1/*
2 * IP multicast forwarding procedures
3 *
4 * Written by David Waitzman, BBN Labs, August 1988.
5 * Modified by Steve Deering, Stanford, February 1989.
6 * Modified by Mark J. Steiglitz, Stanford, May, 1991
7 * Modified by Van Jacobson, LBL, January 1993
8 * Modified by Ajit Thyagarajan, PARC, August 1993
9 * Modified by Bill Fenner, PARC, April 1995
10 *
11 * MROUTING Revision: 3.5
| 1/*
2 * IP multicast forwarding procedures
3 *
4 * Written by David Waitzman, BBN Labs, August 1988.
5 * Modified by Steve Deering, Stanford, February 1989.
6 * Modified by Mark J. Steiglitz, Stanford, May, 1991
7 * Modified by Van Jacobson, LBL, January 1993
8 * Modified by Ajit Thyagarajan, PARC, August 1993
9 * Modified by Bill Fenner, PARC, April 1995
10 *
11 * MROUTING Revision: 3.5
|
12 * $Id: ip_mroute.c,v 1.39 1997/03/24 11:33:28 bde Exp $
| 12 * $Id: ip_mroute.c,v 1.40 1997/07/19 20:07:07 fenner Exp $
|
13 */
14
15#include "opt_mrouting.h"
16
17#include <sys/param.h>
18#include <sys/queue.h>
19#include <sys/systm.h>
20#include <sys/mbuf.h>
21#include <sys/socket.h>
22#include <sys/socketvar.h>
23#include <sys/protosw.h>
24#include <sys/errno.h>
25#include <sys/time.h>
26#include <sys/kernel.h>
27#include <sys/sockio.h>
28#include <sys/syslog.h>
29#include <net/if.h>
30#include <net/route.h>
31#include <netinet/in.h>
32#include <netinet/in_systm.h>
33#include <netinet/ip.h>
34#include <netinet/ip_var.h>
35#include <netinet/in_pcb.h>
36#include <netinet/in_var.h>
37#include <netinet/igmp.h>
38#include <netinet/igmp_var.h>
39#include <netinet/ip_mroute.h>
40#include <netinet/udp.h>
41
42#ifndef NTOHL
43#if BYTE_ORDER != BIG_ENDIAN
44#define NTOHL(d) ((d) = ntohl((d)))
45#define NTOHS(d) ((d) = ntohs((u_short)(d)))
46#define HTONL(d) ((d) = htonl((d)))
47#define HTONS(d) ((d) = htons((u_short)(d)))
48#else
49#define NTOHL(d)
50#define NTOHS(d)
51#define HTONL(d)
52#define HTONS(d)
53#endif
54#endif
55
56#ifndef MROUTING
57extern u_long _ip_mcast_src __P((int vifi));
58extern int _ip_mforward __P((struct ip *ip, struct ifnet *ifp,
59 struct mbuf *m, struct ip_moptions *imo));
60extern int _ip_mrouter_done __P((void));
61extern int _ip_mrouter_get __P((int cmd, struct socket *so,
62 struct mbuf **m));
63extern int _ip_mrouter_set __P((int cmd, struct socket *so,
64 struct mbuf *m));
65extern int _mrt_ioctl __P((int req, caddr_t data, struct proc *p));
66
67/*
68 * Dummy routines and globals used when multicast routing is not compiled in.
69 */
70
71struct socket *ip_mrouter = NULL;
72static u_int ip_mrtproto = 0;
73static struct mrtstat mrtstat;
74u_int rsvpdebug = 0;
75
76int
77_ip_mrouter_set(cmd, so, m)
78 int cmd;
79 struct socket *so;
80 struct mbuf *m;
81{
82 return(EOPNOTSUPP);
83}
84
85int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = _ip_mrouter_set;
86
87
88int
89_ip_mrouter_get(cmd, so, m)
90 int cmd;
91 struct socket *so;
92 struct mbuf **m;
93{
94 return(EOPNOTSUPP);
95}
96
97int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = _ip_mrouter_get;
98
99int
100_ip_mrouter_done()
101{
102 return(0);
103}
104
105int (*ip_mrouter_done)(void) = _ip_mrouter_done;
106
107int
108_ip_mforward(ip, ifp, m, imo)
109 struct ip *ip;
110 struct ifnet *ifp;
111 struct mbuf *m;
112 struct ip_moptions *imo;
113{
114 return(0);
115}
116
117int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
118 struct ip_moptions *) = _ip_mforward;
119
120int
121_mrt_ioctl(int req, caddr_t data, struct proc *p)
122{
123 return EOPNOTSUPP;
124}
125
126int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl;
127
128void
129rsvp_input(m, iphlen) /* XXX must fixup manually */
130 struct mbuf *m;
131 int iphlen;
132{
133 /* Can still get packets with rsvp_on = 0 if there is a local member
134 * of the group to which the RSVP packet is addressed. But in this
135 * case we want to throw the packet away.
136 */
137 if (!rsvp_on) {
138 m_freem(m);
139 return;
140 }
141
142 if (ip_rsvpd != NULL) {
143 if (rsvpdebug)
144 printf("rsvp_input: Sending packet up old-style socket\n");
145 rip_input(m, iphlen);
146 return;
147 }
148 /* Drop the packet */
149 m_freem(m);
150}
151
152void ipip_input(struct mbuf *m, int iphlen) { /* XXX must fixup manually */
153 rip_input(m, iphlen);
154}
155
156int (*legal_vif_num)(int) = 0;
157
158/*
159 * This should never be called, since IP_MULTICAST_VIF should fail, but
160 * just in case it does get called, the code a little lower in ip_output
161 * will assign the packet a local address.
162 */
163u_long
164_ip_mcast_src(int vifi) { return INADDR_ANY; }
165u_long (*ip_mcast_src)(int) = _ip_mcast_src;
166
167int
168ip_rsvp_vif_init(so, m)
169 struct socket *so;
170 struct mbuf *m;
171{
172 return(EINVAL);
173}
174
175int
176ip_rsvp_vif_done(so, m)
177 struct socket *so;
178 struct mbuf *m;
179{
180 return(EINVAL);
181}
182
183void
184ip_rsvp_force_done(so)
185 struct socket *so;
186{
187 return;
188}
189
190#else /* MROUTING */
191
192#define M_HASCL(m) ((m)->m_flags & M_EXT)
193
194#define INSIZ sizeof(struct in_addr)
195#define same(a1, a2) \
196 (bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0)
197
198#define MT_MRTABLE MT_RTABLE /* since nothing else uses it */
199
200/*
201 * Globals. All but ip_mrouter and ip_mrtproto could be static,
202 * except for netstat or debugging purposes.
203 */
204#ifndef MROUTE_LKM
205struct socket *ip_mrouter = NULL;
206struct mrtstat mrtstat;
207
208int ip_mrtproto = IGMP_DVMRP; /* for netstat only */
209#else /* MROUTE_LKM */
210extern void X_ipip_input __P((struct mbuf *m, int iphlen));
211extern struct mrtstat mrtstat;
212static int ip_mrtproto;
213#endif
214
215#define NO_RTE_FOUND 0x1
216#define RTE_FOUND 0x2
217
218static struct mbuf *mfctable[MFCTBLSIZ];
219static u_char nexpire[MFCTBLSIZ];
220static struct vif viftable[MAXVIFS];
221static u_int mrtdebug = 0; /* debug level */
222#define DEBUG_MFC 0x02
223#define DEBUG_FORWARD 0x04
224#define DEBUG_EXPIRE 0x08
225#define DEBUG_XMIT 0x10
226static u_int tbfdebug = 0; /* tbf debug level */
227static u_int rsvpdebug = 0; /* rsvp debug level */
228
| 13 */
14
15#include "opt_mrouting.h"
16
17#include <sys/param.h>
18#include <sys/queue.h>
19#include <sys/systm.h>
20#include <sys/mbuf.h>
21#include <sys/socket.h>
22#include <sys/socketvar.h>
23#include <sys/protosw.h>
24#include <sys/errno.h>
25#include <sys/time.h>
26#include <sys/kernel.h>
27#include <sys/sockio.h>
28#include <sys/syslog.h>
29#include <net/if.h>
30#include <net/route.h>
31#include <netinet/in.h>
32#include <netinet/in_systm.h>
33#include <netinet/ip.h>
34#include <netinet/ip_var.h>
35#include <netinet/in_pcb.h>
36#include <netinet/in_var.h>
37#include <netinet/igmp.h>
38#include <netinet/igmp_var.h>
39#include <netinet/ip_mroute.h>
40#include <netinet/udp.h>
41
42#ifndef NTOHL
43#if BYTE_ORDER != BIG_ENDIAN
44#define NTOHL(d) ((d) = ntohl((d)))
45#define NTOHS(d) ((d) = ntohs((u_short)(d)))
46#define HTONL(d) ((d) = htonl((d)))
47#define HTONS(d) ((d) = htons((u_short)(d)))
48#else
49#define NTOHL(d)
50#define NTOHS(d)
51#define HTONL(d)
52#define HTONS(d)
53#endif
54#endif
55
56#ifndef MROUTING
57extern u_long _ip_mcast_src __P((int vifi));
58extern int _ip_mforward __P((struct ip *ip, struct ifnet *ifp,
59 struct mbuf *m, struct ip_moptions *imo));
60extern int _ip_mrouter_done __P((void));
61extern int _ip_mrouter_get __P((int cmd, struct socket *so,
62 struct mbuf **m));
63extern int _ip_mrouter_set __P((int cmd, struct socket *so,
64 struct mbuf *m));
65extern int _mrt_ioctl __P((int req, caddr_t data, struct proc *p));
66
67/*
68 * Dummy routines and globals used when multicast routing is not compiled in.
69 */
70
71struct socket *ip_mrouter = NULL;
72static u_int ip_mrtproto = 0;
73static struct mrtstat mrtstat;
74u_int rsvpdebug = 0;
75
76int
77_ip_mrouter_set(cmd, so, m)
78 int cmd;
79 struct socket *so;
80 struct mbuf *m;
81{
82 return(EOPNOTSUPP);
83}
84
85int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = _ip_mrouter_set;
86
87
88int
89_ip_mrouter_get(cmd, so, m)
90 int cmd;
91 struct socket *so;
92 struct mbuf **m;
93{
94 return(EOPNOTSUPP);
95}
96
97int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = _ip_mrouter_get;
98
99int
100_ip_mrouter_done()
101{
102 return(0);
103}
104
105int (*ip_mrouter_done)(void) = _ip_mrouter_done;
106
107int
108_ip_mforward(ip, ifp, m, imo)
109 struct ip *ip;
110 struct ifnet *ifp;
111 struct mbuf *m;
112 struct ip_moptions *imo;
113{
114 return(0);
115}
116
117int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
118 struct ip_moptions *) = _ip_mforward;
119
120int
121_mrt_ioctl(int req, caddr_t data, struct proc *p)
122{
123 return EOPNOTSUPP;
124}
125
126int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl;
127
128void
129rsvp_input(m, iphlen) /* XXX must fixup manually */
130 struct mbuf *m;
131 int iphlen;
132{
133 /* Can still get packets with rsvp_on = 0 if there is a local member
134 * of the group to which the RSVP packet is addressed. But in this
135 * case we want to throw the packet away.
136 */
137 if (!rsvp_on) {
138 m_freem(m);
139 return;
140 }
141
142 if (ip_rsvpd != NULL) {
143 if (rsvpdebug)
144 printf("rsvp_input: Sending packet up old-style socket\n");
145 rip_input(m, iphlen);
146 return;
147 }
148 /* Drop the packet */
149 m_freem(m);
150}
151
152void ipip_input(struct mbuf *m, int iphlen) { /* XXX must fixup manually */
153 rip_input(m, iphlen);
154}
155
156int (*legal_vif_num)(int) = 0;
157
158/*
159 * This should never be called, since IP_MULTICAST_VIF should fail, but
160 * just in case it does get called, the code a little lower in ip_output
161 * will assign the packet a local address.
162 */
163u_long
164_ip_mcast_src(int vifi) { return INADDR_ANY; }
165u_long (*ip_mcast_src)(int) = _ip_mcast_src;
166
167int
168ip_rsvp_vif_init(so, m)
169 struct socket *so;
170 struct mbuf *m;
171{
172 return(EINVAL);
173}
174
175int
176ip_rsvp_vif_done(so, m)
177 struct socket *so;
178 struct mbuf *m;
179{
180 return(EINVAL);
181}
182
183void
184ip_rsvp_force_done(so)
185 struct socket *so;
186{
187 return;
188}
189
190#else /* MROUTING */
191
192#define M_HASCL(m) ((m)->m_flags & M_EXT)
193
194#define INSIZ sizeof(struct in_addr)
195#define same(a1, a2) \
196 (bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0)
197
198#define MT_MRTABLE MT_RTABLE /* since nothing else uses it */
199
200/*
201 * Globals. All but ip_mrouter and ip_mrtproto could be static,
202 * except for netstat or debugging purposes.
203 */
204#ifndef MROUTE_LKM
205struct socket *ip_mrouter = NULL;
206struct mrtstat mrtstat;
207
208int ip_mrtproto = IGMP_DVMRP; /* for netstat only */
209#else /* MROUTE_LKM */
210extern void X_ipip_input __P((struct mbuf *m, int iphlen));
211extern struct mrtstat mrtstat;
212static int ip_mrtproto;
213#endif
214
215#define NO_RTE_FOUND 0x1
216#define RTE_FOUND 0x2
217
218static struct mbuf *mfctable[MFCTBLSIZ];
219static u_char nexpire[MFCTBLSIZ];
220static struct vif viftable[MAXVIFS];
221static u_int mrtdebug = 0; /* debug level */
222#define DEBUG_MFC 0x02
223#define DEBUG_FORWARD 0x04
224#define DEBUG_EXPIRE 0x08
225#define DEBUG_XMIT 0x10
226static u_int tbfdebug = 0; /* tbf debug level */
227static u_int rsvpdebug = 0; /* rsvp debug level */
228
|
| 229static struct callout_handle expire_upcalls_ch;
230
|
229#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
230#define UPCALL_EXPIRE 6 /* number of timeouts */
231
232/*
233 * Define the token bucket filter structures
234 * tbftable -> each vif has one of these for storing info
235 */
236
237static struct tbf tbftable[MAXVIFS];
238#define TBF_REPROCESS (hz / 100) /* 100x / second */
239
240/*
241 * 'Interfaces' associated with decapsulator (so we can tell
242 * packets that went through it from ones that get reflected
243 * by a broken gateway). These interfaces are never linked into
244 * the system ifnet list & no routes point to them. I.e., packets
245 * can't be sent this way. They only exist as a placeholder for
246 * multicast source verification.
247 */
248static struct ifnet multicast_decap_if[MAXVIFS];
249
250#define ENCAP_TTL 64
251#define ENCAP_PROTO IPPROTO_IPIP /* 4 */
252
253/* prototype IP hdr for encapsulated packets */
254static struct ip multicast_encap_iphdr = {
255#if BYTE_ORDER == LITTLE_ENDIAN
256 sizeof(struct ip) >> 2, IPVERSION,
257#else
258 IPVERSION, sizeof(struct ip) >> 2,
259#endif
260 0, /* tos */
261 sizeof(struct ip), /* total length */
262 0, /* id */
263 0, /* frag offset */
264 ENCAP_TTL, ENCAP_PROTO,
265 0, /* checksum */
266};
267
268/*
269 * Private variables.
270 */
271static vifi_t numvifs = 0;
272static int have_encap_tunnel = 0;
273
274/*
275 * one-back cache used by ipip_input to locate a tunnel's vif
276 * given a datagram's src ip address.
277 */
278static u_long last_encap_src;
279static struct vif *last_encap_vif;
280
281static u_long X_ip_mcast_src __P((int vifi));
282static int X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo));
283static int X_ip_mrouter_done __P((void));
284static int X_ip_mrouter_get __P((int cmd, struct socket *so, struct mbuf **m));
285static int X_ip_mrouter_set __P((int cmd, struct socket *so, struct mbuf *m));
286static int X_legal_vif_num __P((int vif));
287static int X_mrt_ioctl __P((int cmd, caddr_t data));
288
289static int get_sg_cnt(struct sioc_sg_req *);
290static int get_vif_cnt(struct sioc_vif_req *);
291static int ip_mrouter_init(struct socket *, struct mbuf *);
292static int add_vif(struct vifctl *);
293static int del_vif(vifi_t *);
294static int add_mfc(struct mfcctl *);
295static int del_mfc(struct mfcctl *);
296static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
297static int get_version(struct mbuf *);
298static int get_assert(struct mbuf *);
299static int set_assert(int *);
300static void expire_upcalls(void *);
301static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *,
302 vifi_t);
303static void phyint_send(struct ip *, struct vif *, struct mbuf *);
304static void encap_send(struct ip *, struct vif *, struct mbuf *);
305static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long);
306static void tbf_queue(struct vif *, struct mbuf *);
307static void tbf_process_q(struct vif *);
308static void tbf_reprocess_q(void *);
309static int tbf_dq_sel(struct vif *, struct ip *);
310static void tbf_send_packet(struct vif *, struct mbuf *);
311static void tbf_update_tokens(struct vif *);
312static int priority(struct vif *, struct ip *);
313void multiencap_decap(struct mbuf *);
314
315/*
316 * whether or not special PIM assert processing is enabled.
