
716: {
717: int *i = mtod(m, int *);
718:
719: *i = pim_assert;
720: m->m_len = sizeof(int);
721: return (0);
722: }
723:
724: /*
725: * Configure API capabilities
726: */
727: static int
728: set_api_config(struct mbuf *m)
729: {
730: int i;
731: u_int32_t *apival;
732:
733: if (m == NULL || m->m_len < sizeof(u_int32_t))
734: return (EINVAL);
735:
736: apival = mtod(m, u_int32_t *);
737:
738: /*
739: * We can set the API capabilities only if it is the first operation
740: * after MRT_INIT. I.e.:
741: * - there are no vifs installed
742: * - pim_assert is not enabled
743: * - the MFC table is empty
744: */
745: if (numvifs > 0) {
746: *apival = 0;
747: return (EPERM);
748: }
749: if (pim_assert) {
750: *apival = 0;
751: return (EPERM);
752: }
753: for (i = 0; i < MFCTBLSIZ; i++) {
754: if (LIST_FIRST(&mfchashtbl[i]) != NULL) {
755: *apival = 0;
756: return (EPERM);
757: }
758: }
759:
760: mrt_api_config = *apival & mrt_api_support;
761: *apival = mrt_api_config;
762:
763: return (0);
764: }
765:
766: /*
767: * Get API capabilities
768: */
769: static int
770: get_api_support(struct mbuf *m)
771: {
772: u_int32_t *apival;
773:
774: if (m == NULL || m->m_len < sizeof(u_int32_t))
775: return (EINVAL);
776:
777: apival = mtod(m, u_int32_t *);
778:
779: *apival = mrt_api_support;
780:
781: return (0);
782: }
783:
784: /*
785: * Get API configured capabilities
786: */
787: static int
788: get_api_config(struct mbuf *m)
789: {
790: u_int32_t *apival;
791:
792: if (m == NULL || m->m_len < sizeof(u_int32_t))
793: return (EINVAL);
794:
795: apival = mtod(m, u_int32_t *);
796:
797: *apival = mrt_api_config;
798:
799: return (0);
800: }
801:
802: static struct sockaddr_in sin = { sizeof(sin), AF_INET };
803:
804: /*
805: * Add a vif to the vif table
806: */
807: static int
808: add_vif(struct mbuf *m)
809: {
810: struct vifctl *vifcp;
811: struct vif *vifp;
812: struct ifaddr *ifa;
813: struct ifnet *ifp;
814: struct ifreq ifr;
815: int error, s;
816:
817: if (m == NULL || m->m_len < sizeof(struct vifctl))
818: return (EINVAL);
819:
820: vifcp = mtod(m, struct vifctl *);
821: if (vifcp->vifc_vifi >= MAXVIFS)
822: return (EINVAL);
823: if (in_nullhost(vifcp->vifc_lcl_addr))
824: return (EADDRNOTAVAIL);
825:
826: vifp = &viftable[vifcp->vifc_vifi];
827: if (!in_nullhost(vifp->v_lcl_addr))
828: return (EADDRINUSE);
829:
830: /* Find the interface with an address in AF_INET family. */
831: #ifdef PIM
832: if (vifcp->vifc_flags & VIFF_REGISTER) {
833: /*
834: * XXX: Because VIFF_REGISTER does not really need a valid
835: * local interface (e.g. it could be 127.0.0.2), we don't
836: * check its address.
837: */
838: ifp = NULL;
839: } else
840: #endif
841: {
842: sin.sin_addr = vifcp->vifc_lcl_addr;
843: ifa = ifa_ifwithaddr(sintosa(&sin));
844: if (ifa == NULL)
845: return (EADDRNOTAVAIL);
846: ifp = ifa->ifa_ifp;
847: }
848:
849: if (vifcp->vifc_flags & VIFF_TUNNEL) {
850: /* tunnels are no longer supported use gif(4) instead */
851: return (EOPNOTSUPP);
852: #ifdef PIM
853: } else if (vifcp->vifc_flags & VIFF_REGISTER) {
854: ifp = &multicast_register_if;
855: if (mrtdebug)
856: log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
857: (void *)ifp);
858: if (reg_vif_num == VIFI_INVALID) {
859: bzero(ifp, sizeof(*ifp));
860: snprintf(ifp->if_xname, sizeof ifp->if_xname,
861: "register_vif");
862: ifp->if_flags = IFF_LOOPBACK;
863: bzero(&vifp->v_route, sizeof(vifp->v_route));
864: reg_vif_num = vifcp->vifc_vifi;
865: }
866: #endif
867: } else {
868: /* Use the physical interface associated with the address. */
869: ifp = ifa->ifa_ifp;
870:
871: /* Make sure the interface supports multicast. */
872: if ((ifp->if_flags & IFF_MULTICAST) == 0)
873: return (EOPNOTSUPP);
874:
875: /* Enable promiscuous reception of all IP multicasts. */
876: satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
877: satosin(&ifr.ifr_addr)->sin_family = AF_INET;
878: satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
879: error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
880: if (error)
881: return (error);
882: }
883:
884: s = splsoftnet();
885:
886: /* Define parameters for the tbf structure. */
887: vifp->tbf_q = NULL;
888: vifp->tbf_t = &vifp->tbf_q;
889: microtime(&vifp->tbf_last_pkt_t);
890: vifp->tbf_n_tok = 0;
891: vifp->tbf_q_len = 0;
892: vifp->tbf_max_q_len = MAXQSIZE;
893:
894: vifp->v_flags = vifcp->vifc_flags;
895: vifp->v_threshold = vifcp->vifc_threshold;
896: /* scaling up here allows division by 1024 in critical code */
897: vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000;
898: vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
899: vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
900: vifp->v_ifp = ifp;
901: /* Initialize per vif pkt counters. */
902: vifp->v_pkt_in = 0;
903: vifp->v_pkt_out = 0;
904: vifp->v_bytes_in = 0;
905: vifp->v_bytes_out = 0;
906:
907: timeout_del(&vifp->v_repq_ch);
908:
909: #ifdef RSVP_ISI
910: vifp->v_rsvp_on = 0;
911: vifp->v_rsvpd = NULL;
912: #endif /* RSVP_ISI */
913:
914: splx(s);
915:
916: /* Adjust numvifs up if the vifi is higher than numvifs. */
917: if (numvifs <= vifcp->vifc_vifi)
918: numvifs = vifcp->vifc_vifi + 1;
919:
920: if (mrtdebug)
921: log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, "
922: "thresh %x, rate %d\n",
923: vifcp->vifc_vifi,
924: ntohl(vifcp->vifc_lcl_addr.s_addr),
925: (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
926: ntohl(vifcp->vifc_rmt_addr.s_addr),
927: vifcp->vifc_threshold,
928: vifcp->vifc_rate_limit);
929:
930: return (0);
931: }
932:
933: void
934: reset_vif(struct vif *vifp)
935: {
936: struct mbuf *m, *n;
937: struct ifnet *ifp;
938: struct ifreq ifr;
939:
940: timeout_set(&vifp->v_repq_ch, tbf_reprocess_q, vifp);
941:
942: /*
943: * Free packets queued at the interface
944: */
945: for (m = vifp->tbf_q; m != NULL; m = n) {
946: n = m->m_nextpkt;
947: m_freem(m);
948: }
949:
950: if (vifp->v_flags & VIFF_TUNNEL) {
951: /* empty */
952: } else if (vifp->v_flags & VIFF_REGISTER) {
953: #ifdef PIM
954: reg_vif_num = VIFI_INVALID;
955: #endif
956: } else {
957: satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in);
958: satosin(&ifr.ifr_addr)->sin_family = AF_INET;
959: satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr;
960: ifp = vifp->v_ifp;
961: (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
962: }
963: bzero((caddr_t)vifp, sizeof(*vifp));
964: }
965:
966: /*
967: * Delete a vif from the vif table
968: */
969: static int
970: del_vif(struct mbuf *m)
971: {
972: vifi_t *vifip;
973: struct vif *vifp;
974: vifi_t vifi;
975: int s;
976:
977: if (m == NULL || m->m_len < sizeof(vifi_t))
978: return (EINVAL);
979:
980: vifip = mtod(m, vifi_t *);
981: if (*vifip >= numvifs)
982: return (EINVAL);
983:
984: vifp = &viftable[*vifip];
985: if (in_nullhost(vifp->v_lcl_addr))
986: return (EADDRNOTAVAIL);
987:
988: s = splsoftnet();
989:
990: reset_vif(vifp);
991:
992: /* Adjust numvifs down */
993: for (vifi = numvifs; vifi > 0; vifi--)
994: if (!in_nullhost(viftable[vifi - 1].v_lcl_addr))
995: break;
996: numvifs = vifi;
997:
998: splx(s);
999:
1000: if (mrtdebug)
1001: log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
1002:
1003: return (0);
1004: }
1005:
1006: void
1007: vif_delete(struct ifnet *ifp)
1008: {
1009: int i;
1010: struct vif *vifp;
1011: struct mfc *rt;
1012: struct rtdetq *rte;
1013:
1014: for (i = 0; i < numvifs; i++) {
1015: vifp = &viftable[i];
1016: if (vifp->v_ifp == ifp)
1017: bzero((caddr_t)vifp, sizeof *vifp);
1018: }
1019:
1020: for (i = numvifs; i > 0; i--)
1021: if (!in_nullhost(viftable[i - 1].v_lcl_addr))
1022: break;
1023: numvifs = i;
1024:
1025: for (i = 0; i < MFCTBLSIZ; i++) {
1026: if (nexpire[i] == 0)
1027: continue;
1028: LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) {
1029: for (rte = rt->mfc_stall; rte; rte = rte->next) {
1030: if (rte->ifp == ifp)
1031: rte->ifp = NULL;
1032: }
1033: }
1034: }
1035: }
1036:
1037: /*
1038: * update an mfc entry without resetting counters and S,G addresses.
1039: */
1040: static void
1041: update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1042: {
1043: int i;
1044:
1045: rt->mfc_parent = mfccp->mfcc_parent;
1046: for (i = 0; i < numvifs; i++) {
1047: rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1048: rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
1049: MRT_MFC_FLAGS_ALL;
1050: }
1051: /* set the RP address */
1052: if (mrt_api_config & MRT_MFC_RP)
1053: rt->mfc_rp = mfccp->mfcc_rp;
1054: else
1055: rt->mfc_rp = zeroin_addr;
1056: }
1057:
1058: /*
1059: * fully initialize an mfc entry from the parameter.
