xref: /original-bsd/sys/netinet/ip_output.c (revision 7717c4d4)
1 /*
2  * Copyright (c) 1982, 1986, 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  *	@(#)ip_output.c	7.18 (Berkeley) 04/05/90
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 || (ifp = ro->ro_rt->rt_ifp) == 0) {
125 			if (in_localaddr(ip->ip_dst))
126 				error = EHOSTUNREACH;
127 			else
128 				error = ENETUNREACH;
129 			goto bad;
130 		}
131 		ia = (struct in_ifaddr *)ro->ro_rt->rt_ifa;
132 		ro->ro_rt->rt_use++;
133 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
134 			dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
135 	}
136 #ifndef notdef
137 	/*
138 	 * If source address not specified yet, use address
139 	 * of outgoing interface.
140 	 */
141 	if (ip->ip_src.s_addr == INADDR_ANY)
142 		ip->ip_src = IA_SIN(ia)->sin_addr;
143 #endif
144 	/*
145 	 * Look for broadcast address and
146 	 * and verify user is allowed to send
147 	 * such a packet.
148 	 */
149 	if (in_broadcast(dst->sin_addr)) {
150 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
151 			error = EADDRNOTAVAIL;
152 			goto bad;
153 		}
154 		if ((flags & IP_ALLOWBROADCAST) == 0) {
155 			error = EACCES;
156 			goto bad;
157 		}
158 		/* don't allow broadcast messages to be fragmented */
159 		if (ip->ip_len > ifp->if_mtu) {
160 			error = EMSGSIZE;
161 			goto bad;
162 		}
163 		m->m_flags |= M_BCAST;
164 	}
165 
166 	/*
167 	 * If small enough for interface, can just send directly.
168 	 */
169 	if (ip->ip_len <= ifp->if_mtu) {
170 		ip->ip_len = htons((u_short)ip->ip_len);
171 		ip->ip_off = htons((u_short)ip->ip_off);
172 		ip->ip_sum = 0;
173 		ip->ip_sum = in_cksum(m, hlen);
174 		error = (*ifp->if_output)(ifp, m,
175 				(struct sockaddr *)dst, ro->ro_rt);
176 		goto done;
177 	}
178 	ipstat.ips_fragmented++;
179 	/*
180 	 * Too large for interface; fragment if possible.
181 	 * Must be able to put at least 8 bytes per fragment.
182 	 */
183 	if (ip->ip_off & IP_DF) {
184 		error = EMSGSIZE;
185 		goto bad;
186 	}
187 	len = (ifp->if_mtu - hlen) &~ 7;
188 	if (len < 8) {
189 		error = EMSGSIZE;
190 		goto bad;
191 	}
192 
193     {
194 	int mhlen, firstlen = len;
195 	struct mbuf **mnext = &m->m_nextpkt;
196 
197 	/*
198 	 * Loop through length of segment after first fragment,
199 	 * make new header and copy data of each part and link onto chain.
200 	 */
201 	m0 = m;
202 	mhlen = sizeof (struct ip);
203 	for (off = hlen + len; off < ip->ip_len; off += len) {
204 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
205 		if (m == 0) {
206 			error = ENOBUFS;
207 			goto sendorfree;
208 		}
209 		m->m_data += max_linkhdr;
210 		mhip = mtod(m, struct ip *);
211 		*mhip = *ip;
212 		if (hlen > sizeof (struct ip)) {
213 			mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
214 			mhip->ip_hl = mhlen >> 2;
215 		}
216 		m->m_len = mhlen;
217 		mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF);
218 		if (ip->ip_off & IP_MF)
219 			mhip->ip_off |= IP_MF;
220 		if (off + len >= ip->ip_len)
221 			len = ip->ip_len - off;
222 		else
223 			mhip->ip_off |= IP_MF;
224 		mhip->ip_len = htons((u_short)(len + mhlen));
225 		m->m_next = m_copy(m0, off, len);
226 		if (m->m_next == 0) {
227 			error = ENOBUFS;	/* ??? */
228 			goto sendorfree;
229 		}
230 		m->m_pkthdr.len = mhlen + len;
231 		m->m_pkthdr.rcvif = (struct ifnet *)0;
232 		mhip->ip_off = htons((u_short)mhip->ip_off);
233 		mhip->ip_sum = 0;
234 		mhip->ip_sum = in_cksum(m, mhlen);
235 		*mnext = m;
236 		mnext = &m->m_nextpkt;
237 		ipstat.ips_ofragments++;
238 	}
239 	/*
240 	 * Update first fragment by trimming what's been copied out
241 	 * and updating header, then send each fragment (in order).
