xref: /csrg-svn/sys/netinet/ip_output.c (revision 44371)
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 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  *	@(#)ip_output.c	7.19 (Berkeley) 06/28/90
18  */
19 
20 #include "param.h"
21 #include "malloc.h"
22 #include "mbuf.h"
23 #include "errno.h"
24 #include "protosw.h"
25 #include "socket.h"
26 #include "socketvar.h"
27 
28 #include "../net/if.h"
29 #include "../net/route.h"
30 
31 #include "in.h"
32 #include "in_systm.h"
33 #include "ip.h"
34 #include "in_pcb.h"
35 #include "in_var.h"
36 #include "ip_var.h"
37 
38 #ifdef vax
39 #include "machine/mtpr.h"
40 #endif
41 
42 struct mbuf *ip_insertoptions();
43 
44 /*
45  * IP output.  The packet in mbuf chain m contains a skeletal IP
46  * header (with len, off, ttl, proto, tos, src, dst).
47  * The mbuf chain containing the packet will be freed.
48  * The mbuf opt, if present, will not be freed.
49  */
50 ip_output(m0, opt, ro, flags)
51 	struct mbuf *m0;
52 	struct mbuf *opt;
53 	struct route *ro;
54 	int flags;
55 {
56 	register struct ip *ip, *mhip;
57 	register struct ifnet *ifp;
58 	register struct mbuf *m = m0;
59 	register int hlen = sizeof (struct ip);
60 	int len, off, error = 0;
61 	struct route iproute;
62 	struct sockaddr_in *dst;
63 	struct in_ifaddr *ia;
64 
65 if ((m->m_flags & M_PKTHDR) == 0)
66 panic("ip_output no HDR");
67 	if (opt) {
68 		m = ip_insertoptions(m, opt, &len);
69 		hlen = len;
70 	}
71 	ip = mtod(m, struct ip *);
72 	/*
73 	 * Fill in IP header.
74 	 */
75 	if ((flags & IP_FORWARDING) == 0) {
76 		ip->ip_v = IPVERSION;
77 		ip->ip_off &= IP_DF;
78 		ip->ip_id = htons(ip_id++);
79 		ip->ip_hl = hlen >> 2;
80 	} else {
81 		hlen = ip->ip_hl << 2;
82 		ipstat.ips_localout++;
83 	}
84 	/*
85 	 * Route packet.
86 	 */
87 	if (ro == 0) {
88 		ro = &iproute;
89 		bzero((caddr_t)ro, sizeof (*ro));
90 	}
91 	dst = (struct sockaddr_in *)&ro->ro_dst;
92 	/*
93 	 * If there is a cached route,
94 	 * check that it is to the same destination
95 	 * and is still up.  If not, free it and try again.
96 	 */
97 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
98 	   dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
99 		RTFREE(ro->ro_rt);
100 		ro->ro_rt = (struct rtentry *)0;
101 	}
102 	if (ro->ro_rt == 0) {
103 		dst->sin_family = AF_INET;
104 		dst->sin_len = sizeof(*dst);
105 		dst->sin_addr = ip->ip_dst;
106 	}
107 	/*
108 	 * If routing to interface only,
109 	 * short circuit routing lookup.
110 	 */
111 	if (flags & IP_ROUTETOIF) {
112 
113 		ia = (struct in_ifaddr *)ifa_ifwithdstaddr((struct sockaddr *)dst);
114 		if (ia == 0)
115 			ia = in_iaonnetof(in_netof(ip->ip_dst));
116 		if (ia == 0) {
117 			error = ENETUNREACH;
118 			goto bad;
119 		}
120 		ifp = ia->ia_ifp;
121 	} else {
122 		if (ro->ro_rt == 0)
123 			rtalloc(ro);
124 		if (ro->ro_rt == 0) {
125 			error = EHOSTUNREACH;
126 			goto bad;
127 		}
128 		ia = (struct in_ifaddr *)ro->ro_rt->rt_ifa;
129 		ifp = ro->ro_rt->rt_ifp;
130 		ro->ro_rt->rt_use++;
131 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
132 			dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
133 	}
134 #ifndef notdef
135 	/*
136 	 * If source address not specified yet, use address
137 	 * of outgoing interface.
138 	 */
139 	if (ip->ip_src.s_addr == INADDR_ANY)
140 		ip->ip_src = IA_SIN(ia)->sin_addr;
141 #endif
142 	/*
143 	 * Look for broadcast address and
144 	 * and verify user is allowed to send
145 	 * such a packet.