1060: */
1061: static void
1062: init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1063: {
1064: rt->mfc_origin = mfccp->mfcc_origin;
1065: rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1066:
1067: update_mfc_params(rt, mfccp);
1068:
1069: /* initialize pkt counters per src-grp */
1070: rt->mfc_pkt_cnt = 0;
1071: rt->mfc_byte_cnt = 0;
1072: rt->mfc_wrong_if = 0;
1073: timerclear(&rt->mfc_last_assert);
1074: }
1075:
1076: static void
1077: expire_mfc(struct mfc *rt)
1078: {
1079: struct rtdetq *rte, *nrte;
1080:
1081: free_bw_list(rt->mfc_bw_meter);
1082:
1083: for (rte = rt->mfc_stall; rte != NULL; rte = nrte) {
1084: nrte = rte->next;
1085: m_freem(rte->m);
1086: free(rte, M_MRTABLE);
1087: }
1088:
1089: LIST_REMOVE(rt, mfc_hash);
1090: free(rt, M_MRTABLE);
1091: }
1092:
1093: /*
1094: * Add an mfc entry
1095: */
1096: static int
1097: add_mfc(struct mbuf *m)
1098: {
1099: struct mfcctl2 mfcctl2;
1100: struct mfcctl2 *mfccp;
1101: struct mfc *rt;
1102: u_int32_t hash = 0;
1103: struct rtdetq *rte, *nrte;
1104: u_short nstl;
1105: int s;
1106: int mfcctl_size = sizeof(struct mfcctl);
1107:
1108: if (mrt_api_config & MRT_API_FLAGS_ALL)
1109: mfcctl_size = sizeof(struct mfcctl2);
1110:
1111: if (m == NULL || m->m_len < mfcctl_size)
1112: return (EINVAL);
1113:
1114: /*
1115: * select data size depending on API version.
1116: */
1117: if (mrt_api_config & MRT_API_FLAGS_ALL) {
1118: struct mfcctl2 *mp2 = mtod(m, struct mfcctl2 *);
1119: bcopy(mp2, (caddr_t)&mfcctl2, sizeof(*mp2));
1120: } else {
1121: struct mfcctl *mp = mtod(m, struct mfcctl *);
1122: bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp));
1123: bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl),
1124: sizeof(mfcctl2) - sizeof(struct mfcctl));
1125: }
1126: mfccp = &mfcctl2;
1127:
1128: s = splsoftnet();
1129: rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1130:
1131: /* If an entry already exists, just update the fields */
1132: if (rt) {
1133: if (mrtdebug & DEBUG_MFC)
1134: log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n",
1135: ntohl(mfccp->mfcc_origin.s_addr),
1136: ntohl(mfccp->mfcc_mcastgrp.s_addr),
1137: mfccp->mfcc_parent);
1138:
1139: update_mfc_params(rt, mfccp);
1140:
1141: splx(s);
1142: return (0);
1143: }
1144:
1145: /*
1146: * Find the entry for which the upcall was made and update
1147: */
1148: nstl = 0;
1149: hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1150: LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1151: if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1152: in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1153: rt->mfc_stall != NULL) {
1154: if (nstl++)
1155: log(LOG_ERR, "add_mfc %s o %x g %x "
1156: "p %x dbx %p\n",
1157: "multiple kernel entries",
1158: ntohl(mfccp->mfcc_origin.s_addr),
1159: ntohl(mfccp->mfcc_mcastgrp.s_addr),
1160: mfccp->mfcc_parent, rt->mfc_stall);
1161:
1162: if (mrtdebug & DEBUG_MFC)
1163: log(LOG_DEBUG, "add_mfc o %x g %x "
1164: "p %x dbg %p\n",
1165: ntohl(mfccp->mfcc_origin.s_addr),
1166: ntohl(mfccp->mfcc_mcastgrp.s_addr),
1167: mfccp->mfcc_parent, rt->mfc_stall);
1168:
1169: rte = rt->mfc_stall;
1170: init_mfc_params(rt, mfccp);
1171: rt->mfc_stall = NULL;
1172:
1173: rt->mfc_expire = 0; /* Don't clean this guy up */
1174: nexpire[hash]--;
1175:
1176: /* free packets Qed at the end of this entry */
1177: for (; rte != NULL; rte = nrte) {
1178: nrte = rte->next;
1179: if (rte->ifp) {
1180: #ifdef RSVP_ISI
1181: ip_mdq(rte->m, rte->ifp, rt, -1);
1182: #else
1183: ip_mdq(rte->m, rte->ifp, rt);
1184: #endif /* RSVP_ISI */
1185: }
1186: m_freem(rte->m);
1187: #ifdef UPCALL_TIMING
1188: collate(&rte->t);
1189: #endif /* UPCALL_TIMING */
1190: free(rte, M_MRTABLE);
1191: }
1192: }
1193: }
1194:
1195: /*
1196: * It is possible that an entry is being inserted without an upcall
1197: */
1198: if (nstl == 0) {
1199: /*
1200: * No mfc; make a new one
1201: */
1202: if (mrtdebug & DEBUG_MFC)
1203: log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n",
1204: ntohl(mfccp->mfcc_origin.s_addr),
1205: ntohl(mfccp->mfcc_mcastgrp.s_addr),
1206: mfccp->mfcc_parent);
1207:
1208: LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1209: if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1210: in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1211: init_mfc_params(rt, mfccp);
1212: if (rt->mfc_expire)
1213: nexpire[hash]--;
1214: rt->mfc_expire = 0;
1215: break; /* XXX */
1216: }
1217: }
1218: if (rt == NULL) { /* no upcall, so make a new entry */
1219: rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE,
1220: M_NOWAIT);
1221: if (rt == NULL) {
1222: splx(s);
1223: return (ENOBUFS);
1224: }
1225:
1226: init_mfc_params(rt, mfccp);
1227: rt->mfc_expire = 0;
1228: rt->mfc_stall = NULL;
1229: rt->mfc_bw_meter = NULL;
1230:
1231: /* insert new entry at head of hash chain */
1232: LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1233: }
1234: }
1235:
1236: splx(s);
1237: return (0);
1238: }
1239:
1240: #ifdef UPCALL_TIMING
1241: /*
1242: * collect delay statistics on the upcalls
1243: */
1244: static void
1245: collate(struct timeval *t)
1246: {
1247: u_int32_t d;
1248: struct timeval tp;
1249: u_int32_t delta;
1250:
1251: microtime(&tp);
1252:
1253: if (timercmp(t, &tp, <)) {
1254: TV_DELTA(tp, *t, delta);
1255:
1256: d = delta >> 10;
1257: if (d > 50)
1258: d = 50;
1259:
1260: ++upcall_data[d];
1261: }
1262: }
1263: #endif /* UPCALL_TIMING */
1264:
1265: /*
1266: * Delete an mfc entry
1267: */
1268: static int
1269: del_mfc(struct mbuf *m)
1270: {
1271: struct mfcctl2 mfcctl2;
1272: struct mfcctl2 *mfccp;
1273: struct mfc *rt;
1274: int s;
1275: int mfcctl_size = sizeof(struct mfcctl);
1276: struct mfcctl *mp = mtod(m, struct mfcctl *);
1277:
1278: /*
1279: * XXX: for deleting MFC entries the information in entries
1280: * of size "struct mfcctl" is sufficient.
1281: */
1282:
1283: if (m == NULL || m->m_len < mfcctl_size)
1284: return (EINVAL);
1285:
1286: bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp));
1287: bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl),
1288: sizeof(mfcctl2) - sizeof(struct mfcctl));
1289:
1290: mfccp = &mfcctl2;
1291:
1292: if (mrtdebug & DEBUG_MFC)
1293: log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n",
1294: ntohl(mfccp->mfcc_origin.s_addr),
1295: ntohl(mfccp->mfcc_mcastgrp.s_addr));
1296:
1297: s = splsoftnet();
1298:
1299: rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1300: if (rt == NULL) {
1301: splx(s);
1302: return (EADDRNOTAVAIL);
1303: }
1304:
1305: /*
1306: * free the bw_meter entries
1307: */
1308: free_bw_list(rt->mfc_bw_meter);
1309: rt->mfc_bw_meter = NULL;
1310:
1311: LIST_REMOVE(rt, mfc_hash);
1312: free(rt, M_MRTABLE);
1313:
1314: splx(s);
1315: return (0);
1316: }
1317:
1318: static int
1319: socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1320: {
1321: if (s != NULL) {
1322: if (sbappendaddr(&s->so_rcv, sintosa(src), mm,
1323: (struct mbuf *)NULL) != 0) {
1324: sorwakeup(s);
1325: return (0);
1326: }
1327: }
1328: m_freem(mm);
1329: return (-1);
1330: }
1331:
1332: /*
1333: * IP multicast forwarding function. This function assumes that the packet
1334: * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1335: * pointed to by "ifp", and the packet is to be relayed to other networks
1336: * that have members of the packet's destination IP multicast group.
1337: *
1338: * The packet is returned unscathed to the caller, unless it is
1339: * erroneous, in which case a non-zero return value tells the caller to
1340: * discard it.
1341: */
1342:
1343: #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */
1344: #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1345:
1346: int
1347: #ifdef RSVP_ISI
1348: ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo)
1349: #else
1350: ip_mforward(struct mbuf *m, struct ifnet *ifp)
1351: #endif /* RSVP_ISI */
1352: {
1353: struct ip *ip = mtod(m, struct ip *);
1354: struct mfc *rt;
1355: static int srctun = 0;
1356: struct mbuf *mm;
1357: int s;
1358: vifi_t vifi;
1359:
1360: if (mrtdebug & DEBUG_FORWARD)
1361: log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n",
1362: ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1363:
1364: if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1365: ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1366: /*
1367: * Packet arrived via a physical interface or
1368: * an encapsulated tunnel or a register_vif.
1369: */
1370: } else {
1371: /*
1372: * Packet arrived through a source-route tunnel.
1373: * Source-route tunnels are no longer supported.
1374: */
1375: if ((srctun++ % 1000) == 0)
1376: log(LOG_ERR, "ip_mforward: received source-routed "
1377: "packet from %x\n", ntohl(ip->ip_src.s_addr));
1378:
1379: return (1);
1380: }
1381:
1382: #ifdef RSVP_ISI
1383: if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1384: if (ip->ip_ttl < 255) {
1385: /* compensate for -1 in *_send routines */
1386: ip->ip_ttl++;
1387: }
1388: if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1389: struct vif *vifp = viftable + vifi;
1390: printf("Sending IPPROTO_RSVP from %x to %x on "
1391: "vif %d (%s%s)\n",
1392: ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi,
1393: (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1394: vifp->v_ifp->if_xname);
1395: }
1396: return (ip_mdq(m, ifp, (struct mfc *)NULL, vifi));
1397: }
1398: if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1399: printf("Warning: IPPROTO_RSVP from %x to %x without "
1400: "vif option\n", ntohl(ip->ip_src), ntohl(ip->ip_dst));
1401: }
1402: #endif /* RSVP_ISI */
1403:
1404: /*
1405: * Don't forward a packet with time-to-live of zero or one,
1406: * or a packet destined to a local-only group.
