xref: /netbsd-src/sys/netinet/ip_output.c (revision cda4f8f6ee55684e8d311b86c99ea59191e6b74f)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1990 Regents of the University of California.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	from: @(#)ip_output.c	7.23 (Berkeley) 11/12/90
34  *	$Id: ip_output.c,v 1.3 1993/05/22 11:42:35 cgd Exp $
35  */
36 
37 #include "param.h"
38 #include "malloc.h"
39 #include "mbuf.h"
40 #include "errno.h"
41 #include "protosw.h"
42 #include "socket.h"
43 #include "socketvar.h"
44 
45 #include "../net/if.h"
46 #include "../net/route.h"
47 
48 #include "in.h"
49 #include "in_systm.h"
50 #include "ip.h"
51 #include "in_pcb.h"
52 #include "in_var.h"
53 #include "ip_var.h"
54 
55 #ifdef vax
56 #include "machine/mtpr.h"
57 #endif
58 
59 struct mbuf *ip_insertoptions();
60 
61 /*
62  * IP output.  The packet in mbuf chain m contains a skeletal IP
63  * header (with len, off, ttl, proto, tos, src, dst).
64  * The mbuf chain containing the packet will be freed.
65  * The mbuf opt, if present, will not be freed.
66  */
67 ip_output(m0, opt, ro, flags)
68 	struct mbuf *m0;
69 	struct mbuf *opt;
70 	struct route *ro;
71 	int flags;
72 {
73 	register struct ip *ip, *mhip;
74 	register struct ifnet *ifp;
75 	register struct mbuf *m = m0;
76 	register int hlen = sizeof (struct ip);
77 	int len, off, error = 0;
78 	struct route iproute;
79 	struct sockaddr_in *dst;
80 	struct in_ifaddr *ia;
81 
82 #ifdef	DIAGNOSTIC
83 	if ((m->m_flags & M_PKTHDR) == 0)
84 		panic("ip_output no HDR");
85 #endif
86 	if (opt) {
87 		m = ip_insertoptions(m, opt, &len);
88 		hlen = len;
89 	}
90 	ip = mtod(m, struct ip *);
91 	/*
92 	 * Fill in IP header.
93 	 */
94 	if ((flags & IP_FORWARDING) == 0) {
95 		ip->ip_v = IPVERSION;
96 		ip->ip_off &= IP_DF;
97 		ip->ip_id = htons(ip_id++);
98 		ip->ip_hl = hlen >> 2;
99 	} else {
100 		hlen = ip->ip_hl << 2;
101 		ipstat.ips_localout++;
102 	}
103 	/*
104 	 * Route packet.
105 	 */
106 	if (ro == 0) {
107 		ro = &iproute;
108 		bzero((caddr_t)ro, sizeof (*ro));
109 	}
110 	dst = (struct sockaddr_in *)&ro->ro_dst;
111 	/*
112 	 * If there is a cached route,
113 	 * check that it is to the same destination
114 	 * and is still up.  If not, free it and try again.
115 	 */
116 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
117 	   dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
118 		RTFREE(ro->ro_rt);
119 		ro->ro_rt = (struct rtentry *)0;
120 	}
121 	if (ro->ro_rt == 0) {
122 		dst->sin_family = AF_INET;
123 		dst->sin_len = sizeof(*dst);
124 		dst->sin_addr = ip->ip_dst;
125 	}
126 	/*
127 	 * If routing to interface only,
128 	 * short circuit routing lookup.
129 	 */
130 	if (flags & IP_ROUTETOIF) {
131 
132 		ia = (struct in_ifaddr *)ifa_ifwithdstaddr((struct sockaddr *)dst);
133 		if (ia == 0)
134 			ia = in_iaonnetof(in_netof(ip->ip_dst));
135 		if (ia == 0) {
136 			error = ENETUNREACH;
137 			goto bad;
138 		}
139 		ifp = ia->ia_ifp;
140 	} else {
141 		if (ro->ro_rt == 0)
142 			rtalloc(ro);
143 		if (ro->ro_rt == 0) {
144 			error = EHOSTUNREACH;
145 			goto bad;
146 		}
147 		ia = (struct in_ifaddr *)ro->ro_rt->rt_ifa;
148 		ifp = ro->ro_rt->rt_ifp;
149 		ro->ro_rt->rt_use++;
150 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
151 			dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
152 	}
153 #ifndef notdef
154 	/*
155 	 * If source address not specified yet, use address
156 	 * of outgoing interface.
157 	 */
158 	if (ip->ip_src.s_addr == INADDR_ANY)
159 		ip->ip_src = IA_SIN(ia)->sin_addr;
160 #endif
161 	/*
162 	 * Look for broadcast address and
163 	 * and verify user is allowed to send
164 	 * such a packet.