146 	 */
147 	if (in_broadcast(dst->sin_addr)) {
148 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
149 			error = EADDRNOTAVAIL;
150 			goto bad;
151 		}
152 		if ((flags & IP_ALLOWBROADCAST) == 0) {
153 			error = EACCES;
154 			goto bad;
155 		}
156 		/* don't allow broadcast messages to be fragmented */
157 		if (ip->ip_len > ifp->if_mtu) {
158 			error = EMSGSIZE;
159 			goto bad;
160 		}
161 		m->m_flags |= M_BCAST;
162 	}
163 
164 	/*
165 	 * If small enough for interface, can just send directly.
166 	 */
167 	if (ip->ip_len <= ifp->if_mtu) {
168 		ip->ip_len = htons((u_short)ip->ip_len);
169 		ip->ip_off = htons((u_short)ip->ip_off);
170 		ip->ip_sum = 0;
171 		ip->ip_sum = in_cksum(m, hlen);
172 		error = (*ifp->if_output)(ifp, m,
173 				(struct sockaddr *)dst, ro->ro_rt);
174 		goto done;
175 	}
176 	ipstat.ips_fragmented++;
177 	/*
178 	 * Too large for interface; fragment if possible.
179 	 * Must be able to put at least 8 bytes per fragment.
180 	 */
181 	if (ip->ip_off & IP_DF) {
182 		error = EMSGSIZE;
183 		goto bad;
184 	}
185 	len = (ifp->if_mtu - hlen) &~ 7;
186 	if (len < 8) {
187 		error = EMSGSIZE;
188 		goto bad;
189 	}
190 
191     {
192 	int mhlen, firstlen = len;
193 	struct mbuf **mnext = &m->m_nextpkt;
194 
195 	/*
196 	 * Loop through length of segment after first fragment,
197 	 * make new header and copy data of each part and link onto chain.
198 	 */
199 	m0 = m;
200 	mhlen = sizeof (struct ip);
201 	for (off = hlen + len; off < ip->ip_len; off += len) {
202 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
203 		if (m == 0) {
204 			error = ENOBUFS;
205 			goto sendorfree;
206 		}
207 		m->m_data += max_linkhdr;
208 		mhip = mtod(m, struct ip *);
209 		*mhip = *ip;
210 		if (hlen > sizeof (struct ip)) {
211 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
212 			mhip->ip_hl = mhlen >> 2;
213 		}
214 		m->m_len = mhlen;
215 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
216 		if (ip->ip_off & IP_MF)
217 			mhip->ip_off |= IP_MF;
218 		if (off + len >= ip->ip_len)
219 			len = ip->ip_len - off;
220 		else
221 			mhip->ip_off |= IP_MF;
222 		mhip->ip_len = htons((u_short)(len + mhlen));
223 		m->m_next = m_copy(m0, off, len);
224 		if (m->m_next == 0) {
225 			error = ENOBUFS;	/* ??? */
226 			goto sendorfree;
227 		}
228 		m->m_pkthdr.len = mhlen + len;
229 		m->m_pkthdr.rcvif = (struct ifnet *)0;
230 		mhip->ip_off = htons((u_short)mhip->ip_off);
231 		mhip->ip_sum = 0;
232 		mhip->ip_sum = in_cksum(m, mhlen);
233 		*mnext = m;
234 		mnext = &m->m_nextpkt;
235 		ipstat.ips_ofragments++;
236 	}
237 	/*
238 	 * Update first fragment by trimming what's been copied out
239 	 * and updating header, then send each fragment (in order).
240 	 */
241 	m = m0;
242 	m_adj(m, hlen + firstlen - ip->ip_len);
243 	m->m_pkthdr.len = hlen + firstlen;
244 	ip->ip_len = htons((u_short)m->m_pkthdr.len);
245 	ip->ip_off = htons((u_short)(ip->ip_off | IP_MF));
246 	ip->ip_sum = 0;
247 	ip->ip_sum = in_cksum(m, hlen);
248 sendorfree:
249 	for (m = m0; m; m = m0) {
250 		m0 = m->m_nextpkt;
251 		m->m_nextpkt = 0;
252 		if (error == 0)
253 			error = (*ifp->if_output)(ifp, m,
254 			    (struct sockaddr *)dst, ro->ro_rt);
255 		else
256 			m_freem(m);
257 	}
258     }
259 done:
260 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt)
261 		RTFREE(ro->ro_rt);
262 	return (error);
263 bad:
264 	m_freem(m0);
265 	goto done;
266 }
267 
268 /*
269  * Insert IP options into preformed packet.