1407: */
1408: if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr))
1409: return (0);
1410:
1411: /*
1412: * Determine forwarding vifs from the forwarding cache table
1413: */
1414: s = splsoftnet();
1415: ++mrtstat.mrts_mfc_lookups;
1416: rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1417:
1418: /* Entry exists, so forward if necessary */
1419: if (rt != NULL) {
1420: splx(s);
1421: #ifdef RSVP_ISI
1422: return (ip_mdq(m, ifp, rt, -1));
1423: #else
1424: return (ip_mdq(m, ifp, rt));
1425: #endif /* RSVP_ISI */
1426: } else {
1427: /*
1428: * If we don't have a route for packet's origin,
1429: * Make a copy of the packet & send message to routing daemon
1430: */
1431:
1432: struct mbuf *mb0;
1433: struct rtdetq *rte;
1434: u_int32_t hash;
1435: int hlen = ip->ip_hl << 2;
1436: #ifdef UPCALL_TIMING
1437: struct timeval tp;
1438:
1439: microtime(&tp);
1440: #endif /* UPCALL_TIMING */
1441:
1442: ++mrtstat.mrts_mfc_misses;
1443:
1444: mrtstat.mrts_no_route++;
1445: if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1446: log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1447: ntohl(ip->ip_src.s_addr),
1448: ntohl(ip->ip_dst.s_addr));
1449:
1450: /*
1451: * Allocate mbufs early so that we don't do extra work if we are
1452: * just going to fail anyway. Make sure to pullup the header so
1453: * that other people can't step on it.
1454: */
1455: rte = (struct rtdetq *)malloc(sizeof(*rte),
1456: M_MRTABLE, M_NOWAIT);
1457: if (rte == NULL) {
1458: splx(s);
1459: return (ENOBUFS);
1460: }
1461: mb0 = m_copy(m, 0, M_COPYALL);
1462: M_PULLUP(mb0, hlen);
1463: if (mb0 == NULL) {
1464: free(rte, M_MRTABLE);
1465: splx(s);
1466: return (ENOBUFS);
1467: }
1468:
1469: /* is there an upcall waiting for this flow? */
1470: hash = MFCHASH(ip->ip_src, ip->ip_dst);
1471: LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) {
1472: if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1473: in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1474: rt->mfc_stall != NULL)
1475: break;
1476: }
1477:
1478: if (rt == NULL) {
1479: int i;
1480: struct igmpmsg *im;
1481:
1482: /*
1483: * Locate the vifi for the incoming interface for
1484: * this packet.
1485: * If none found, drop packet.
1486: */
1487: for (vifi = 0; vifi < numvifs &&
1488: viftable[vifi].v_ifp != ifp; vifi++)
1489: ;
1490: if (vifi >= numvifs) /* vif not found, drop packet */
1491: goto non_fatal;
1492:
1493: /* no upcall, so make a new entry */
1494: rt = (struct mfc *)malloc(sizeof(*rt),
1495: M_MRTABLE, M_NOWAIT);
1496: if (rt == NULL)
1497: goto fail;
1498: /*
1499: * Make a copy of the header to send to the user level
1500: * process
1501: */
1502: mm = m_copy(m, 0, hlen);
1503: M_PULLUP(mm, hlen);
1504: if (mm == NULL)
1505: goto fail1;
1506:
1507: /*
1508: * Send message to routing daemon to install
1509: * a route into the kernel table
1510: */
1511:
1512: im = mtod(mm, struct igmpmsg *);
1513: im->im_msgtype = IGMPMSG_NOCACHE;
1514: im->im_mbz = 0;
1515: im->im_vif = vifi;
1516:
1517: mrtstat.mrts_upcalls++;
1518:
1519: sin.sin_addr = ip->ip_src;
1520: if (socket_send(ip_mrouter, mm, &sin) < 0) {
1521: log(LOG_WARNING, "ip_mforward: ip_mrouter "
1522: "socket queue full\n");
1523: ++mrtstat.mrts_upq_sockfull;
1524: fail1:
1525: free(rt, M_MRTABLE);
1526: fail:
1527: free(rte, M_MRTABLE);
1528: m_freem(mb0);
1529: splx(s);
1530: return (ENOBUFS);
1531: }
1532:
1533: /* insert new entry at head of hash chain */
1534: rt->mfc_origin = ip->ip_src;
1535: rt->mfc_mcastgrp = ip->ip_dst;
1536: rt->mfc_pkt_cnt = 0;
1537: rt->mfc_byte_cnt = 0;
1538: rt->mfc_wrong_if = 0;
1539: rt->mfc_expire = UPCALL_EXPIRE;
1540: nexpire[hash]++;
1541: for (i = 0; i < numvifs; i++) {
1542: rt->mfc_ttls[i] = 0;
1543: rt->mfc_flags[i] = 0;
1544: }
1545: rt->mfc_parent = -1;
1546:
1547: /* clear the RP address */
1548: rt->mfc_rp = zeroin_addr;
1549:
1550: rt->mfc_bw_meter = NULL;
1551:
1552: /* link into table */
1553: LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash);
1554: /* Add this entry to the end of the queue */
1555: rt->mfc_stall = rte;
1556: } else {
1557: /* determine if q has overflowed */
1558: struct rtdetq **p;
1559: int npkts = 0;
1560:
1561: /*
1562: * XXX ouch! we need to append to the list, but we
1563: * only have a pointer to the front, so we have to
1564: * scan the entire list every time.
1565: */
1566: for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1567: if (++npkts > MAX_UPQ) {
1568: mrtstat.mrts_upq_ovflw++;
1569: non_fatal:
1570: free(rte, M_MRTABLE);
1571: m_freem(mb0);
1572: splx(s);
1573: return (0);
1574: }
1575:
1576: /* Add this entry to the end of the queue */
1577: *p = rte;
1578: }
1579:
1580: rte->next = NULL;
1581: rte->m = mb0;
1582: rte->ifp = ifp;
1583: #ifdef UPCALL_TIMING
1584: rte->t = tp;
1585: #endif /* UPCALL_TIMING */
1586:
1587: splx(s);
1588:
1589: return (0);
1590: }
1591: }
1592:
1593:
1594: /*ARGSUSED*/
1595: static void
1596: expire_upcalls(void *v)
1597: {
1598: int i;
1599: int s;
1600:
1601: s = splsoftnet();
1602:
1603: for (i = 0; i < MFCTBLSIZ; i++) {
1604: struct mfc *rt, *nrt;
1605:
1606: if (nexpire[i] == 0)
1607: continue;
1608:
1609: for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) {
1610: nrt = LIST_NEXT(rt, mfc_hash);
1611:
1612: if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1613: continue;
1614: nexpire[i]--;
1615:
1616: /*
1617: * free the bw_meter entries
1618: */
1619: while (rt->mfc_bw_meter != NULL) {
1620: struct bw_meter *x = rt->mfc_bw_meter;
1621:
1622: rt->mfc_bw_meter = x->bm_mfc_next;
1623: free(x, M_BWMETER);
1624: }
1625:
1626: ++mrtstat.mrts_cache_cleanups;
1627: if (mrtdebug & DEBUG_EXPIRE)
1628: log(LOG_DEBUG,
1629: "expire_upcalls: expiring (%x %x)\n",
1630: ntohl(rt->mfc_origin.s_addr),
1631: ntohl(rt->mfc_mcastgrp.s_addr));
1632:
1633: expire_mfc(rt);
1634: }
1635: }
1636:
1637: splx(s);
1638: timeout_add(&expire_upcalls_ch, EXPIRE_TIMEOUT);
1639: }
1640:
1641: /*
1642: * Packet forwarding routine once entry in the cache is made
1643: */
1644: static int
1645: #ifdef RSVP_ISI
1646: ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1647: #else
1648: ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt)
1649: #endif /* RSVP_ISI */
1650: {
1651: struct ip *ip = mtod(m, struct ip *);
1652: vifi_t vifi;
1653: struct vif *vifp;
1654: int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2);
1655:
1656: /*
1657: * Macro to send packet on vif. Since RSVP packets don't get counted on
1658: * input, they shouldn't get counted on output, so statistics keeping is
1659: * separate.
1660: */
1661: #define MC_SEND(ip, vifp, m) do { \
1662: if ((vifp)->v_flags & VIFF_TUNNEL) \
1663: encap_send((ip), (vifp), (m)); \
1664: else \
1665: phyint_send((ip), (vifp), (m)); \
1666: } while (/*CONSTCOND*/ 0)
1667:
1668: #ifdef RSVP_ISI
1669: /*
1670: * If xmt_vif is not -1, send on only the requested vif.
1671: *
1672: * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.
1673: */
1674: if (xmt_vif < numvifs) {
1675: #ifdef PIM
1676: if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1677: pim_register_send(ip, viftable + xmt_vif, m, rt);
1678: else
1679: #endif
1680: MC_SEND(ip, viftable + xmt_vif, m);
1681: return (1);
1682: }
1683: #endif /* RSVP_ISI */
1684:
1685: /*
1686: * Don't forward if it didn't arrive from the parent vif for its origin.
1687: */
1688: vifi = rt->mfc_parent;
1689: if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1690: /* came in the wrong interface */
1691: if (mrtdebug & DEBUG_FORWARD)
1692: log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1693: ifp, vifi,
1694: vifi >= numvifs ? 0 : viftable[vifi].v_ifp);
1695: ++mrtstat.mrts_wrong_if;
1696: ++rt->mfc_wrong_if;
1697: /*
1698: * If we are doing PIM assert processing, send a message
1699: * to the routing daemon.
1700: *
1701: * XXX: A PIM-SM router needs the WRONGVIF detection so it
1702: * can complete the SPT switch, regardless of the type
1703: * of interface (broadcast media, GRE tunnel, etc).
1704: */
1705: if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1706: struct timeval now;
1707: u_int32_t delta;
1708:
1709: #ifdef PIM
1710: if (ifp == &multicast_register_if)
1711: pimstat.pims_rcv_registers_wrongiif++;
1712: #endif
1713:
1714: /* Get vifi for the incoming packet */
1715: for (vifi = 0;
1716: vifi < numvifs && viftable[vifi].v_ifp != ifp;
1717: vifi++)
1718: ;
1719: if (vifi >= numvifs) {
1720: /* The iif is not found: ignore the packet. */
1721: return (0);
1722: }
1723:
1724: if (rt->mfc_flags[vifi] &
1725: MRT_MFC_FLAGS_DISABLE_WRONGVIF) {
1726: /* WRONGVIF disabled: ignore the packet */
1727: return (0);
1728: }
1729:
1730: microtime(&now);
1731:
1732: TV_DELTA(rt->mfc_last_assert, now, delta);
1733:
1734: if (delta > ASSERT_MSG_TIME) {
1735: struct igmpmsg *im;
1736: int hlen = ip->ip_hl << 2;
1737: struct mbuf *mm = m_copy(m, 0, hlen);
1738:
1739: M_PULLUP(mm, hlen);
1740: if (mm == NULL)
1741: return (ENOBUFS);
1742:
1743: rt->mfc_last_assert = now;
1744:
1745: im = mtod(mm, struct igmpmsg *);
1746: im->im_msgtype = IGMPMSG_WRONGVIF;
1747: im->im_mbz = 0;
1748: im->im_vif = vifi;
1749:
1750: mrtstat.mrts_upcalls++;
1751:
1752: sin.sin_addr = im->im_src;
1753: if (socket_send(ip_mrouter, mm, &sin) < 0) {
1754: log(LOG_WARNING, "ip_mforward: "
1755: "ip_mrouter socket queue full\n");
1756: ++mrtstat.mrts_upq_sockfull;
1757: return (ENOBUFS);
1758: }
1759: }
1760: }
1761: return (0);
1762: }
1763:
1764: /* If I sourced this packet, it counts as output, else it was input. */
1765: if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) {
1766: viftable[vifi].v_pkt_out++;
1767: viftable[vifi].v_bytes_out += plen;
1768: } else {
1769: viftable[vifi].v_pkt_in++;
1770: viftable[vifi].v_bytes_in += plen;
1771: }
1772: rt->mfc_pkt_cnt++;
1773: rt->mfc_byte_cnt += plen;
1774:
1775: /*
1776: * For each vif, decide if a copy of the packet should be forwarded.
