xref: /csrg-svn/sys/net/rtsock.c (revision 42356)
1 /*
2  * Copyright (c) 1988 Regents of the University of California.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms are permitted
6  * provided that the above copyright notice and this paragraph are
7  * duplicated in all such forms and that any documentation,
8  * advertising materials, and other materials related to such
9  * distribution and use acknowledge that the software was developed
10  * by the University of California, Berkeley.  The name of the
11  * University may not be used to endorse or promote products derived
12  * from this software without specific prior written permission.
13  * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
14  * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
15  * WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.
16  *
17  *	@(#)rtsock.c	7.9 (Berkeley) 05/25/90
18  */
19 
20 #include "param.h"
21 #include "mbuf.h"
22 #include "user.h"
23 #include "proc.h"
24 #include "socket.h"
25 #include "socketvar.h"
26 #include "domain.h"
27 #include "protosw.h"
28 #include "errno.h"
29 
30 #include "af.h"
31 #include "if.h"
32 #include "route.h"
33 #include "raw_cb.h"
34 
35 #include "machine/mtpr.h"
36 
37 struct sockaddr route_dst = { 0, PF_ROUTE, };
38 struct sockaddr route_src = { 0, PF_ROUTE, };
39 struct sockproto route_proto = { PF_ROUTE, };
40 
41 /*ARGSUSED*/
42 route_usrreq(so, req, m, nam, control)
43 	register struct socket *so;
44 	int req;
45 	struct mbuf *m, *nam, *control;
46 {
47 	register int error = 0;
48 	register struct rawcb *rp = sotorawcb(so);
49 	if (req == PRU_ATTACH) {
50 		MALLOC(rp, struct rawcb *, sizeof(*rp), M_PCB, M_WAITOK);
51 		if (so->so_pcb = (caddr_t)rp)
52 			bzero(so->so_pcb, sizeof(*rp));
53 
54 	}
55 	if (req == PRU_DETACH && rp) {
56 		int af = rp->rcb_proto.sp_protocol;
57 		if (af == AF_INET)
58 			route_cb.ip_count--;
59 		else if (af == AF_NS)
60 			route_cb.ns_count--;
61 		else if (af == AF_ISO)
62 			route_cb.iso_count--;
63 		route_cb.any_count--;
64 	}
65 	error = raw_usrreq(so, req, m, nam, control);
66 	rp = sotorawcb(so);
67 	if (req == PRU_ATTACH && rp) {
68 		int af = rp->rcb_proto.sp_protocol;
69 		if (error) {
70 			free((caddr_t)rp, M_PCB);
71 			return (error);
72 		}
73 		if (af == AF_INET)
74 			route_cb.ip_count++;
75 		else if (af == AF_NS)
76 			route_cb.ns_count++;
77 		else if (af == AF_ISO)
78 			route_cb.iso_count++;
79 		rp->rcb_faddr = &route_src;
80 		route_cb.any_count++;
81 		soisconnected(so);
82 		so->so_options |= SO_USELOOPBACK;
83 	}
84 	return (error);
85 }
86 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
87 
88 /*ARGSUSED*/
89 route_output(m, so)
90 	register struct mbuf *m;
91 	struct socket *so;
92 {
93 	register struct rt_msghdr *rtm = 0;
94 	register struct rtentry *rt = 0;
95 	struct rtentry *saved_nrt = 0;
96 	struct sockaddr *dst = 0, *gate = 0, *netmask = 0, *genmask = 0;
97 	caddr_t cp, lim;
98 	int len, error = 0;
99 
100 #define senderr(e) { error = e; goto flush;}
101 	if (m == 0 || m->m_len < sizeof(long))
102 		return (ENOBUFS);
103 	if ((m = m_pullup(m, sizeof(long))) == 0)
104 		return (ENOBUFS);
105 	if ((m->m_flags & M_PKTHDR) == 0)
106 		panic("route_output");
107 	len = m->m_pkthdr.len;
108 	rtm = mtod(m, struct rt_msghdr *);
109 	if (len < rtm->rtm_msglen)
110 		senderr(EINVAL);
111 	R_Malloc(rtm, struct rt_msghdr *, len);
112 	if (rtm == 0)
113 		senderr(ENOBUFS);
114 	m_copydata(m, 0, len, (caddr_t)rtm);
115 	if (rtm->rtm_version != RTM_VERSION)