242 	 */
243 	m = m0;
244 	m_adj(m, hlen + firstlen - ip->ip_len);
245 	m->m_pkthdr.len = hlen + firstlen;
246 	ip->ip_len = htons((u_short)m->m_pkthdr.len);
247 	ip->ip_off = htons((u_short)(ip->ip_off | IP_MF));
248 	ip->ip_sum = 0;
249 	ip->ip_sum = in_cksum(m, hlen);
250 sendorfree:
251 	for (m = m0; m; m = m0) {
252 		m0 = m->m_nextpkt;
253 		m->m_nextpkt = 0;
254 		if (error == 0)
255 			error = (*ifp->if_output)(ifp, m,
256 			    (struct sockaddr *)dst, ro->ro_rt);
257 		else
258 			m_freem(m);
259 	}
260     }
261 done:
262 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt)
263 		RTFREE(ro->ro_rt);
264 	return (error);
265 bad:
266 	m_freem(m0);
267 	goto done;
268 }
269 
270 /*
271  * Insert IP options into preformed packet.
272  * Adjust IP destination as required for IP source routing,
273  * as indicated by a non-zero in_addr at the start of the options.
274  */
275 struct mbuf *
276 ip_insertoptions(m, opt, phlen)
277 	register struct mbuf *m;
278 	struct mbuf *opt;
279 	int *phlen;
280 {
281 	register struct ipoption *p = mtod(opt, struct ipoption *);
282 	struct mbuf *n;
283 	register struct ip *ip = mtod(m, struct ip *);
284 	unsigned optlen;
285 
286 	optlen = opt->m_len - sizeof(p->ipopt_dst);
287 	if (p->ipopt_dst.s_addr)
288 		ip->ip_dst = p->ipopt_dst;
289 	if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
290 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
291 		if (n == 0)
292 			return (m);
293 		n->m_pkthdr.len = m->m_pkthdr.len + optlen;
294 		m->m_len -= sizeof(struct ip);
295 		m->m_data += sizeof(struct ip);
296 		n->m_next = m;
297 		m = n;
298 		m->m_len = optlen + sizeof(struct ip);
299 		m->m_data += max_linkhdr;
300 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
301 	} else {
302 		m->m_data -= optlen;
303 		m->m_len += optlen;
304 		m->m_pkthdr.len += optlen;
305 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
306 	}
307 	ip = mtod(m, struct ip *);
308 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
309 	*phlen = sizeof(struct ip) + optlen;
310 	ip->ip_len += optlen;
311 	return (m);
312 }
313 
314 /*
315  * Copy options from ip to jp,
316  * omitting those not copied during fragmentation.
317  */
318 ip_optcopy(ip, jp)
319 	struct ip *ip, *jp;
320 {
321 	register u_char *cp, *dp;
322 	int opt, optlen, cnt;
323 
324 	cp = (u_char *)(ip + 1);
325 	dp = (u_char *)(jp + 1);
326 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
327 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
328 		opt = cp[0];
329 		if (opt == IPOPT_EOL)
330 			break;
331 		if (opt == IPOPT_NOP)
332 			optlen = 1;
333 		else
334 			optlen = cp[IPOPT_OLEN];
335 		/* bogus lengths should have been caught by ip_dooptions */
336 		if (optlen > cnt)
337 			optlen = cnt;
338 		if (IPOPT_COPIED(opt)) {
339 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
340 			dp += optlen;
341 		}
342 	}
343 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
344 		*dp++ = IPOPT_EOL;
345 	return (optlen);
346 }
347 
348 /*
349  * IP socket option processing.