1777: * Forward if:
1778: * - the ttl exceeds the vif's threshold
1779: * - there are group members downstream on interface
1780: */
1781: for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1782: if ((rt->mfc_ttls[vifi] > 0) &&
1783: (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1784: vifp->v_pkt_out++;
1785: vifp->v_bytes_out += plen;
1786: #ifdef PIM
1787: if (vifp->v_flags & VIFF_REGISTER)
1788: pim_register_send(ip, vifp, m, rt);
1789: else
1790: #endif
1791: MC_SEND(ip, vifp, m);
1792: }
1793:
1794: /*
1795: * Perform upcall-related bw measuring.
1796: */
1797: if (rt->mfc_bw_meter != NULL) {
1798: struct bw_meter *x;
1799: struct timeval now;
1800:
1801: microtime(&now);
1802: for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1803: bw_meter_receive_packet(x, plen, &now);
1804: }
1805:
1806: return (0);
1807: }
1808:
1809: #ifdef RSVP_ISI
1810: /*
1811: * check if a vif number is legal/ok. This is used by ip_output.
1812: */
1813: int
1814: legal_vif_num(int vif)
1815: {
1816: if (vif >= 0 && vif < numvifs)
1817: return (1);
1818: else
1819: return (0);
1820: }
1821: #endif /* RSVP_ISI */
1822:
1823: static void
1824: phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1825: {
1826: struct mbuf *mb_copy;
1827: int hlen = ip->ip_hl << 2;
1828:
1829: /*
1830: * Make a new reference to the packet; make sure that
1831: * the IP header is actually copied, not just referenced,
1832: * so that ip_output() only scribbles on the copy.
1833: */
1834: mb_copy = m_copy(m, 0, M_COPYALL);
1835: M_PULLUP(mb_copy, hlen);
1836: if (mb_copy == NULL)
1837: return;
1838:
1839: if (vifp->v_rate_limit <= 0)
1840: tbf_send_packet(vifp, mb_copy);
1841: else
1842: tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *),
1843: ntohs(ip->ip_len));
1844: }
1845:
1846: static void
1847: encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1848: {
1849: struct mbuf *mb_copy;
1850: struct ip *ip_copy;
1851: int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr);
1852:
1853: /* Take care of delayed checksums */
1854: if (m->m_pkthdr.csum_flags & (M_TCPV4_CSUM_OUT | M_UDPV4_CSUM_OUT)) {
1855: in_delayed_cksum(m);
1856: m->m_pkthdr.csum_flags &=
1857: ~(M_UDPV4_CSUM_OUT | M_TCPV4_CSUM_OUT);
1858: }
1859:
1860: /*
1861: * copy the old packet & pullup its IP header into the
1862: * new mbuf so we can modify it. Try to fill the new
1863: * mbuf since if we don't the ethernet driver will.
1864: */
1865: MGETHDR(mb_copy, M_DONTWAIT, MT_DATA);
1866: if (mb_copy == NULL)
1867: return;
1868: mb_copy->m_data += max_linkhdr;
1869: mb_copy->m_pkthdr.len = len;
1870: mb_copy->m_len = sizeof(multicast_encap_iphdr);
1871:
1872: if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1873: m_freem(mb_copy);
1874: return;
1875: }
1876: i = MHLEN - max_linkhdr;
1877: if (i > len)
1878: i = len;
1879: mb_copy = m_pullup(mb_copy, i);
1880: if (mb_copy == NULL)
1881: return;
1882:
1883: /*
1884: * fill in the encapsulating IP header.
1885: */
1886: ip_copy = mtod(mb_copy, struct ip *);
1887: *ip_copy = multicast_encap_iphdr;
1888: ip_copy->ip_id = htons(ip_randomid());
1889: ip_copy->ip_len = htons(len);
1890: ip_copy->ip_src = vifp->v_lcl_addr;
1891: ip_copy->ip_dst = vifp->v_rmt_addr;
1892:
1893: /*
1894: * turn the encapsulated IP header back into a valid one.
1895: */
1896: ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1897: --ip->ip_ttl;
1898: ip->ip_sum = 0;
1899: mb_copy->m_data += sizeof(multicast_encap_iphdr);
1900: ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1901: mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1902:
1903: if (vifp->v_rate_limit <= 0)
1904: tbf_send_packet(vifp, mb_copy);
1905: else
1906: tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len));
1907: }
1908:
1909: /*
1910: * Token bucket filter module
1911: */
1912: static void
1913: tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len)
1914: {
1915:
1916: if (len > MAX_BKT_SIZE) {
1917: /* drop if packet is too large */
1918: mrtstat.mrts_pkt2large++;
1919: m_freem(m);
1920: return;
1921: }
1922:
1923: tbf_update_tokens(vifp);
1924:
1925: /*
1926: * If there are enough tokens, and the queue is empty, send this packet
1927: * out immediately. Otherwise, try to insert it on this vif's queue.
1928: */
1929: if (vifp->tbf_q_len == 0) {
1930: if (len <= vifp->tbf_n_tok) {
1931: vifp->tbf_n_tok -= len;
1932: tbf_send_packet(vifp, m);
1933: } else {
1934: /* queue packet and timeout till later */
1935: tbf_queue(vifp, m);
1936: timeout_add(&vifp->v_repq_ch, TBF_REPROCESS);
1937: }
1938: } else {
1939: if (vifp->tbf_q_len >= vifp->tbf_max_q_len &&
1940: !tbf_dq_sel(vifp, ip)) {
1941: /* queue full, and couldn't make room */
1942: mrtstat.mrts_q_overflow++;
1943: m_freem(m);
1944: } else {
1945: /* queue length low enough, or made room */
1946: tbf_queue(vifp, m);
1947: tbf_process_q(vifp);
1948: }
1949: }
1950: }
1951:
1952: /*
1953: * adds a packet to the queue at the interface
1954: */
1955: static void
1956: tbf_queue(struct vif *vifp, struct mbuf *m)
1957: {
1958: int s = splsoftnet();
1959:
1960: /* insert at tail */
1961: *vifp->tbf_t = m;
1962: vifp->tbf_t = &m->m_nextpkt;
1963: vifp->tbf_q_len++;
1964:
1965: splx(s);
1966: }
1967:
1968:
1969: /*
1970: * processes the queue at the interface
1971: */
1972: static void
1973: tbf_process_q(struct vif *vifp)
1974: {
1975: struct mbuf *m;
1976: int len;
1977: int s = splsoftnet();
1978:
1979: /*
1980: * Loop through the queue at the interface and send as many packets
1981: * as possible.
1982: */
1983: for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) {
1984: len = ntohs(mtod(m, struct ip *)->ip_len);
1985:
1986: /* determine if the packet can be sent */
1987: if (len <= vifp->tbf_n_tok) {
1988: /* if so,
1989: * reduce no of tokens, dequeue the packet,
1990: * send the packet.
1991: */
1992: if ((vifp->tbf_q = m->m_nextpkt) == NULL)
1993: vifp->tbf_t = &vifp->tbf_q;
1994: --vifp->tbf_q_len;
1995:
1996: m->m_nextpkt = NULL;
1997: vifp->tbf_n_tok -= len;
1998: tbf_send_packet(vifp, m);
1999: } else
2000: break;
2001: }
2002: splx(s);
2003: }
2004:
2005: static void
2006: tbf_reprocess_q(void *arg)
2007: {
2008: struct vif *vifp = arg;
2009:
2010: if (ip_mrouter == NULL)
2011: return;
2012:
2013: tbf_update_tokens(vifp);
2014: tbf_process_q(vifp);
2015:
2016: if (vifp->tbf_q_len != 0)
2017: timeout_add(&vifp->v_repq_ch, TBF_REPROCESS);
2018: }
2019:
2020: /* function that will selectively discard a member of the queue
2021: * based on the precedence value and the priority
2022: */
2023: static int
2024: tbf_dq_sel(struct vif *vifp, struct ip *ip)
2025: {
2026: u_int p;
2027: struct mbuf **mp, *m;
2028: int s = splsoftnet();
2029:
2030: p = priority(vifp, ip);
2031:
2032: for (mp = &vifp->tbf_q, m = *mp;
2033: m != NULL;
2034: mp = &m->m_nextpkt, m = *mp) {
2035: if (p > priority(vifp, mtod(m, struct ip *))) {
2036: if ((*mp = m->m_nextpkt) == NULL)
2037: vifp->tbf_t = mp;
2038: --vifp->tbf_q_len;
2039:
2040: m_freem(m);
2041: mrtstat.mrts_drop_sel++;
2042: splx(s);
2043: return (1);
2044: }
2045: }
2046: splx(s);
2047: return (0);
2048: }
2049:
2050: static void
2051: tbf_send_packet(struct vif *vifp, struct mbuf *m)
2052: {
2053: int error;
2054: int s = splsoftnet();
2055:
2056: if (vifp->v_flags & VIFF_TUNNEL) {
2057: /* If tunnel options */
2058: ip_output(m, (struct mbuf *)NULL, &vifp->v_route,
2059: IP_FORWARDING, (struct ip_moptions *)NULL,
2060: (struct inpcb *)NULL);
2061: } else {
2062: /*
2063: * if physical interface option, extract the options
2064: * and then send
2065: */
2066: struct ip_moptions imo;
2067:
2068: imo.imo_multicast_ifp = vifp->v_ifp;
2069: imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
2070: imo.imo_multicast_loop = 1;
2071: #ifdef RSVP_ISI
2072: imo.imo_multicast_vif = -1;
2073: #endif
2074:
2075: error = ip_output(m, (struct mbuf *)NULL, (struct route *)NULL,
2076: IP_FORWARDING|IP_MULTICASTOPTS, &imo,
2077: (struct inpcb *)NULL);
2078:
2079: if (mrtdebug & DEBUG_XMIT)
2080: log(LOG_DEBUG, "phyint_send on vif %ld err %d\n",
2081: (long)(vifp - viftable), error);
2082: }
2083: splx(s);
2084: }
2085:
2086: /* determine the current time and then
2087: * the elapsed time (between the last time and time now)
2088: * in milliseconds & update the no. of tokens in the bucket
2089: */
2090: static void
2091: tbf_update_tokens(struct vif *vifp)
2092: {
2093: struct timeval tp;
2094: u_int32_t tm;
2095: int s = splsoftnet();
2096:
2097: microtime(&tp);
2098:
2099: TV_DELTA(tp, vifp->tbf_last_pkt_t, tm);
2100:
2101: /*
2102: * This formula is actually
2103: * "time in seconds" * "bytes/second".