116 		senderr(EPROTONOSUPPORT);
117 	rtm->rtm_pid = u.u_procp->p_pid;
118 	lim = len + (caddr_t) rtm;
119 	cp = (caddr_t) (rtm + 1);
120 	if (rtm->rtm_addrs & RTA_DST) {
121 		dst = (struct sockaddr *)cp;
122 		cp += ROUNDUP(dst->sa_len);
123 	} else
124 		senderr(EINVAL);
125 	if ((rtm->rtm_addrs & RTA_GATEWAY) && cp < lim)  {
126 		gate = (struct sockaddr *)cp;
127 		cp += ROUNDUP(gate->sa_len);
128 	}
129 	if ((rtm->rtm_addrs & RTA_NETMASK) && cp < lim)  {
130 		netmask = (struct sockaddr *)cp;
131 		if (*cp)
132 			cp += ROUNDUP(netmask->sa_len);
133 		else
134 			cp += sizeof(long);
135 
136 	}
137 	if ((rtm->rtm_addrs & RTA_GENMASK) && cp < lim)  {
138 		genmask = (struct sockaddr *)cp;
139 	}
140 	switch (rtm->rtm_type) {
141 	case RTM_ADD:
142 		if (gate == 0)
143 			senderr(EINVAL);
144 		error = rtrequest(RTM_ADD, dst, gate, netmask,
145 					rtm->rtm_flags, &saved_nrt);
146 		if (error == 0 && saved_nrt) {
147 			rt_setmetrics(rtm->rtm_inits,
148 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
149 			saved_nrt->rt_refcnt--;
150 		}
151 		break;
152 
153 	case RTM_DELETE:
154 		error = rtrequest(RTM_DELETE, dst, gate, netmask,
155 				rtm->rtm_flags, (struct rtentry **)0);
156 		break;
157 
158 	case RTM_GET:
159 	case RTM_CHANGE:
160 	case RTM_LOCK:
161 		rt = rtalloc1(dst, 0);
162 		if (rt == 0)
163 			senderr(ESRCH);
164 		switch(rtm->rtm_type) {
165 			 struct	sockaddr *outmask;
166 
167 		case RTM_GET:
168 			netmask = rt_mask(rt);
169 			len = sizeof(*rtm) + ROUNDUP(rt_key(rt)->sa_len);
170 			rtm->rtm_addrs = RTA_DST;
171 			if (rt->rt_gateway) {
172 				len += ROUNDUP(rt->rt_gateway->sa_len);
173 				rtm->rtm_addrs |= RTA_GATEWAY;
174 			}
175 			if (netmask) {
176 				len += netmask->sa_len;
177 				rtm->rtm_addrs |= RTA_NETMASK;
178 			}
179 			if (len > rtm->rtm_msglen) {
180 				struct rt_msghdr *new_rtm;
181 				R_Malloc(new_rtm, struct rt_msghdr *, len);
182 				if (new_rtm == 0)
183 					senderr(ENOBUFS);
184 				Bcopy(rtm, new_rtm, rtm->rtm_msglen);
185 				Free(rtm); rtm = new_rtm;
186 				gate = (struct sockaddr *)
187 				    (ROUNDUP(rt->rt_gateway->sa_len)
188 								+ (char *)dst);
189 				Bcopy(&rt->rt_gateway, gate,
190 						rt->rt_gateway->sa_len);
191 				rtm->rtm_flags = rt->rt_flags;
192 				if (netmask) {
193 				    outmask = (struct sockaddr *)
194 				       (ROUNDUP(netmask->sa_len)+(char *)gate);
195 				    Bcopy(netmask, outmask, netmask->sa_len);
196 				}
197 			}
198 			break;
199 
200 		case RTM_CHANGE:
201 			if (gate == 0)
202 				senderr(EINVAL);
203 			if (gate->sa_len > (len = rt->rt_gateway->sa_len))
204 				senderr(EDQUOT);
205 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
206 				rt->rt_ifa->ifa_rtrequest(RTM_CHANGE, rt, gate);
207 			Bcopy(gate, rt->rt_gateway, len);
208 			rt->rt_gateway->sa_len = len;
209 
210 			rt_setmetrics(rtm->rtm_inits,
211 				&rtm->rtm_rmx, &rt->rt_rmx);
212 			/*
213 			 * Fall into
214 			 */
215 		case RTM_LOCK:
216 			rt->rt_rmx.rmx_locks |=
217 				(rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
218 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
219 			break;
220 		}
221 		goto cleanup;
222 
223 	default:
224 		senderr(EOPNOTSUPP);
225 	}
226 
227 flush:
228 	if (rtm) {
229 		if (error)
230 			rtm->rtm_errno = error;
231 		else
232 			rtm->rtm_flags |= RTF_DONE;
233 	}
234 cleanup:
235 	if (rt)
236 		rtfree(rt);
237     {
238 	register struct rawcb *rp = 0;
239 	/*
240 	 * Check to see if we don't want our own messages.
