xref: /csrg-svn/sys/netinet/ip_output.c (revision 29923)
1 /*
2  * Copyright (c) 1982, 1986 Regents of the University of California.
3  * All rights reserved.  The Berkeley software License Agreement
4  * specifies the terms and conditions for redistribution.
5  *
6  *	@(#)ip_output.c	7.2 (Berkeley) 10/28/86
7  */
8 
9 #include "param.h"
10 #include "mbuf.h"
11 #include "errno.h"
12 #include "protosw.h"
13 #include "socket.h"
14 #include "socketvar.h"
15 
16 #include "../net/if.h"
17 #include "../net/route.h"
18 
19 #include "in.h"
20 #include "in_pcb.h"
21 #include "in_systm.h"
22 #include "in_var.h"
23 #include "ip.h"
24 #include "ip_var.h"
25 
26 #ifdef vax
27 #include "../machine/mtpr.h"
28 #endif
29 
30 struct mbuf *ip_insertoptions();
31 
32 /*
33  * IP output.  The packet in mbuf chain m contains a skeletal IP
34  * header (as ipovly).  The mbuf chain containing the packet will
35  * be freed.  The mbuf opt, if present, will not be freed.
36  */
37 ip_output(m, opt, ro, flags)
38 	struct mbuf *m;
39 	struct mbuf *opt;
40 	struct route *ro;
41 	int flags;
42 {
43 	register struct ip *ip;
44 	register struct ifnet *ifp;
45 	int len, hlen = sizeof (struct ip), off, error = 0;
46 	struct route iproute;
47 	struct sockaddr_in *dst;
48 
49 	if (opt)
50 		m = ip_insertoptions(m, opt, &hlen);
51 	ip = mtod(m, struct ip *);
52 	/*
53 	 * Fill in IP header.
54 	 */
55 	if ((flags & IP_FORWARDING) == 0) {
56 		ip->ip_v = IPVERSION;
57 		ip->ip_off &= IP_DF;
58 		ip->ip_id = htons(ip_id++);
59 		ip->ip_hl = hlen >> 2;
60 	} else
61 		hlen = ip->ip_hl << 2;
62 
63 	/*
64 	 * Route packet.
65 	 */
66 	if (ro == 0) {
67 		ro = &iproute;
68 		bzero((caddr_t)ro, sizeof (*ro));
69 	}
70 	dst = (struct sockaddr_in *)&ro->ro_dst;
71 	/*
72 	 * If there is a cached route,
73 	 * check that it is to the same destination
74 	 * and is still up.  If not, free it and try again.
75 	 */
76 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
77 	   dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
78 		RTFREE(ro->ro_rt);
79 		ro->ro_rt = (struct rtentry *)0;
80 	}
81 	if (ro->ro_rt == 0) {
82 		dst->sin_family = AF_INET;
83 		dst->sin_addr = ip->ip_dst;
84 	}
85 	/*
86 	 * If routing to interface only,
87 	 * short circuit routing lookup.
88 	 */
89 	if (flags & IP_ROUTETOIF) {
90 		struct in_ifaddr *ia;
91 
92 		ia = (struct in_ifaddr *)ifa_ifwithdstaddr(dst);
93 		if (ia == 0)
94 			ia = in_iaonnetof(in_netof(ip->ip_dst));
95 		if (ia == 0) {
96 			error = ENETUNREACH;
97 			goto bad;
98 		}
99 		ifp = ia->ia_ifp;
100 	} else {
101 		if (ro->ro_rt == 0)
102 			rtalloc(ro);
103 		if (ro->ro_rt == 0 || (ifp = ro->ro_rt->rt_ifp) == 0) {
104 			error = ENETUNREACH;
105 			goto bad;
106 		}
107 		ro->ro_rt->rt_use++;
108 		if (ro->ro_rt->rt_flags & (RTF_GATEWAY|RTF_HOST))
109 			dst = (struct sockaddr_in *)&ro->ro_rt->rt_gateway;
110 	}
111 #ifndef notdef
112 	/*
113 	 * If source address not specified yet, use address
114 	 * of outgoing interface.