317 */
318static int pim_assert;
319/*
320 * Rate limit for assert notification messages, in usec
321 */
322#define ASSERT_MSG_TIME 3000000
323
324/*
325 * Hash function for a source, group entry
326 */
327#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
328 ((g) >> 20) ^ ((g) >> 10) ^ (g))
329
330/*
331 * Find a route for a given origin IP address and Multicast group address
332 * Type of service parameter to be added in the future!!!
333 */
334
335#define MFCFIND(o, g, rt) { \
336 register struct mbuf *_mb_rt = mfctable[MFCHASH(o,g)]; \
337 register struct mfc *_rt = NULL; \
338 rt = NULL; \
339 ++mrtstat.mrts_mfc_lookups; \
340 while (_mb_rt) { \
341 _rt = mtod(_mb_rt, struct mfc *); \
342 if ((_rt->mfc_origin.s_addr == o) && \
343 (_rt->mfc_mcastgrp.s_addr == g) && \
344 (_mb_rt->m_act == NULL)) { \
345 rt = _rt; \
346 break; \
347 } \
348 _mb_rt = _mb_rt->m_next; \
349 } \
350 if (rt == NULL) { \
351 ++mrtstat.mrts_mfc_misses; \
352 } \
353}
354
355
356/*
357 * Macros to compute elapsed time efficiently
358 * Borrowed from Van Jacobson's scheduling code
359 */
360#define TV_DELTA(a, b, delta) { \
361 register int xxs; \
362 \
363 delta = (a).tv_usec - (b).tv_usec; \
364 if ((xxs = (a).tv_sec - (b).tv_sec)) { \
365 switch (xxs) { \
366 case 2: \
367 delta += 1000000; \
368 /* fall through */ \
369 case 1: \
370 delta += 1000000; \
371 break; \
372 default: \
373 delta += (1000000 * xxs); \
374 } \
375 } \
376}
377
378#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
379 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
380
381#ifdef UPCALL_TIMING
382u_long upcall_data[51];
383static void collate(struct timeval *);
384#endif /* UPCALL_TIMING */
385
386
387/*
388 * Handle MRT setsockopt commands to modify the multicast routing tables.
389 */
390static int
391X_ip_mrouter_set(cmd, so, m)
392 int cmd;
393 struct socket *so;
394 struct mbuf *m;
395{
396 if (cmd != MRT_INIT && so != ip_mrouter) return EACCES;
397
398 switch (cmd) {
399 case MRT_INIT: return ip_mrouter_init(so, m);
400 case MRT_DONE: return ip_mrouter_done();
401 case MRT_ADD_VIF: return add_vif (mtod(m, struct vifctl *));
402 case MRT_DEL_VIF: return del_vif (mtod(m, vifi_t *));
403 case MRT_ADD_MFC: return add_mfc (mtod(m, struct mfcctl *));
404 case MRT_DEL_MFC: return del_mfc (mtod(m, struct mfcctl *));
405 case MRT_ASSERT: return set_assert(mtod(m, int *));
406 default: return EOPNOTSUPP;
407 }
408}
409
410#ifndef MROUTE_LKM
411int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = X_ip_mrouter_set;
412#endif
413
414/*
415 * Handle MRT getsockopt commands
416 */
417static int
418X_ip_mrouter_get(cmd, so, m)
419 int cmd;
420 struct socket *so;
421 struct mbuf **m;
422{
423 struct mbuf *mb;
424
425 if (so != ip_mrouter) return EACCES;
426
427 *m = mb = m_get(M_WAIT, MT_SOOPTS);
428
429 switch (cmd) {
430 case MRT_VERSION: return get_version(mb);
431 case MRT_ASSERT: return get_assert(mb);
432 default: return EOPNOTSUPP;
433 }
434}
435
436#ifndef MROUTE_LKM
437int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = X_ip_mrouter_get;
438#endif
439
440/*
441 * Handle ioctl commands to obtain information from the cache
442 */
443static int
444X_mrt_ioctl(cmd, data)
445 int cmd;
446 caddr_t data;
447{
448 int error = 0;
449
450 switch (cmd) {
451 case (SIOCGETVIFCNT):
452 return (get_vif_cnt((struct sioc_vif_req *)data));
453 break;
454 case (SIOCGETSGCNT):
455 return (get_sg_cnt((struct sioc_sg_req *)data));
456 break;
457 default:
458 return (EINVAL);
459 break;
460 }
461 return error;
462}
463
464#ifndef MROUTE_LKM
465int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl;
466#endif
467
468/*
469 * returns the packet, byte, rpf-failure count for the source group provided
470 */
471static int
472get_sg_cnt(req)
473 register struct sioc_sg_req *req;
474{
475 register struct mfc *rt;
476 int s;
477
478 s = splnet();
479 MFCFIND(req->src.s_addr, req->grp.s_addr, rt);
480 splx(s);
481 if (rt != NULL) {
482 req->pktcnt = rt->mfc_pkt_cnt;
483 req->bytecnt = rt->mfc_byte_cnt;
484 req->wrong_if = rt->mfc_wrong_if;
485 } else
486 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
487
488 return 0;
489}
490
491/*
492 * returns the input and output packet and byte counts on the vif provided
493 */
494static int
495get_vif_cnt(req)
496 register struct sioc_vif_req *req;
497{
498 register vifi_t vifi = req->vifi;
499
500 if (vifi >= numvifs) return EINVAL;
501
502 req->icount = viftable[vifi].v_pkt_in;
503 req->ocount = viftable[vifi].v_pkt_out;
504 req->ibytes = viftable[vifi].v_bytes_in;
505 req->obytes = viftable[vifi].v_bytes_out;
506
507 return 0;
508}
509
510/*
511 * Enable multicast routing
512 */
513static int
514ip_mrouter_init(so, m)
515 struct socket *so;
516 struct mbuf *m;
517{
518 int *v;
519
520 if (mrtdebug)
521 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
522 so->so_type, so->so_proto->pr_protocol);
523
524 if (so->so_type != SOCK_RAW ||
525 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP;
526
527 if (!m || (m->m_len != sizeof(int *)))
528 return ENOPROTOOPT;
529
530 v = mtod(m, int *);
531 if (*v != 1)
532 return ENOPROTOOPT;
533
534 if (ip_mrouter != NULL) return EADDRINUSE;
535
536 ip_mrouter = so;
537
538 bzero((caddr_t)mfctable, sizeof(mfctable));
539 bzero((caddr_t)nexpire, sizeof(nexpire));
540
541 pim_assert = 0;
542
| 231#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
232#define UPCALL_EXPIRE 6 /* number of timeouts */
233
234/*
235 * Define the token bucket filter structures
236 * tbftable -> each vif has one of these for storing info
237 */
238
239static struct tbf tbftable[MAXVIFS];
240#define TBF_REPROCESS (hz / 100) /* 100x / second */
241
242/*
243 * 'Interfaces' associated with decapsulator (so we can tell
244 * packets that went through it from ones that get reflected
245 * by a broken gateway). These interfaces are never linked into
246 * the system ifnet list & no routes point to them. I.e., packets
247 * can't be sent this way. They only exist as a placeholder for
248 * multicast source verification.
249 */
250static struct ifnet multicast_decap_if[MAXVIFS];
251
252#define ENCAP_TTL 64
253#define ENCAP_PROTO IPPROTO_IPIP /* 4 */
254
255/* prototype IP hdr for encapsulated packets */
256static struct ip multicast_encap_iphdr = {
257#if BYTE_ORDER == LITTLE_ENDIAN
258 sizeof(struct ip) >> 2, IPVERSION,
259#else
260 IPVERSION, sizeof(struct ip) >> 2,
261#endif
262 0, /* tos */
263 sizeof(struct ip), /* total length */
264 0, /* id */
265 0, /* frag offset */
266 ENCAP_TTL, ENCAP_PROTO,
267 0, /* checksum */
268};
269
270/*
271 * Private variables.
272 */
273static vifi_t numvifs = 0;
274static int have_encap_tunnel = 0;
275
276/*
277 * one-back cache used by ipip_input to locate a tunnel's vif
278 * given a datagram's src ip address.
279 */
280static u_long last_encap_src;
281static struct vif *last_encap_vif;
282
283static u_long X_ip_mcast_src __P((int vifi));
284static int X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo));
285static int X_ip_mrouter_done __P((void));
286static int X_ip_mrouter_get __P((int cmd, struct socket *so, struct mbuf **m));
287static int X_ip_mrouter_set __P((int cmd, struct socket *so, struct mbuf *m));
288static int X_legal_vif_num __P((int vif));
289static int X_mrt_ioctl __P((int cmd, caddr_t data));
290
291static int get_sg_cnt(struct sioc_sg_req *);
292static int get_vif_cnt(struct sioc_vif_req *);
293static int ip_mrouter_init(struct socket *, struct mbuf *);
294static int add_vif(struct vifctl *);
295static int del_vif(vifi_t *);
296static int add_mfc(struct mfcctl *);
297static int del_mfc(struct mfcctl *);
298static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
299static int get_version(struct mbuf *);
300static int get_assert(struct mbuf *);
301static int set_assert(int *);
302static void expire_upcalls(void *);
303static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *,
304 vifi_t);
305static void phyint_send(struct ip *, struct vif *, struct mbuf *);
306static void encap_send(struct ip *, struct vif *, struct mbuf *);
307static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long);
308static void tbf_queue(struct vif *, struct mbuf *);
309static void tbf_process_q(struct vif *);
310static void tbf_reprocess_q(void *);
311static int tbf_dq_sel(struct vif *, struct ip *);
312static void tbf_send_packet(struct vif *, struct mbuf *);
313static void tbf_update_tokens(struct vif *);
314static int priority(struct vif *, struct ip *);
315void multiencap_decap(struct mbuf *);
316
317/*
318 * whether or not special PIM assert processing is enabled.
319 */
320static int pim_assert;
321/*
322 * Rate limit for assert notification messages, in usec
323 */
324#define ASSERT_MSG_TIME 3000000
325
326/*
327 * Hash function for a source, group entry
328 */
329#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
330 ((g) >> 20) ^ ((g) >> 10) ^ (g))
331
332/*
333 * Find a route for a given origin IP address and Multicast group address
334 * Type of service parameter to be added in the future!!!
335 */
336
337#define MFCFIND(o, g, rt) { \
338 register struct mbuf *_mb_rt = mfctable[MFCHASH(o,g)]; \
339 register struct mfc *_rt = NULL; \
340 rt = NULL; \
341 ++mrtstat.mrts_mfc_lookups; \
342 while (_mb_rt) { \
343 _rt = mtod(_mb_rt, struct mfc *); \
344 if ((_rt->mfc_origin.s_addr == o) && \
345 (_rt->mfc_mcastgrp.s_addr == g) && \
346 (_mb_rt->m_act == NULL)) { \
347 rt = _rt; \
348 break; \
349 } \
350 _mb_rt = _mb_rt->m_next; \
351 } \
352 if (rt == NULL) { \
353 ++mrtstat.mrts_mfc_misses; \
354 } \
355}
356
357
358/*
359 * Macros to compute elapsed time efficiently
360 * Borrowed from Van Jacobson's scheduling code
361 */
362#define TV_DELTA(a, b, delta) { \
363 register int xxs; \
364 \
365 delta = (a).tv_usec - (b).tv_usec; \
366 if ((xxs = (a).tv_sec - (b).tv_sec)) { \
367 switch (xxs) { \
368 case 2: \
369 delta += 1000000; \
370 /* fall through */ \
371 case 1: \
372 delta += 1000000; \
373 break; \
374 default: \
375 delta += (1000000 * xxs); \
376 } \
377 } \
378}
379
380#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
381 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
382
383#ifdef UPCALL_TIMING
384u_long upcall_data[51];
385static void collate(struct timeval *);
386#endif /* UPCALL_TIMING */
387
388
389/*
390 * Handle MRT setsockopt commands to modify the multicast routing tables.
391 */
392static int
393X_ip_mrouter_set(cmd, so, m)
394 int cmd;
395 struct socket *so;
396 struct mbuf *m;
397{
398 if (cmd != MRT_INIT && so != ip_mrouter) return EACCES;
399
400 switch (cmd) {
401 case MRT_INIT: return ip_mrouter_init(so, m);
402 case MRT_DONE: return ip_mrouter_done();
403 case MRT_ADD_VIF: return add_vif (mtod(m, struct vifctl *));
404 case MRT_DEL_VIF: return del_vif (mtod(m, vifi_t *));
405 case MRT_ADD_MFC: return add_mfc (mtod(m, struct mfcctl *));
406 case MRT_DEL_MFC: return del_mfc (mtod(m, struct mfcctl *));
407 case MRT_ASSERT: return set_assert(mtod(m, int *));
408 default: return EOPNOTSUPP;
409 }
410}
411
412#ifndef MROUTE_LKM
413int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = X_ip_mrouter_set;
414#endif
415
416/*
417 * Handle MRT getsockopt commands
418 */
419static int
420X_ip_mrouter_get(cmd, so, m)
421 int cmd;
422 struct socket *so;
423 struct mbuf **m;
424{
425 struct mbuf *mb;
426
427 if (so != ip_mrouter) return EACCES;
428
429 *m = mb = m_get(M_WAIT, MT_SOOPTS);
430
431 switch (cmd) {
432 case MRT_VERSION: return get_version(mb);
433 case MRT_ASSERT: return get_assert(mb);
434 default: return EOPNOTSUPP;
435 }
436}
437
438#ifndef MROUTE_LKM
439int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = X_ip_mrouter_get;
440#endif
441
442/*
443 * Handle ioctl commands to obtain information from the cache
444 */
445static int
446X_mrt_ioctl(cmd, data)
447 int cmd;
448 caddr_t data;
449{
450 int error = 0;
451
452 switch (cmd) {
453 case (SIOCGETVIFCNT):
454 return (get_vif_cnt((struct sioc_vif_req *)data));
455 break;
456 case (SIOCGETSGCNT):
457 return (get_sg_cnt((struct sioc_sg_req *)data));
458 break;
459 default:
460 return (EINVAL);
461 break;
462 }
463 return error;
464}
465
466#ifndef MROUTE_LKM
467int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl;
468#endif
469
470/*
471 * returns the packet, byte, rpf-failure count for the source group provided
472 */
473static int
474get_sg_cnt(req)
475 register struct sioc_sg_req *req;
476{
477 register struct mfc *rt;
478 int s;
479
480 s = splnet();
481 MFCFIND(req->src.s_addr, req->grp.s_addr, rt);
482 splx(s);
483 if (rt != NULL) {
484 req->pktcnt = rt->mfc_pkt_cnt;
485 req->bytecnt = rt->mfc_byte_cnt;
486 req->wrong_if = rt->mfc_wrong_if;
487 } else
488 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
489
490 return 0;
491}
492
493/*
494 * returns the input and output packet and byte counts on the vif provided
495 */
496static int
497get_vif_cnt(req)
498 register struct sioc_vif_req *req;
499{
500 register vifi_t vifi = req->vifi;
501
502 if (vifi >= numvifs) return EINVAL;
503
504 req->icount = viftable[vifi].v_pkt_in;
505 req->ocount = viftable[vifi].v_pkt_out;
506 req->ibytes = viftable[vifi].v_bytes_in;
507 req->obytes = viftable[vifi].v_bytes_out;
508
509 return 0;
510}
511
512/*
513 * Enable multicast routing
514 */
515static int
516ip_mrouter_init(so, m)
517 struct socket *so;
518 struct mbuf *m;
519{
520 int *v;
521
522 if (mrtdebug)
523 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
524 so->so_type, so->so_proto->pr_protocol);
525
526 if (so->so_type != SOCK_RAW ||
527 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP;
528
529 if (!m || (m->m_len != sizeof(int *)))
530 return ENOPROTOOPT;
531
532 v = mtod(m, int *);
533 if (*v != 1)
534 return ENOPROTOOPT;
535
536 if (ip_mrouter != NULL) return EADDRINUSE;
537
538 ip_mrouter = so;
539
540 bzero((caddr_t)mfctable, sizeof(mfctable));
541 bzero((caddr_t)nexpire, sizeof(nexpire));
542
543 pim_assert = 0;
544
|
543 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
| 545 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
|
544
545 if (mrtdebug)
546 log(LOG_DEBUG, "ip_mrouter_init\n");
547
548 return 0;
549}
550
551/*
552 * Disable multicast routing
553 */
554static int
555X_ip_mrouter_done()
556{
557 vifi_t vifi;
558 int i;
559 struct ifnet *ifp;
560 struct ifreq ifr;
561 struct mbuf *mb_rt;
562 struct mbuf *m;
563 struct rtdetq *rte;
564 int s;
565
566 s = splnet();
567
568 /*
569 * For each phyint in use, disable promiscuous reception of all IP
570 * multicasts.
571 */
572 for (vifi = 0; vifi < numvifs; vifi++) {
573 if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
574 !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
575 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
576 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr
577 = INADDR_ANY;
578 ifp = viftable[vifi].v_ifp;
579 if_allmulti(ifp, 0);
580 }
581 }
582 bzero((caddr_t)tbftable, sizeof(tbftable));
583 bzero((caddr_t)viftable, sizeof(viftable));
584 numvifs = 0;
585 pim_assert = 0;
586
| 546
547 if (mrtdebug)
548 log(LOG_DEBUG, "ip_mrouter_init\n");
549
550 return 0;
551}
552
553/*
554 * Disable multicast routing
555 */
556static int
557X_ip_mrouter_done()
558{
559 vifi_t vifi;
560 int i;
561 struct ifnet *ifp;
562 struct ifreq ifr;
563 struct mbuf *mb_rt;
564 struct mbuf *m;
565 struct rtdetq *rte;
566 int s;
567
568 s = splnet();
569
570 /*
571 * For each phyint in use, disable promiscuous reception of all IP
572 * multicasts.