241 	 */
242 	if ((so->so_options & SO_USELOOPBACK) == 0) {
243 		if (route_cb.any_count <= 1) {
244 			if (rtm)
245 				Free(rtm);
246 			m_freem(m);
247 			return (error);
248 		}
249 		/* There is another listener, so construct message */
250 		rp = sotorawcb(so);
251 	}
252 	if (cp = (caddr_t)rtm) {
253 		m_copyback(m, 0, len, cp);
254 		Free(rtm);
255 	}
256 	if (rp)
257 		rp->rcb_proto.sp_family = 0; /* Avoid us */
258 	route_proto.sp_protocol = dst->sa_family;
259 	raw_input(m, &route_proto, &route_src, &route_dst);
260 	if (rp)
261 		rp->rcb_proto.sp_family = PF_ROUTE;
262     }
263 	return (error);
264 }
265 
266 static rt_setmetrics(which, in, out)
267 	u_long which;
268 	register struct rt_metrics *in, *out;
269 {
270 #define metric(f, e) if (which & (f)) out->e = in->e;
271 	metric(RTV_RPIPE, rmx_recvpipe);
272 	metric(RTV_SPIPE, rmx_sendpipe);
273 	metric(RTV_SSTHRESH, rmx_ssthresh);
274 	metric(RTV_RTT, rmx_rtt);
275 	metric(RTV_RTTVAR, rmx_rttvar);
276 	metric(RTV_HOPCOUNT, rmx_hopcount);
277 	metric(RTV_MTU, rmx_mtu);
278 #undef metric
279 }
280 
281 /*
282  * Copy data from a buffer back into the indicated mbuf chain,
283  * starting "off" bytes from the beginning, extending the mbuf
284  * chain if necessary.
285  */
286 m_copyback(m0, off, len, cp)
287 	struct	mbuf *m0;
288 	register int off;
289 	register int len;
290 	caddr_t cp;
291 
292 {
293 	register int mlen;
294 	register struct mbuf *m = m0, *n;
295 	int totlen = 0;
296 
297 	if (m0 == 0)
298 		return;
299 	while (off >= (mlen = m->m_len)) {
300 		off -= mlen;
301 		totlen += mlen;
302 		if (m->m_next == 0) {
303 			n = m_getclr(M_DONTWAIT, m->m_type);
304 			if (n == 0)
305 				goto out;
306 			n->m_len = min(MLEN, len + off);
307 			m->m_next = n;
308 		}
309 		m = m->m_next;
310 	}
311 	while (len > 0) {
312 		mlen = min (m->m_len - off, len);
313 		bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen);
314 		cp += mlen;
315 		len -= mlen;
316 		mlen += off;
317 		off = 0;
318 		totlen += mlen;
319 		if (len == 0)
320 			break;
321 		if (m->m_next == 0) {
322 			n = m_get(M_DONTWAIT, m->m_type);
323 			if (n == 0)
324 				break;
325 			n->m_len = min(MLEN, len);
326 			m->m_next = n;
327 		}
328 		m = m->m_next;
329 	}
330 out:	if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen))
331 		m->m_pkthdr.len = totlen;
332 }
333 
334 /*
335  * The miss message and losing message are very similar.