115 	 */
116 	if (ip->ip_src.s_addr == INADDR_ANY) {
117 		register struct in_ifaddr *ia;
118 
119 		for (ia = in_ifaddr; ia; ia = ia->ia_next)
120 			if (ia->ia_ifp == ifp) {
121 				ip->ip_src = IA_SIN(ia)->sin_addr;
122 				break;
123 			}
124 	}
125 #endif
126 	/*
127 	 * Look for broadcast address and
128 	 * and verify user is allowed to send
129 	 * such a packet.
130 	 */
131 	if (in_broadcast(dst->sin_addr)) {
132 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
133 			error = EADDRNOTAVAIL;
134 			goto bad;
135 		}
136 		if ((flags & IP_ALLOWBROADCAST) == 0) {
137 			error = EACCES;
138 			goto bad;
139 		}
140 		/* don't allow broadcast messages to be fragmented */
141 		if (ip->ip_len > ifp->if_mtu) {
142 			error = EMSGSIZE;
143 			goto bad;
144 		}
145 	}
146 
147 	/*
148 	 * If small enough for interface, can just send directly.
149 	 */
150 	if (ip->ip_len <= ifp->if_mtu) {
151 		ip->ip_len = htons((u_short)ip->ip_len);
152 		ip->ip_off = htons((u_short)ip->ip_off);
153 		ip->ip_sum = 0;
154 		ip->ip_sum = in_cksum(m, hlen);
155 		error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst);
156 		goto done;
157 	}
158 
159 	/*
160 	 * Too large for interface; fragment if possible.
161 	 * Must be able to put at least 8 bytes per fragment.
162 	 */
163 	if (ip->ip_off & IP_DF) {
164 		error = EMSGSIZE;
165 		goto bad;
166 	}
167 	len = (ifp->if_mtu - hlen) &~ 7;
168 	if (len < 8) {
169 		error = EMSGSIZE;
170 		goto bad;
171 	}
172 
173 	/*
174 	 * Discard IP header from logical mbuf for m_copy's sake.
175 	 * Loop through length of segment, make a copy of each
176 	 * part and output.
177 	 */
178 	m->m_len -= sizeof (struct ip);
179 	m->m_off += sizeof (struct ip);
180 	for (off = 0; off < ip->ip_len-hlen; off += len) {
181 		struct mbuf *mh = m_get(M_DONTWAIT, MT_HEADER);
182 		struct ip *mhip;
183 
184 		if (mh == 0) {
185 			error = ENOBUFS;
186 			goto bad;
187 		}
188 		mh->m_off = MMAXOFF - hlen;
189 		mhip = mtod(mh, struct ip *);
190 		*mhip = *ip;
191 		if (hlen > sizeof (struct ip)) {
192 			int olen = ip_optcopy(ip, mhip, off);
193 			mh->m_len = sizeof (struct ip) + olen;
194 		} else
195 			mh->m_len = sizeof (struct ip);
196 		mhip->ip_off = (off >> 3) + (ip->ip_off & ~IP_MF);
197 		if (ip->ip_off & IP_MF)
198 			mhip->ip_off |= IP_MF;
199 		if (off + len >= ip->ip_len-hlen)
200 			len = mhip->ip_len = ip->ip_len - hlen - off;
201 		else {
202 			mhip->ip_len = len;
203 			mhip->ip_off |= IP_MF;
204 		}
205 		mhip->ip_len += sizeof (struct ip);
206 		mhip->ip_len = htons((u_short)mhip->ip_len);
207 		mh->m_next = m_copy(m, off, len);
208 		if (mh->m_next == 0) {
209 			(void) m_free(mh);
210 			error = ENOBUFS;	/* ??? */
211 			goto bad;
212 		}
213 		mhip->ip_off = htons((u_short)mhip->ip_off);
214 		mhip->ip_sum = 0;
215 		mhip->ip_sum = in_cksum(mh, hlen);
216 		if (error = (*ifp->if_output)(ifp, mh, (struct sockaddr *)dst))
217 			break;
218 	}
219 bad:
220 	m_freem(m);
221 done:
222 	if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt)
223 		RTFREE(ro->ro_rt);
224 	return (error);
225 }
226 
227 /*
228  * Insert IP options into preformed packet.