573 */
574 for (vifi = 0; vifi < numvifs; vifi++) {
575 if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
576 !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
577 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
578 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr
579 = INADDR_ANY;
580 ifp = viftable[vifi].v_ifp;
581 if_allmulti(ifp, 0);
582 }
583 }
584 bzero((caddr_t)tbftable, sizeof(tbftable));
585 bzero((caddr_t)viftable, sizeof(viftable));
586 numvifs = 0;
587 pim_assert = 0;
588
|
587 untimeout(expire_upcalls, (caddr_t)NULL);
| 589 untimeout(expire_upcalls, (caddr_t)NULL, expire_upcalls_ch);
|
588
589 /*
590 * Free all multicast forwarding cache entries.
591 */
592 for (i = 0; i < MFCTBLSIZ; i++) {
593 mb_rt = mfctable[i];
594 while (mb_rt) {
595 if (mb_rt->m_act != NULL) {
596 while (mb_rt->m_act) {
597 m = mb_rt->m_act;
598 mb_rt->m_act = m->m_act;
599 rte = mtod(m, struct rtdetq *);
600 m_freem(rte->m);
601 m_free(m);
602 }
603 }
604 mb_rt = m_free(mb_rt);
605 }
606 }
607
608 bzero((caddr_t)mfctable, sizeof(mfctable));
609
610 /*
611 * Reset de-encapsulation cache
612 */
613 last_encap_src = 0;
614 last_encap_vif = NULL;
615 have_encap_tunnel = 0;
616
617 ip_mrouter = NULL;
618
619 splx(s);
620
621 if (mrtdebug)
622 log(LOG_DEBUG, "ip_mrouter_done\n");
623
624 return 0;
625}
626
627#ifndef MROUTE_LKM
628int (*ip_mrouter_done)(void) = X_ip_mrouter_done;
629#endif
630
631static int
632get_version(mb)
633 struct mbuf *mb;
634{
635 int *v;
636
637 v = mtod(mb, int *);
638
639 *v = 0x0305; /* XXX !!!! */
640 mb->m_len = sizeof(int);
641
642 return 0;
643}
644
645/*
646 * Set PIM assert processing global
647 */
648static int
649set_assert(i)
650 int *i;
651{
652 if ((*i != 1) && (*i != 0))
653 return EINVAL;
654
655 pim_assert = *i;
656
657 return 0;
658}
659
660/*
661 * Get PIM assert processing global
662 */
663static int
664get_assert(m)
665 struct mbuf *m;
666{
667 int *i;
668
669 i = mtod(m, int *);
670
671 *i = pim_assert;
672
673 return 0;
674}
675
676/*
677 * Add a vif to the vif table
678 */
679static int
680add_vif(vifcp)
681 register struct vifctl *vifcp;
682{
683 register struct vif *vifp = viftable + vifcp->vifc_vifi;
684 static struct sockaddr_in sin = {sizeof sin, AF_INET};
685 struct ifaddr *ifa;
686 struct ifnet *ifp;
687 struct ifreq ifr;
688 int error, s;
689 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
690
691 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL;
692 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE;
693
694 /* Find the interface with an address in AF_INET family */
695 sin.sin_addr = vifcp->vifc_lcl_addr;
696 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
697 if (ifa == 0) return EADDRNOTAVAIL;
698 ifp = ifa->ifa_ifp;
699
700 if (vifcp->vifc_flags & VIFF_TUNNEL) {
701 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
702 /*
703 * An encapsulating tunnel is wanted. Tell ipip_input() to
704 * start paying attention to encapsulated packets.
705 */
706 if (have_encap_tunnel == 0) {
707 have_encap_tunnel = 1;
708 for (s = 0; s < MAXVIFS; ++s) {
709 multicast_decap_if[s].if_name = "mdecap";
710 multicast_decap_if[s].if_unit = s;
711 }
712 }
713 /*
714 * Set interface to fake encapsulator interface
715 */
716 ifp = &multicast_decap_if[vifcp->vifc_vifi];
717 /*
718 * Prepare cached route entry
719 */
720 bzero(&vifp->v_route, sizeof(vifp->v_route));
721 } else {
722 log(LOG_ERR, "source routed tunnels not supported\n");
723 return EOPNOTSUPP;
724 }
725 } else {
726 /* Make sure the interface supports multicast */
727 if ((ifp->if_flags & IFF_MULTICAST) == 0)
728 return EOPNOTSUPP;
729
730 /* Enable promiscuous reception of all IP multicasts from the if */
731 s = splnet();
732 error = if_allmulti(ifp, 1);
733 splx(s);
734 if (error)
735 return error;
736 }
737
738 s = splnet();
739 /* define parameters for the tbf structure */
740 vifp->v_tbf = v_tbf;
741 GET_TIME(vifp->v_tbf->tbf_last_pkt_t);
742 vifp->v_tbf->tbf_n_tok = 0;
743 vifp->v_tbf->tbf_q_len = 0;
744 vifp->v_tbf->tbf_max_q_len = MAXQSIZE;
745 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL;
746
747 vifp->v_flags = vifcp->vifc_flags;
748 vifp->v_threshold = vifcp->vifc_threshold;
749 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
750 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
751 vifp->v_ifp = ifp;
752 /* scaling up here allows division by 1024 in critical code */
753 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000;
754 vifp->v_rsvp_on = 0;
755 vifp->v_rsvpd = NULL;
756 /* initialize per vif pkt counters */
757 vifp->v_pkt_in = 0;
758 vifp->v_pkt_out = 0;
759 vifp->v_bytes_in = 0;
760 vifp->v_bytes_out = 0;
761 splx(s);
762
763 /* Adjust numvifs up if the vifi is higher than numvifs */
764 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
765
766 if (mrtdebug)
767 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
768 vifcp->vifc_vifi,
769 ntohl(vifcp->vifc_lcl_addr.s_addr),
770 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
771 ntohl(vifcp->vifc_rmt_addr.s_addr),
772 vifcp->vifc_threshold,
773 vifcp->vifc_rate_limit);
774
775 return 0;
776}
777
778/*
779 * Delete a vif from the vif table
780 */
781static int
782del_vif(vifip)
783 vifi_t *vifip;
784{
785 register struct vif *vifp = viftable + *vifip;
786 register vifi_t vifi;
787 register struct mbuf *m;
788 struct ifnet *ifp;
789 struct ifreq ifr;
790 int s;
791
792 if (*vifip >= numvifs) return EINVAL;
793 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL;
794
795 s = splnet();
796
797 if (!(vifp->v_flags & VIFF_TUNNEL)) {
798 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
799 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY;
800 ifp = vifp->v_ifp;
801 if_allmulti(ifp, 0);
802 }
803
804 if (vifp == last_encap_vif) {
805 last_encap_vif = 0;
806 last_encap_src = 0;
807 }
808
809 /*
810 * Free packets queued at the interface
811 */
812 while (vifp->v_tbf->tbf_q) {
813 m = vifp->v_tbf->tbf_q;
814 vifp->v_tbf->tbf_q = m->m_act;
815 m_freem(m);
816 }
817
818 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
819 bzero((caddr_t)vifp, sizeof (*vifp));
820
821 /* Adjust numvifs down */
822 for (vifi = numvifs; vifi > 0; vifi--)
823 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break;
824 numvifs = vifi;
825
826 splx(s);
827
828 if (mrtdebug)
829 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
830
831 return 0;
832}
833
834/*
835 * Add an mfc entry
836 */
837static int
838add_mfc(mfccp)
839 struct mfcctl *mfccp;
840{
841 struct mfc *rt;
842 register struct mbuf *mb_rt;
843 u_long hash;
844 struct mbuf *mb_ntry;
845 struct rtdetq *rte;
846 register u_short nstl;
847 int s;
848 int i;
849
850 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt);
851
852 /* If an entry already exists, just update the fields */
853 if (rt) {
854 if (mrtdebug & DEBUG_MFC)
855 log(LOG_DEBUG,"add_mfc update o %x g %x p %x\n",
856 ntohl(mfccp->mfcc_origin.s_addr),
857 ntohl(mfccp->mfcc_mcastgrp.s_addr),
858 mfccp->mfcc_parent);
859
860 s = splnet();
861 rt->mfc_parent = mfccp->mfcc_parent;
862 for (i = 0; i < numvifs; i++)
863 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
864 splx(s);
865 return 0;
866 }
867
868 /*
869 * Find the entry for which the upcall was made and update
870 */
871 s = splnet();
872 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
873 for (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) {
874
875 rt = mtod(mb_rt, struct mfc *);
876 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
877 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
878 (mb_rt->m_act != NULL)) {
879
880 if (nstl++)
881 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %x\n",
882 "multiple kernel entries",
883 ntohl(mfccp->mfcc_origin.s_addr),
884 ntohl(mfccp->mfcc_mcastgrp.s_addr),
885 mfccp->mfcc_parent, mb_rt->m_act);
886
887 if (mrtdebug & DEBUG_MFC)
888 log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %x\n",
889 ntohl(mfccp->mfcc_origin.s_addr),
890 ntohl(mfccp->mfcc_mcastgrp.s_addr),
891 mfccp->mfcc_parent, mb_rt->m_act);
892
893 rt->mfc_origin = mfccp->mfcc_origin;
894 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
895 rt->mfc_parent = mfccp->mfcc_parent;
896 for (i = 0; i < numvifs; i++)
897 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
898 /* initialize pkt counters per src-grp */
899 rt->mfc_pkt_cnt = 0;
900 rt->mfc_byte_cnt = 0;
901 rt->mfc_wrong_if = 0;
902 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
903
904 rt->mfc_expire = 0; /* Don't clean this guy up */
905 nexpire[hash]--;
906
907 /* free packets Qed at the end of this entry */
908 while (mb_rt->m_act) {
909 mb_ntry = mb_rt->m_act;
910 rte = mtod(mb_ntry, struct rtdetq *);
911/* #ifdef RSVP_ISI */
912 ip_mdq(rte->m, rte->ifp, rt, -1);
913/* #endif */
914 mb_rt->m_act = mb_ntry->m_act;
915 m_freem(rte->m);
916#ifdef UPCALL_TIMING
917 collate(&(rte->t));
918#endif /* UPCALL_TIMING */
919 m_free(mb_ntry);
920 }
921 }
922 }
923
924 /*
925 * It is possible that an entry is being inserted without an upcall
926 */
927 if (nstl == 0) {
928 if (mrtdebug & DEBUG_MFC)
929 log(LOG_DEBUG,"add_mfc no upcall h %d o %x g %x p %x\n",
930 hash, ntohl(mfccp->mfcc_origin.s_addr),
931 ntohl(mfccp->mfcc_mcastgrp.s_addr),
932 mfccp->mfcc_parent);
933
934 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) {
935
936 rt = mtod(mb_rt, struct mfc *);
937 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
938 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
939
940 rt->mfc_origin = mfccp->mfcc_origin;
941 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
942 rt->mfc_parent = mfccp->mfcc_parent;
943 for (i = 0; i < numvifs; i++)
944 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
945 /* initialize pkt counters per src-grp */
946 rt->mfc_pkt_cnt = 0;
947 rt->mfc_byte_cnt = 0;
948 rt->mfc_wrong_if = 0;
949 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
950 if (rt->mfc_expire)
951 nexpire[hash]--;
952 rt->mfc_expire = 0;
953 }
954 }
955 if (mb_rt == NULL) {
956 /* no upcall, so make a new entry */
957 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE);
958 if (mb_rt == NULL) {
959 splx(s);
960 return ENOBUFS;
961 }
962
963 rt = mtod(mb_rt, struct mfc *);
964
965 /* insert new entry at head of hash chain */
966 rt->mfc_origin = mfccp->mfcc_origin;
967 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
968 rt->mfc_parent = mfccp->mfcc_parent;
969 for (i = 0; i < numvifs; i++)
970 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
971 /* initialize pkt counters per src-grp */
972 rt->mfc_pkt_cnt = 0;
973 rt->mfc_byte_cnt = 0;
974 rt->mfc_wrong_if = 0;
975 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
976 rt->mfc_expire = 0;
977
978 /* link into table */
979 mb_rt->m_next = mfctable[hash];
980 mfctable[hash] = mb_rt;
981 mb_rt->m_act = NULL;
982 }
983 }
984 splx(s);
985 return 0;
986}
987
988#ifdef UPCALL_TIMING
989/*
990 * collect delay statistics on the upcalls
991 */
992static void collate(t)
993register struct timeval *t;
994{
995 register u_long d;
996 register struct timeval tp;
997 register u_long delta;
998
999 GET_TIME(tp);
1000
1001 if (TV_LT(*t, tp))
1002 {
1003 TV_DELTA(tp, *t, delta);
1004
1005 d = delta >> 10;
1006 if (d > 50)
1007 d = 50;
1008
1009 ++upcall_data[d];
1010 }
1011}
1012#endif /* UPCALL_TIMING */
1013
1014/*
1015 * Delete an mfc entry
1016 */
1017static int
1018del_mfc(mfccp)
1019 struct mfcctl *mfccp;
1020{
1021 struct in_addr origin;
1022 struct in_addr mcastgrp;
1023 struct mfc *rt;
1024 struct mbuf *mb_rt;
1025 struct mbuf **nptr;
1026 u_long hash;
1027 int s;
1028
1029 origin = mfccp->mfcc_origin;
1030 mcastgrp = mfccp->mfcc_mcastgrp;
1031 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1032
1033 if (mrtdebug & DEBUG_MFC)
1034 log(LOG_DEBUG,"del_mfc orig %x mcastgrp %x\n",
1035 ntohl(origin.s_addr), ntohl(mcastgrp.s_addr));
1036
1037 s = splnet();
1038
1039 nptr = &mfctable[hash];
1040 while ((mb_rt = *nptr) != NULL) {
1041 rt = mtod(mb_rt, struct mfc *);
1042 if (origin.s_addr == rt->mfc_origin.s_addr &&
1043 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1044 mb_rt->m_act == NULL)
1045 break;
1046
1047 nptr = &mb_rt->m_next;
1048 }
1049 if (mb_rt == NULL) {
1050 splx(s);
1051 return EADDRNOTAVAIL;
1052 }
1053
1054 MFREE(mb_rt, *nptr);
1055
1056 splx(s);
1057
1058 return 0;
1059}
1060
1061/*
1062 * Send a message to mrouted on the multicast routing socket
1063 */
1064static int
1065socket_send(s, mm, src)
1066 struct socket *s;
1067 struct mbuf *mm;
1068 struct sockaddr_in *src;
1069{
1070 if (s) {
1071 if (sbappendaddr(&s->so_rcv,
1072 (struct sockaddr *)src,
1073 mm, (struct mbuf *)0) != 0) {
1074 sorwakeup(s);
1075 return 0;
1076 }
1077 }
1078 m_freem(mm);
1079 return -1;
1080}
1081
1082/*
1083 * IP multicast forwarding function. This function assumes that the packet
1084 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1085 * pointed to by "ifp", and the packet is to be relayed to other networks
1086 * that have members of the packet's destination IP multicast group.
1087 *
1088 * The packet is returned unscathed to the caller, unless it is
1089 * erroneous, in which case a non-zero return value tells the caller to
1090 * discard it.
1091 */
1092
1093#define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */
1094#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1095
1096static int
1097X_ip_mforward(ip, ifp, m, imo)
1098 register struct ip *ip;
1099 struct ifnet *ifp;
1100 struct mbuf *m;
1101 struct ip_moptions *imo;
1102{
1103 register struct mfc *rt;
1104 register u_char *ipoptions;
1105 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1106 static int srctun = 0;
1107 register struct mbuf *mm;
1108 int s;
1109 vifi_t vifi;
1110 struct vif *vifp;
1111
1112 if (mrtdebug & DEBUG_FORWARD)
1113 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %x\n",
1114 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1115
1116 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1117 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1118 /*
1119 * Packet arrived via a physical interface or
1120 * an encapsulated tunnel.
1121 */
1122 } else {
1123 /*
1124 * Packet arrived through a source-route tunnel.
1125 * Source-route tunnels are no longer supported.
1126 */
1127 if ((srctun++ % 1000) == 0)
1128 log(LOG_ERR, "ip_mforward: received source-routed packet from %x\n",
1129 ntohl(ip->ip_src.s_addr));
1130
1131 return 1;
1132 }
1133
1134 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1135 if (ip->ip_ttl < 255)
1136 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1137 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1138 vifp = viftable + vifi;
1139 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n",
1140 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi,
1141 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1142 vifp->v_ifp->if_name, vifp->v_ifp->if_unit);
1143 }
1144 return (ip_mdq(m, ifp, NULL, vifi));
1145 }
1146 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1147 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n",
1148 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr));
1149 if(!imo)
1150 printf("In fact, no options were specified at all\n");
1151 }
1152
1153 /*
1154 * Don't forward a packet with time-to-live of zero or one,
1155 * or a packet destined to a local-only group.
