xref: /csrg-svn/sys/net/rtsock.c (revision 40786)
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	7.8 (Berkeley) 04/05/90
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 	}
83 	return (error);
84 }
85 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
86 
87 /*ARGSUSED*/
88 route_output(m, so)
89 	register struct mbuf *m;
90 	struct socket *so;
91 {
92 	register struct rt_msghdr *rtm = 0;
93 	register struct rtentry *rt = 0;
94 	struct rtentry *saved_nrt = 0;
95 	struct sockaddr *dst = 0, *gate = 0, *netmask = 0, *genmask = 0;
96 	caddr_t cp, lim;
97 	int len, error = 0;
98 
99 #define senderr(e) { error = e; goto flush;}
100 	if (m == 0 || m->m_len < sizeof(long))
101 		return (ENOBUFS);
102 	if ((m = m_pullup(m, sizeof(long))) == 0)
103 		return (ENOBUFS);
104 	if ((m->m_flags & M_PKTHDR) == 0)
105 		panic("route_output");
106 	len = m->m_pkthdr.len;
107 	rtm = mtod(m, struct rt_msghdr *);
108 	if (len < rtm->rtm_msglen)
109 		senderr(EINVAL);
110 	R_Malloc(rtm, struct rt_msghdr *, len);
111 	if (rtm == 0)
112 		senderr(ENOBUFS);
113 	m_copydata(m, 0, len, (caddr_t)rtm);
114 	if (rtm->rtm_version != RTM_VERSION)
115 		senderr(EPROTONOSUPPORT);
116 	rtm->rtm_pid = u.u_procp->p_pid;
117 	lim = len + (caddr_t) rtm;
118 	cp = (caddr_t) (rtm + 1);
119 	if (rtm->rtm_addrs & RTA_DST) {
120 		dst = (struct sockaddr *)cp;
121 		cp += ROUNDUP(dst->sa_len);
122 	} else
123 		senderr(EINVAL);
124 	if ((rtm->rtm_addrs & RTA_GATEWAY) && cp < lim)  {
125 		gate = (struct sockaddr *)cp;
126 		cp += ROUNDUP(gate->sa_len);
127 	}
128 	if ((rtm->rtm_addrs & RTA_NETMASK) && cp < lim)  {
129 		netmask = (struct sockaddr *)cp;
130 		if (*cp)
131 			cp += ROUNDUP(netmask->sa_len);
132 		else
133 			cp += sizeof(long);
134 
135 	}
136 	if ((rtm->rtm_addrs & RTA_GENMASK) && cp < lim)  {
137 		genmask = (struct sockaddr *)cp;
138 	}
139 	switch (rtm->rtm_type) {
140 	case RTM_ADD:
141 		if (gate == 0)
142 			senderr(EINVAL);
143 		error = rtrequest(RTM_ADD, dst, gate, netmask,
144 					rtm->rtm_flags, &saved_nrt);
145 		if (error == 0 && saved_nrt) {
146 			rt_setmetrics(rtm->rtm_inits,
147 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
148 			saved_nrt->rt_refcnt--;
149 		}
150 		break;
151 
152 	case RTM_DELETE:
153 		error = rtrequest(RTM_DELETE, dst, gate, netmask,
154 				rtm->rtm_flags, (struct rtentry **)0);
155 		break;
156 
157 	case RTM_GET:
158 	case RTM_CHANGE:
159 	case RTM_LOCK:
160 		rt = rtalloc1(dst, 0);
161 		if (rt == 0)
162 			senderr(ESRCH);
163 		switch(rtm->rtm_type) {
164 			 struct	sockaddr *outmask;
165 
166 		case RTM_GET:
167 			netmask = rt_mask(rt);
168 			len = sizeof(*rtm) + ROUNDUP(rt_key(rt)->sa_len);
169 			rtm->rtm_addrs = RTA_DST;
170 			if (rt->rt_gateway) {
171 				len += ROUNDUP(rt->rt_gateway->sa_len);
172 				rtm->rtm_addrs |= RTA_GATEWAY;
173 			}
174 			if (netmask) {
175 				len += netmask->sa_len;
176 				rtm->rtm_addrs |= RTA_NETMASK;
177 			}
178 			if (len > rtm->rtm_msglen) {
179 				struct rt_msghdr *new_rtm;
180 				R_Malloc(new_rtm, struct rt_msghdr *, len);
181 				if (new_rtm == 0)
182 					senderr(ENOBUFS);
183 				Bcopy(rtm, new_rtm, rtm->rtm_msglen);
184 				Free(rtm); rtm = new_rtm;
185 				gate = (struct sockaddr *)
186 				    (ROUNDUP(rt->rt_gateway->sa_len)