350  */
351 ip_ctloutput(op, so, level, optname, m)
352 	int op;
353 	struct socket *so;
354 	int level, optname;
355 	struct mbuf **m;
356 {
357 	int error = 0;
358 	struct inpcb *inp = sotoinpcb(so);
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 			return (ip_pcbopts(&inp->inp_options, *m));
368 
369 		default:
370 			error = EINVAL;
371 			break;
372 		}
373 		break;
374 
375 	case PRCO_GETOPT:
376 		switch (optname) {
377 		case IP_OPTIONS:
378 			*m = m_get(M_WAIT, MT_SOOPTS);
379 			if (inp->inp_options) {
380 				(*m)->m_len = inp->inp_options->m_len;
381 				bcopy(mtod(inp->inp_options, caddr_t),
382 				    mtod(*m, caddr_t), (unsigned)(*m)->m_len);
383 			} else
384 				(*m)->m_len = 0;
385 			break;
386 		default:
387 			error = EINVAL;
388 			break;
389 		}
390 		break;
391 	}
392 	if (op == PRCO_SETOPT && *m)
393 		(void)m_free(*m);
394 	return (error);
395 }
396 
397 /*
398  * Set up IP options in pcb for insertion in output packets.
399  * Store in mbuf with pointer in pcbopt, adding pseudo-option
400  * with destination address if source routed.
401  */
402 ip_pcbopts(pcbopt, m)
403 	struct mbuf **pcbopt;
404 	register struct mbuf *m;
405 {
406 	register cnt, optlen;
407 	register u_char *cp;
408 	u_char opt;
409 
410 	/* turn off any old options */
411 	if (*pcbopt)
412 		(void)m_free(*pcbopt);
413 	*pcbopt = 0;
414 	if (m == (struct mbuf *)0 || m->m_len == 0) {
415 		/*
416 		 * Only turning off any previous options.
417 		 */
418 		if (m)
419 			(void)m_free(m);
420 		return (0);
421 	}
422 
423 #ifndef	vax
424 	if (m->m_len % sizeof(long))
425 		goto bad;
426 #endif
427 	/*
428 	 * IP first-hop destination address will be stored before
429 	 * actual options; move other options back
430 	 * and clear it when none present.
431 	 */
432 	if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
433 		goto bad;
434 	cnt = m->m_len;
435 	m->m_len += sizeof(struct in_addr);
436 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
437 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
438 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
439 
440 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
441 		opt = cp[IPOPT_OPTVAL];
442 		if (opt == IPOPT_EOL)
443 			break;
444 		if (opt == IPOPT_NOP)
445 			optlen = 1;
446 		else {
447 			optlen = cp[IPOPT_OLEN];
448 			if (optlen <= IPOPT_OLEN || optlen > cnt)
449 				goto bad;
450 		}
451 		switch (opt) {
452 
453 		default:
454 			break;
455 
456 		case IPOPT_LSRR:
457 		case IPOPT_SSRR:
458 			/*
459 			 * user process specifies route as:
460 			 *	->A->B->C->D
461 			 * D must be our final destination (but we can't
462 			 * check that since we may not have connected yet).
463 			 * A is first hop destination, which doesn't appear in
464 			 * actual IP option, but is stored before the options.
465 			 */
466 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
467 				goto bad;
468 			m->m_len -= sizeof(struct in_addr);
469 			cnt -= sizeof(struct in_addr);
470 			optlen -= sizeof(struct in_addr);
471 			cp[IPOPT_OLEN] = optlen;
472 			/*
473 			 * Move first hop before start of options.
474 			 */
475 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
476 			    sizeof(struct in_addr));
477 			/*
478 			 * Then copy rest of options back
479 			 * to close up the deleted entry.
480 			 */
481 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
482 			    sizeof(struct in_addr)),
483 			    (caddr_t)&cp[IPOPT_OFFSET+1],
484 			    (unsigned)cnt + sizeof(struct in_addr));
485 			break;
486 		}
487 	}
488 	if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
489 		goto bad;
490 	*pcbopt = m;
491 	return (0);
492 
493 bad:
494 	(void)m_free(m);
495 	return (EINVAL);
496 }
497