1156 */
1157 if (ip->ip_ttl <= 1 ||
1158 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP)
1159 return 0;
1160
1161 /*
1162 * Determine forwarding vifs from the forwarding cache table
1163 */
1164 s = splnet();
1165 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt);
1166
1167 /* Entry exists, so forward if necessary */
1168 if (rt != NULL) {
1169 splx(s);
1170 return (ip_mdq(m, ifp, rt, -1));
1171 } else {
1172 /*
1173 * If we don't have a route for packet's origin,
1174 * Make a copy of the packet &
1175 * send message to routing daemon
1176 */
1177
1178 register struct mbuf *mb_rt;
1179 register struct mbuf *mb_ntry;
1180 register struct mbuf *mb0;
1181 register struct rtdetq *rte;
1182 register struct mbuf *rte_m;
1183 register u_long hash;
1184 register int npkts;
1185 int hlen = ip->ip_hl << 2;
1186#ifdef UPCALL_TIMING
1187 struct timeval tp;
1188
1189 GET_TIME(tp);
1190#endif
1191
1192 mrtstat.mrts_no_route++;
1193 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1194 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1195 ntohl(ip->ip_src.s_addr),
1196 ntohl(ip->ip_dst.s_addr));
1197
1198 /*
1199 * Allocate mbufs early so that we don't do extra work if we are
1200 * just going to fail anyway. Make sure to pullup the header so
1201 * that other people can't step on it.
1202 */
1203 MGET(mb_ntry, M_DONTWAIT, MT_DATA);
1204 if (mb_ntry == NULL) {
1205 splx(s);
1206 return ENOBUFS;
1207 }
1208 mb0 = m_copy(m, 0, M_COPYALL);
1209 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1210 mb0 = m_pullup(mb0, hlen);
1211 if (mb0 == NULL) {
1212 m_free(mb_ntry);
1213 splx(s);
1214 return ENOBUFS;
1215 }
1216
1217 /* is there an upcall waiting for this packet? */
1218 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1219 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) {
1220 rt = mtod(mb_rt, struct mfc *);
1221 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1222 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1223 (mb_rt->m_act != NULL))
1224 break;
1225 }
1226
1227 if (mb_rt == NULL) {
1228 int i;
1229 struct igmpmsg *im;
1230
1231 /* no upcall, so make a new entry */
1232 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE);
1233 if (mb_rt == NULL) {
1234 m_free(mb_ntry);
1235 m_freem(mb0);
1236 splx(s);
1237 return ENOBUFS;
1238 }
1239 /* Make a copy of the header to send to the user level process */
1240 mm = m_copy(mb0, 0, hlen);
1241 if (mm == NULL) {
1242 m_free(mb_ntry);
1243 m_freem(mb0);
1244 m_free(mb_rt);
1245 splx(s);
1246 return ENOBUFS;
1247 }
1248
1249 /*
1250 * Send message to routing daemon to install
1251 * a route into the kernel table
1252 */
1253 k_igmpsrc.sin_addr = ip->ip_src;
1254
1255 im = mtod(mm, struct igmpmsg *);
1256 im->im_msgtype = IGMPMSG_NOCACHE;
1257 im->im_mbz = 0;
1258
1259 mrtstat.mrts_upcalls++;
1260
1261 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1262 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1263 ++mrtstat.mrts_upq_sockfull;
1264 m_free(mb_ntry);
1265 m_freem(mb0);
1266 m_free(mb_rt);
1267 splx(s);
1268 return ENOBUFS;
1269 }
1270
1271 rt = mtod(mb_rt, struct mfc *);
1272
1273 /* insert new entry at head of hash chain */
1274 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1275 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1276 rt->mfc_expire = UPCALL_EXPIRE;
1277 nexpire[hash]++;
1278 for (i = 0; i < numvifs; i++)
1279 rt->mfc_ttls[i] = 0;
1280 rt->mfc_parent = -1;
1281
1282 /* link into table */
1283 mb_rt->m_next = mfctable[hash];
1284 mfctable[hash] = mb_rt;
1285 mb_rt->m_act = NULL;
1286
1287 rte_m = mb_rt;
1288 } else {
1289 /* determine if q has overflowed */
1290 for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act)
1291 npkts++;
1292
1293 if (npkts > MAX_UPQ) {
1294 mrtstat.mrts_upq_ovflw++;
1295 m_free(mb_ntry);
1296 m_freem(mb0);
1297 splx(s);
1298 return 0;
1299 }
1300 }
1301
1302 mb_ntry->m_act = NULL;
1303 rte = mtod(mb_ntry, struct rtdetq *);
1304
1305 rte->m = mb0;
1306 rte->ifp = ifp;
1307#ifdef UPCALL_TIMING
1308 rte->t = tp;
1309#endif
1310
1311 /* Add this entry to the end of the queue */
1312 rte_m->m_act = mb_ntry;
1313
1314 splx(s);
1315
1316 return 0;
1317 }
1318}
1319
1320#ifndef MROUTE_LKM
1321int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
1322 struct ip_moptions *) = X_ip_mforward;
1323#endif
1324
1325/*
1326 * Clean up the cache entry if upcall is not serviced
1327 */
1328static void
1329expire_upcalls(void *unused)
1330{
1331 struct mbuf *mb_rt, *m, **nptr;
1332 struct rtdetq *rte;
1333 struct mfc *mfc;
1334 int i;
1335 int s;
1336
1337 s = splnet();
1338 for (i = 0; i < MFCTBLSIZ; i++) {
1339 if (nexpire[i] == 0)
1340 continue;
1341 nptr = &mfctable[i];
1342 for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) {
1343 mfc = mtod(mb_rt, struct mfc *);
1344
1345 /*
1346 * Skip real cache entries
1347 * Make sure it wasn't marked to not expire (shouldn't happen)
1348 * If it expires now
1349 */
1350 if (mb_rt->m_act != NULL &&
1351 mfc->mfc_expire != 0 &&
1352 --mfc->mfc_expire == 0) {
1353 if (mrtdebug & DEBUG_EXPIRE)
1354 log(LOG_DEBUG, "expire_upcalls: expiring (%x %x)\n",
1355 ntohl(mfc->mfc_origin.s_addr),
1356 ntohl(mfc->mfc_mcastgrp.s_addr));
1357 /*
1358 * drop all the packets
1359 * free the mbuf with the pkt, if, timing info
1360 */
1361 while (mb_rt->m_act) {
1362 m = mb_rt->m_act;
1363 mb_rt->m_act = m->m_act;
1364
1365 rte = mtod(m, struct rtdetq *);
1366 m_freem(rte->m);
1367 m_free(m);
1368 }
1369 ++mrtstat.mrts_cache_cleanups;
1370 nexpire[i]--;
1371
1372 MFREE(mb_rt, *nptr);
1373 } else {
1374 nptr = &mb_rt->m_next;
1375 }
1376 }
1377 }
1378 splx(s);
| 590
591 /*
592 * Free all multicast forwarding cache entries.
593 */
594 for (i = 0; i < MFCTBLSIZ; i++) {
595 mb_rt = mfctable[i];
596 while (mb_rt) {
597 if (mb_rt->m_act != NULL) {
598 while (mb_rt->m_act) {
599 m = mb_rt->m_act;
600 mb_rt->m_act = m->m_act;
601 rte = mtod(m, struct rtdetq *);
602 m_freem(rte->m);
603 m_free(m);
604 }
605 }
606 mb_rt = m_free(mb_rt);
607 }
608 }
609
610 bzero((caddr_t)mfctable, sizeof(mfctable));
611
612 /*
613 * Reset de-encapsulation cache
614 */
615 last_encap_src = 0;
616 last_encap_vif = NULL;
617 have_encap_tunnel = 0;
618
619 ip_mrouter = NULL;
620
621 splx(s);
622
623 if (mrtdebug)
624 log(LOG_DEBUG, "ip_mrouter_done\n");
625
626 return 0;
627}
628
629#ifndef MROUTE_LKM
630int (*ip_mrouter_done)(void) = X_ip_mrouter_done;
631#endif
632
633static int
634get_version(mb)
635 struct mbuf *mb;
636{
637 int *v;
638
639 v = mtod(mb, int *);
640
641 *v = 0x0305; /* XXX !!!! */
642 mb->m_len = sizeof(int);
643
644 return 0;
645}
646
647/*
648 * Set PIM assert processing global
649 */
650static int
651set_assert(i)
652 int *i;
653{
654 if ((*i != 1) && (*i != 0))
655 return EINVAL;
656
657 pim_assert = *i;
658
659 return 0;
660}
661
662/*
663 * Get PIM assert processing global
664 */
665static int
666get_assert(m)
667 struct mbuf *m;
668{
669 int *i;
670
671 i = mtod(m, int *);
672
673 *i = pim_assert;
674
675 return 0;
676}
677
678/*
679 * Add a vif to the vif table
680 */
681static int
682add_vif(vifcp)
683 register struct vifctl *vifcp;
684{
685 register struct vif *vifp = viftable + vifcp->vifc_vifi;
686 static struct sockaddr_in sin = {sizeof sin, AF_INET};
687 struct ifaddr *ifa;
688 struct ifnet *ifp;
689 struct ifreq ifr;
690 int error, s;
691 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
692
693 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL;
694 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE;
695
696 /* Find the interface with an address in AF_INET family */
697 sin.sin_addr = vifcp->vifc_lcl_addr;
698 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
699 if (ifa == 0) return EADDRNOTAVAIL;
700 ifp = ifa->ifa_ifp;
701
702 if (vifcp->vifc_flags & VIFF_TUNNEL) {
703 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
704 /*
705 * An encapsulating tunnel is wanted. Tell ipip_input() to
706 * start paying attention to encapsulated packets.
707 */
708 if (have_encap_tunnel == 0) {
709 have_encap_tunnel = 1;
710 for (s = 0; s < MAXVIFS; ++s) {
711 multicast_decap_if[s].if_name = "mdecap";
712 multicast_decap_if[s].if_unit = s;
713 }
714 }
715 /*
716 * Set interface to fake encapsulator interface
717 */
718 ifp = &multicast_decap_if[vifcp->vifc_vifi];
719 /*
720 * Prepare cached route entry
721 */
722 bzero(&vifp->v_route, sizeof(vifp->v_route));
723 } else {
724 log(LOG_ERR, "source routed tunnels not supported\n");
725 return EOPNOTSUPP;
726 }
727 } else {
728 /* Make sure the interface supports multicast */
729 if ((ifp->if_flags & IFF_MULTICAST) == 0)
730 return EOPNOTSUPP;
731
732 /* Enable promiscuous reception of all IP multicasts from the if */
733 s = splnet();
734 error = if_allmulti(ifp, 1);
735 splx(s);
736 if (error)
737 return error;
738 }
739
740 s = splnet();
741 /* define parameters for the tbf structure */
742 vifp->v_tbf = v_tbf;
743 GET_TIME(vifp->v_tbf->tbf_last_pkt_t);
744 vifp->v_tbf->tbf_n_tok = 0;
745 vifp->v_tbf->tbf_q_len = 0;
746 vifp->v_tbf->tbf_max_q_len = MAXQSIZE;
747 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL;
748
749 vifp->v_flags = vifcp->vifc_flags;
750 vifp->v_threshold = vifcp->vifc_threshold;
751 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
752 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
753 vifp->v_ifp = ifp;
754 /* scaling up here allows division by 1024 in critical code */
755 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000;
756 vifp->v_rsvp_on = 0;
757 vifp->v_rsvpd = NULL;
758 /* initialize per vif pkt counters */
759 vifp->v_pkt_in = 0;
760 vifp->v_pkt_out = 0;
761 vifp->v_bytes_in = 0;
762 vifp->v_bytes_out = 0;
763 splx(s);
764
765 /* Adjust numvifs up if the vifi is higher than numvifs */
766 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
767
768 if (mrtdebug)
769 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n",
770 vifcp->vifc_vifi,
771 ntohl(vifcp->vifc_lcl_addr.s_addr),
772 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
773 ntohl(vifcp->vifc_rmt_addr.s_addr),
774 vifcp->vifc_threshold,
775 vifcp->vifc_rate_limit);
776
777 return 0;
778}
779
780/*
781 * Delete a vif from the vif table
782 */
783static int
784del_vif(vifip)
785 vifi_t *vifip;
786{
787 register struct vif *vifp = viftable + *vifip;
788 register vifi_t vifi;
789 register struct mbuf *m;
790 struct ifnet *ifp;
791 struct ifreq ifr;
792 int s;
793
794 if (*vifip >= numvifs) return EINVAL;
795 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL;
796
797 s = splnet();
798
799 if (!(vifp->v_flags & VIFF_TUNNEL)) {
800 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET;
801 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY;
802 ifp = vifp->v_ifp;
803 if_allmulti(ifp, 0);
804 }
805
806 if (vifp == last_encap_vif) {
807 last_encap_vif = 0;
808 last_encap_src = 0;
809 }
810
811 /*
812 * Free packets queued at the interface
813 */
814 while (vifp->v_tbf->tbf_q) {
815 m = vifp->v_tbf->tbf_q;
816 vifp->v_tbf->tbf_q = m->m_act;
817 m_freem(m);
818 }
819
820 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
821 bzero((caddr_t)vifp, sizeof (*vifp));
822
823 /* Adjust numvifs down */
824 for (vifi = numvifs; vifi > 0; vifi--)
825 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break;
826 numvifs = vifi;
827
828 splx(s);
829
830 if (mrtdebug)
831 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs);
832
833 return 0;
834}
835
836/*
837 * Add an mfc entry
838 */
839static int
840add_mfc(mfccp)
841 struct mfcctl *mfccp;
842{
843 struct mfc *rt;
844 register struct mbuf *mb_rt;
845 u_long hash;
846 struct mbuf *mb_ntry;
847 struct rtdetq *rte;
848 register u_short nstl;
849 int s;
850 int i;
851
852 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt);
853
854 /* If an entry already exists, just update the fields */
855 if (rt) {
856 if (mrtdebug & DEBUG_MFC)
857 log(LOG_DEBUG,"add_mfc update o %x g %x p %x\n",
858 ntohl(mfccp->mfcc_origin.s_addr),
859 ntohl(mfccp->mfcc_mcastgrp.s_addr),
860 mfccp->mfcc_parent);
861
862 s = splnet();
863 rt->mfc_parent = mfccp->mfcc_parent;
864 for (i = 0; i < numvifs; i++)
865 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
866 splx(s);
867 return 0;
868 }
869
870 /*
871 * Find the entry for which the upcall was made and update
872 */
873 s = splnet();
874 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
875 for (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) {
876
877 rt = mtod(mb_rt, struct mfc *);
878 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
879 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
880 (mb_rt->m_act != NULL)) {
881
882 if (nstl++)
883 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %x\n",
884 "multiple kernel entries",
885 ntohl(mfccp->mfcc_origin.s_addr),
886 ntohl(mfccp->mfcc_mcastgrp.s_addr),
887 mfccp->mfcc_parent, mb_rt->m_act);
888
889 if (mrtdebug & DEBUG_MFC)
890 log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %x\n",
891 ntohl(mfccp->mfcc_origin.s_addr),
892 ntohl(mfccp->mfcc_mcastgrp.s_addr),
893 mfccp->mfcc_parent, mb_rt->m_act);
894
895 rt->mfc_origin = mfccp->mfcc_origin;
896 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
897 rt->mfc_parent = mfccp->mfcc_parent;
898 for (i = 0; i < numvifs; i++)
899 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
900 /* initialize pkt counters per src-grp */
901 rt->mfc_pkt_cnt = 0;
902 rt->mfc_byte_cnt = 0;
903 rt->mfc_wrong_if = 0;
904 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
905
906 rt->mfc_expire = 0; /* Don't clean this guy up */
907 nexpire[hash]--;
908
909 /* free packets Qed at the end of this entry */
910 while (mb_rt->m_act) {
911 mb_ntry = mb_rt->m_act;
912 rte = mtod(mb_ntry, struct rtdetq *);
913/* #ifdef RSVP_ISI */
914 ip_mdq(rte->m, rte->ifp, rt, -1);
915/* #endif */
916 mb_rt->m_act = mb_ntry->m_act;
917 m_freem(rte->m);
918#ifdef UPCALL_TIMING
919 collate(&(rte->t));
920#endif /* UPCALL_TIMING */
921 m_free(mb_ntry);
922 }
923 }
924 }
925
926 /*
927 * It is possible that an entry is being inserted without an upcall
928 */
929 if (nstl == 0) {
930 if (mrtdebug & DEBUG_MFC)
931 log(LOG_DEBUG,"add_mfc no upcall h %d o %x g %x p %x\n",
932 hash, ntohl(mfccp->mfcc_origin.s_addr),
933 ntohl(mfccp->mfcc_mcastgrp.s_addr),
934 mfccp->mfcc_parent);
935
936 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) {
937
938 rt = mtod(mb_rt, struct mfc *);
939 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
940 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
941
942 rt->mfc_origin = mfccp->mfcc_origin;
943 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
944 rt->mfc_parent = mfccp->mfcc_parent;
945 for (i = 0; i < numvifs; i++)
946 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
947 /* initialize pkt counters per src-grp */
948 rt->mfc_pkt_cnt = 0;
949 rt->mfc_byte_cnt = 0;
950 rt->mfc_wrong_if = 0;
951 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
952 if (rt->mfc_expire)
953 nexpire[hash]--;
954 rt->mfc_expire = 0;
955 }
956 }
957 if (mb_rt == NULL) {
958 /* no upcall, so make a new entry */
959 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE);
960 if (mb_rt == NULL) {
961 splx(s);
962 return ENOBUFS;
963 }
964
965 rt = mtod(mb_rt, struct mfc *);
966
967 /* insert new entry at head of hash chain */
968 rt->mfc_origin = mfccp->mfcc_origin;
969 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
970 rt->mfc_parent = mfccp->mfcc_parent;
971 for (i = 0; i < numvifs; i++)
972 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
973 /* initialize pkt counters per src-grp */
974 rt->mfc_pkt_cnt = 0;
975 rt->mfc_byte_cnt = 0;
976 rt->mfc_wrong_if = 0;
977 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
978 rt->mfc_expire = 0;
979
980 /* link into table */
981 mb_rt->m_next = mfctable[hash];
982 mfctable[hash] = mb_rt;
983 mb_rt->m_act = NULL;
984 }
985 }
986 splx(s);
987 return 0;
988}
989
990#ifdef UPCALL_TIMING
991/*
992 * collect delay statistics on the upcalls
993 */
994static void collate(t)
995register struct timeval *t;
996{
997 register u_long d;
998 register struct timeval tp;
999 register u_long delta;
1000
1001 GET_TIME(tp);
1002
1003 if (TV_LT(*t, tp))
1004 {
1005 TV_DELTA(tp, *t, delta);
1006
1007 d = delta >> 10;
1008 if (d > 50)
1009 d = 50;
1010
1011 ++upcall_data[d];
1012 }
1013}
1014#endif /* UPCALL_TIMING */
1015
1016/*
1017 * Delete an mfc entry
1018 */
1019static int
1020del_mfc(mfccp)
1021 struct mfcctl *mfccp;
1022{
1023 struct in_addr origin;
1024 struct in_addr mcastgrp;
1025 struct mfc *rt;
1026 struct mbuf *mb_rt;
1027 struct mbuf **nptr;
1028 u_long hash;
1029 int s;
1030
1031 origin = mfccp->mfcc_origin;
1032 mcastgrp = mfccp->mfcc_mcastgrp;
1033 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1034
1035 if (mrtdebug & DEBUG_MFC)
1036 log(LOG_DEBUG,"del_mfc orig %x mcastgrp %x\n",
1037 ntohl(origin.s_addr), ntohl(mcastgrp.s_addr));
1038
1039 s = splnet();
1040
1041 nptr = &mfctable[hash];
1042 while ((mb_rt = *nptr) != NULL) {
1043 rt = mtod(mb_rt, struct mfc *);
1044 if (origin.s_addr == rt->mfc_origin.s_addr &&
1045 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1046 mb_rt->m_act == NULL)
1047 break;
1048
1049 nptr = &mb_rt->m_next;
1050 }
1051 if (mb_rt == NULL) {
1052 splx(s);
1053 return EADDRNOTAVAIL;
1054 }
1055
1056 MFREE(mb_rt, *nptr);
1057
1058 splx(s);
1059
1060 return 0;
1061}
1062
1063/*
1064 * Send a message to mrouted on the multicast routing socket
1065 */
1066static int
1067socket_send(s, mm, src)
1068 struct socket *s;
1069 struct mbuf *mm;
1070 struct sockaddr_in *src;
1071{
1072 if (s) {
1073 if (sbappendaddr(&s->so_rcv,
1074 (struct sockaddr *)src,
1075 mm, (struct mbuf *)0) != 0) {
1076 sorwakeup(s);
1077 return 0;
1078 }
1079 }
1080 m_freem(mm);
1081 return -1;
1082}
1083
1084/*
1085 * IP multicast forwarding function. This function assumes that the packet
1086 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1087 * pointed to by "ifp", and the packet is to be relayed to other networks
1088 * that have members of the packet's destination IP multicast group.