2104: *
2105: * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
2106: *
2107: * The (1000/1024) was introduced in add_vif to optimize
2108: * this divide into a shift.
2109: */
2110: vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192;
2111: vifp->tbf_last_pkt_t = tp;
2112:
2113: if (vifp->tbf_n_tok > MAX_BKT_SIZE)
2114: vifp->tbf_n_tok = MAX_BKT_SIZE;
2115:
2116: splx(s);
2117: }
2118:
2119: static int
2120: priority(struct vif *vifp, struct ip *ip)
2121: {
2122: int prio = 50; /* the lowest priority -- default case */
2123:
2124: /* temporary hack; may add general packet classifier some day */
2125:
2126: /*
2127: * The UDP port space is divided up into four priority ranges:
2128: * [0, 16384) : unclassified - lowest priority
2129: * [16384, 32768) : audio - highest priority
2130: * [32768, 49152) : whiteboard - medium priority
2131: * [49152, 65536) : video - low priority
2132: */
2133: if (ip->ip_p == IPPROTO_UDP) {
2134: struct udphdr *udp =
2135: (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2136:
2137: switch (ntohs(udp->uh_dport) & 0xc000) {
2138: case 0x4000:
2139: prio = 70;
2140: break;
2141: case 0x8000:
2142: prio = 60;
2143: break;
2144: case 0xc000:
2145: prio = 55;
2146: break;
2147: }
2148:
2149: if (tbfdebug > 1)
2150: log(LOG_DEBUG, "port %x prio %d\n",
2151: ntohs(udp->uh_dport), prio);
2152: }
2153:
2154: return (prio);
2155: }
2156:
2157: /*
2158: * End of token bucket filter modifications
2159: */
2160: #ifdef RSVP_ISI
2161: int
2162: ip_rsvp_vif_init(struct socket *so, struct mbuf *m)
2163: {
2164: int vifi, s;
2165:
2166: if (rsvpdebug)
2167: printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2168: so->so_type, so->so_proto->pr_protocol);
2169:
2170: if (so->so_type != SOCK_RAW ||
2171: so->so_proto->pr_protocol != IPPROTO_RSVP)
2172: return (EOPNOTSUPP);
2173:
2174: /* Check mbuf. */
2175: if (m == NULL || m->m_len != sizeof(int)) {
2176: return (EINVAL);
2177: }
2178: vifi = *(mtod(m, int *));
2179:
2180: if (rsvpdebug)
2181: printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",
2182: vifi, rsvp_on);
2183:
2184: s = splsoftnet();
2185:
2186: /* Check vif. */
2187: if (!legal_vif_num(vifi)) {
2188: splx(s);
2189: return (EADDRNOTAVAIL);
2190: }
2191:
2192: /* Check if socket is available. */
2193: if (viftable[vifi].v_rsvpd != NULL) {
2194: splx(s);
2195: return (EADDRINUSE);
2196: }
2197:
2198: viftable[vifi].v_rsvpd = so;
2199: /* This may seem silly, but we need to be sure we don't over-increment
2200: * the RSVP counter, in case something slips up.
2201: */
2202: if (!viftable[vifi].v_rsvp_on) {
2203: viftable[vifi].v_rsvp_on = 1;
2204: rsvp_on++;
2205: }
2206:
2207: splx(s);
2208: return (0);
2209: }
2210:
2211: int
2212: ip_rsvp_vif_done(struct socket *so, struct mbuf *m)
2213: {
2214: int vifi, s;
2215:
2216: if (rsvpdebug)
2217: printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2218: so->so_type, so->so_proto->pr_protocol);
2219:
2220: if (so->so_type != SOCK_RAW ||
2221: so->so_proto->pr_protocol != IPPROTO_RSVP)
2222: return (EOPNOTSUPP);
2223:
2224: /* Check mbuf. */
2225: if (m == NULL || m->m_len != sizeof(int)) {
2226: return (EINVAL);
2227: }
2228: vifi = *(mtod(m, int *));
2229:
2230: s = splsoftnet();
2231:
2232: /* Check vif. */
2233: if (!legal_vif_num(vifi)) {
2234: splx(s);
2235: return (EADDRNOTAVAIL);
2236: }
2237:
2238: if (rsvpdebug)
2239: printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n",
2240: viftable[vifi].v_rsvpd, so);
2241:
2242: viftable[vifi].v_rsvpd = NULL;
2243: /*
2244: * This may seem silly, but we need to be sure we don't over-decrement
2245: * the RSVP counter, in case something slips up.
2246: */
2247: if (viftable[vifi].v_rsvp_on) {
2248: viftable[vifi].v_rsvp_on = 0;
2249: rsvp_on--;
2250: }
2251:
2252: splx(s);
2253: return (0);
2254: }
2255:
2256: void
2257: ip_rsvp_force_done(struct socket *so)
2258: {
2259: int vifi, s;
2260:
2261: /* Don't bother if it is not the right type of socket. */
2262: if (so->so_type != SOCK_RAW ||
2263: so->so_proto->pr_protocol != IPPROTO_RSVP)
2264: return;
2265:
2266: s = splsoftnet();
2267:
2268: /*
2269: * The socket may be attached to more than one vif...this
2270: * is perfectly legal.
2271: */
2272: for (vifi = 0; vifi < numvifs; vifi++) {
2273: if (viftable[vifi].v_rsvpd == so) {
2274: viftable[vifi].v_rsvpd = NULL;
2275: /*
2276: * This may seem silly, but we need to be sure we don't
2277: * over-decrement the RSVP counter, in case something
2278: * slips up.
2279: */
2280: if (viftable[vifi].v_rsvp_on) {
2281: viftable[vifi].v_rsvp_on = 0;
2282: rsvp_on--;
2283: }
2284: }
2285: }
2286:
2287: splx(s);
2288: return;
2289: }
2290:
2291: void
2292: rsvp_input(struct mbuf *m, struct ifnet *ifp)
2293: {
2294: int vifi, s;
2295: struct ip *ip = mtod(m, struct ip *);
2296: static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET };
2297:
2298: if (rsvpdebug)
2299: printf("rsvp_input: rsvp_on %d\n", rsvp_on);
2300:
2301: /*
2302: * Can still get packets with rsvp_on = 0 if there is a local member
2303: * of the group to which the RSVP packet is addressed. But in this
2304: * case we want to throw the packet away.
2305: */
2306: if (!rsvp_on) {
2307: m_freem(m);
2308: return;
2309: }
2310:
2311: /*
2312: * If the old-style non-vif-associated socket is set, then use
2313: * it and ignore the new ones.
2314: */
2315: if (ip_rsvpd != NULL) {
2316: if (rsvpdebug)
2317: printf("rsvp_input: "
2318: "Sending packet up old-style socket\n");
2319: rip_input(m, 0); /*XXX*/
2320: return;
2321: }
2322:
2323: s = splsoftnet();
2324:
2325: if (rsvpdebug)
2326: printf("rsvp_input: check vifs\n");
2327:
2328: /* Find which vif the packet arrived on. */
2329: for (vifi = 0; vifi < numvifs; vifi++) {
2330: if (viftable[vifi].v_ifp == ifp)
2331: break;
2332: }
2333:
2334: if (vifi == numvifs) {
2335: /* Can't find vif packet arrived on. Drop packet. */
2336: if (rsvpdebug)
2337: printf("rsvp_input: "
2338: "Can't find vif for packet...dropping it.\n");
2339: m_freem(m);
2340: splx(s);
2341: return;
2342: }
2343:
2344: if (rsvpdebug)
2345: printf("rsvp_input: check socket\n");
2346:
2347: if (viftable[vifi].v_rsvpd == NULL) {
2348: /*
2349: * drop packet, since there is no specific socket for this
2350: * interface
2351: */
2352: if (rsvpdebug)
2353: printf("rsvp_input: No socket defined for vif %d\n",
2354: vifi);
2355: m_freem(m);
2356: splx(s);
2357: return;
2358: }
2359:
2360: rsvp_src.sin_addr = ip->ip_src;
2361:
2362: if (rsvpdebug && m)
2363: printf("rsvp_input: m->m_len = %d, sbspace() = %d\n",
2364: m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv));
2365:
2366: if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2367: if (rsvpdebug)
2368: printf("rsvp_input: Failed to append to socket\n");
2369: else
2370: if (rsvpdebug)
2371: printf("rsvp_input: send packet up\n");
2372:
2373: splx(s);
2374: }
2375: #endif /* RSVP_ISI */
2376:
2377: /*
2378: * Code for bandwidth monitors
2379: */
2380:
2381: /*
2382: * Define common interface for timeval-related methods
2383: */
2384: #define BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp)
2385: #define BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp))
2386: #define BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp))
2387:
2388: static uint32_t
2389: compute_bw_meter_flags(struct bw_upcall *req)
2390: {
2391: uint32_t flags = 0;
2392:
2393: if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2394: flags |= BW_METER_UNIT_PACKETS;
2395: if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2396: flags |= BW_METER_UNIT_BYTES;
2397: if (req->bu_flags & BW_UPCALL_GEQ)
2398: flags |= BW_METER_GEQ;
2399: if (req->bu_flags & BW_UPCALL_LEQ)
2400: flags |= BW_METER_LEQ;
2401:
2402: return (flags);
2403: }
2404:
2405: /*
2406: * Add a bw_meter entry
2407: */
2408: static int
2409: add_bw_upcall(struct mbuf *m)
2410: {
2411: int s;
2412: struct mfc *mfc;
2413: struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2414: BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2415: struct timeval now;
2416: struct bw_meter *x;
2417: uint32_t flags;
2418: struct bw_upcall *req;
2419:
2420: if (m == NULL || m->m_len < sizeof(struct bw_upcall))
2421: return (EINVAL);
2422:
2423: req = mtod(m, struct bw_upcall *);
2424:
2425: if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2426: return (EOPNOTSUPP);
2427:
2428: /* Test if the flags are valid */
2429: if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2430: return (EINVAL);
2431: if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2432: return (EINVAL);
2433: if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2434: == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2435: return (EINVAL);
2436:
2437: /* Test if the threshold time interval is valid */
2438: if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2439: return (EINVAL);
2440:
2441: flags = compute_bw_meter_flags(req);
2442:
2443: /* Find if we have already same bw_meter entry */
2444: s = splsoftnet();
2445: mfc = mfc_find(&req->bu_src, &req->bu_dst);
2446: if (mfc == NULL) {
2447: splx(s);
2448: return (EADDRNOTAVAIL);
2449: }
2450: for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2451: if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2452: &req->bu_threshold.b_time, ==)) &&
2453: (x->bm_threshold.b_packets ==
2454: req->bu_threshold.b_packets) &&
2455: (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2456: (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
2457: splx(s);
2458: return (0); /* XXX Already installed */
2459: }
2460: }
2461:
2462: /* Allocate the new bw_meter entry */
2463: x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
2464: if (x == NULL) {
2465: splx(s);
2466: return (ENOBUFS);
2467: }
2468:
2469: /* Set the new bw_meter entry */
2470: x->bm_threshold.b_time = req->bu_threshold.b_time;
2471: microtime(&now);
2472: x->bm_start_time = now;
2473: x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2474: x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2475: x->bm_measured.b_packets = 0;
