xref: /csrg-svn/sys/netinet/ip_input.c (revision 4801)
1 /* ip_input.c 1.12 81/11/08 */
2 
3 #include "../h/param.h"
4 #include "../h/systm.h"
5 #include "../h/clock.h"
6 #include "../h/mbuf.h"
7 #include "../net/inet_cksum.h"
8 #include "../net/inet.h"
9 #include "../net/inet_systm.h"
10 #include "../net/imp.h"
11 #include "../net/ip.h"			/* belongs before inet.h */
12 #include "../net/ip_icmp.h"
13 #include "../net/tcp.h"
14 
15 /*
16  * Ip initialization.
17  */
18 ip_init()
19 {
20 
21 	ipq.next = ipq.prev = &ipq;
22 	ip_id = time & 0xffff;
23 }
24 
25 int	ipcksum = 1;
26 
27 struct	ip *ip_reass();
28 
29 /*
30  * Ip input routines.
31  */
32 
33 /*
34  * Ip input routine.  Checksum and byte swap header.  If fragmented
35  * try to reassamble.  If complete and fragment queue exists, discard.
36  * Process options.  Pass to next level.
37  */
38 ip_input(m0)
39 	struct mbuf *m0;
40 {
41 	register struct ip *ip;		/* known to be r11 in CKSUM below */
42 	register struct mbuf *m = m0;
43 	register int i;
44 	register struct ipq *q;
45 	register struct ipq *fp;
46 	int hlen;
47 
48 COUNT(IP_INPUT);
49 	/*
50 	 * Check header and byteswap.
51 	 */
52 	ip = mtod(m, struct ip *);
53 	if ((hlen = ip->ip_hl << 2) > m->m_len) {
54 		printf("ip hdr ovflo\n");
55 		m_freem(m);
56 		return;
57 	}
58 	CKSUM_IPCHK(m, ip, r11, hlen);
59 	if (ip->ip_sum) {
60 		printf("ip_sum %x\n", ip->ip_sum);
61 		netstat.ip_badsum++;
62 		if (ipcksum) {
63 			m_freem(m);
64 			return;
65 		}
66 	}
67 	ip->ip_len = ntohs(ip->ip_len);
68 	ip->ip_id = ntohs(ip->ip_id);
69 	ip->ip_off = ntohs(ip->ip_off);
70 
71 	/*
72 	 * Check that the amount of data in the buffers
73 	 * is as at least much as the IP header would have us expect.
74 	 * Trim mbufs if longer than we expect.
75 	 * Drop packet if shorter than we expect.
76 	 */
77 	i = 0;
78 	for (; m != NULL; m = m->m_next)
79 		i += m->m_len;
80 	m = m0;
81 	if (i != ip->ip_len) {
82 		if (i < ip->ip_len) {
83 			printf("ip_input: short packet\n");
84 			m_freem(m);
85 			return;
86 		}
87 		m_adj(m, ip->ip_len - i);
88 	}
89 
90 	/*
91 	 * Process options and, if not destined for us,
92 	 * ship it on.
93 	 */
94 	if (hlen > sizeof (struct ip))
95 		ip_dooptions(ip, hlen);
96 	if (ip->ip_dst.s_addr != n_lhost.s_addr) {
97 		if (--ip->ip_ttl == 0) {
98 			icmp_error(ip, ICMP_TIMXCEED);
99 			return;
100 		}
101 		ip_output(dtom(ip));
102 		return;
103 	}
104 
105 	/*
106 	 * Look for queue of fragments
107 	 * of this datagram.
108 	 */
109 	for (fp = ipq.next; fp != &ipq; fp = fp->next)
110 		if (ip->ip_id == fp->ipq_id &&
111 		    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
112 		    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
113 		    ip->ip_p == fp->ipq_p)
114 			goto found;
115 	fp = 0;
116 found:
117 
118 	/*
119 	 * Adjust ip_len to not reflect header,
120 	 * set ip_mff if more fragments are expected,
121 	 * convert offset of this to bytes.
122 	 */
123 	ip->ip_len -= hlen;
124 	ip->ip_mff = 0;
125 	if (ip->ip_off & IP_MF)
126 		ip->ip_mff = 1;
127 	ip->ip_off <<= 3;
128 
129 	/*
130 	 * If datagram marked as having more fragments
131 	 * or if this is not the first fragment,
132 	 * attempt reassembly; if it succeeds, proceed.
133 	 */
134 	if (ip->ip_mff || ip->ip_off) {
135 		ip = ip_reass(ip, fp);
136 		if (ip == 0)
137 			return;
138 		hlen = ip->ip_hl << 2;
139 		m = dtom(ip);
140 	} else
141 		if (fp)
142 			(void) ip_freef(fp);
143 
144 	/*
145 	 * Switch out to protocol specific routine.