229  * Adjust IP destination as required for IP source routing,
230  * as indicated by a non-zero in_addr at the start of the options.
231  */
232 struct mbuf *
233 ip_insertoptions(m, opt, phlen)
234 	register struct mbuf *m;
235 	struct mbuf *opt;
236 	int *phlen;
237 {
238 	register struct ipoption *p = mtod(opt, struct ipoption *);
239 	struct mbuf *n;
240 	register struct ip *ip = mtod(m, struct ip *);
241 	unsigned optlen;
242 
243 	optlen = opt->m_len - sizeof(p->ipopt_dst);
244 	if (p->ipopt_dst.s_addr)
245 		ip->ip_dst = p->ipopt_dst;
246 	if (m->m_off >= MMAXOFF || MMINOFF + optlen > m->m_off) {
247 		MGET(n, M_DONTWAIT, MT_HEADER);
248 		if (n == 0)
249 			return (m);
250 		m->m_len -= sizeof(struct ip);
251 		m->m_off += sizeof(struct ip);
252 		n->m_next = m;
253 		m = n;
254 		m->m_off = MMAXOFF - sizeof(struct ip) - optlen;
255 		m->m_len = optlen + sizeof(struct ip);
256 		bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
257 	} else {
258 		m->m_off -= optlen;
259 		m->m_len += optlen;
260 		ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
261 	}
262 	ip = mtod(m, struct ip *);
263 	bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
264 	*phlen = sizeof(struct ip) + optlen;
265 	ip->ip_len += optlen;
266 	return (m);
267 }
268 
269 /*
270  * Copy options from ip to jp.
271  * If off is 0 all options are copied
272  * otherwise copy selectively.
273  */
274 ip_optcopy(ip, jp, off)
275 	struct ip *ip, *jp;
276 	int off;
277 {
278 	register u_char *cp, *dp;
279 	int opt, optlen, cnt;
280 
281 	cp = (u_char *)(ip + 1);
282 	dp = (u_char *)(jp + 1);
283 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
284 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
285 		opt = cp[0];
286 		if (opt == IPOPT_EOL)
287 			break;
288 		if (opt == IPOPT_NOP)
289 			optlen = 1;
290 		else
291 			optlen = cp[IPOPT_OLEN];
292 		if (optlen > cnt)			/* XXX */
293 			optlen = cnt;			/* XXX */
294 		if (off == 0 || IPOPT_COPIED(opt)) {
295 			bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
296 			dp += optlen;
297 		}
298 	}
299 	for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
300 		*dp++ = IPOPT_EOL;
301 	return (optlen);
302 }
303 
304 /*
305  * IP socket option processing.
306  */
307 ip_ctloutput(op, so, level, optname, m)
308 	int op;
309 	struct socket *so;
310 	int level, optname;
311 	struct mbuf **m;
312 {
313 	int error = 0;
314 	struct inpcb *inp = sotoinpcb(so);
315 
316 	if (level != IPPROTO_IP)
317 		error = EINVAL;
318 	else switch (op) {
319 
320 	case PRCO_SETOPT:
321 		switch (optname) {
322 		case IP_OPTIONS:
323 			return (ip_pcbopts(&inp->inp_options, *m));
324 
325 		default:
326 			error = EINVAL;
327 			break;
328 		}
329 		break;
330 
331 	case PRCO_GETOPT:
332 		switch (optname) {
333 		case IP_OPTIONS:
334 			*m = m_get(M_WAIT, MT_SOOPTS);
335 			if (inp->inp_options) {
336 				(*m)->m_off = inp->inp_options->m_off;
337 				(*m)->m_len = inp->inp_options->m_len;
338 				bcopy(mtod(inp->inp_options, caddr_t),
339 				    mtod(*m, caddr_t), (unsigned)(*m)->m_len);
340 			} else
341 				(*m)->m_len = 0;
342 			break;
343 		default:
344 			error = EINVAL;
345 			break;
346 		}
347 		break;
348 	}
349 	if (op == PRCO_SETOPT)
350 		(void)m_free(*m);
351 	return (error);
352 }
353 
354 /*
355  * Set up IP options in pcb for insertion in output packets.