336  */
337 
338 rt_missmsg(type, dst, gate, mask, src, flags, error)
339 register struct sockaddr *dst;
340 struct sockaddr *gate, *mask, *src;
341 {
342 	register struct rt_msghdr *rtm;
343 	register struct mbuf *m;
344 	int dlen = ROUNDUP(dst->sa_len);
345 	int len = dlen + sizeof(*rtm);
346 
347 	if (route_cb.any_count == 0)
348 		return;
349 	m = m_gethdr(M_DONTWAIT, MT_DATA);
350 	if (m == 0)
351 		return;
352 	m->m_pkthdr.len = m->m_len = min(len, MHLEN);
353 	m->m_pkthdr.rcvif = 0;
354 	rtm = mtod(m, struct rt_msghdr *);
355 	bzero((caddr_t)rtm, sizeof(*rtm)); /*XXX assumes sizeof(*rtm) < MHLEN*/
356 	rtm->rtm_flags = RTF_DONE | flags;
357 	rtm->rtm_msglen = len;
358 	rtm->rtm_version = RTM_VERSION;
359 	rtm->rtm_type = type;
360 	rtm->rtm_addrs = RTA_DST;
361 	if (type == RTM_OLDADD || type == RTM_OLDDEL) {
362 		rtm->rtm_pid = u.u_procp->p_pid;
363 	}
364 	m_copyback(m, sizeof (*rtm), dlen, (caddr_t)dst);
365 	if (gate) {
366 		dlen = ROUNDUP(gate->sa_len);
367 		m_copyback(m, len ,  dlen, (caddr_t)gate);
368 		len += dlen;
369 		rtm->rtm_addrs |= RTA_GATEWAY;
370 	}
371 	if (mask) {
372 		if (mask->sa_len)
373 			dlen = ROUNDUP(mask->sa_len);
374 		else
375 			dlen = sizeof(long);
376 		m_copyback(m, len ,  dlen, (caddr_t)mask);
377 		len += dlen;
378 		rtm->rtm_addrs |= RTA_NETMASK;
379 	}
380 	if (src) {
381 		dlen = ROUNDUP(src->sa_len);
382 		m_copyback(m, len ,  dlen, (caddr_t)src);
383 		len += dlen;
384 		rtm->rtm_addrs |= RTA_AUTHOR;
385 	}
386 	if (m->m_pkthdr.len != len) {
387 		m_freem(m);
388 		return;
389 	}
390 	rtm->rtm_errno = error;
391 	rtm->rtm_msglen = len;
392 	route_proto.sp_protocol = dst->sa_family;
393 	raw_input(m, &route_proto, &route_src, &route_dst);
394 }
395 
396 #include "kinfo.h"
397 struct walkarg {
398 	int	w_op, w_arg;
399 	int	w_given, w_needed;
400 	caddr_t	w_where;
401 	struct	{
402 		struct rt_msghdr m_rtm;
403 		char	m_sabuf[128];
404 	} w_m;
405 #define w_rtm w_m.m_rtm
406 };
407 /*
408  * This is used in dumping the kernel table via getkinfo().
409  */
410 rt_dumpentry(rn, w)
411 	struct radix_node *rn;
412 	register struct walkarg *w;
413 {
414 	register struct sockaddr *sa;
415 	int n, error;
416 
417     for (; rn; rn = rn->rn_dupedkey) {
418 	int count = 0, size = sizeof(w->w_rtm);
419 	register struct rtentry *rt = (struct rtentry *)rn;
420 
421 	if (rn->rn_flags & RNF_ROOT)
422 		continue;
423 	if (w->w_op == KINFO_RT_FLAGS && !(rt->rt_flags & w->w_arg))
424 		continue;
425 #define next(a, l) {size += (l); w->w_rtm.rtm_addrs |= (a); }
426 	w->w_rtm.rtm_addrs = 0;
427 	if (sa = rt_key(rt))
428 		next(RTA_DST, ROUNDUP(sa->sa_len));
429 	if (sa = rt->rt_gateway)
430 		next(RTA_GATEWAY, ROUNDUP(sa->sa_len));
431 	if (sa = rt_mask(rt))
432 		next(RTA_NETMASK,
433 			sa->sa_len ? ROUNDUP(sa->sa_len) : sizeof(long));
434 	if (sa = rt->rt_genmask)
435 		next(RTA_GENMASK, ROUNDUP(sa->sa_len));
436 	w->w_needed += size;
437 	if (w->w_where == NULL || w->w_needed > 0)
438 		continue;
439 	w->w_rtm.rtm_msglen = size;
440 	w->w_rtm.rtm_flags = rt->rt_flags;
441 	w->w_rtm.rtm_use = rt->rt_use;
442 	w->w_rtm.rtm_rmx = rt->rt_rmx;
443 	w->w_rtm.rtm_index = rt->rt_ifp->if_index;
444 #undef next
445 #define next(l) {n = (l); Bcopy(sa, cp, n); cp += n;}
446 	if (size <= sizeof(w->w_m)) {