270  * Adjust IP destination as required for IP source routing,
271  * as indicated by a non-zero in_addr at the start of the options.
272  */
273 struct mbuf *
274 ip_insertoptions(m, opt, phlen)
275 	register struct mbuf *m;
276 	struct mbuf *opt;
277 	int *phlen;
278 {
279 	register struct ipoption *p = mtod(opt, struct ipoption *);
280 	struct mbuf *n;
281 	register struct ip *ip = mtod(m, struct ip *);
282 	unsigned optlen;
283 
284 	optlen = opt->m_len - sizeof(p->ipopt_dst);
285 	if (p->ipopt_dst.s_addr)
286 		ip->ip_dst = p->ipopt_dst;
287 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
288 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
289 		if (n == 0)
290 			return (m);
291 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
292 		m->m_len -= sizeof(struct ip);
293 		m->m_data += sizeof(struct ip);
294 		n->m_next = m;
295 		m = n;
296 		m->m_len = optlen + sizeof(struct ip);
297 		m->m_data += max_linkhdr;
298 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
299 	} else {
300 		m->m_data -= optlen;
301 		m->m_len += optlen;
302 		m->m_pkthdr.len += optlen;
303 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
304 	}
305 	ip = mtod(m, struct ip *);
306 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
307 	*phlen = sizeof(struct ip) + optlen;
308 	ip->ip_len += optlen;
309 	return (m);
310 }
311 
312 /*
313  * Copy options from ip to jp,
314  * omitting those not copied during fragmentation.
315  */
316 ip_optcopy(ip, jp)
317 	struct ip *ip, *jp;
318 {
319 	register u_char *cp, *dp;
320 	int opt, optlen, cnt;
321 
322 	cp = (u_char *)(ip + 1);
323 	dp = (u_char *)(jp + 1);
324 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
325 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
326 		opt = cp[0];
327 		if (opt == IPOPT_EOL)
328 			break;
329 		if (opt == IPOPT_NOP)
330 			optlen = 1;
331 		else
332 			optlen = cp[IPOPT_OLEN];
333 		/* bogus lengths should have been caught by ip_dooptions */
334 		if (optlen > cnt)
335 			optlen = cnt;
336 		if (IPOPT_COPIED(opt)) {
337 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
338 			dp += optlen;
339 		}
340 	}
341 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
342 		*dp++ = IPOPT_EOL;
343 	return (optlen);
344 }
345 
346 /*
347  * IP socket option processing.
348  */
349 ip_ctloutput(op, so, level, optname, mp)
350 	int op;
351 	struct socket *so;
352 	int level, optname;
353 	struct mbuf **mp;
354 {
355 	register struct inpcb *inp = sotoinpcb(so);
356 	register struct mbuf *m = *mp;
357 	register int optval;
358 	int error = 0;
359 
360 	if (level != IPPROTO_IP)
361 		error = EINVAL;
362 	else switch (op) {
363 
364 	case PRCO_SETOPT:
365 		switch (optname) {
366 		case IP_OPTIONS:
367 		case IP_RETOPTS:
368 			return (ip_pcbopts(optname, &inp->inp_options, m));
369 
370 		case IP_TOS:
371 		case IP_TTL:
372 		case IP_RECVOPTS:
373 		case IP_RECVRETOPTS:
374 		case IP_RECVDSTADDR:
375 			if (m->m_len != sizeof(int))
376 				error = EINVAL;
377 			else {
378 				optval = *mtod(m, int *);
379 				switch (op) {
380 
381 				case IP_TOS:
382 					inp->inp_ip.ip_tos = optval;
383 					break;
384 
385 				case IP_TTL:
386 					inp->inp_ip.ip_tos = optval;
387 					break;
388 #define	OPTSET(bit) \
389 	if (optval) \
390 		inp->inp_flags |= bit; \
391 	else \
392 		inp->inp_flags &= ~bit;
393 
394 				case IP_RECVOPTS:
395 					OPTSET(INP_RECVOPTS);
396 					break;
397 
398 				case IP_RECVRETOPTS:
399 					OPTSET(INP_RECVRETOPTS);
400 					break;
401 
402 				case IP_RECVDSTADDR:
403 					OPTSET(INP_RECVDSTADDR);
404 					break;
405 				}
406 			}
407 			break;
408 #undef OPTSET
409 
410 		default:
411 			error = EINVAL;
412 			break;
413 		}
414 		if (m)
415 			(void)m_free(m);
416 		break;
417 
418 	case PRCO_GETOPT:
419 		switch (optname) {
420 		case IP_OPTIONS:
421 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
422 			if (inp->inp_options) {