2476: x->bm_measured.b_bytes = 0;
2477: x->bm_flags = flags;
2478: x->bm_time_next = NULL;
2479: x->bm_time_hash = BW_METER_BUCKETS;
2480:
2481: /* Add the new bw_meter entry to the front of entries for this MFC */
2482: x->bm_mfc = mfc;
2483: x->bm_mfc_next = mfc->mfc_bw_meter;
2484: mfc->mfc_bw_meter = x;
2485: schedule_bw_meter(x, &now);
2486: splx(s);
2487:
2488: return (0);
2489: }
2490:
2491: static void
2492: free_bw_list(struct bw_meter *list)
2493: {
2494: while (list != NULL) {
2495: struct bw_meter *x = list;
2496:
2497: list = list->bm_mfc_next;
2498: unschedule_bw_meter(x);
2499: free(x, M_BWMETER);
2500: }
2501: }
2502:
2503: /*
2504: * Delete one or multiple bw_meter entries
2505: */
2506: static int
2507: del_bw_upcall(struct mbuf *m)
2508: {
2509: int s;
2510: struct mfc *mfc;
2511: struct bw_meter *x;
2512: struct bw_upcall *req;
2513:
2514: if (m == NULL || m->m_len < sizeof(struct bw_upcall))
2515: return (EINVAL);
2516:
2517: req = mtod(m, struct bw_upcall *);
2518:
2519: if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2520: return (EOPNOTSUPP);
2521:
2522: s = splsoftnet();
2523: /* Find the corresponding MFC entry */
2524: mfc = mfc_find(&req->bu_src, &req->bu_dst);
2525: if (mfc == NULL) {
2526: splx(s);
2527: return (EADDRNOTAVAIL);
2528: } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2529: /* Delete all bw_meter entries for this mfc */
2530: struct bw_meter *list;
2531:
2532: list = mfc->mfc_bw_meter;
2533: mfc->mfc_bw_meter = NULL;
2534: free_bw_list(list);
2535: splx(s);
2536: return (0);
2537: } else { /* Delete a single bw_meter entry */
2538: struct bw_meter *prev;
2539: uint32_t flags = 0;
2540:
2541: flags = compute_bw_meter_flags(req);
2542:
2543: /* Find the bw_meter entry to delete */
2544: for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2545: prev = x, x = x->bm_mfc_next) {
2546: if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2547: &req->bu_threshold.b_time, ==)) &&
2548: (x->bm_threshold.b_packets ==
2549: req->bu_threshold.b_packets) &&
2550: (x->bm_threshold.b_bytes ==
2551: req->bu_threshold.b_bytes) &&
2552: (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2553: break;
2554: }
2555: if (x != NULL) { /* Delete entry from the list for this MFC */
2556: if (prev != NULL) {
2557: /* remove from middle */
2558: prev->bm_mfc_next = x->bm_mfc_next;
2559: } else {
2560: /* new head of list */
2561: x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;
2562: }
2563:
2564: unschedule_bw_meter(x);
2565: splx(s);
2566: /* Free the bw_meter entry */
2567: free(x, M_BWMETER);
2568: return (0);
2569: } else {
2570: splx(s);
2571: return (EINVAL);
2572: }
2573: }
2574: /* NOTREACHED */
2575: }
2576:
2577: /*
2578: * Perform bandwidth measurement processing that may result in an upcall
2579: */
2580: static void
2581: bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2582: {
2583: struct timeval delta;
2584:
2585: delta = *nowp;
2586: BW_TIMEVALDECR(&delta, &x->bm_start_time);
2587:
2588: if (x->bm_flags & BW_METER_GEQ) {
2589: /* Processing for ">=" type of bw_meter entry */
2590: if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2591: /* Reset the bw_meter entry */
2592: x->bm_start_time = *nowp;
2593: x->bm_measured.b_packets = 0;
2594: x->bm_measured.b_bytes = 0;
2595: x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2596: }
2597:
2598: /* Record that a packet is received */
2599: x->bm_measured.b_packets++;
2600: x->bm_measured.b_bytes += plen;
2601:
2602: /* Test if we should deliver an upcall */
2603: if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2604: if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2605: (x->bm_measured.b_packets >=
2606: x->bm_threshold.b_packets)) ||
2607: ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2608: (x->bm_measured.b_bytes >=
2609: x->bm_threshold.b_bytes))) {
2610: /* Prepare an upcall for delivery */
2611: bw_meter_prepare_upcall(x, nowp);
2612: x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2613: }
2614: }
2615: } else if (x->bm_flags & BW_METER_LEQ) {
2616: /* Processing for "<=" type of bw_meter entry */
2617: if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2618: /*
2619: * We are behind time with the multicast forwarding
2620: * table scanning for "<=" type of bw_meter entries,
2621: * so test now if we should deliver an upcall.
2622: */
2623: if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2624: (x->bm_measured.b_packets <=
2625: x->bm_threshold.b_packets)) ||
2626: ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2627: (x->bm_measured.b_bytes <=
2628: x->bm_threshold.b_bytes))) {
2629: /* Prepare an upcall for delivery */
2630: bw_meter_prepare_upcall(x, nowp);
2631: }
2632: /* Reschedule the bw_meter entry */
2633: unschedule_bw_meter(x);
2634: schedule_bw_meter(x, nowp);
2635: }
2636:
2637: /* Record that a packet is received */
2638: x->bm_measured.b_packets++;
2639: x->bm_measured.b_bytes += plen;
2640:
2641: /* Test if we should restart the measuring interval */
2642: if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2643: x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2644: (x->bm_flags & BW_METER_UNIT_BYTES &&
2645: x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2646: /* Don't restart the measuring interval */
2647: } else {
2648: /* Do restart the measuring interval */
2649: /*
2650: * XXX: note that we don't unschedule and schedule,
2651: * because this might be too much overhead per packet.
2652: * Instead, when we process all entries for a given
2653: * timer hash bin, we check whether it is really a
2654: * timeout. If not, we reschedule at that time.
2655: */
2656: x->bm_start_time = *nowp;
2657: x->bm_measured.b_packets = 0;
2658: x->bm_measured.b_bytes = 0;
2659: x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2660: }
2661: }
2662: }
2663:
2664: /*
2665: * Prepare a bandwidth-related upcall
2666: */
2667: static void
2668: bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2669: {
2670: struct timeval delta;
2671: struct bw_upcall *u;
2672:
2673: /* Compute the measured time interval */
2674: delta = *nowp;
2675: BW_TIMEVALDECR(&delta, &x->bm_start_time);
2676:
2677: /* If there are too many pending upcalls, deliver them now */
2678: if (bw_upcalls_n >= BW_UPCALLS_MAX)
2679: bw_upcalls_send();
2680:
2681: /* Set the bw_upcall entry */
2682: u = &bw_upcalls[bw_upcalls_n++];
2683: u->bu_src = x->bm_mfc->mfc_origin;
2684: u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2685: u->bu_threshold.b_time = x->bm_threshold.b_time;
2686: u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2687: u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2688: u->bu_measured.b_time = delta;
2689: u->bu_measured.b_packets = x->bm_measured.b_packets;
2690: u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2691: u->bu_flags = 0;
2692: if (x->bm_flags & BW_METER_UNIT_PACKETS)
2693: u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2694: if (x->bm_flags & BW_METER_UNIT_BYTES)
2695: u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2696: if (x->bm_flags & BW_METER_GEQ)
2697: u->bu_flags |= BW_UPCALL_GEQ;
2698: if (x->bm_flags & BW_METER_LEQ)
2699: u->bu_flags |= BW_UPCALL_LEQ;
2700: }
2701:
2702: /*
2703: * Send the pending bandwidth-related upcalls
2704: */
2705: static void
2706: bw_upcalls_send(void)
2707: {
2708: struct mbuf *m;
2709: int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2710: struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2711: static struct igmpmsg igmpmsg = {
2712: 0, /* unused1 */
2713: 0, /* unused2 */
2714: IGMPMSG_BW_UPCALL, /* im_msgtype */
2715: 0, /* im_mbz */
2716: 0, /* im_vif */
2717: 0, /* unused3 */
2718: { 0 }, /* im_src */
2719: { 0 } }; /* im_dst */
2720:
2721: if (bw_upcalls_n == 0)
2722: return; /* No pending upcalls */
2723:
2724: bw_upcalls_n = 0;
2725:
2726: /*
2727: * Allocate a new mbuf, initialize it with the header and
2728: * the payload for the pending calls.
2729: */
2730: MGETHDR(m, M_DONTWAIT, MT_HEADER);
2731: if (m == NULL) {
2732: log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2733: return;
2734: }
2735:
2736: m->m_len = m->m_pkthdr.len = 0;
2737: m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2738: m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]);
2739:
2740: /*
2741: * Send the upcalls
2742: * XXX do we need to set the address in k_igmpsrc ?
2743: */
2744: mrtstat.mrts_upcalls++;
2745: if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2746: log(LOG_WARNING,
2747: "bw_upcalls_send: ip_mrouter socket queue full\n");
2748: ++mrtstat.mrts_upq_sockfull;
2749: }
2750: }
2751:
2752: /*
2753: * Compute the timeout hash value for the bw_meter entries
2754: */
2755: #define BW_METER_TIMEHASH(bw_meter, hash) do { \
2756: struct timeval next_timeval = (bw_meter)->bm_start_time; \
2757: \
2758: BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2759: (hash) = next_timeval.tv_sec; \
2760: if (next_timeval.tv_usec) \
2761: (hash)++; /* XXX: make sure we don't timeout early */ \
2762: (hash) %= BW_METER_BUCKETS; \
2763: } while (/*CONSTCOND*/ 0)
2764:
2765: /*
2766: * Schedule a timer to process periodically bw_meter entry of type "<="
2767: * by linking the entry in the proper hash bucket.
2768: */
2769: static void
2770: schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2771: {
2772: int time_hash;
2773:
2774: if (!(x->bm_flags & BW_METER_LEQ))
2775: return; /* XXX: we schedule timers only for "<=" entries */
2776:
2777: /* Reset the bw_meter entry */
2778: x->bm_start_time = *nowp;
2779: x->bm_measured.b_packets = 0;
2780: x->bm_measured.b_bytes = 0;
2781: x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2782:
2783: /* Compute the timeout hash value and insert the entry */
2784: BW_METER_TIMEHASH(x, time_hash);
2785: x->bm_time_next = bw_meter_timers[time_hash];
2786: bw_meter_timers[time_hash] = x;
2787: x->bm_time_hash = time_hash;
2788: }
2789:
2790: /*
2791: * Unschedule the periodic timer that processes bw_meter entry of type "<="
2792: * by removing the entry from the proper hash bucket.