1089 *
1090 * The packet is returned unscathed to the caller, unless it is
1091 * erroneous, in which case a non-zero return value tells the caller to
1092 * discard it.
1093 */
1094
1095#define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */
1096#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1097
1098static int
1099X_ip_mforward(ip, ifp, m, imo)
1100 register struct ip *ip;
1101 struct ifnet *ifp;
1102 struct mbuf *m;
1103 struct ip_moptions *imo;
1104{
1105 register struct mfc *rt;
1106 register u_char *ipoptions;
1107 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1108 static int srctun = 0;
1109 register struct mbuf *mm;
1110 int s;
1111 vifi_t vifi;
1112 struct vif *vifp;
1113
1114 if (mrtdebug & DEBUG_FORWARD)
1115 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %x\n",
1116 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp);
1117
1118 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 ||
1119 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1120 /*
1121 * Packet arrived via a physical interface or
1122 * an encapsulated tunnel.
1123 */
1124 } else {
1125 /*
1126 * Packet arrived through a source-route tunnel.
1127 * Source-route tunnels are no longer supported.
1128 */
1129 if ((srctun++ % 1000) == 0)
1130 log(LOG_ERR, "ip_mforward: received source-routed packet from %x\n",
1131 ntohl(ip->ip_src.s_addr));
1132
1133 return 1;
1134 }
1135
1136 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1137 if (ip->ip_ttl < 255)
1138 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1139 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1140 vifp = viftable + vifi;
1141 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n",
1142 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi,
1143 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1144 vifp->v_ifp->if_name, vifp->v_ifp->if_unit);
1145 }
1146 return (ip_mdq(m, ifp, NULL, vifi));
1147 }
1148 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1149 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n",
1150 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr));
1151 if(!imo)
1152 printf("In fact, no options were specified at all\n");
1153 }
1154
1155 /*
1156 * Don't forward a packet with time-to-live of zero or one,
1157 * or a packet destined to a local-only group.
1158 */
1159 if (ip->ip_ttl <= 1 ||
1160 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP)
1161 return 0;
1162
1163 /*
1164 * Determine forwarding vifs from the forwarding cache table
1165 */
1166 s = splnet();
1167 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt);
1168
1169 /* Entry exists, so forward if necessary */
1170 if (rt != NULL) {
1171 splx(s);
1172 return (ip_mdq(m, ifp, rt, -1));
1173 } else {
1174 /*
1175 * If we don't have a route for packet's origin,
1176 * Make a copy of the packet &
1177 * send message to routing daemon
1178 */
1179
1180 register struct mbuf *mb_rt;
1181 register struct mbuf *mb_ntry;
1182 register struct mbuf *mb0;
1183 register struct rtdetq *rte;
1184 register struct mbuf *rte_m;
1185 register u_long hash;
1186 register int npkts;
1187 int hlen = ip->ip_hl << 2;
1188#ifdef UPCALL_TIMING
1189 struct timeval tp;
1190
1191 GET_TIME(tp);
1192#endif
1193
1194 mrtstat.mrts_no_route++;
1195 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1196 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n",
1197 ntohl(ip->ip_src.s_addr),
1198 ntohl(ip->ip_dst.s_addr));
1199
1200 /*
1201 * Allocate mbufs early so that we don't do extra work if we are
1202 * just going to fail anyway. Make sure to pullup the header so
1203 * that other people can't step on it.
1204 */
1205 MGET(mb_ntry, M_DONTWAIT, MT_DATA);
1206 if (mb_ntry == NULL) {
1207 splx(s);
1208 return ENOBUFS;
1209 }
1210 mb0 = m_copy(m, 0, M_COPYALL);
1211 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1212 mb0 = m_pullup(mb0, hlen);
1213 if (mb0 == NULL) {
1214 m_free(mb_ntry);
1215 splx(s);
1216 return ENOBUFS;
1217 }
1218
1219 /* is there an upcall waiting for this packet? */
1220 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1221 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) {
1222 rt = mtod(mb_rt, struct mfc *);
1223 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1224 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1225 (mb_rt->m_act != NULL))
1226 break;
1227 }
1228
1229 if (mb_rt == NULL) {
1230 int i;
1231 struct igmpmsg *im;
1232
1233 /* no upcall, so make a new entry */
1234 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE);
1235 if (mb_rt == NULL) {
1236 m_free(mb_ntry);
1237 m_freem(mb0);
1238 splx(s);
1239 return ENOBUFS;
1240 }
1241 /* Make a copy of the header to send to the user level process */
1242 mm = m_copy(mb0, 0, hlen);
1243 if (mm == NULL) {
1244 m_free(mb_ntry);
1245 m_freem(mb0);
1246 m_free(mb_rt);
1247 splx(s);
1248 return ENOBUFS;
1249 }
1250
1251 /*
1252 * Send message to routing daemon to install
1253 * a route into the kernel table
1254 */
1255 k_igmpsrc.sin_addr = ip->ip_src;
1256
1257 im = mtod(mm, struct igmpmsg *);
1258 im->im_msgtype = IGMPMSG_NOCACHE;
1259 im->im_mbz = 0;
1260
1261 mrtstat.mrts_upcalls++;
1262
1263 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1264 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1265 ++mrtstat.mrts_upq_sockfull;
1266 m_free(mb_ntry);
1267 m_freem(mb0);
1268 m_free(mb_rt);
1269 splx(s);
1270 return ENOBUFS;
1271 }
1272
1273 rt = mtod(mb_rt, struct mfc *);
1274
1275 /* insert new entry at head of hash chain */
1276 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1277 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1278 rt->mfc_expire = UPCALL_EXPIRE;
1279 nexpire[hash]++;
1280 for (i = 0; i < numvifs; i++)
1281 rt->mfc_ttls[i] = 0;
1282 rt->mfc_parent = -1;
1283
1284 /* link into table */
1285 mb_rt->m_next = mfctable[hash];
1286 mfctable[hash] = mb_rt;
1287 mb_rt->m_act = NULL;
1288
1289 rte_m = mb_rt;
1290 } else {
1291 /* determine if q has overflowed */
1292 for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act)
1293 npkts++;
1294
1295 if (npkts > MAX_UPQ) {
1296 mrtstat.mrts_upq_ovflw++;
1297 m_free(mb_ntry);
1298 m_freem(mb0);
1299 splx(s);
1300 return 0;
1301 }
1302 }
1303
1304 mb_ntry->m_act = NULL;
1305 rte = mtod(mb_ntry, struct rtdetq *);
1306
1307 rte->m = mb0;
1308 rte->ifp = ifp;
1309#ifdef UPCALL_TIMING
1310 rte->t = tp;
1311#endif
1312
1313 /* Add this entry to the end of the queue */
1314 rte_m->m_act = mb_ntry;
1315
1316 splx(s);
1317
1318 return 0;
1319 }
1320}
1321
1322#ifndef MROUTE_LKM
1323int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
1324 struct ip_moptions *) = X_ip_mforward;
1325#endif
1326
1327/*
1328 * Clean up the cache entry if upcall is not serviced
1329 */
1330static void
1331expire_upcalls(void *unused)
1332{
1333 struct mbuf *mb_rt, *m, **nptr;
1334 struct rtdetq *rte;
1335 struct mfc *mfc;
1336 int i;
1337 int s;
1338
1339 s = splnet();
1340 for (i = 0; i < MFCTBLSIZ; i++) {
1341 if (nexpire[i] == 0)
1342 continue;
1343 nptr = &mfctable[i];
1344 for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) {
1345 mfc = mtod(mb_rt, struct mfc *);
1346
1347 /*
1348 * Skip real cache entries
1349 * Make sure it wasn't marked to not expire (shouldn't happen)
1350 * If it expires now
1351 */
1352 if (mb_rt->m_act != NULL &&
1353 mfc->mfc_expire != 0 &&
1354 --mfc->mfc_expire == 0) {
1355 if (mrtdebug & DEBUG_EXPIRE)
1356 log(LOG_DEBUG, "expire_upcalls: expiring (%x %x)\n",
1357 ntohl(mfc->mfc_origin.s_addr),
1358 ntohl(mfc->mfc_mcastgrp.s_addr));
1359 /*
1360 * drop all the packets
1361 * free the mbuf with the pkt, if, timing info
1362 */
1363 while (mb_rt->m_act) {
1364 m = mb_rt->m_act;
1365 mb_rt->m_act = m->m_act;
1366
1367 rte = mtod(m, struct rtdetq *);
1368 m_freem(rte->m);
1369 m_free(m);
1370 }
1371 ++mrtstat.mrts_cache_cleanups;
1372 nexpire[i]--;
1373
1374 MFREE(mb_rt, *nptr);
1375 } else {
1376 nptr = &mb_rt->m_next;
1377 }
1378 }
1379 }
1380 splx(s);
|
1379 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
| 1381 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
|
1380}
1381
1382/*
1383 * Packet forwarding routine once entry in the cache is made
1384 */
1385static int
1386ip_mdq(m, ifp, rt, xmt_vif)
1387 register struct mbuf *m;
1388 register struct ifnet *ifp;
1389 register struct mfc *rt;
1390 register vifi_t xmt_vif;
1391{
1392 register struct ip *ip = mtod(m, struct ip *);
1393 register vifi_t vifi;
1394 register struct vif *vifp;
1395 register int plen = ntohs(ip->ip_len);
1396
1397/*
1398 * Macro to send packet on vif. Since RSVP packets don't get counted on
1399 * input, they shouldn't get counted on output, so statistics keeping is
1400 * seperate.
1401 */
1402#define MC_SEND(ip,vifp,m) { \
1403 if ((vifp)->v_flags & VIFF_TUNNEL) \
1404 encap_send((ip), (vifp), (m)); \
1405 else \
1406 phyint_send((ip), (vifp), (m)); \
1407}
1408
1409 /*
1410 * If xmt_vif is not -1, send on only the requested vif.
1411 *
1412 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1413 */
1414 if (xmt_vif < numvifs) {
1415 MC_SEND(ip, viftable + xmt_vif, m);
1416 return 1;
1417 }
1418
1419 /*
1420 * Don't forward if it didn't arrive from the parent vif for its origin.
1421 */
1422 vifi = rt->mfc_parent;
1423 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1424 /* came in the wrong interface */
1425 if (mrtdebug & DEBUG_FORWARD)
1426 log(LOG_DEBUG, "wrong if: ifp %x vifi %d vififp %x\n",
1427 ifp, vifi, viftable[vifi].v_ifp);
1428 ++mrtstat.mrts_wrong_if;
1429 ++rt->mfc_wrong_if;
1430 /*
1431 * If we are doing PIM assert processing, and we are forwarding
1432 * packets on this interface, and it is a broadcast medium
1433 * interface (and not a tunnel), send a message to the routing daemon.
1434 */
1435 if (pim_assert && rt->mfc_ttls[vifi] &&
1436 (ifp->if_flags & IFF_BROADCAST) &&
1437 !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
1438 struct sockaddr_in k_igmpsrc;
1439 struct mbuf *mm;
1440 struct igmpmsg *im;
1441 int hlen = ip->ip_hl << 2;
1442 struct timeval now;
1443 register u_long delta;
1444
1445 GET_TIME(now);
1446
1447 TV_DELTA(rt->mfc_last_assert, now, delta);
1448
1449 if (delta > ASSERT_MSG_TIME) {
1450 mm = m_copy(m, 0, hlen);
1451 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1452 mm = m_pullup(mm, hlen);
1453 if (mm == NULL) {
1454 return ENOBUFS;
1455 }
1456
1457 rt->mfc_last_assert = now;
1458
1459 im = mtod(mm, struct igmpmsg *);
1460 im->im_msgtype = IGMPMSG_WRONGVIF;
1461 im->im_mbz = 0;
1462 im->im_vif = vifi;
1463
1464 k_igmpsrc.sin_addr = im->im_src;
1465
1466 socket_send(ip_mrouter, mm, &k_igmpsrc);
1467 }
1468 }
1469 return 0;
1470 }
1471
1472 /* If I sourced this packet, it counts as output, else it was input. */
1473 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1474 viftable[vifi].v_pkt_out++;
1475 viftable[vifi].v_bytes_out += plen;
1476 } else {
1477 viftable[vifi].v_pkt_in++;
1478 viftable[vifi].v_bytes_in += plen;
1479 }
1480 rt->mfc_pkt_cnt++;
1481 rt->mfc_byte_cnt += plen;
1482
1483 /*
1484 * For each vif, decide if a copy of the packet should be forwarded.
1485 * Forward if:
1486 * - the ttl exceeds the vif's threshold
1487 * - there are group members downstream on interface
1488 */
1489 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1490 if ((rt->mfc_ttls[vifi] > 0) &&
1491 (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1492 vifp->v_pkt_out++;
1493 vifp->v_bytes_out += plen;
1494 MC_SEND(ip, vifp, m);
1495 }
1496
1497 return 0;
1498}
1499
1500/*
1501 * check if a vif number is legal/ok. This is used by ip_output, to export
1502 * numvifs there,
1503 */
1504static int
1505X_legal_vif_num(vif)
1506 int vif;
1507{
1508 if (vif >= 0 && vif < numvifs)
1509 return(1);
1510 else
1511 return(0);
1512}
1513
1514#ifndef MROUTE_LKM
1515int (*legal_vif_num)(int) = X_legal_vif_num;
1516#endif
1517
1518/*
1519 * Return the local address used by this vif
1520 */
1521static u_long
1522X_ip_mcast_src(vifi)
1523 int vifi;
1524{
1525 if (vifi >= 0 && vifi < numvifs)
1526 return viftable[vifi].v_lcl_addr.s_addr;
1527 else
1528 return INADDR_ANY;
1529}
1530
1531#ifndef MROUTE_LKM
1532u_long (*ip_mcast_src)(int) = X_ip_mcast_src;
1533#endif
1534
1535static void
1536phyint_send(ip, vifp, m)
1537 struct ip *ip;
1538 struct vif *vifp;
1539 struct mbuf *m;
1540{
1541 register struct mbuf *mb_copy;
1542 register int hlen = ip->ip_hl << 2;
1543
1544 /*
1545 * Make a new reference to the packet; make sure that
1546 * the IP header is actually copied, not just referenced,
1547 * so that ip_output() only scribbles on the copy.