165 	 */
166 	if (in_broadcast(dst->sin_addr)) {
167 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
168 			error = EADDRNOTAVAIL;
169 			goto bad;
170 		}
171 		if ((flags & IP_ALLOWBROADCAST) == 0) {
172 			error = EACCES;
173 			goto bad;
174 		}
175 		/* don't allow broadcast messages to be fragmented */
176 		if ((u_short)ip->ip_len > ifp->if_mtu) {
177 			error = EMSGSIZE;
178 			goto bad;
179 		}
180 		m->m_flags |= M_BCAST;
181 	}
182 
183 	/*
184 	 * If small enough for interface, can just send directly.
185 	 */
186 	if ((u_short)ip->ip_len <= ifp->if_mtu) {
187 		ip->ip_len = htons((u_short)ip->ip_len);
188 		ip->ip_off = htons((u_short)ip->ip_off);
189 		ip->ip_sum = 0;
190 		ip->ip_sum = in_cksum(m, hlen);
191 		error = (*ifp->if_output)(ifp, m,
192 				(struct sockaddr *)dst, ro->ro_rt);
193 		goto done;
194 	}
195 	ipstat.ips_fragmented++;
196 	/*
197 	 * Too large for interface; fragment if possible.
198 	 * Must be able to put at least 8 bytes per fragment.
199 	 */
200 	if (ip->ip_off & IP_DF) {
201 		error = EMSGSIZE;
202 		goto bad;
203 	}
204 	len = (ifp->if_mtu - hlen) &~ 7;
205 	if (len < 8) {
206 		error = EMSGSIZE;
207 		goto bad;
208 	}
209 
210     {
211 	int mhlen, firstlen = len;
212 	struct mbuf **mnext = &m->m_nextpkt;
213 
214 	/*
215 	 * Loop through length of segment after first fragment,
216 	 * make new header and copy data of each part and link onto chain.
217 	 */
218 	m0 = m;
219 	mhlen = sizeof (struct ip);
220 	for (off = hlen + len; off < (u_short)ip->ip_len; off += len) {
221 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
222 		if (m == 0) {
223 			error = ENOBUFS;
224 			goto sendorfree;
225 		}
226 		m->m_data += max_linkhdr;
227 		mhip = mtod(m, struct ip *);
228 		*mhip = *ip;
229 		if (hlen > sizeof (struct ip)) {
230 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
231 			mhip->ip_hl = mhlen >> 2;
232 		}
233 		m->m_len = mhlen;
234 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
235 		if (ip->ip_off & IP_MF)
236 			mhip->ip_off |= IP_MF;
237 		if (off + len >= (u_short)ip->ip_len)
238 			len = (u_short)ip->ip_len - off;
239 		else
240 			mhip->ip_off |= IP_MF;
241 		mhip->ip_len = htons((u_short)(len + mhlen));
242 		m->m_next = m_copy(m0, off, len);
243 		if (m->m_next == 0) {
244 			error = ENOBUFS;	/* ??? */
245 			goto sendorfree;
246 		}
247 		m->m_pkthdr.len = mhlen + len;
248 		m->m_pkthdr.rcvif = (struct ifnet *)0;
249 		mhip->ip_off = htons((u_short)mhip->ip_off);
250 		mhip->ip_sum = 0;
251 		mhip->ip_sum = in_cksum(m, mhlen);
252 		*mnext = m;
253 		mnext = &m->m_nextpkt;
254 		ipstat.ips_ofragments++;
255 	}
256 	/*
257 	 * Update first fragment by trimming what's been copied out
258 	 * and updating header, then send each fragment (in order).
259 	 */
260 	m = m0;
261 	m_adj(m, hlen + firstlen - (u_short)ip->ip_len);
262 	m->m_pkthdr.len = hlen + firstlen;
263 	ip->ip_len = htons((u_short)m->m_pkthdr.len);
264 	ip->ip_off = htons((u_short)(ip->ip_off | IP_MF));
265 	ip->ip_sum = 0;
266 	ip->ip_sum = in_cksum(m, hlen);
267 sendorfree:
268 	for (m = m0; m; m = m0) {
269 		m0 = m->m_nextpkt;
270 		m->m_nextpkt = 0;
271 		if (error == 0)
272 			error = (*ifp->if_output)(ifp, m,
273 			    (struct sockaddr *)dst, ro->ro_rt);
274 		else
275 			m_freem(m);
276 	}
277     }
278 done:
279 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt)
280 		RTFREE(ro->ro_rt);
281 	return (error);
282 bad:
283 	m_freem(m0);
284 	goto done;
285 }
286 
287 /*
288  * Insert IP options into preformed packet.