2793: */
2794: static void
2795: unschedule_bw_meter(struct bw_meter *x)
2796: {
2797: int time_hash;
2798: struct bw_meter *prev, *tmp;
2799:
2800: if (!(x->bm_flags & BW_METER_LEQ))
2801: return; /* XXX: we schedule timers only for "<=" entries */
2802:
2803: /* Compute the timeout hash value and delete the entry */
2804: time_hash = x->bm_time_hash;
2805: if (time_hash >= BW_METER_BUCKETS)
2806: return; /* Entry was not scheduled */
2807:
2808: for (prev = NULL, tmp = bw_meter_timers[time_hash];
2809: tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2810: if (tmp == x)
2811: break;
2812:
2813: if (tmp == NULL)
2814: panic("unschedule_bw_meter: bw_meter entry not found");
2815:
2816: if (prev != NULL)
2817: prev->bm_time_next = x->bm_time_next;
2818: else
2819: bw_meter_timers[time_hash] = x->bm_time_next;
2820:
2821: x->bm_time_next = NULL;
2822: x->bm_time_hash = BW_METER_BUCKETS;
2823: }
2824:
2825: /*
2826: * Process all "<=" type of bw_meter that should be processed now,
2827: * and for each entry prepare an upcall if necessary. Each processed
2828: * entry is rescheduled again for the (periodic) processing.
2829: *
2830: * This is run periodically (once per second normally). On each round,
2831: * all the potentially matching entries are in the hash slot that we are
2832: * looking at.
2833: */
2834: static void
2835: bw_meter_process()
2836: {
2837: int s;
2838: static uint32_t last_tv_sec; /* last time we processed this */
2839:
2840: uint32_t loops;
2841: int i;
2842: struct timeval now, process_endtime;
2843:
2844: microtime(&now);
2845: if (last_tv_sec == now.tv_sec)
2846: return; /* nothing to do */
2847:
2848: loops = now.tv_sec - last_tv_sec;
2849: last_tv_sec = now.tv_sec;
2850: if (loops > BW_METER_BUCKETS)
2851: loops = BW_METER_BUCKETS;
2852:
2853: s = splsoftnet();
2854: /*
2855: * Process all bins of bw_meter entries from the one after the last
2856: * processed to the current one. On entry, i points to the last bucket
2857: * visited, so we need to increment i at the beginning of the loop.
2858: */
2859: for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2860: struct bw_meter *x, *tmp_list;
2861:
2862: if (++i >= BW_METER_BUCKETS)
2863: i = 0;
2864:
2865: /* Disconnect the list of bw_meter entries from the bin */
2866: tmp_list = bw_meter_timers[i];
2867: bw_meter_timers[i] = NULL;
2868:
2869: /* Process the list of bw_meter entries */
2870: while (tmp_list != NULL) {
2871: x = tmp_list;
2872: tmp_list = tmp_list->bm_time_next;
2873:
2874: /* Test if the time interval is over */
2875: process_endtime = x->bm_start_time;
2876: BW_TIMEVALADD(&process_endtime,
2877: &x->bm_threshold.b_time);
2878: if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2879: /* Not yet: reschedule, but don't reset */
2880: int time_hash;
2881:
2882: BW_METER_TIMEHASH(x, time_hash);
2883: if (time_hash == i &&
2884: process_endtime.tv_sec == now.tv_sec) {
2885: /*
2886: * XXX: somehow the bin processing is
2887: * a bit ahead of time. Put the entry
2888: * in the next bin.
2889: */
2890: if (++time_hash >= BW_METER_BUCKETS)
2891: time_hash = 0;
2892: }
2893: x->bm_time_next = bw_meter_timers[time_hash];
2894: bw_meter_timers[time_hash] = x;
2895: x->bm_time_hash = time_hash;
2896:
2897: continue;
2898: }
2899:
2900: /* Test if we should deliver an upcall */
2901: if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2902: (x->bm_measured.b_packets <=
2903: x->bm_threshold.b_packets)) ||
2904: ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2905: (x->bm_measured.b_bytes <=
2906: x->bm_threshold.b_bytes))) {
2907: /* Prepare an upcall for delivery */
2908: bw_meter_prepare_upcall(x, &now);
2909: }
2910:
2911: /* Reschedule for next processing */
2912: schedule_bw_meter(x, &now);
2913: }
2914: }
2915:
2916: /* Send all upcalls that are pending delivery */
2917: bw_upcalls_send();
2918:
2919: splx(s);
2920: }
2921:
2922: /*
2923: * A periodic function for sending all upcalls that are pending delivery
2924: */
2925: static void
2926: expire_bw_upcalls_send(void *unused)
2927: {
2928: int s;
2929:
2930: s = splsoftnet();
2931: bw_upcalls_send();
2932: splx(s);
2933:
2934: timeout_add(&bw_upcalls_ch, BW_UPCALLS_PERIOD);
2935: }
2936:
2937: /*
2938: * A periodic function for periodic scanning of the multicast forwarding
2939: * table for processing all "<=" bw_meter entries.
2940: */
2941: static void
2942: expire_bw_meter_process(void *unused)
2943: {
2944: if (mrt_api_config & MRT_MFC_BW_UPCALL)
2945: bw_meter_process();
2946:
2947: timeout_add(&bw_meter_ch, BW_METER_PERIOD);
2948: }
2949:
2950: /*
2951: * End of bandwidth monitoring code
2952: */
2953:
2954: #ifdef PIM
2955: /*
2956: * Send the packet up to the user daemon, or eventually do kernel encapsulation
2957: */
2958: static int
2959: pim_register_send(struct ip *ip, struct vif *vifp,
2960: struct mbuf *m, struct mfc *rt)
2961: {
2962: struct mbuf *mb_copy, *mm;
2963:
2964: if (mrtdebug & DEBUG_PIM)
2965: log(LOG_DEBUG, "pim_register_send: ");
2966:
2967: mb_copy = pim_register_prepare(ip, m);
2968: if (mb_copy == NULL)
2969: return (ENOBUFS);
2970:
2971: /*
2972: * Send all the fragments. Note that the mbuf for each fragment
2973: * is freed by the sending machinery.
2974: */
2975: for (mm = mb_copy; mm; mm = mb_copy) {
2976: mb_copy = mm->m_nextpkt;
2977: mm->m_nextpkt = NULL;
2978: mm = m_pullup(mm, sizeof(struct ip));
2979: if (mm != NULL) {
2980: ip = mtod(mm, struct ip *);
2981: if ((mrt_api_config & MRT_MFC_RP) &&
2982: !in_nullhost(rt->mfc_rp)) {
2983: pim_register_send_rp(ip, vifp, mm, rt);
2984: } else {
2985: pim_register_send_upcall(ip, vifp, mm, rt);
2986: }
2987: }
2988: }
2989:
2990: return (0);
2991: }
2992:
2993: /*
2994: * Return a copy of the data packet that is ready for PIM Register
2995: * encapsulation.
2996: * XXX: Note that in the returned copy the IP header is a valid one.
2997: */
2998: static struct mbuf *
2999: pim_register_prepare(struct ip *ip, struct mbuf *m)
3000: {
3001: struct mbuf *mb_copy = NULL;
3002: int mtu;
3003:
3004: /* Take care of delayed checksums */
3005: if (m->m_pkthdr.csum_flags & (M_TCPV4_CSUM_OUT | M_UDPV4_CSUM_OUT)) {
3006: in_delayed_cksum(m);
3007: m->m_pkthdr.csum_flags &=
3008: ~(M_UDPV4_CSUM_OUT | M_TCPV4_CSUM_OUT);
3009: }
3010:
3011: /*
3012: * Copy the old packet & pullup its IP header into the
3013: * new mbuf so we can modify it.
3014: */
3015: mb_copy = m_copy(m, 0, M_COPYALL);
3016: if (mb_copy == NULL)
3017: return (NULL);
3018: mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
3019: if (mb_copy == NULL)
3020: return (NULL);
3021:
3022: /* take care of the TTL */
3023: ip = mtod(mb_copy, struct ip *);
3024: --ip->ip_ttl;
3025:
3026: /* Compute the MTU after the PIM Register encapsulation */
3027: mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
3028:
3029: if (ntohs(ip->ip_len) <= mtu) {
3030: /* Turn the IP header into a valid one */
3031: ip->ip_sum = 0;
3032: ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
3033: } else {
3034: /* Fragment the packet */
3035: if (ip_fragment(mb_copy, NULL, mtu) != 0) {
3036: /* XXX: mb_copy was freed by ip_fragment() */
3037: return (NULL);
3038: }
3039: }
3040: return (mb_copy);
3041: }
3042:
3043: /*
3044: * Send an upcall with the data packet to the user-level process.
3045: */
3046: static int
3047: pim_register_send_upcall(struct ip *ip, struct vif *vifp,
3048: struct mbuf *mb_copy, struct mfc *rt)
3049: {
3050: struct mbuf *mb_first;
3051: int len = ntohs(ip->ip_len);
3052: struct igmpmsg *im;
3053: struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
3054:
3055: /* Add a new mbuf with an upcall header */
3056: MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3057: if (mb_first == NULL) {
3058: m_freem(mb_copy);
3059: return (ENOBUFS);
3060: }
3061: mb_first->m_data += max_linkhdr;
3062: mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
3063: mb_first->m_len = sizeof(struct igmpmsg);
3064: mb_first->m_next = mb_copy;
3065:
3066: /* Send message to routing daemon */
3067: im = mtod(mb_first, struct igmpmsg *);
3068: im->im_msgtype = IGMPMSG_WHOLEPKT;
3069: im->im_mbz = 0;
3070: im->im_vif = vifp - viftable;
3071: im->im_src = ip->ip_src;
3072: im->im_dst = ip->ip_dst;
3073:
3074: k_igmpsrc.sin_addr = ip->ip_src;
3075:
3076: mrtstat.mrts_upcalls++;
3077:
3078: if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
3079: if (mrtdebug & DEBUG_PIM)
3080: log(LOG_WARNING, "mcast: pim_register_send_upcall: "
3081: "ip_mrouter socket queue full");
3082: ++mrtstat.mrts_upq_sockfull;
3083: return (ENOBUFS);
3084: }
3085:
3086: /* Keep statistics */
3087: pimstat.pims_snd_registers_msgs++;
3088: pimstat.pims_snd_registers_bytes += len;
3089:
3090: return (0);
3091: }
3092:
3093: /*
3094: * Encapsulate the data packet in PIM Register message and send it to the RP.