146 	 * SHOULD GO THROUGH PROTOCOL SWITCH TABLE
147 	 */
148 	switch (ip->ip_p) {
149 
150 	case IPPROTO_ICMP:
151 		icmp_input(m);
152 		break;
153 
154 	case IPPROTO_TCP:
155 		if (hlen > sizeof (struct ip))
156 			ip_stripoptions(ip, hlen);
157 		tcp_input(m);
158 		break;
159 
160 	case IPPROTO_UDP:
161 		if (hlen > sizeof (struct ip))
162 			ip_stripoptions(ip, hlen);
163 		udp_input(m);
164 		break;
165 
166 	default:
167 		ri_input(m);
168 		break;
169 	}
170 }
171 
172 /*
173  * Take incoming datagram fragment and try to
174  * reassamble it into whole datagram.  If a chain for
175  * reassembly of this datagram already exists, then it
176  * is given as fp; otherwise have to make a chain.
177  */
178 struct ip *
179 ip_reass(ip, fp)
180 	register struct ip *ip;
181 	register struct ipq *fp;
182 {
183 	register struct mbuf *m = dtom(ip);
184 	register struct ip *q;
185 	struct mbuf *t;
186 	int hlen = ip->ip_hl << 2;
187 	int i, next;
188 
189 	/*
190 	 * Presence of header sizes in mbufs
191 	 * would confuse code below.
192 	 */
193 	m->m_off += hlen;
194 	m->m_len -= hlen;
195 
196 	/*
197 	 * If first fragment to arrive, create a reassembly queue.
198 	 */
199 	if (fp == 0) {
200 		if ((t = m_get(1)) == NULL)
201 			goto dropfrag;
202 		t->m_off = MMINOFF;
203 		fp = mtod(t, struct ipq *);
204 		insque(fp, &ipq);
205 		fp->ipq_ttl = IPFRAGTTL;
206 		fp->ipq_p = ip->ip_p;
207 		fp->ipq_id = ip->ip_id;
208 		fp->ipq_next = fp->ipq_prev = (struct ip *)fp;
209 		fp->ipq_src = ip->ip_src;
210 		fp->ipq_dst = ip->ip_dst;
211 	}
212 
213 	/*
214 	 * Find a segment which begins after this one does.
215 	 */
216 	for (q = fp->ipq_next; q != (struct ip *)fp; q = q->ip_next)
217 		if (q->ip_off > ip->ip_off)
218 			break;
219 
220 	/*
221 	 * If there is a preceding segment, it may provide some of
222 	 * our data already.  If so, drop the data from the incoming
223 	 * segment.  If it provides all of our data, drop us.
224 	 */
225 	if (q->ip_prev != (struct ip *)fp) {
226 		i = q->ip_prev->ip_off + q->ip_prev->ip_len - ip->ip_off;
227 		if (i > 0) {
228 			if (i >= ip->ip_len)
229 				goto dropfrag;
230 			m_adj(dtom(ip), i);
231 			ip->ip_off += i;
232 			ip->ip_len -= i;
233 		}
234 	}
235 
236 	/*
237 	 * While we overlap succeeding segments trim them or,
238 	 * if they are completely covered, dequeue them.
239 	 */
240 	while (q != (struct ip *)fp && ip->ip_off + ip->ip_len > q->ip_off) {
241 		i = (ip->ip_off + ip->ip_len) - q->ip_off;
242 		if (i < q->ip_len) {
243 			q->ip_len -= i;
244 			m_adj(dtom(q), i);
245 			break;
246 		}
247 		q = q->ip_next;
248 		m_freem(dtom(q->ip_prev));
249 		ip_deq(q->ip_prev);
250 	}
251 
252 	/*
253 	 * Stick new segment in its place;
254 	 * check for complete reassembly.
255 	 */
256 	ip_enq(ip, q->ip_prev);
257 	next = 0;
258 	for (q = fp->ipq_next; q != (struct ip *)fp; q = q->ip_next) {
259 		if (q->ip_off != next)
260 			return (0);
261 		next += q->ip_len;
262 	}
263 	if (q->ip_prev->ip_mff)
264 		return (0);
265 
266 	/*
267 	 * Reassembly is complete; concatenate fragments.
268 	 */
269 	q = fp->ipq_next;
270 	m = dtom(q);
271 	t = m->m_next;
272 	m->m_next = 0;
273 	m_cat(m, t);
274 	while ((q = q->ip_next) != (struct ip *)fp)
275 		m_cat(m, dtom(q));
276 
277 	/*
278 	 * Create header for new ip packet by
279 	 * modifying header of first packet;
280 	 * dequeue and discard fragment reassembly header.
281 	 * Make header visible.
282 	 */
283 	ip = fp->ipq_next;
284 	ip->ip_len = next;
285 	ip->ip_src = fp->ipq_src;
286 	ip->ip_dst = fp->ipq_dst;
287 	remque(fp);
288 	m_free(dtom(fp));
289 	m = dtom(ip);
290 	m->m_len += sizeof (struct ip);
291 	m->m_off -= sizeof (struct ip);
292 	return (ip);
293 
294 dropfrag:
295 	m_freem(m);
296 	return (0);
297 }
298 
299 /*
300  * Free a fragment reassembly header and all
301  * associated datagrams.