289  * Adjust IP destination as required for IP source routing,
290  * as indicated by a non-zero in_addr at the start of the options.
291  */
292 struct mbuf *
293 ip_insertoptions(m, opt, phlen)
294 	register struct mbuf *m;
295 	struct mbuf *opt;
296 	int *phlen;
297 {
298 	register struct ipoption *p = mtod(opt, struct ipoption *);
299 	struct mbuf *n;
300 	register struct ip *ip = mtod(m, struct ip *);
301 	unsigned optlen;
302 
303 	optlen = opt->m_len - sizeof(p->ipopt_dst);
304 	if (optlen + (u_short)ip->ip_len > IP_MAXPACKET)
305 		return (m);		/* XXX should fail */
306 	if (p->ipopt_dst.s_addr)
307 		ip->ip_dst = p->ipopt_dst;
308 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
309 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
310 		if (n == 0)
311 			return (m);
312 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
313 		m->m_len -= sizeof(struct ip);
314 		m->m_data += sizeof(struct ip);
315 		n->m_next = m;
316 		m = n;
317 		m->m_len = optlen + sizeof(struct ip);
318 		m->m_data += max_linkhdr;
319 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
320 	} else {
321 		m->m_data -= optlen;
322 		m->m_len += optlen;
323 		m->m_pkthdr.len += optlen;
324 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
325 	}
326 	ip = mtod(m, struct ip *);
327 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
328 	*phlen = sizeof(struct ip) + optlen;
329 	ip->ip_len += optlen;
330 	return (m);
331 }
332 
333 /*
334  * Copy options from ip to jp,
335  * omitting those not copied during fragmentation.
336  */
337 ip_optcopy(ip, jp)
338 	struct ip *ip, *jp;
339 {
340 	register u_char *cp, *dp;
341 	int opt, optlen, cnt;
342 
343 	cp = (u_char *)(ip + 1);
344 	dp = (u_char *)(jp + 1);
345 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
346 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
347 		opt = cp[0];
348 		if (opt == IPOPT_EOL)
349 			break;
350 		if (opt == IPOPT_NOP)
351 			optlen = 1;
352 		else
353 			optlen = cp[IPOPT_OLEN];
354 		/* bogus lengths should have been caught by ip_dooptions */
355 		if (optlen > cnt)
356 			optlen = cnt;
357 		if (IPOPT_COPIED(opt)) {
358 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
359 			dp += optlen;
360 		}
361 	}
362 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
363 		*dp++ = IPOPT_EOL;
364 	return (optlen);
365 }
366 
367 /*
368  * IP socket option processing.
369  */
370 ip_ctloutput(op, so, level, optname, mp)
371 	int op;
372 	struct socket *so;
373 	int level, optname;
374 	struct mbuf **mp;
375 {
376 	register struct inpcb *inp = sotoinpcb(so);
377 	register struct mbuf *m = *mp;
378 	register int optval;
379 	int error = 0;
380 
381 	if (level != IPPROTO_IP)
382 		error = EINVAL;
383 	else switch (op) {
384 
385 	case PRCO_SETOPT:
386 		switch (optname) {
387 		case IP_OPTIONS:
388 #ifdef notyet
389 		case IP_RETOPTS:
390 			return (ip_pcbopts(optname, &inp->inp_options, m));
391 #else
392 			return (ip_pcbopts(&inp->inp_options, m));
393 #endif
394 
395 		case IP_TOS:
396 		case IP_TTL:
397 		case IP_RECVOPTS:
398 		case IP_RECVRETOPTS:
399 		case IP_RECVDSTADDR:
400 			if (m->m_len != sizeof(int))
401 				error = EINVAL;
402 			else {
403 				optval = *mtod(m, int *);
404 				switch (optname) {
405 
406 				case IP_TOS:
407 					inp->inp_ip.ip_tos = optval;
408 					break;
409 
410 				case IP_TTL:
411 					inp->inp_ip.ip_ttl = optval;
412 					break;
413 #define	OPTSET(bit) \
414 	if (optval) \
415 		inp->inp_flags |= bit; \
416 	else \
417 		inp->inp_flags &= ~bit;
418 
419 				case IP_RECVOPTS:
420 					OPTSET(INP_RECVOPTS);
421 					break;
422 
423 				case IP_RECVRETOPTS:
424 					OPTSET(INP_RECVRETOPTS);
425 					break;
426 
427 				case IP_RECVDSTADDR:
428 					OPTSET(INP_RECVDSTADDR);
429 					break;
430 				}
431 			}
432 			break;
433 #undef OPTSET
434 
435 		default:
436 			error = EINVAL;
437 			break;
438 		}
439 		if (m)
440 			(void)m_free(m);
441 		break;
442 
443 	case PRCO_GETOPT:
444 		switch (optname) {
445 		case IP_OPTIONS:
446 		case IP_RETOPTS:
447 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
448 			if (inp->inp_options) {
449 				m->m_len = inp->inp_options->m_len;
450 				bcopy(mtod(inp->inp_options, caddr_t),
451 				    mtod(m, caddr_t), (unsigned)m->m_len);
452 			} else
453 				m->m_len = 0;
454 			break;
455 
456 		case IP_TOS:
457 		case IP_TTL:
458 		case IP_RECVOPTS:
459 		case IP_RECVRETOPTS:
460 		case IP_RECVDSTADDR:
461 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
462 			m->m_len = sizeof(int);
463 			switch (optname) {
464 
465 			case IP_TOS:
466 				optval = inp->inp_ip.ip_tos;
467 				break;
468 
469 			case IP_TTL:
470 				optval = inp->inp_ip.ip_ttl;
471 				break;
472 
473 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
474 
475 			case IP_RECVOPTS:
476 				optval = OPTBIT(INP_RECVOPTS);
477 				break;
478 
479 			case IP_RECVRETOPTS:
480 				optval = OPTBIT(INP_RECVRETOPTS);
481 				break;
482 
483 			case IP_RECVDSTADDR:
484 				optval = OPTBIT(INP_RECVDSTADDR);
485 				break;
486 			}
487 			*mtod(m, int *) = optval;
488 			break;
489 
490 		default:
491 			error = EINVAL;
492 			break;
493 		}
494 		break;
495 	}
496 	return (error);
497 }
498 
499 /*
500  * Set up IP options in pcb for insertion in output packets.
501  * Store in mbuf with pointer in pcbopt, adding pseudo-option
502  * with destination address if source routed.
503  */
504 #ifdef notyet
505 ip_pcbopts(optname, pcbopt, m)
506 	int optname;
507 #else
508 ip_pcbopts(pcbopt, m)
509 #endif
510 	struct mbuf **pcbopt;
511 	register struct mbuf *m;
512 {
513 	register cnt, optlen;
514 	register u_char *cp;
515 	u_char opt;
516 
517 	/* turn off any old options */
518 	if (*pcbopt)
519 		(void)m_free(*pcbopt);
520 	*pcbopt = 0;
521 	if (m == (struct mbuf *)0 || m->m_len == 0) {
522 		/*
523 		 * Only turning off any previous options.
524 		 */
525 		if (m)
526 			(void)m_free(m);
527 		return (0);
528 	}
529 
530 #ifndef	vax
531 	if (m->m_len % sizeof(long))
532 		goto bad;
533 #endif
534 	/*
535 	 * IP first-hop destination address will be stored before
536 	 * actual options; move other options back
537 	 * and clear it when none present.
538 	 */
539 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
540 		goto bad;
541 	cnt = m->m_len;
542 	m->m_len += sizeof(struct in_addr);
543 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
544 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
545 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
546 
547 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
548 		opt = cp[IPOPT_OPTVAL];
549 		if (opt == IPOPT_EOL)
550 			break;
551 		if (opt == IPOPT_NOP)
552 			optlen = 1;
553 		else {
554 			optlen = cp[IPOPT_OLEN];
555 			if (optlen <= IPOPT_OLEN || optlen > cnt)
556 				goto bad;
557 		}
558 		switch (opt) {
559 
560 		default:
561 			break;
562 
563 		case IPOPT_LSRR:
564 		case IPOPT_SSRR:
565 			/*
566 			 * user process specifies route as:
567 			 *	->A->B->C->D
568 			 * D must be our final destination (but we can't
569 			 * check that since we may not have connected yet).
570 			 * A is first hop destination, which doesn't appear in
571 			 * actual IP option, but is stored before the options.
572 			 */
573 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
574 				goto bad;
575 			m->m_len -= sizeof(struct in_addr);
576 			cnt -= sizeof(struct in_addr);
577 			optlen -= sizeof(struct in_addr);
578 			cp[IPOPT_OLEN] = optlen;
579 			/*
580 			 * Move first hop before start of options.
581 			 */
582 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
583 			    sizeof(struct in_addr));
584 			/*
585 			 * Then copy rest of options back
586 			 * to close up the deleted entry.
587 			 */
588 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
589 			    sizeof(struct in_addr)),
590 			    (caddr_t)&cp[IPOPT_OFFSET+1],
591 			    (unsigned)cnt + sizeof(struct in_addr));
592 			break;
593 		}
594 	}
595 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
596 		goto bad;
597 	*pcbopt = m;
598 	return (0);
599 
600 bad:
601 	(void)m_free(m);
602 	return (EINVAL);
603 }
604