356  * Store in mbuf with pointer in pcbopt, adding pseudo-option
357  * with destination address if source routed.
358  */
359 ip_pcbopts(pcbopt, m)
360 	struct mbuf **pcbopt;
361 	register struct mbuf *m;
362 {
363 	register cnt, optlen;
364 	register u_char *cp;
365 	u_char opt;
366 
367 	/* turn off any old options */
368 	if (*pcbopt)
369 		(void)m_free(*pcbopt);
370 	*pcbopt = 0;
371 	if (m == (struct mbuf *)0 || m->m_len == 0) {
372 		/*
373 		 * Only turning off any previous options.
374 		 */
375 		if (m)
376 			(void)m_free(m);
377 		return (0);
378 	}
379 
380 #ifndef	vax
381 	if (m->m_len % sizeof(long))
382 		goto bad;
383 #endif
384 	/*
385 	 * IP first-hop destination address will be stored before
386 	 * actual options; move other options back
387 	 * and clear it when none present.
388 	 */
389 #if	MAX_IPOPTLEN >= MMAXOFF - MMINOFF
390 	if (m->m_off + m->m_len + sizeof(struct in_addr) > MAX_IPOPTLEN)
391 		goto bad;
392 #else
393 	if (m->m_off + m->m_len + sizeof(struct in_addr) > MMAXOFF)
394 		goto bad;
395 #endif
396 	cnt = m->m_len;
397 	m->m_len += sizeof(struct in_addr);
398 	cp = mtod(m, u_char *) + sizeof(struct in_addr);
399 	ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
400 	bzero(mtod(m, caddr_t), sizeof(struct in_addr));
401 
402 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
403 		opt = cp[IPOPT_OPTVAL];
404 		if (opt == IPOPT_EOL)
405 			break;
406 		if (opt == IPOPT_NOP)
407 			optlen = 1;
408 		else {
409 			optlen = cp[IPOPT_OLEN];
410 			if (optlen <= IPOPT_OLEN || optlen > cnt)
411 				goto bad;
412 		}
413 		switch (opt) {
414 
415 		default:
416 			break;
417 
418 		case IPOPT_LSRR:
419 		case IPOPT_SSRR:
420 			/*
421 			 * user process specifies route as:
422 			 *	->A->B->C->D
423 			 * D must be our final destination (but we can't
424 			 * check that since we may not have connected yet).
425 			 * A is first hop destination, which doesn't appear in
426 			 * actual IP option, but is stored before the options.
427 			 */
428 			if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
429 				goto bad;
430 			m->m_len -= sizeof(struct in_addr);
431 			cnt -= sizeof(struct in_addr);
432 			optlen -= sizeof(struct in_addr);
433 			cp[IPOPT_OLEN] = optlen;
434 			/*
435 			 * Move first hop before start of options.
436 			 */
437 			bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
438 			    sizeof(struct in_addr));
439 			/*
440 			 * Then copy rest of options back
441 			 * to close up the deleted entry.
442 			 */
443 			ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
444 			    sizeof(struct in_addr)),
445 			    (caddr_t)&cp[IPOPT_OFFSET+1],
446 			    (unsigned)cnt + sizeof(struct in_addr));
447 			break;
448 		}
449 	}
450 	*pcbopt = m;
451 	return (0);
452 
453 bad:
454 	(void)m_free(m);
455 	return (EINVAL);
456 }
457