423 				m->m_len = inp->inp_options->m_len;
424 				bcopy(mtod(inp->inp_options, caddr_t),
425 				    mtod(m, caddr_t), (unsigned)m->m_len);
426 			} else
427 				m->m_len = 0;
428 			break;
429 
430 		case IP_TOS:
431 		case IP_TTL:
432 		case IP_RECVOPTS:
433 		case IP_RECVRETOPTS:
434 		case IP_RECVDSTADDR:
435 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
436 			m->m_len = sizeof(int);
437 			switch (op) {
438 
439 			case IP_TOS:
440 				optval = inp->inp_ip.ip_tos;
441 				break;
442 
443 			case IP_TTL:
444 				optval = inp->inp_ip.ip_tos;
445 				break;
446 
447 #define	OPTBIT(bit)	(inp->inp_flags & bit ? 1 : 0)
448 
449 			case IP_RECVOPTS:
450 				optval = OPTBIT(INP_RECVOPTS);
451 				break;
452 
453 			case IP_RECVRETOPTS:
454 				optval = OPTBIT(INP_RECVRETOPTS);
455 				break;
456 
457 			case IP_RECVDSTADDR:
458 				optval = OPTBIT(INP_RECVDSTADDR);
459 				break;
460 			}
461 			*mtod(m, int *) = optval;
462 			break;
463 
464 		default:
465 			error = EINVAL;
466 			break;
467 		}
468 		break;
469 	}
470 	return (error);
471 }
472 
473 /*
474  * Set up IP options in pcb for insertion in output packets.
475  * Store in mbuf with pointer in pcbopt, adding pseudo-option
476  * with destination address if source routed.
477  */
478 ip_pcbopts(pcbopt, m)
479 	struct mbuf **pcbopt;
480 	register struct mbuf *m;
481 {
482 	register cnt, optlen;
483 	register u_char *cp;
484 	u_char opt;
485 
486 	/* turn off any old options */
487 	if (*pcbopt)
488 		(void)m_free(*pcbopt);
489 	*pcbopt = 0;
490 	if (m == (struct mbuf *)0 || m->m_len == 0) {
491 		/*
492 		 * Only turning off any previous options.
493 		 */
494 		if (m)
495 			(void)m_free(m);
496 		return (0);
497 	}
498 
499 #ifndef	vax
500 	if (m->m_len % sizeof(long))
501 		goto bad;
502 #endif
503 	/*
504 	 * IP first-hop destination address will be stored before
505 	 * actual options; move other options back
506 	 * and clear it when none present.
507 	 */
508 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
509 		goto bad;
510 	cnt = m->m_len;
511 	m->m_len += sizeof(struct in_addr);
512 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
513 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
514 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
515 
516 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
517 		opt = cp[IPOPT_OPTVAL];
518 		if (opt == IPOPT_EOL)
519 			break;
520 		if (opt == IPOPT_NOP)
521 			optlen = 1;
522 		else {
523 			optlen = cp[IPOPT_OLEN];
524 			if (optlen <= IPOPT_OLEN || optlen > cnt)
525 				goto bad;
526 		}
527 		switch (opt) {
528 
529 		default:
530 			break;
531 
532 		case IPOPT_LSRR:
533 		case IPOPT_SSRR:
534 			/*
535 			 * user process specifies route as:
536 			 *	->A->B->C->D
537 			 * D must be our final destination (but we can't
538 			 * check that since we may not have connected yet).
539 			 * A is first hop destination, which doesn't appear in
540 			 * actual IP option, but is stored before the options.
541 			 */
542 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
543 				goto bad;
544 			m->m_len -= sizeof(struct in_addr);
545 			cnt -= sizeof(struct in_addr);
546 			optlen -= sizeof(struct in_addr);
547 			cp[IPOPT_OLEN] = optlen;
548 			/*
549 			 * Move first hop before start of options.
550 			 */
551 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
552 			    sizeof(struct in_addr));
553 			/*
554 			 * Then copy rest of options back
555 			 * to close up the deleted entry.
556 			 */
557 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
558 			    sizeof(struct in_addr)),
559 			    (caddr_t)&cp[IPOPT_OFFSET+1],
560 			    (unsigned)cnt + sizeof(struct in_addr));
561 			break;
562 		}
563 	}
564 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
565 		goto bad;
566 	*pcbopt = m;
567 	return (0);
568 
569 bad:
570 	(void)m_free(m);
571 	return (EINVAL);
572 }
573