447 		register caddr_t cp = (caddr_t)(w->w_m.m_sabuf);
448 		if (sa = rt_key(rt))
449 			next(ROUNDUP(sa->sa_len));
450 		if (sa = rt->rt_gateway)
451 			next(ROUNDUP(sa->sa_len));
452 		if (sa = rt_mask(rt))
453 			next(sa->sa_len ? ROUNDUP(sa->sa_len) : sizeof(long));
454 		if (sa = rt->rt_genmask)
455 			next(ROUNDUP(sa->sa_len));
456 #undef next
457 #define next(s, l) {n = (l); \
458     if (error = copyout((caddr_t)(s), w->w_where, n)) return (error); \
459     w->w_where += n;}
460 
461 		next(&w->w_m, size); /* Copy rtmsg and sockaddrs back */
462 		continue;
463 	}
464 	next(&w->w_rtm, sizeof(w->w_rtm));
465 	if (sa = rt_key(rt))
466 		next(sa, ROUNDUP(sa->sa_len));
467 	if (sa = rt->rt_gateway)
468 		next(sa, ROUNDUP(sa->sa_len));
469 	if (sa = rt_mask(rt))
470 		next(sa, sa->sa_len ? ROUNDUP(sa->sa_len) : sizeof(long));
471 	if (sa = rt->rt_genmask)
472 		next(sa, ROUNDUP(sa->sa_len));
473     }
474 	return (0);
475 #undef next
476 }
477 
478 kinfo_rtable(op, where, given, arg, needed)
479 	int	op, arg;
480 	caddr_t	where;
481 	int	*given, *needed;
482 {
483 	register struct radix_node_head *rnh;
484 	int	s, error = 0;
485 	u_char  af = ki_af(op);
486 	struct	walkarg w;
487 
488 	op &= 0xffff;
489 	if (op != KINFO_RT_DUMP && op != KINFO_RT_FLAGS)
490 		return (EINVAL);
491 
492 	Bzero(&w, sizeof(w));
493 	if ((w.w_where = where) && given)
494 		w.w_given = *given;
495 	w.w_needed = 0 - w.w_given;
496 	w.w_arg = arg;
497 	w.w_op = op;
498 	w.w_rtm.rtm_version = RTM_VERSION;
499 	w.w_rtm.rtm_type = RTM_GET;
500 
501 	s = splnet();
502 	for (rnh = radix_node_head; rnh; rnh = rnh->rnh_next) {
503 		if (rnh->rnh_af == 0)
504 			continue;
505 		if (af && af != rnh->rnh_af)
506 			continue;
507 		error = rt_walk(rnh->rnh_treetop, rt_dumpentry, &w);
508 		if (error)
509 			break;
510 	}
511 	w.w_needed += w.w_given;
512 	if (where && given)
513 		*given = w.w_where - where;
514 	else
515 		w.w_needed = (11 * w.w_needed) / 10;
516 	*needed = w.w_needed;
517 	splx(s);
518 	return (error);
519 }
520 
521 rt_walk(rn, f, w)
522 	register struct radix_node *rn;
523 	register int (*f)();
524 	struct walkarg *w;
525 {
526 	int error;
527 	for (;;) {
528 		while (rn->rn_b >= 0)
529 			rn = rn->rn_l;	/* First time through node, go left */
530 		if (error = (*f)(rn, w))
531 			return (error);	/* Process Leaf */
532 		while (rn->rn_p->rn_r == rn) {	/* if coming back from right */
533 			rn = rn->rn_p;		/* go back up */
534 			if (rn->rn_flags & RNF_ROOT)
535 				return 0;
536 		}
537 		rn = rn->rn_p->rn_r;		/* otherwise, go right*/
538 	}
539 }
540 
541 /*
542  * Definitions of protocols supported in the ROUTE domain.
543  */
544 
545 int	raw_init(),raw_usrreq(),raw_input(),raw_ctlinput();
546 extern	struct domain routedomain;		/* or at least forward */
547 
548 struct protosw routesw[] = {
549 { SOCK_RAW,	&routedomain,	0,		PR_ATOMIC|PR_ADDR,
550   raw_input,	route_output,	raw_ctlinput,	0,
551   route_usrreq,
552   raw_init,	0,		0,		0,
553 },
554 { 0,		0,		0,		0,
555   raw_input,	0,		raw_ctlinput,	0,
556   raw_usrreq,
557   raw_init,	0,		0,		0,
558 }
559 };
560 
561 int	unp_externalize(), unp_dispose();
562 
563 struct domain routedomain =
564     { PF_ROUTE, "route", 0, 0, 0,
565       routesw, &routesw[sizeof(routesw)/sizeof(routesw[0])] };
566