187 								+ (char *)dst);
188 				Bcopy(&rt->rt_gateway, gate,
189 						rt->rt_gateway->sa_len);
190 				rtm->rtm_flags = rt->rt_flags;
191 				if (netmask) {
192 				    outmask = (struct sockaddr *)
193 				       (ROUNDUP(netmask->sa_len)+(char *)gate);
194 				    Bcopy(netmask, outmask, netmask->sa_len);
195 				}
196 			}
197 			break;
198 
199 		case RTM_CHANGE:
200 			if (gate == 0)
201 				senderr(EINVAL);
202 			if (gate->sa_len > (len = rt->rt_gateway->sa_len))
203 				senderr(EDQUOT);
204 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
205 				rt->rt_ifa->ifa_rtrequest(RTM_CHANGE, rt, gate);
206 			Bcopy(gate, rt->rt_gateway, len);
207 			rt->rt_gateway->sa_len = len;
208 
209 			rt_setmetrics(rtm->rtm_inits,
210 				&rtm->rtm_rmx, &rt->rt_rmx);
211 			/*
212 			 * Fall into
213 			 */
214 		case RTM_LOCK:
215 			rt->rt_rmx.rmx_locks |=
216 				(rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
217 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
218 			break;
219 		}
220 		goto cleanup;
221 
222 	default:
223 		senderr(EOPNOTSUPP);
224 	}
225 
226 flush:
227 	if (rtm) {
228 		if (error)
229 			rtm->rtm_errno = error;
230 		else
231 			rtm->rtm_flags |= RTF_DONE;
232 	}
233 cleanup:
234 	if (rt)
235 		rtfree(rt);
236 	if (cp = (caddr_t)rtm) {
237 		m_copyback(m, 0, len, cp);
238 		Free(rtm);
239 	}
240 	route_proto.sp_protocol = dst->sa_family;
241 	raw_input(m, &route_proto, &route_src, &route_dst);
242 	return (error);
243 }
244 
245 static rt_setmetrics(which, in, out)
246 	u_long which;
247 	register struct rt_metrics *in, *out;
248 {
249 #define metric(f, e) if (which & (f)) out->e = in->e;
250 	metric(RTV_RPIPE, rmx_recvpipe);
251 	metric(RTV_SPIPE, rmx_sendpipe);
252 	metric(RTV_SSTHRESH, rmx_ssthresh);
253 	metric(RTV_RTT, rmx_rtt);
254 	metric(RTV_RTTVAR, rmx_rttvar);
255 	metric(RTV_HOPCOUNT, rmx_hopcount);
256 	metric(RTV_MTU, rmx_mtu);
257 #undef metric
258 }
259 
260 /*
261  * Copy data from a buffer back into the indicated mbuf chain,
262  * starting "off" bytes from the beginning, extending the mbuf
263  * chain if necessary.
264  */
265 m_copyback(m0, off, len, cp)
266 	struct	mbuf *m0;
267 	register int off;
268 	register int len;
269 	caddr_t cp;
270 
271 {
272 	register int mlen;
273 	register struct mbuf *m = m0, *n;
274 	int totlen = 0;
275 
276 	if (m0 == 0)
277 		return;
278 	while (off >= (mlen = m->m_len)) {
279 		off -= mlen;
280 		totlen += mlen;
281 		if (m->m_next == 0) {
282 			n = m_getclr(M_DONTWAIT, m->m_type);
283 			if (n == 0)
284 				goto out;
285 			n->m_len = min(MLEN, len + off);
286 			m->m_next = n;
287 		}
288 		m = m->m_next;
289 	}
290 	while (len > 0) {
291 		mlen = min (m->m_len - off, len);
292 		bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen);
293 		cp += mlen;
294 		len -= mlen;
295 		mlen += off;
296 		off = 0;
297 		totlen += mlen;
298 		if (len == 0)
299 			break;
300 		if (m->m_next == 0) {
301 			n = m_get(M_DONTWAIT, m->m_type);
302 			if (n == 0)
303 				break;
304 			n->m_len = min(MLEN, len);
305 			m->m_next = n;
306 		}
307 		m = m->m_next;
308 	}
309 out:	if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen))
310 		m->m_pkthdr.len = totlen;
311 }
312 
313 /*
314  * The miss message and losing message are very similar.