3095: */
3096: static int
3097: pim_register_send_rp(struct ip *ip, struct vif *vifp,
3098: struct mbuf *mb_copy, struct mfc *rt)
3099: {
3100: struct mbuf *mb_first;
3101: struct ip *ip_outer;
3102: struct pim_encap_pimhdr *pimhdr;
3103: int len = ntohs(ip->ip_len);
3104: vifi_t vifi = rt->mfc_parent;
3105:
3106: if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) {
3107: m_freem(mb_copy);
3108: return (EADDRNOTAVAIL); /* The iif vif is invalid */
3109: }
3110:
3111: /* Add a new mbuf with the encapsulating header */
3112: MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
3113: if (mb_first == NULL) {
3114: m_freem(mb_copy);
3115: return (ENOBUFS);
3116: }
3117: mb_first->m_data += max_linkhdr;
3118: mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
3119: mb_first->m_next = mb_copy;
3120:
3121: mb_first->m_pkthdr.len = len + mb_first->m_len;
3122:
3123: /* Fill in the encapsulating IP and PIM header */
3124: ip_outer = mtod(mb_first, struct ip *);
3125: *ip_outer = pim_encap_iphdr;
3126: ip_outer->ip_id = htons(ip_randomid());
3127: ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
3128: sizeof(pim_encap_pimhdr));
3129: ip_outer->ip_src = viftable[vifi].v_lcl_addr;
3130: ip_outer->ip_dst = rt->mfc_rp;
3131: /*
3132: * Copy the inner header TOS to the outer header, and take care of the
3133: * IP_DF bit.
3134: */
3135: ip_outer->ip_tos = ip->ip_tos;
3136: if (ntohs(ip->ip_off) & IP_DF)
3137: ip_outer->ip_off |= htons(IP_DF);
3138: pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
3139: + sizeof(pim_encap_iphdr));
3140: *pimhdr = pim_encap_pimhdr;
3141: /* If the iif crosses a border, set the Border-bit */
3142: if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3143: pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3144:
3145: mb_first->m_data += sizeof(pim_encap_iphdr);
3146: pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3147: mb_first->m_data -= sizeof(pim_encap_iphdr);
3148:
3149: if (vifp->v_rate_limit == 0)
3150: tbf_send_packet(vifp, mb_first);
3151: else
3152: tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len));
3153:
3154: /* Keep statistics */
3155: pimstat.pims_snd_registers_msgs++;
3156: pimstat.pims_snd_registers_bytes += len;
3157:
3158: return (0);
3159: }
3160:
3161: /*
3162: * PIM-SMv2 and PIM-DM messages processing.
3163: * Receives and verifies the PIM control messages, and passes them
3164: * up to the listening socket, using rip_input().
3165: * The only message with special processing is the PIM_REGISTER message
3166: * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3167: * is passed to if_simloop().
3168: */
3169: void
3170: pim_input(struct mbuf *m, ...)
3171: {
3172: struct ip *ip = mtod(m, struct ip *);
3173: struct pim *pim;
3174: int minlen;
3175: int datalen;
3176: int ip_tos;
3177: int iphlen;
3178: va_list ap;
3179:
3180: va_start(ap, m);
3181: iphlen = va_arg(ap, int);
3182: va_end(ap);
3183:
3184: datalen = ntohs(ip->ip_len) - iphlen;
3185:
3186: /* Keep statistics */
3187: pimstat.pims_rcv_total_msgs++;
3188: pimstat.pims_rcv_total_bytes += datalen;
3189:
3190: /* Validate lengths */
3191: if (datalen < PIM_MINLEN) {
3192: pimstat.pims_rcv_tooshort++;
3193: log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3194: datalen, (u_long)ip->ip_src.s_addr);
3195: m_freem(m);
3196: return;
3197: }
3198:
3199: /*
3200: * If the packet is at least as big as a REGISTER, go agead
3201: * and grab the PIM REGISTER header size, to avoid another
3202: * possible m_pullup() later.
3203: *
3204: * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
3205: * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3206: */
3207: minlen = iphlen + (datalen >= PIM_REG_MINLEN ?
3208: PIM_REG_MINLEN : PIM_MINLEN);
3209: /*
3210: * Get the IP and PIM headers in contiguous memory, and
3211: * possibly the PIM REGISTER header.
3212: */
3213: if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3214: (m = m_pullup(m, minlen)) == NULL) {
3215: log(LOG_ERR, "pim_input: m_pullup failure\n");
3216: return;
3217: }
3218: /* m_pullup() may have given us a new mbuf so reset ip. */
3219: ip = mtod(m, struct ip *);
3220: ip_tos = ip->ip_tos;
3221:
3222: /* adjust mbuf to point to the PIM header */
3223: m->m_data += iphlen;
3224: m->m_len -= iphlen;
3225: pim = mtod(m, struct pim *);
3226:
3227: /*
3228: * Validate checksum. If PIM REGISTER, exclude the data packet.
3229: *
3230: * XXX: some older PIMv2 implementations don't make this distinction,
3231: * so for compatibility reason perform the checksum over part of the
3232: * message, and if error, then over the whole message.
3233: */
3234: if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER &&
3235: in_cksum(m, PIM_MINLEN) == 0) {
3236: /* do nothing, checksum okay */
3237: } else if (in_cksum(m, datalen)) {
3238: pimstat.pims_rcv_badsum++;
3239: if (mrtdebug & DEBUG_PIM)
3240: log(LOG_DEBUG, "pim_input: invalid checksum");
3241: m_freem(m);
3242: return;
3243: }
3244:
3245: /* PIM version check */
3246: if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3247: pimstat.pims_rcv_badversion++;
3248: log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3249: PIM_VT_V(pim->pim_vt), PIM_VERSION);
3250: m_freem(m);
3251: return;
3252: }
3253:
3254: /* restore mbuf back to the outer IP */
3255: m->m_data -= iphlen;
3256: m->m_len += iphlen;
3257:
3258: if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3259: /*
3260: * Since this is a REGISTER, we'll make a copy of the register
3261: * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3262: * routing daemon.
3263: */
3264: int s;
3265: struct sockaddr_in dst = { sizeof(dst), AF_INET };
3266: struct mbuf *mcp;
3267: struct ip *encap_ip;
3268: u_int32_t *reghdr;
3269: struct ifnet *vifp;
3270:
3271: s = splsoftnet();
3272: if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3273: splx(s);
3274: if (mrtdebug & DEBUG_PIM)
3275: log(LOG_DEBUG, "pim_input: register vif "
3276: "not set: %d\n", reg_vif_num);
3277: m_freem(m);
3278: return;
3279: }
3280: /* XXX need refcnt? */
3281: vifp = viftable[reg_vif_num].v_ifp;
3282: splx(s);
3283:
3284: /* Validate length */
3285: if (datalen < PIM_REG_MINLEN) {
3286: pimstat.pims_rcv_tooshort++;
3287: pimstat.pims_rcv_badregisters++;
3288: log(LOG_ERR, "pim_input: register packet size "
3289: "too small %d from %lx\n",
3290: datalen, (u_long)ip->ip_src.s_addr);
3291: m_freem(m);
3292: return;
3293: }
3294:
3295: reghdr = (u_int32_t *)(pim + 1);
3296: encap_ip = (struct ip *)(reghdr + 1);
3297:
3298: if (mrtdebug & DEBUG_PIM) {
3299: log(LOG_DEBUG, "pim_input[register], encap_ip: "
3300: "%lx -> %lx, encap_ip len %d\n",
3301: (u_long)ntohl(encap_ip->ip_src.s_addr),
3302: (u_long)ntohl(encap_ip->ip_dst.s_addr),
3303: ntohs(encap_ip->ip_len));
3304: }
3305:
3306: /* verify the version number of the inner packet */
3307: if (encap_ip->ip_v != IPVERSION) {
3308: pimstat.pims_rcv_badregisters++;
3309: if (mrtdebug & DEBUG_PIM) {
3310: log(LOG_DEBUG, "pim_input: invalid IP version"
3311: " (%d) of the inner packet\n",
3312: encap_ip->ip_v);
3313: }
3314: m_freem(m);
3315: return;
3316: }
3317:
3318: /* verify the inner packet is destined to a mcast group */
3319: if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) {
3320: pimstat.pims_rcv_badregisters++;
3321: if (mrtdebug & DEBUG_PIM)
3322: log(LOG_DEBUG,
3323: "pim_input: inner packet of register is"
3324: " not multicast %lx\n",
3325: (u_long)ntohl(encap_ip->ip_dst.s_addr));
3326: m_freem(m);
3327: return;
3328: }
3329:
3330: /* If a NULL_REGISTER, pass it to the daemon */
3331: if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3332: goto pim_input_to_daemon;
3333:
3334: /*
3335: * Copy the TOS from the outer IP header to the inner
3336: * IP header.
3337: */
3338: if (encap_ip->ip_tos != ip_tos) {
3339: /* Outer TOS -> inner TOS */
3340: encap_ip->ip_tos = ip_tos;
3341: /* Recompute the inner header checksum. Sigh... */
3342:
3343: /* adjust mbuf to point to the inner IP header */
3344: m->m_data += (iphlen + PIM_MINLEN);
3345: m->m_len -= (iphlen + PIM_MINLEN);
3346:
3347: encap_ip->ip_sum = 0;
3348: encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3349:
3350: /* restore mbuf to point back to the outer IP header */
3351: m->m_data -= (iphlen + PIM_MINLEN);
3352: m->m_len += (iphlen + PIM_MINLEN);
3353: }
3354:
3355: /*
3356: * Decapsulate the inner IP packet and loopback to forward it
3357: * as a normal multicast packet. Also, make a copy of the
3358: * outer_iphdr + pimhdr + reghdr + encap_iphdr
3359: * to pass to the daemon later, so it can take the appropriate
3360: * actions (e.g., send back PIM_REGISTER_STOP).
3361: * XXX: here m->m_data points to the outer IP header.
3362: */
3363: mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
3364: if (mcp == NULL) {
3365: log(LOG_ERR, "pim_input: pim register: could not "
3366: "copy register head\n");
3367: m_freem(m);
3368: return;
3369: }
3370:
3371: /* Keep statistics */
3372: /* XXX: registers_bytes include only the encap. mcast pkt */
3373: pimstat.pims_rcv_registers_msgs++;
3374: pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3375:
3376: /* forward the inner ip packet; point m_data at the inner ip. */
3377: m_adj(m, iphlen + PIM_MINLEN);
3378:
3379: if (mrtdebug & DEBUG_PIM) {
3380: log(LOG_DEBUG,
3381: "pim_input: forwarding decapsulated register: "
3382: "src %lx, dst %lx, vif %d\n",
3383: (u_long)ntohl(encap_ip->ip_src.s_addr),
3384: (u_long)ntohl(encap_ip->ip_dst.s_addr),
3385: reg_vif_num);
3386: }
3387: /* NB: vifp was collected above; can it change on us? */
3388: looutput(vifp, m, (struct sockaddr *)&dst,
3389: (struct rtentry *)NULL);
3390:
3391: /* prepare the register head to send to the mrouting daemon */
3392: m = mcp;
3393: }
3394:
3395: pim_input_to_daemon:
3396: /*
3397: * Pass the PIM message up to the daemon; if it is a Register message,
3398: * pass the 'head' only up to the daemon. This includes the
3399: * outer IP header, PIM header, PIM-Register header and the
3400: * inner IP header.
3401: * XXX: the outer IP header pkt size of a Register is not adjust to
3402: * reflect the fact that the inner multicast data is truncated.
3403: */
3404: rip_input(m);
3405:
3406: return;
3407: }
3408: #endif /* PIM */
CVSweb