1548 */
1549 mb_copy = m_copy(m, 0, M_COPYALL);
1550 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1551 mb_copy = m_pullup(mb_copy, hlen);
1552 if (mb_copy == NULL)
1553 return;
1554
1555 if (vifp->v_rate_limit <= 0)
1556 tbf_send_packet(vifp, mb_copy);
1557 else
1558 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1559}
1560
1561static void
1562encap_send(ip, vifp, m)
1563 register struct ip *ip;
1564 register struct vif *vifp;
1565 register struct mbuf *m;
1566{
1567 register struct mbuf *mb_copy;
1568 register struct ip *ip_copy;
1569 register int i, len = ip->ip_len;
1570
1571 /*
1572 * copy the old packet & pullup it's IP header into the
1573 * new mbuf so we can modify it. Try to fill the new
1574 * mbuf since if we don't the ethernet driver will.
1575 */
1576 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER);
1577 if (mb_copy == NULL)
1578 return;
1579 mb_copy->m_data += max_linkhdr;
1580 mb_copy->m_len = sizeof(multicast_encap_iphdr);
1581
1582 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1583 m_freem(mb_copy);
1584 return;
1585 }
1586 i = MHLEN - M_LEADINGSPACE(mb_copy);
1587 if (i > len)
1588 i = len;
1589 mb_copy = m_pullup(mb_copy, i);
1590 if (mb_copy == NULL)
1591 return;
1592 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1593
1594 /*
1595 * fill in the encapsulating IP header.
1596 */
1597 ip_copy = mtod(mb_copy, struct ip *);
1598 *ip_copy = multicast_encap_iphdr;
1599 ip_copy->ip_id = htons(ip_id++);
1600 ip_copy->ip_len += len;
1601 ip_copy->ip_src = vifp->v_lcl_addr;
1602 ip_copy->ip_dst = vifp->v_rmt_addr;
1603
1604 /*
1605 * turn the encapsulated IP header back into a valid one.
1606 */
1607 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1608 --ip->ip_ttl;
1609 HTONS(ip->ip_len);
1610 HTONS(ip->ip_off);
1611 ip->ip_sum = 0;
1612 mb_copy->m_data += sizeof(multicast_encap_iphdr);
1613 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1614 mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1615
1616 if (vifp->v_rate_limit <= 0)
1617 tbf_send_packet(vifp, mb_copy);
1618 else
1619 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1620}
1621
1622/*
1623 * De-encapsulate a packet and feed it back through ip input (this
1624 * routine is called whenever IP gets a packet with proto type
1625 * ENCAP_PROTO and a local destination address).
1626 */
1627void
1628#ifdef MROUTE_LKM
1629X_ipip_input(m, iphlen)
1630#else
1631ipip_input(m, iphlen)
1632#endif
1633 register struct mbuf *m;
1634 int iphlen;
1635{
1636 struct ifnet *ifp = m->m_pkthdr.rcvif;
1637 register struct ip *ip = mtod(m, struct ip *);
1638 register int hlen = ip->ip_hl << 2;
1639 register int s;
1640 register struct ifqueue *ifq;
1641 register struct vif *vifp;
1642
1643 if (!have_encap_tunnel) {
1644 rip_input(m, iphlen);
1645 return;
1646 }
1647 /*
1648 * dump the packet if it's not to a multicast destination or if
1649 * we don't have an encapsulating tunnel with the source.
1650 * Note: This code assumes that the remote site IP address
1651 * uniquely identifies the tunnel (i.e., that this site has
1652 * at most one tunnel with the remote site).
1653 */
1654 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) {
1655 ++mrtstat.mrts_bad_tunnel;
1656 m_freem(m);
1657 return;
1658 }
1659 if (ip->ip_src.s_addr != last_encap_src) {
1660 register struct vif *vife;
1661
1662 vifp = viftable;
1663 vife = vifp + numvifs;
1664 last_encap_src = ip->ip_src.s_addr;
1665 last_encap_vif = 0;
1666 for ( ; vifp < vife; ++vifp)
1667 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
1668 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT))
1669 == VIFF_TUNNEL)
1670 last_encap_vif = vifp;
1671 break;
1672 }
1673 }
1674 if ((vifp = last_encap_vif) == 0) {
1675 last_encap_src = 0;
1676 mrtstat.mrts_cant_tunnel++; /*XXX*/
1677 m_freem(m);
1678 if (mrtdebug)
1679 log(LOG_DEBUG, "ip_mforward: no tunnel with %x\n",
1680 ntohl(ip->ip_src.s_addr));
1681 return;
1682 }
1683 ifp = vifp->v_ifp;
1684
1685 if (hlen > IP_HDR_LEN)
1686 ip_stripoptions(m, (struct mbuf *) 0);
1687 m->m_data += IP_HDR_LEN;
1688 m->m_len -= IP_HDR_LEN;
1689 m->m_pkthdr.len -= IP_HDR_LEN;
1690 m->m_pkthdr.rcvif = ifp;
1691
1692 ifq = &ipintrq;
1693 s = splimp();
1694 if (IF_QFULL(ifq)) {
1695 IF_DROP(ifq);
1696 m_freem(m);
1697 } else {
1698 IF_ENQUEUE(ifq, m);
1699 /*
1700 * normally we would need a "schednetisr(NETISR_IP)"
1701 * here but we were called by ip_input and it is going
1702 * to loop back & try to dequeue the packet we just
1703 * queued as soon as we return so we avoid the
1704 * unnecessary software interrrupt.
1705 */
1706 }
1707 splx(s);
1708}
1709
1710/*
1711 * Token bucket filter module
1712 */
1713
1714static void
1715tbf_control(vifp, m, ip, p_len)
1716 register struct vif *vifp;
1717 register struct mbuf *m;
1718 register struct ip *ip;
1719 register u_long p_len;
1720{
1721 register struct tbf *t = vifp->v_tbf;
1722
1723 if (p_len > MAX_BKT_SIZE) {
1724 /* drop if packet is too large */
1725 mrtstat.mrts_pkt2large++;
1726 m_freem(m);
1727 return;
1728 }
1729
1730 tbf_update_tokens(vifp);
1731
1732 /* if there are enough tokens,
1733 * and the queue is empty,
1734 * send this packet out
1735 */
1736
1737 if (t->tbf_q_len == 0) {
1738 /* queue empty, send packet if enough tokens */
1739 if (p_len <= t->tbf_n_tok) {
1740 t->tbf_n_tok -= p_len;
1741 tbf_send_packet(vifp, m);
1742 } else {
1743 /* queue packet and timeout till later */
1744 tbf_queue(vifp, m);
1745 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1746 }
1747 } else if (t->tbf_q_len < t->tbf_max_q_len) {
1748 /* finite queue length, so queue pkts and process queue */
1749 tbf_queue(vifp, m);
1750 tbf_process_q(vifp);
1751 } else {
1752 /* queue length too much, try to dq and queue and process */
1753 if (!tbf_dq_sel(vifp, ip)) {
1754 mrtstat.mrts_q_overflow++;
1755 m_freem(m);
1756 return;
1757 } else {
1758 tbf_queue(vifp, m);
1759 tbf_process_q(vifp);
1760 }
1761 }
1762 return;
1763}
1764
1765/*
1766 * adds a packet to the queue at the interface
1767 */
1768static void
1769tbf_queue(vifp, m)
1770 register struct vif *vifp;
1771 register struct mbuf *m;
1772{
1773 register int s = splnet();
1774 register struct tbf *t = vifp->v_tbf;
1775
1776 if (t->tbf_t == NULL) {
1777 /* Queue was empty */
1778 t->tbf_q = m;
1779 } else {
1780 /* Insert at tail */
1781 t->tbf_t->m_act = m;
1782 }
1783
1784 /* Set new tail pointer */
1785 t->tbf_t = m;
1786
1787#ifdef DIAGNOSTIC
1788 /* Make sure we didn't get fed a bogus mbuf */
1789 if (m->m_act)
1790 panic("tbf_queue: m_act");
1791#endif
1792 m->m_act = NULL;
1793
1794 t->tbf_q_len++;
1795
1796 splx(s);
1797}
1798
1799
1800/*
1801 * processes the queue at the interface
1802 */
1803static void
1804tbf_process_q(vifp)
1805 register struct vif *vifp;
1806{
1807 register struct mbuf *m;
1808 register int len;
1809 register int s = splnet();
1810 register struct tbf *t = vifp->v_tbf;
1811
1812 /* loop through the queue at the interface and send as many packets
1813 * as possible
1814 */
1815 while (t->tbf_q_len > 0) {
1816 m = t->tbf_q;
1817
1818 len = mtod(m, struct ip *)->ip_len;
1819
1820 /* determine if the packet can be sent */
1821 if (len <= t->tbf_n_tok) {
1822 /* if so,
1823 * reduce no of tokens, dequeue the packet,
1824 * send the packet.
1825 */
1826 t->tbf_n_tok -= len;
1827
1828 t->tbf_q = m->m_act;
1829 if (--t->tbf_q_len == 0)
1830 t->tbf_t = NULL;
1831
1832 m->m_act = NULL;
1833 tbf_send_packet(vifp, m);
1834
1835 } else break;
1836 }
1837 splx(s);
1838}
1839
1840static void
1841tbf_reprocess_q(xvifp)
1842 void *xvifp;
1843{
1844 register struct vif *vifp = xvifp;
1845 if (ip_mrouter == NULL)
1846 return;
1847
1848 tbf_update_tokens(vifp);
1849
1850 tbf_process_q(vifp);
1851
1852 if (vifp->v_tbf->tbf_q_len)
1853 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1854}
1855
1856/* function that will selectively discard a member of the queue
1857 * based on the precedence value and the priority
1858 */
1859static int
1860tbf_dq_sel(vifp, ip)
1861 register struct vif *vifp;
1862 register struct ip *ip;
1863{
1864 register int s = splnet();
1865 register u_int p;
1866 register struct mbuf *m, *last;
1867 register struct mbuf **np;
1868 register struct tbf *t = vifp->v_tbf;
1869
1870 p = priority(vifp, ip);
1871
1872 np = &t->tbf_q;
1873 last = NULL;
1874 while ((m = *np) != NULL) {
1875 if (p > priority(vifp, mtod(m, struct ip *))) {
1876 *np = m->m_act;
1877 /* If we're removing the last packet, fix the tail pointer */
1878 if (m == t->tbf_t)
1879 t->tbf_t = last;
1880 m_freem(m);
1881 /* it's impossible for the queue to be empty, but
1882 * we check anyway. */
1883 if (--t->tbf_q_len == 0)
1884 t->tbf_t = NULL;
1885 splx(s);
1886 mrtstat.mrts_drop_sel++;
1887 return(1);
1888 }
1889 np = &m->m_act;
1890 last = m;
1891 }
1892 splx(s);
1893 return(0);
1894}
1895
1896static void
1897tbf_send_packet(vifp, m)
1898 register struct vif *vifp;
1899 register struct mbuf *m;
1900{
1901 struct ip_moptions imo;
1902 int error;
1903 static struct route ro;
1904 int s = splnet();
1905
1906 if (vifp->v_flags & VIFF_TUNNEL) {
1907 /* If tunnel options */
1908 ip_output(m, (struct mbuf *)0, &vifp->v_route,
1909 IP_FORWARDING, (struct ip_moptions *)0);
1910 } else {
1911 imo.imo_multicast_ifp = vifp->v_ifp;
1912 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1913 imo.imo_multicast_loop = 1;
1914 imo.imo_multicast_vif = -1;
1915
1916 /*
1917 * Re-entrancy should not be a problem here, because
1918 * the packets that we send out and are looped back at us
1919 * should get rejected because they appear to come from
1920 * the loopback interface, thus preventing looping.
1921 */
1922 error = ip_output(m, (struct mbuf *)0, &ro,
1923 IP_FORWARDING, &imo);
1924
1925 if (mrtdebug & DEBUG_XMIT)
1926 log(LOG_DEBUG, "phyint_send on vif %d err %d\n",
1927 vifp - viftable, error);
1928 }
1929 splx(s);
1930}
1931
1932/* determine the current time and then
1933 * the elapsed time (between the last time and time now)
1934 * in milliseconds & update the no. of tokens in the bucket
1935 */
1936static void
1937tbf_update_tokens(vifp)
1938 register struct vif *vifp;
1939{
1940 struct timeval tp;
1941 register u_long tm;
1942 register int s = splnet();
1943 register struct tbf *t = vifp->v_tbf;
1944
1945 GET_TIME(tp);
1946
1947 TV_DELTA(tp, t->tbf_last_pkt_t, tm);
1948
1949 /*
1950 * This formula is actually
1951 * "time in seconds" * "bytes/second".
1952 *
1953 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
1954 *
1955 * The (1000/1024) was introduced in add_vif to optimize
1956 * this divide into a shift.
1957 */
1958 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8;
1959 t->tbf_last_pkt_t = tp;
1960
1961 if (t->tbf_n_tok > MAX_BKT_SIZE)
1962 t->tbf_n_tok = MAX_BKT_SIZE;
1963
1964 splx(s);
1965}
1966
1967static int
1968priority(vifp, ip)
1969 register struct vif *vifp;
1970 register struct ip *ip;
1971{
1972 register int prio;
1973
1974 /* temporary hack; may add general packet classifier some day */
1975
1976 /*
1977 * The UDP port space is divided up into four priority ranges:
1978 * [0, 16384) : unclassified - lowest priority
1979 * [16384, 32768) : audio - highest priority
1980 * [32768, 49152) : whiteboard - medium priority
1981 * [49152, 65536) : video - low priority
1982 */
1983 if (ip->ip_p == IPPROTO_UDP) {
1984 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
1985 switch (ntohs(udp->uh_dport) & 0xc000) {
1986 case 0x4000:
1987 prio = 70;
1988 break;
1989 case 0x8000:
1990 prio = 60;
1991 break;
1992 case 0xc000:
1993 prio = 55;
1994 break;
1995 default:
1996 prio = 50;
1997 break;
1998 }
1999 if (tbfdebug > 1)
2000 log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio);
2001 } else {
2002 prio = 50;
2003 }
2004 return prio;
2005}
2006
2007/*
2008 * End of token bucket filter modifications
2009 */
2010
2011int
2012ip_rsvp_vif_init(so, m)
2013 struct socket *so;
2014 struct mbuf *m;
2015{
2016 int i;
2017 register int s;
2018
2019 if (rsvpdebug)
2020 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2021 so->so_type, so->so_proto->pr_protocol);
2022
2023 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2024 return EOPNOTSUPP;
2025
2026 /* Check mbuf. */
2027 if (m == NULL || m->m_len != sizeof(int)) {
2028 return EINVAL;
2029 }
2030 i = *(mtod(m, int *));
2031
2032 if (rsvpdebug)
2033 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on);
2034
2035 s = splnet();
2036
2037 /* Check vif. */
2038 if (!legal_vif_num(i)) {
2039 splx(s);
2040 return EADDRNOTAVAIL;
2041 }
2042
2043 /* Check if socket is available. */
2044 if (viftable[i].v_rsvpd != NULL) {
2045 splx(s);
2046 return EADDRINUSE;
2047 }
2048
2049 viftable[i].v_rsvpd = so;
2050 /* This may seem silly, but we need to be sure we don't over-increment
2051 * the RSVP counter, in case something slips up.
2052 */
2053 if (!viftable[i].v_rsvp_on) {
2054 viftable[i].v_rsvp_on = 1;
2055 rsvp_on++;
2056 }
2057
2058 splx(s);
2059 return 0;
2060}
2061
2062int
2063ip_rsvp_vif_done(so, m)
2064 struct socket *so;
2065 struct mbuf *m;
2066{
2067 int i;
2068 register int s;
2069
2070 if (rsvpdebug)
2071 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2072 so->so_type, so->so_proto->pr_protocol);
2073
2074 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2075 return EOPNOTSUPP;
2076
2077 /* Check mbuf. */
2078 if (m == NULL || m->m_len != sizeof(int)) {
2079 return EINVAL;
2080 }
2081 i = *(mtod(m, int *));
2082
2083 s = splnet();
2084
2085 /* Check vif. */
2086 if (!legal_vif_num(i)) {
2087 splx(s);
2088 return EADDRNOTAVAIL;
2089 }
2090
2091 if (rsvpdebug)
2092 printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n",
2093 viftable[i].v_rsvpd, so);
2094
2095 viftable[i].v_rsvpd = NULL;
2096 /* This may seem silly, but we need to be sure we don't over-decrement
2097 * the RSVP counter, in case something slips up.