302  */
303 struct ipq *
304 ip_freef(fp)
305 	struct ipq *fp;
306 {
307 	register struct ip *q;
308 	struct mbuf *m;
309 
310 	for (q = fp->ipq_next; q != (struct ip *)fp; q = q->ip_next)
311 		m_freem(dtom(q));
312 	m = dtom(fp);
313 	fp = fp->next;
314 	remque(fp->prev);
315 	m_free(m);
316 	return (fp);
317 }
318 
319 /*
320  * Put an ip fragment on a reassembly chain.
321  * Like insque, but pointers in middle of structure.
322  */
323 ip_enq(p, prev)
324 	register struct ip *p;
325 	register struct ip *prev;
326 {
327 COUNT(IP_ENQ);
328 
329 	p->ip_prev = prev;
330 	p->ip_next = prev->ip_next;
331 	prev->ip_next->ip_prev = p;
332 	prev->ip_next = p;
333 }
334 
335 /*
336  * To ip_enq as remque is to insque.
337  */
338 ip_deq(p)
339 	register struct ip *p;
340 {
341 COUNT(IP_DEQ);
342 
343 	p->ip_prev->ip_next = p->ip_next;
344 	p->ip_next->ip_prev = p->ip_prev;
345 }
346 
347 /*
348  * IP timer processing;
349  * if a timer expires on a reassembly
350  * queue, discard it.
351  */
352 ip_slowtimo()
353 {
354 	register struct ip *q;
355 	register struct ipq *fp;
356 	int s = splnet();
357 COUNT(IP_SLOWTIMO);
358 
359 	for (fp = ipq.next; fp != &ipq; )
360 		if (--fp->ipq_ttl == 0)
361 			fp = ip_freef(fp);
362 		else
363 			fp = fp->next;
364 	splx(s);
365 }
366 
367 ip_drain()
368 {
369 
370 }
371 /*
372  * Do option processing on a datagram,
373  * possibly discarding it if bad options
374  * are encountered.
375  */
376 ip_dooptions(ip)
377 	struct ip *ip;
378 {
379 	register u_char *cp;
380 	int opt, optlen, cnt, s;
381 	struct ip_addr *sp;
382 	register struct ip_timestamp *ipt;
383 	int x;
384 
385 	cp = (u_char *)(ip + 1);
386 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
387 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
388 		opt = cp[0];
389 		if (opt == IPOPT_EOL)
390 			break;
391 		if (opt == IPOPT_NOP)
392 			optlen = 1;
393 		else
394 			optlen = cp[1];
395 		switch (opt) {
396 
397 		default:
398 			break;
399 
400 		case IPOPT_LSRR:
401 		case IPOPT_SSRR:
402 			if (cp[2] < 4 || cp[2] > optlen - (sizeof (long) - 1))
403 				break;
404 			sp = (struct ip_addr *)(cp + cp[2]);
405 			if (n_lhost.s_addr == *(u_long *)sp) {
406 				if (opt == IPOPT_SSRR) {
407 					/* MAKE SURE *SP DIRECTLY ACCESSIBLE */
408 				}
409 				ip->ip_dst = *sp;
410 				*sp = n_lhost;
411 				cp[2] += 4;
412 			}
413 			break;
414 
415 		case IPOPT_TS:
416 			ipt = (struct ip_timestamp *)cp;
417 			if (ipt->ipt_len < 5)
418 				goto bad;
419 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (long)) {
420 				if (++ipt->ipt_oflw == 0)
421 					goto bad;
422 				break;
423 			}
424 			sp = (struct ip_addr *)(cp+cp[2]);
425 			switch (ipt->ipt_flg) {
426 
427 			case IPOPT_TS_TSONLY:
428 				break;
429 
430 			case IPOPT_TS_TSANDADDR:
431 				if (ipt->ipt_ptr + 8 > ipt->ipt_len)
432 					goto bad;
433 				*(struct ip_addr *)sp++ = n_lhost;
434 				break;
435 
436 			case IPOPT_TS_PRESPEC:
437 				if (*(u_long *)sp != n_lhost.s_addr)
438 					break;
439 				if (ipt->ipt_ptr + 8 > ipt->ipt_len)
440 					goto bad;
441 				ipt->ipt_ptr += 4;
442 				break;
443 
444 			default:
445 				goto bad;
446 			}
447 			*(n_time *)sp = ip_time();
448 			ipt->ipt_ptr += 4;
449 		}
450 	}
451 	return (0);
452 bad:
453 	/* SHOULD FORCE ICMP MESSAGE */
454 	return (-1);
455 }
456 
457 /*
458  * Strip out IP options, e.g. before passing
459  * to higher level protocol in the kernel.
460  */
461 ip_stripoptions(ip)
462 	struct ip *ip;
463 {
464 	register int i;
465 	register struct mbuf *m;
466 	char *op;
467 	int olen;
468 COUNT(IP_OPT);
469 
470 	olen = (ip->ip_hl<<2) - sizeof (struct ip);
471 	op = (caddr_t)ip + olen;
472 	m = dtom(++ip);
473 	i = m->m_len - (sizeof (struct ip) + olen);
474 	bcopy((caddr_t)ip+olen, (caddr_t)ip, i);
475 	m->m_len -= i;
476 }
477