315  */
316 
317 rt_missmsg(type, dst, gate, mask, src, flags, error)
318 register struct sockaddr *dst;
319 struct sockaddr *gate, *mask, *src;
320 {
321 	register struct rt_msghdr *rtm;
322 	register struct mbuf *m;
323 	int dlen = ROUNDUP(dst->sa_len);
324 	int len = dlen + sizeof(*rtm);
325 
326 	if (route_cb.any_count == 0)
327 		return;
328 	m = m_gethdr(M_DONTWAIT, MT_DATA);
329 	if (m == 0)
330 		return;
331 	m->m_pkthdr.len = m->m_len = min(len, MHLEN);
332 	m->m_pkthdr.rcvif = 0;
333 	rtm = mtod(m, struct rt_msghdr *);
334 	bzero((caddr_t)rtm, sizeof(*rtm)); /*XXX assumes sizeof(*rtm) < MHLEN*/
335 	rtm->rtm_flags = RTF_DONE | flags;
336 	rtm->rtm_msglen = len;
337 	rtm->rtm_version = RTM_VERSION;
338 	rtm->rtm_type = type;
339 	rtm->rtm_addrs = RTA_DST;
340 	if (type == RTM_OLDADD || type == RTM_OLDDEL) {
341 		rtm->rtm_pid = u.u_procp->p_pid;
342 	}
343 	m_copyback(m, sizeof (*rtm), dlen, (caddr_t)dst);
344 	if (gate) {
345 		dlen = ROUNDUP(gate->sa_len);
346 		m_copyback(m, len ,  dlen, (caddr_t)gate);
347 		len += dlen;
348 		rtm->rtm_addrs |= RTA_GATEWAY;
349 	}
350 	if (mask) {
351 		if (mask->sa_len)
352 			dlen = ROUNDUP(mask->sa_len);
353 		else
354 			dlen = sizeof(long);
355 		m_copyback(m, len ,  dlen, (caddr_t)mask);
356 		len += dlen;
357 		rtm->rtm_addrs |= RTA_NETMASK;
358 	}
359 	if (src) {
360 		dlen = ROUNDUP(src->sa_len);
361 		m_copyback(m, len ,  dlen, (caddr_t)src);
362 		len += dlen;
363 		rtm->rtm_addrs |= RTA_AUTHOR;
364 	}
365 	if (m->m_pkthdr.len != len) {
366 		m_freem(m);
367 		return;
368 	}
369 	rtm->rtm_errno = error;
370 	rtm->rtm_msglen = len;
371 	route_proto.sp_protocol = dst->sa_family;
372 	raw_input(m, &route_proto, &route_src, &route_dst);
373 }
374 
375 #include "kinfo.h"
376 struct walkarg {
377 	int	w_op, w_arg;
378 	int	w_given, w_needed;
379 	caddr_t	w_where;
380 	struct	{
381 		struct rt_msghdr m_rtm;
382 		char	m_sabuf[128];
383 	} w_m;
384 #define w_rtm w_m.m_rtm
385 };
386 /*
387  * This is used in dumping the kernel table via getkinfo().
388  */
389 rt_dumpentry(rn, w)
390 	struct radix_node *rn;
391 	register struct walkarg *w;
392 {
393 	register struct sockaddr *sa;
394 	int n, error;
395 
396     for (; rn && !(rn->rn_flags & RNF_ROOT); rn = rn->rn_dupedkey) {
397 	int count = 0, size = sizeof(w->w_rtm);
398 	register struct rtentry *rt = (struct rtentry *)rn;
399 
400 	if (w->w_op == KINFO_RT_FLAGS && !(rt->rt_flags & w->w_arg))
401 		continue;
402 #define next(a, l) {size += (l); w->w_rtm.rtm_addrs |= (a); }
403 	w->w_rtm.rtm_addrs = 0;
404 	if (sa = rt_key(rt))
405 		next(RTA_DST, ROUNDUP(sa->sa_len));
406 	if (sa = rt->rt_gateway)
407 		next(RTA_GATEWAY, ROUNDUP(sa->sa_len));
408 	if (sa = rt_mask(rt))
409 		next(RTA_NETMASK,
410 			sa->sa_len ? ROUNDUP(sa->sa_len) : sizeof(long));
411 	if (sa = rt->rt_genmask)
412 		next(RTA_GENMASK, ROUNDUP(sa->sa_len));
413 	w->w_needed += size;
414 	if (w->w_where == NULL || w->w_needed > 0)
415 		continue;
416 	w->w_rtm.rtm_msglen = size;
417 	w->w_rtm.rtm_flags = rt->rt_flags;
418 	w->w_rtm.rtm_use = rt->rt_use;
419 	w->w_rtm.rtm_rmx = rt->rt_rmx;
420 	w->w_rtm.rtm_index = rt->rt_ifp->if_index;
421 #undef next
422 #define next(l) {n = (l); Bcopy(sa, cp, n); cp += n;}
423 	if (size <= sizeof(w->w_m)) {