2098 */
2099 if (viftable[i].v_rsvp_on) {
2100 viftable[i].v_rsvp_on = 0;
2101 rsvp_on--;
2102 }
2103
2104 splx(s);
2105 return 0;
2106}
2107
2108void
2109ip_rsvp_force_done(so)
2110 struct socket *so;
2111{
2112 int vifi;
2113 register int s;
2114
2115 /* Don't bother if it is not the right type of socket. */
2116 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2117 return;
2118
2119 s = splnet();
2120
2121 /* The socket may be attached to more than one vif...this
2122 * is perfectly legal.
2123 */
2124 for (vifi = 0; vifi < numvifs; vifi++) {
2125 if (viftable[vifi].v_rsvpd == so) {
2126 viftable[vifi].v_rsvpd = NULL;
2127 /* This may seem silly, but we need to be sure we don't
2128 * over-decrement the RSVP counter, in case something slips up.
2129 */
2130 if (viftable[vifi].v_rsvp_on) {
2131 viftable[vifi].v_rsvp_on = 0;
2132 rsvp_on--;
2133 }
2134 }
2135 }
2136
2137 splx(s);
2138 return;
2139}
2140
2141void
2142rsvp_input(m, iphlen)
2143 struct mbuf *m;
2144 int iphlen;
2145{
2146 int vifi;
2147 register struct ip *ip = mtod(m, struct ip *);
2148 static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET };
2149 register int s;
2150 struct ifnet *ifp;
2151
2152 if (rsvpdebug)
2153 printf("rsvp_input: rsvp_on %d\n",rsvp_on);
2154
2155 /* Can still get packets with rsvp_on = 0 if there is a local member
2156 * of the group to which the RSVP packet is addressed. But in this
2157 * case we want to throw the packet away.
2158 */
2159 if (!rsvp_on) {
2160 m_freem(m);
2161 return;
2162 }
2163
2164 /* If the old-style non-vif-associated socket is set, then use
2165 * it and ignore the new ones.
2166 */
2167 if (ip_rsvpd != NULL) {
2168 if (rsvpdebug)
2169 printf("rsvp_input: Sending packet up old-style socket\n");
2170 rip_input(m, iphlen);
2171 return;
2172 }
2173
2174 s = splnet();
2175
2176 if (rsvpdebug)
2177 printf("rsvp_input: check vifs\n");
2178
2179#ifdef DIAGNOSTIC
2180 if (!(m->m_flags & M_PKTHDR))
2181 panic("rsvp_input no hdr");
2182#endif
2183
2184 ifp = m->m_pkthdr.rcvif;
2185 /* Find which vif the packet arrived on. */
2186 for (vifi = 0; vifi < numvifs; vifi++) {
2187 if (viftable[vifi].v_ifp == ifp)
2188 break;
2189 }
2190
2191 if (vifi == numvifs) {
2192 /* Can't find vif packet arrived on. Drop packet. */
2193 if (rsvpdebug)
2194 printf("rsvp_input: Can't find vif for packet...dropping it.\n");
2195 m_freem(m);
2196 splx(s);
2197 return;
2198 }
2199
2200 if (rsvpdebug)
2201 printf("rsvp_input: check socket\n");
2202
2203 if (viftable[vifi].v_rsvpd == NULL) {
2204 /* drop packet, since there is no specific socket for this
2205 * interface */
2206 if (rsvpdebug)
2207 printf("rsvp_input: No socket defined for vif %d\n",vifi);
2208 m_freem(m);
2209 splx(s);
2210 return;
2211 }
2212 rsvp_src.sin_addr = ip->ip_src;
2213
2214 if (rsvpdebug && m)
2215 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n",
2216 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv)));
2217
2218 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2219 if (rsvpdebug)
2220 printf("rsvp_input: Failed to append to socket\n");
2221 else
2222 if (rsvpdebug)
2223 printf("rsvp_input: send packet up\n");
2224
2225 splx(s);
2226}
2227
2228#ifdef MROUTE_LKM
2229#include <sys/conf.h>
2230#include <sys/exec.h>
2231#include <sys/sysent.h>
2232#include <sys/lkm.h>
2233
2234MOD_MISC("ip_mroute_mod")
2235
2236static int
2237ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd)
2238{
2239 int i;
2240 struct lkm_misc *args = lkmtp->private.lkm_misc;
2241 int err = 0;
2242
2243 switch(cmd) {
2244 static int (*old_ip_mrouter_cmd)();
2245 static int (*old_ip_mrouter_done)();
2246 static int (*old_ip_mforward)();
2247 static int (*old_mrt_ioctl)();
2248 static void (*old_proto4_input)();
2249 static int (*old_legal_vif_num)();
2250 extern struct protosw inetsw[];
2251
2252 case LKM_E_LOAD:
2253 if(lkmexists(lkmtp) || ip_mrtproto)
2254 return(EEXIST);
2255 old_ip_mrouter_cmd = ip_mrouter_cmd;
2256 ip_mrouter_cmd = X_ip_mrouter_cmd;
2257 old_ip_mrouter_done = ip_mrouter_done;
2258 ip_mrouter_done = X_ip_mrouter_done;
2259 old_ip_mforward = ip_mforward;
2260 ip_mforward = X_ip_mforward;
2261 old_mrt_ioctl = mrt_ioctl;
2262 mrt_ioctl = X_mrt_ioctl;
2263 old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input;
2264 inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input;
2265 old_legal_vif_num = legal_vif_num;
2266 legal_vif_num = X_legal_vif_num;
2267 ip_mrtproto = IGMP_DVMRP;
2268
2269 printf("\nIP multicast routing loaded\n");
2270 break;
2271
2272 case LKM_E_UNLOAD:
2273 if (ip_mrouter)
2274 return EINVAL;
2275
2276 ip_mrouter_cmd = old_ip_mrouter_cmd;
2277 ip_mrouter_done = old_ip_mrouter_done;
2278 ip_mforward = old_ip_mforward;
2279 mrt_ioctl = old_mrt_ioctl;
2280 inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input;
2281 legal_vif_num = old_legal_vif_num;
2282 ip_mrtproto = 0;
2283 break;
2284
2285 default:
2286 err = EINVAL;
2287 break;
2288 }
2289
2290 return(err);
2291}
2292
2293int
2294ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) {
2295 DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle,
2296 nosys);
2297}
2298
2299#endif /* MROUTE_LKM */
2300#endif /* MROUTING */
| 1382}
1383
1384/*
1385 * Packet forwarding routine once entry in the cache is made
1386 */
1387static int
1388ip_mdq(m, ifp, rt, xmt_vif)
1389 register struct mbuf *m;
1390 register struct ifnet *ifp;
1391 register struct mfc *rt;
1392 register vifi_t xmt_vif;
1393{
1394 register struct ip *ip = mtod(m, struct ip *);
1395 register vifi_t vifi;
1396 register struct vif *vifp;
1397 register int plen = ntohs(ip->ip_len);
1398
1399/*
1400 * Macro to send packet on vif. Since RSVP packets don't get counted on
1401 * input, they shouldn't get counted on output, so statistics keeping is
1402 * seperate.
1403 */
1404#define MC_SEND(ip,vifp,m) { \
1405 if ((vifp)->v_flags & VIFF_TUNNEL) \
1406 encap_send((ip), (vifp), (m)); \
1407 else \
1408 phyint_send((ip), (vifp), (m)); \
1409}
1410
1411 /*
1412 * If xmt_vif is not -1, send on only the requested vif.
1413 *
1414 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1415 */
1416 if (xmt_vif < numvifs) {
1417 MC_SEND(ip, viftable + xmt_vif, m);
1418 return 1;
1419 }
1420
1421 /*
1422 * Don't forward if it didn't arrive from the parent vif for its origin.
1423 */
1424 vifi = rt->mfc_parent;
1425 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1426 /* came in the wrong interface */
1427 if (mrtdebug & DEBUG_FORWARD)
1428 log(LOG_DEBUG, "wrong if: ifp %x vifi %d vififp %x\n",
1429 ifp, vifi, viftable[vifi].v_ifp);
1430 ++mrtstat.mrts_wrong_if;
1431 ++rt->mfc_wrong_if;
1432 /*
1433 * If we are doing PIM assert processing, and we are forwarding
1434 * packets on this interface, and it is a broadcast medium
1435 * interface (and not a tunnel), send a message to the routing daemon.
1436 */
1437 if (pim_assert && rt->mfc_ttls[vifi] &&
1438 (ifp->if_flags & IFF_BROADCAST) &&
1439 !(viftable[vifi].v_flags & VIFF_TUNNEL)) {
1440 struct sockaddr_in k_igmpsrc;
1441 struct mbuf *mm;
1442 struct igmpmsg *im;
1443 int hlen = ip->ip_hl << 2;
1444 struct timeval now;
1445 register u_long delta;
1446
1447 GET_TIME(now);
1448
1449 TV_DELTA(rt->mfc_last_assert, now, delta);
1450
1451 if (delta > ASSERT_MSG_TIME) {
1452 mm = m_copy(m, 0, hlen);
1453 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1454 mm = m_pullup(mm, hlen);
1455 if (mm == NULL) {
1456 return ENOBUFS;
1457 }
1458
1459 rt->mfc_last_assert = now;
1460
1461 im = mtod(mm, struct igmpmsg *);
1462 im->im_msgtype = IGMPMSG_WRONGVIF;
1463 im->im_mbz = 0;
1464 im->im_vif = vifi;
1465
1466 k_igmpsrc.sin_addr = im->im_src;
1467
1468 socket_send(ip_mrouter, mm, &k_igmpsrc);
1469 }
1470 }
1471 return 0;
1472 }
1473
1474 /* If I sourced this packet, it counts as output, else it was input. */
1475 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1476 viftable[vifi].v_pkt_out++;
1477 viftable[vifi].v_bytes_out += plen;
1478 } else {
1479 viftable[vifi].v_pkt_in++;
1480 viftable[vifi].v_bytes_in += plen;
1481 }
1482 rt->mfc_pkt_cnt++;
1483 rt->mfc_byte_cnt += plen;
1484
1485 /*
1486 * For each vif, decide if a copy of the packet should be forwarded.
1487 * Forward if:
1488 * - the ttl exceeds the vif's threshold
1489 * - there are group members downstream on interface
1490 */
1491 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++)
1492 if ((rt->mfc_ttls[vifi] > 0) &&
1493 (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1494 vifp->v_pkt_out++;
1495 vifp->v_bytes_out += plen;
1496 MC_SEND(ip, vifp, m);
1497 }
1498
1499 return 0;
1500}
1501
1502/*
1503 * check if a vif number is legal/ok. This is used by ip_output, to export
1504 * numvifs there,
1505 */
1506static int
1507X_legal_vif_num(vif)
1508 int vif;
1509{
1510 if (vif >= 0 && vif < numvifs)
1511 return(1);
1512 else
1513 return(0);
1514}
1515
1516#ifndef MROUTE_LKM
1517int (*legal_vif_num)(int) = X_legal_vif_num;
1518#endif
1519
1520/*
1521 * Return the local address used by this vif
1522 */
1523static u_long
1524X_ip_mcast_src(vifi)
1525 int vifi;
1526{
1527 if (vifi >= 0 && vifi < numvifs)
1528 return viftable[vifi].v_lcl_addr.s_addr;
1529 else
1530 return INADDR_ANY;
1531}
1532
1533#ifndef MROUTE_LKM
1534u_long (*ip_mcast_src)(int) = X_ip_mcast_src;
1535#endif
1536
1537static void
1538phyint_send(ip, vifp, m)
1539 struct ip *ip;
1540 struct vif *vifp;
1541 struct mbuf *m;
1542{
1543 register struct mbuf *mb_copy;
1544 register int hlen = ip->ip_hl << 2;
1545
1546 /*
1547 * Make a new reference to the packet; make sure that
1548 * the IP header is actually copied, not just referenced,
1549 * so that ip_output() only scribbles on the copy.
1550 */
1551 mb_copy = m_copy(m, 0, M_COPYALL);
1552 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1553 mb_copy = m_pullup(mb_copy, hlen);
1554 if (mb_copy == NULL)
1555 return;
1556
1557 if (vifp->v_rate_limit <= 0)
1558 tbf_send_packet(vifp, mb_copy);
1559 else
1560 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1561}
1562
1563static void
1564encap_send(ip, vifp, m)
1565 register struct ip *ip;
1566 register struct vif *vifp;
1567 register struct mbuf *m;
1568{
1569 register struct mbuf *mb_copy;
1570 register struct ip *ip_copy;
1571 register int i, len = ip->ip_len;
1572
1573 /*
1574 * copy the old packet & pullup it's IP header into the
1575 * new mbuf so we can modify it. Try to fill the new
1576 * mbuf since if we don't the ethernet driver will.
1577 */
1578 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER);
1579 if (mb_copy == NULL)
1580 return;
1581 mb_copy->m_data += max_linkhdr;
1582 mb_copy->m_len = sizeof(multicast_encap_iphdr);
1583
1584 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1585 m_freem(mb_copy);
1586 return;
1587 }
1588 i = MHLEN - M_LEADINGSPACE(mb_copy);
1589 if (i > len)
1590 i = len;
1591 mb_copy = m_pullup(mb_copy, i);
1592 if (mb_copy == NULL)
1593 return;
1594 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1595
1596 /*
1597 * fill in the encapsulating IP header.
1598 */
1599 ip_copy = mtod(mb_copy, struct ip *);
1600 *ip_copy = multicast_encap_iphdr;
1601 ip_copy->ip_id = htons(ip_id++);
1602 ip_copy->ip_len += len;
1603 ip_copy->ip_src = vifp->v_lcl_addr;
1604 ip_copy->ip_dst = vifp->v_rmt_addr;
1605
1606 /*
1607 * turn the encapsulated IP header back into a valid one.
1608 */
1609 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1610 --ip->ip_ttl;
1611 HTONS(ip->ip_len);
1612 HTONS(ip->ip_off);
1613 ip->ip_sum = 0;
1614 mb_copy->m_data += sizeof(multicast_encap_iphdr);
1615 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1616 mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1617
1618 if (vifp->v_rate_limit <= 0)
1619 tbf_send_packet(vifp, mb_copy);
1620 else
1621 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1622}
1623
1624/*
1625 * De-encapsulate a packet and feed it back through ip input (this
1626 * routine is called whenever IP gets a packet with proto type
1627 * ENCAP_PROTO and a local destination address).
1628 */
1629void
1630#ifdef MROUTE_LKM
1631X_ipip_input(m, iphlen)
1632#else
1633ipip_input(m, iphlen)
1634#endif
1635 register struct mbuf *m;
1636 int iphlen;
1637{
1638 struct ifnet *ifp = m->m_pkthdr.rcvif;
1639 register struct ip *ip = mtod(m, struct ip *);
1640 register int hlen = ip->ip_hl << 2;
1641 register int s;
1642 register struct ifqueue *ifq;
1643 register struct vif *vifp;
1644
1645 if (!have_encap_tunnel) {
1646 rip_input(m, iphlen);
1647 return;
1648 }
1649 /*
1650 * dump the packet if it's not to a multicast destination or if
1651 * we don't have an encapsulating tunnel with the source.
1652 * Note: This code assumes that the remote site IP address
1653 * uniquely identifies the tunnel (i.e., that this site has
1654 * at most one tunnel with the remote site).
1655 */
1656 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) {
1657 ++mrtstat.mrts_bad_tunnel;
1658 m_freem(m);
1659 return;
1660 }
1661 if (ip->ip_src.s_addr != last_encap_src) {
1662 register struct vif *vife;
1663
1664 vifp = viftable;
1665 vife = vifp + numvifs;
1666 last_encap_src = ip->ip_src.s_addr;
1667 last_encap_vif = 0;
1668 for ( ; vifp < vife; ++vifp)
1669 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
1670 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT))
1671 == VIFF_TUNNEL)
1672 last_encap_vif = vifp;
1673 break;
1674 }
1675 }
1676 if ((vifp = last_encap_vif) == 0) {
1677 last_encap_src = 0;
1678 mrtstat.mrts_cant_tunnel++; /*XXX*/
1679 m_freem(m);
1680 if (mrtdebug)
1681 log(LOG_DEBUG, "ip_mforward: no tunnel with %x\n",
1682 ntohl(ip->ip_src.s_addr));
1683 return;
1684 }
1685 ifp = vifp->v_ifp;
1686
1687 if (hlen > IP_HDR_LEN)
1688 ip_stripoptions(m, (struct mbuf *) 0);
1689 m->m_data += IP_HDR_LEN;
1690 m->m_len -= IP_HDR_LEN;
1691 m->m_pkthdr.len -= IP_HDR_LEN;
1692 m->m_pkthdr.rcvif = ifp;
1693
1694 ifq = &ipintrq;
1695 s = splimp();
1696 if (IF_QFULL(ifq)) {
1697 IF_DROP(ifq);
1698 m_freem(m);
1699 } else {
1700 IF_ENQUEUE(ifq, m);
1701 /*
1702 * normally we would need a "schednetisr(NETISR_IP)"
1703 * here but we were called by ip_input and it is going
1704 * to loop back & try to dequeue the packet we just
1705 * queued as soon as we return so we avoid the
1706 * unnecessary software interrrupt.
1707 */
1708 }
1709 splx(s);
1710}
1711
1712/*
1713 * Token bucket filter module
1714 */
1715
1716static void
1717tbf_control(vifp, m, ip, p_len)
1718 register struct vif *vifp;
1719 register struct mbuf *m;
1720 register struct ip *ip;
1721 register u_long p_len;
1722{
1723 register struct tbf *t = vifp->v_tbf;
1724
1725 if (p_len > MAX_BKT_SIZE) {
1726 /* drop if packet is too large */
1727 mrtstat.mrts_pkt2large++;
1728 m_freem(m);
1729 return;
1730 }
1731
1732 tbf_update_tokens(vifp);
1733
1734 /* if there are enough tokens,
1735 * and the queue is empty,
1736 * send this packet out
1737 */
1738
1739 if (t->tbf_q_len == 0) {
1740 /* queue empty, send packet if enough tokens */
1741 if (p_len <= t->tbf_n_tok) {
1742 t->tbf_n_tok -= p_len;
1743 tbf_send_packet(vifp, m);
1744 } else {
1745 /* queue packet and timeout till later */
1746 tbf_queue(vifp, m);
1747 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1748 }
1749 } else if (t->tbf_q_len < t->tbf_max_q_len) {
1750 /* finite queue length, so queue pkts and process queue */
1751 tbf_queue(vifp, m);
1752 tbf_process_q(vifp);
1753 } else {
1754 /* queue length too much, try to dq and queue and process */
1755 if (!tbf_dq_sel(vifp, ip)) {
1756 mrtstat.mrts_q_overflow++;
1757 m_freem(m);
1758 return;
1759 } else {
1760 tbf_queue(vifp, m);
1761 tbf_process_q(vifp);
1762 }
1763 }
1764 return;
1765}
1766
1767/*
1768 * adds a packet to the queue at the interface
1769 */
1770static void
1771tbf_queue(vifp, m)
1772 register struct vif *vifp;
1773 register struct mbuf *m;
1774{
1775 register int s = splnet();
1776 register struct tbf *t = vifp->v_tbf;
1777
1778 if (t->tbf_t == NULL) {
1779 /* Queue was empty */
1780 t->tbf_q = m;
1781 } else {
1782 /* Insert at tail */
1783 t->tbf_t->m_act = m;
1784 }
1785
1786 /* Set new tail pointer */
1787 t->tbf_t = m;
1788
1789#ifdef DIAGNOSTIC
1790 /* Make sure we didn't get fed a bogus mbuf */
1791 if (m->m_act)
1792 panic("tbf_queue: m_act");
1793#endif
1794 m->m_act = NULL;
1795
1796 t->tbf_q_len++;
1797
1798 splx(s);
1799}
1800
1801
1802/*
1803 * processes the queue at the interface
1804 */
1805static void
1806tbf_process_q(vifp)
1807 register struct vif *vifp;
1808{
1809 register struct mbuf *m;
1810 register int len;
1811 register int s = splnet();
1812 register struct tbf *t = vifp->v_tbf;
1813
1814 /* loop through the queue at the interface and send as many packets
1815 * as possible
1816 */
1817 while (t->tbf_q_len > 0) {
1818 m = t->tbf_q;
1819
1820 len = mtod(m, struct ip *)->ip_len;
1821
1822 /* determine if the packet can be sent */
1823 if (len <= t->tbf_n_tok) {
1824 /* if so,
1825 * reduce no of tokens, dequeue the packet,
1826 * send the packet.
1827 */
1828 t->tbf_n_tok -= len;
1829
1830 t->tbf_q = m->m_act;
1831 if (--t->tbf_q_len == 0)
1832 t->tbf_t = NULL;
1833
1834 m->m_act = NULL;
1835 tbf_send_packet(vifp, m);
1836
1837 } else break;
1838 }
1839 splx(s);
1840}
1841
1842static void
1843tbf_reprocess_q(xvifp)
1844 void *xvifp;
1845{
1846 register struct vif *vifp = xvifp;
1847 if (ip_mrouter == NULL)
1848 return;
1849
1850 tbf_update_tokens(vifp);
1851
1852 tbf_process_q(vifp);
1853
1854 if (vifp->v_tbf->tbf_q_len)
1855 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
1856}
1857
1858/* function that will selectively discard a member of the queue
1859 * based on the precedence value and the priority
1860 */
1861static int
1862tbf_dq_sel(vifp, ip)
1863 register struct vif *vifp;
1864 register struct ip *ip;
1865{
1866 register int s = splnet();
1867 register u_int p;
1868 register struct mbuf *m, *last;
1869 register struct mbuf **np;
1870 register struct tbf *t = vifp->v_tbf;
1871
1872 p = priority(vifp, ip);
1873
1874 np = &t->tbf_q;
1875 last = NULL;
1876 while ((m = *np) != NULL) {
1877 if (p > priority(vifp, mtod(m, struct ip *))) {
1878 *np = m->m_act;
1879 /* If we're removing the last packet, fix the tail pointer */
1880 if (m == t->tbf_t)
1881 t->tbf_t = last;
1882 m_freem(m);
1883 /* it's impossible for the queue to be empty, but
1884 * we check anyway. */
1885 if (--t->tbf_q_len == 0)
1886 t->tbf_t = NULL;
1887 splx(s);
1888 mrtstat.mrts_drop_sel++;
1889 return(1);
1890 }
1891 np = &m->m_act;
1892 last = m;
1893 }
1894 splx(s);
1895 return(0);
1896}
1897
1898static void
1899tbf_send_packet(vifp, m)
1900 register struct vif *vifp;
1901 register struct mbuf *m;
1902{
1903 struct ip_moptions imo;
1904 int error;
1905 static struct route ro;
1906 int s = splnet();
1907
1908 if (vifp->v_flags & VIFF_TUNNEL) {
1909 /* If tunnel options */
1910 ip_output(m, (struct mbuf *)0, &vifp->v_route,
1911 IP_FORWARDING, (struct ip_moptions *)0);
1912 } else {
1913 imo.imo_multicast_ifp = vifp->v_ifp;
1914 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1915 imo.imo_multicast_loop = 1;
1916 imo.imo_multicast_vif = -1;
1917
1918 /*
1919 * Re-entrancy should not be a problem here, because
1920 * the packets that we send out and are looped back at us
1921 * should get rejected because they appear to come from
1922 * the loopback interface, thus preventing looping.
1923 */
1924 error = ip_output(m, (struct mbuf *)0, &ro,
1925 IP_FORWARDING, &imo);
1926
1927 if (mrtdebug & DEBUG_XMIT)
1928 log(LOG_DEBUG, "phyint_send on vif %d err %d\n",
1929 vifp - viftable, error);
1930 }
1931 splx(s);
1932}
1933
1934/* determine the current time and then
1935 * the elapsed time (between the last time and time now)
1936 * in milliseconds & update the no. of tokens in the bucket
1937 */
1938static void
1939tbf_update_tokens(vifp)
1940 register struct vif *vifp;
1941{
1942 struct timeval tp;
1943 register u_long tm;
1944 register int s = splnet();
1945 register struct tbf *t = vifp->v_tbf;
1946
1947 GET_TIME(tp);
1948
1949 TV_DELTA(tp, t->tbf_last_pkt_t, tm);
1950
1951 /*
1952 * This formula is actually
1953 * "time in seconds" * "bytes/second".
1954 *
1955 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
1956 *
1957 * The (1000/1024) was introduced in add_vif to optimize
1958 * this divide into a shift.
1959 */
1960 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8;
1961 t->tbf_last_pkt_t = tp;
1962
1963 if (t->tbf_n_tok > MAX_BKT_SIZE)
1964 t->tbf_n_tok = MAX_BKT_SIZE;
1965
1966 splx(s);
1967}
1968
1969static int
1970priority(vifp, ip)
1971 register struct vif *vifp;
1972 register struct ip *ip;
1973{
1974 register int prio;
1975
1976 /* temporary hack; may add general packet classifier some day */
1977
1978 /*
1979 * The UDP port space is divided up into four priority ranges:
1980 * [0, 16384) : unclassified - lowest priority
1981 * [16384, 32768) : audio - highest priority
1982 * [32768, 49152) : whiteboard - medium priority
1983 * [49152, 65536) : video - low priority
1984 */
1985 if (ip->ip_p == IPPROTO_UDP) {
1986 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
1987 switch (ntohs(udp->uh_dport) & 0xc000) {
1988 case 0x4000:
1989 prio = 70;
1990 break;
1991 case 0x8000:
1992 prio = 60;
1993 break;
1994 case 0xc000:
1995 prio = 55;
1996 break;
1997 default:
1998 prio = 50;
1999 break;
2000 }
2001 if (tbfdebug > 1)
2002 log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio);
2003 } else {
2004 prio = 50;
2005 }
2006 return prio;
2007}
2008
2009/*
2010 * End of token bucket filter modifications
2011 */
2012
2013int
2014ip_rsvp_vif_init(so, m)
2015 struct socket *so;
2016 struct mbuf *m;
2017{
2018 int i;
2019 register int s;
2020
2021 if (rsvpdebug)
2022 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n",
2023 so->so_type, so->so_proto->pr_protocol);
2024
2025 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2026 return EOPNOTSUPP;
2027
2028 /* Check mbuf. */
2029 if (m == NULL || m->m_len != sizeof(int)) {
2030 return EINVAL;
2031 }
2032 i = *(mtod(m, int *));
2033
2034 if (rsvpdebug)
2035 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on);
2036
2037 s = splnet();
2038
2039 /* Check vif. */
2040 if (!legal_vif_num(i)) {
2041 splx(s);
2042 return EADDRNOTAVAIL;
2043 }
2044
2045 /* Check if socket is available. */
2046 if (viftable[i].v_rsvpd != NULL) {
2047 splx(s);
2048 return EADDRINUSE;
2049 }
2050
2051 viftable[i].v_rsvpd = so;
2052 /* This may seem silly, but we need to be sure we don't over-increment
2053 * the RSVP counter, in case something slips up.
2054 */
2055 if (!viftable[i].v_rsvp_on) {
2056 viftable[i].v_rsvp_on = 1;
2057 rsvp_on++;
2058 }
2059
2060 splx(s);
2061 return 0;
2062}
2063
2064int
2065ip_rsvp_vif_done(so, m)
2066 struct socket *so;
2067 struct mbuf *m;
2068{
2069 int i;
2070 register int s;
2071
2072 if (rsvpdebug)
2073 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n",
2074 so->so_type, so->so_proto->pr_protocol);
2075
2076 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2077 return EOPNOTSUPP;
2078
2079 /* Check mbuf. */
2080 if (m == NULL || m->m_len != sizeof(int)) {
2081 return EINVAL;
2082 }
2083 i = *(mtod(m, int *));
2084
2085 s = splnet();
2086
2087 /* Check vif. */
2088 if (!legal_vif_num(i)) {
2089 splx(s);
2090 return EADDRNOTAVAIL;
2091 }
2092
2093 if (rsvpdebug)
2094 printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n",
2095 viftable[i].v_rsvpd, so);
2096
2097 viftable[i].v_rsvpd = NULL;
2098 /* This may seem silly, but we need to be sure we don't over-decrement
2099 * the RSVP counter, in case something slips up.
2100 */
2101 if (viftable[i].v_rsvp_on) {
2102 viftable[i].v_rsvp_on = 0;
2103 rsvp_on--;
2104 }
2105
2106 splx(s);
2107 return 0;
2108}
2109
2110void
2111ip_rsvp_force_done(so)
2112 struct socket *so;
2113{
2114 int vifi;
2115 register int s;
2116
2117 /* Don't bother if it is not the right type of socket. */
2118 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2119 return;
2120
2121 s = splnet();
2122
2123 /* The socket may be attached to more than one vif...this
2124 * is perfectly legal.
2125 */
2126 for (vifi = 0; vifi < numvifs; vifi++) {
2127 if (viftable[vifi].v_rsvpd == so) {
2128 viftable[vifi].v_rsvpd = NULL;
2129 /* This may seem silly, but we need to be sure we don't
2130 * over-decrement the RSVP counter, in case something slips up.
2131 */
2132 if (viftable[vifi].v_rsvp_on) {
2133 viftable[vifi].v_rsvp_on = 0;
2134 rsvp_on--;
2135 }
2136 }
2137 }
2138
2139 splx(s);
2140 return;
2141}
2142
2143void
2144rsvp_input(m, iphlen)
2145 struct mbuf *m;
2146 int iphlen;
2147{
2148 int vifi;
2149 register struct ip *ip = mtod(m, struct ip *);
2150 static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET };
2151 register int s;
2152 struct ifnet *ifp;
2153
2154 if (rsvpdebug)
2155 printf("rsvp_input: rsvp_on %d\n",rsvp_on);
2156
2157 /* Can still get packets with rsvp_on = 0 if there is a local member
2158 * of the group to which the RSVP packet is addressed. But in this
2159 * case we want to throw the packet away.
2160 */
2161 if (!rsvp_on) {
2162 m_freem(m);
2163 return;
2164 }
2165
2166 /* If the old-style non-vif-associated socket is set, then use
2167 * it and ignore the new ones.
2168 */
2169 if (ip_rsvpd != NULL) {
2170 if (rsvpdebug)
2171 printf("rsvp_input: Sending packet up old-style socket\n");
2172 rip_input(m, iphlen);
2173 return;
2174 }
2175
2176 s = splnet();
2177
2178 if (rsvpdebug)
2179 printf("rsvp_input: check vifs\n");
2180
2181#ifdef DIAGNOSTIC
2182 if (!(m->m_flags & M_PKTHDR))
2183 panic("rsvp_input no hdr");
2184#endif
2185
2186 ifp = m->m_pkthdr.rcvif;
2187 /* Find which vif the packet arrived on. */
2188 for (vifi = 0; vifi < numvifs; vifi++) {
2189 if (viftable[vifi].v_ifp == ifp)
2190 break;
2191 }
2192
2193 if (vifi == numvifs) {
2194 /* Can't find vif packet arrived on. Drop packet. */
2195 if (rsvpdebug)
2196 printf("rsvp_input: Can't find vif for packet...dropping it.\n");
2197 m_freem(m);
2198 splx(s);
2199 return;
2200 }
2201
2202 if (rsvpdebug)
2203 printf("rsvp_input: check socket\n");
2204
2205 if (viftable[vifi].v_rsvpd == NULL) {
2206 /* drop packet, since there is no specific socket for this
2207 * interface */
2208 if (rsvpdebug)
2209 printf("rsvp_input: No socket defined for vif %d\n",vifi);
2210 m_freem(m);
2211 splx(s);
2212 return;
2213 }
2214 rsvp_src.sin_addr = ip->ip_src;
2215
2216 if (rsvpdebug && m)
2217 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n",
2218 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv)));
2219
2220 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0)
2221 if (rsvpdebug)
2222 printf("rsvp_input: Failed to append to socket\n");
2223 else
2224 if (rsvpdebug)
2225 printf("rsvp_input: send packet up\n");
2226
2227 splx(s);
2228}
2229
2230#ifdef MROUTE_LKM
2231#include <sys/conf.h>
2232#include <sys/exec.h>
2233#include <sys/sysent.h>
2234#include <sys/lkm.h>
2235
2236MOD_MISC("ip_mroute_mod")
2237
2238static int
2239ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd)
2240{
2241 int i;
2242 struct lkm_misc *args = lkmtp->private.lkm_misc;
2243 int err = 0;
2244
2245 switch(cmd) {
2246 static int (*old_ip_mrouter_cmd)();
2247 static int (*old_ip_mrouter_done)();
2248 static int (*old_ip_mforward)();
2249 static int (*old_mrt_ioctl)();
2250 static void (*old_proto4_input)();
2251 static int (*old_legal_vif_num)();
2252 extern struct protosw inetsw[];
2253
2254 case LKM_E_LOAD:
2255 if(lkmexists(lkmtp) || ip_mrtproto)
2256 return(EEXIST);
2257 old_ip_mrouter_cmd = ip_mrouter_cmd;
2258 ip_mrouter_cmd = X_ip_mrouter_cmd;
2259 old_ip_mrouter_done = ip_mrouter_done;
2260 ip_mrouter_done = X_ip_mrouter_done;
2261 old_ip_mforward = ip_mforward;
2262 ip_mforward = X_ip_mforward;
2263 old_mrt_ioctl = mrt_ioctl;
2264 mrt_ioctl = X_mrt_ioctl;
2265 old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input;
2266 inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input;
2267 old_legal_vif_num = legal_vif_num;
2268 legal_vif_num = X_legal_vif_num;
2269 ip_mrtproto = IGMP_DVMRP;
2270
2271 printf("\nIP multicast routing loaded\n");
2272 break;
2273
2274 case LKM_E_UNLOAD:
2275 if (ip_mrouter)
2276 return EINVAL;
2277
2278 ip_mrouter_cmd = old_ip_mrouter_cmd;
2279 ip_mrouter_done = old_ip_mrouter_done;
2280 ip_mforward = old_ip_mforward;
2281 mrt_ioctl = old_mrt_ioctl;
2282 inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input;
2283 legal_vif_num = old_legal_vif_num;
2284 ip_mrtproto = 0;
2285 break;
2286
2287 default:
2288 err = EINVAL;
2289 break;
2290 }
2291
2292 return(err);
2293}
2294
2295int
2296ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) {
2297 DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle,
2298 nosys);
2299}
2300
2301#endif /* MROUTE_LKM */
2302#endif /* MROUTING */
|