424 		register caddr_t cp = (caddr_t)(w->w_m.m_sabuf);
425 		if (sa = rt_key(rt))
426 			next(ROUNDUP(sa->sa_len));
427 		if (sa = rt->rt_gateway)
428 			next(ROUNDUP(sa->sa_len));
429 		if (sa = rt_mask(rt))
430 			next(sa->sa_len ? ROUNDUP(sa->sa_len) : sizeof(long));
431 		if (sa = rt->rt_genmask)
432 			next(ROUNDUP(sa->sa_len));
433 #undef next
434 #define next(s, l) {n = (l); \
435     if (error = copyout((caddr_t)(s), w->w_where, n)) return (error); \
436     w->w_where += n;}
437 
438 		next(&w->w_m, size); /* Copy rtmsg and sockaddrs back */
439 		continue;
440 	}
441 	next(&w->w_rtm, sizeof(w->w_rtm));
442 	if (sa = rt_key(rt))
443 		next(sa, ROUNDUP(sa->sa_len));
444 	if (sa = rt->rt_gateway)
445 		next(sa, ROUNDUP(sa->sa_len));
446 	if (sa = rt_mask(rt))
447 		next(sa, sa->sa_len ? ROUNDUP(sa->sa_len) : sizeof(long));
448 	if (sa = rt->rt_genmask)
449 		next(sa, ROUNDUP(sa->sa_len));
450     }
451 	return (0);
452 #undef next
453 }
454 
455 kinfo_rtable(op, where, given, arg, needed)
456 	int	op, arg;
457 	caddr_t	where;
458 	int	*given, *needed;
459 {
460 	register struct radix_node_head *rnh;
461 	int	s, error = 0;
462 	u_char  af = ki_af(op);
463 	struct	walkarg w;
464 
465 	op &= 0xffff;
466 	if (op != KINFO_RT_DUMP && op != KINFO_RT_FLAGS)
467 		return (EINVAL);
468 
469 	Bzero(&w, sizeof(w));
470 	if ((w.w_where = where) && given)
471 		w.w_given = *given;
472 	w.w_needed = 0 - w.w_given;
473 	w.w_arg = arg;
474 	w.w_op = op;
475 	w.w_rtm.rtm_version = RTM_VERSION;
476 	w.w_rtm.rtm_type = RTM_GET;
477 
478 	s = splnet();
479 	for (rnh = radix_node_head; rnh; rnh = rnh->rnh_next) {
480 		if (rnh->rnh_af == 0)
481 			continue;
482 		if (af && af != rnh->rnh_af)
483 			continue;
484 		error = rt_walk(rnh->rnh_treetop, rt_dumpentry, &w);
485 		if (error)
486 			break;
487 	}
488 	w.w_needed += w.w_given;
489 	if (where && given)
490 		*given = w.w_where - where;
491 	else
492 		w.w_needed = (11 * w.w_needed) / 10;
493 	*needed = w.w_needed;
494 	splx(s);
495 	return (error);
496 }
497 
498 rt_walk(rn, f, w)
499 	register struct radix_node *rn;
500 	register int (*f)();
501 	struct walkarg *w;
502 {
503 	int error;
504 	for (;;) {
505 		while (rn->rn_b >= 0)
506 			rn = rn->rn_l;	/* First time through node, go left */
507 		if (error = (*f)(rn, w))
508 			return (error);	/* Process Leaf */
509 		while (rn->rn_p->rn_r == rn) {	/* if coming back from right */
510 			rn = rn->rn_p;		/* go back up */
511 			if (rn->rn_flags & RNF_ROOT)
512 				return 0;
513 		}
514 		rn = rn->rn_p->rn_r;		/* otherwise, go right*/
515 	}
516 }
517 
518 /*
519  * Definitions of protocols supported in the ROUTE domain.
520  */
521 
522 int	route_output();
523 int	raw_init(),raw_usrreq(),raw_input(),raw_ctlinput();
524 extern	struct domain routedomain;		/* or at least forward */
525 
526 struct protosw routesw[] = {
527 { SOCK_RAW,	&routedomain,	0,		PR_ATOMIC|PR_ADDR,
528   raw_input,	route_output,	raw_ctlinput,	0,
529   route_usrreq,
530   raw_init,	0,		0,		0,
531 },
532 { 0,		0,		0,		0,
533   raw_input,	0,		raw_ctlinput,	0,
534   raw_usrreq,
535   raw_init,	0,		0,		0,
536 }
537 };
538 
539 int	unp_externalize(), unp_dispose();
540 
541 struct domain routedomain =
542     { PF_ROUTE, "route", 0, 0, 0,
543       routesw, &routesw[sizeof(routesw)/sizeof(routesw[0])] };
544