xref: /csrg-svn/sys/netinet/ip_input.c (revision 4898)
1 /* ip_input.c 1.13 81/11/15 */
2 
3 #include "../h/param.h"
4 #include "../h/systm.h"
5 #include "../h/clock.h"
6 #include "../h/mbuf.h"
7 #include "../h/protocol.h"
8 #include "../h/protosw.h"
9 #include "../net/inet_cksum.h"
10 #include "../net/inet.h"
11 #include "../net/inet_systm.h"
12 #include "../net/imp.h"
13 #include "../net/ip.h"			/* belongs before inet.h */
14 #include "../net/ip_var.h"
15 #include "../net/ip_icmp.h"
16 #include "../net/tcp.h"
17 
18 u_char	ip_protox[IPPROTO_MAX];
19 
20 /*
21  * Ip initialization.
22  */
23 ip_init()
24 {
25 	register struct protosw *pr;
26 	register int i;
27 	int raw;
28 
29 	pr = pffindproto(PF_INET, IPPROTO_RAW);
30 	if (pr == 0)
31 		panic("ip_init");
32 	for (i = 0; i < IPPROTO_MAX; i++)
33 		ip_protox[i] = pr - protosw;
34 	for (pr = protosw; pr <= protoswLAST; pr++)
35 		if (pr->pr_family == PF_INET &&
36 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
37 			ip_protox[pr->pr_protocol] = pr - protosw;
38 	ipq.next = ipq.prev = &ipq;
39 	ip_id = time & 0xffff;
40 }
41 
42 u_char	ipcksum = 1;
43 struct	ip *ip_reass();
44 
45 /*
46  * Ip input routines.
47  */
48 
49 /*
50  * Ip input routine.  Checksum and byte swap header.  If fragmented
51  * try to reassamble.  If complete and fragment queue exists, discard.
52  * Process options.  Pass to next level.
53  */
54 ip_input(m0)
55 	struct mbuf *m0;
56 {
57 	register struct ip *ip;		/* known to be r11 in CKSUM below */
58 	register struct mbuf *m = m0;
59 	register int i;
60 	register struct ipq *q;
61 	register struct ipq *fp;
62 	int hlen;
63 
64 COUNT(IP_INPUT);
65 	/*
66 	 * Check header and byteswap.
67 	 */
68 	ip = mtod(m, struct ip *);
69 	if ((hlen = ip->ip_hl << 2) > m->m_len) {
70 		printf("ip hdr ovflo\n");
71 		m_freem(m);
72 		return;
73 	}
74 	CKSUM_IPCHK(m, ip, r11, hlen);
75 	if (ip->ip_sum) {
76 		printf("ip_sum %x\n", ip->ip_sum);
77 		netstat.ip_badsum++;
78 		if (ipcksum) {
79 			m_freem(m);
80 			return;
81 		}
82 	}
83 	ip->ip_len = ntohs(ip->ip_len);
84 	ip->ip_id = ntohs(ip->ip_id);
85 	ip->ip_off = ntohs(ip->ip_off);
86 
87 	/*
88 	 * Check that the amount of data in the buffers
89 	 * is as at least much as the IP header would have us expect.
90 	 * Trim mbufs if longer than we expect.
91 	 * Drop packet if shorter than we expect.
92 	 */
93 	i = 0;
94 	for (; m != NULL; m = m->m_next)
95 		i += m->m_len;
96 	m = m0;
97 	if (i != ip->ip_len) {
98 		if (i < ip->ip_len) {
99 			printf("ip_input: short packet\n");
100 			m_freem(m);
101 			return;
102 		}
103 		m_adj(m, ip->ip_len - i);
104 	}
105 
106 	/*
107 	 * Process options and, if not destined for us,
108 	 * ship it on.
109 	 */
110 	if (hlen > sizeof (struct ip))
111 		ip_dooptions(ip, hlen);
112 	if (ip->ip_dst.s_addr != n_lhost.s_addr) {
113 		if (--ip->ip_ttl == 0) {
114 			icmp_error(ip, ICMP_TIMXCEED);
115 			return;
116 		}
117 		ip_output(dtom(ip));
118 		return;
119 	}
120 
121 	/*
122 	 * Look for queue of fragments
123 	 * of this datagram.
124 	 */
125 	for (fp = ipq.next; fp != &ipq; fp = fp->next)
126 		if (ip->ip_id == fp->ipq_id &&
127 		    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
128 		    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
129 		    ip->ip_p == fp->ipq_p)
130 			goto found;
131 	fp = 0;
132 found:
133 
134 	/*
135 	 * Adjust ip_len to not reflect header,
136 	 * set ip_mff if more fragments are expected,
137 	 * convert offset of this to bytes.
138 	 */
139 	ip->ip_len -= hlen;
140 	((struct ipasfrag *)ip)->ipf_mff = 0;
141 	if (ip->ip_off & IP_MF)
142 		((struct ipasfrag *)ip)->ipf_mff = 1;
143 	ip->ip_off <<= 3;
144 
145 	/*
146 	 * If datagram marked as having more fragments
147 	 * or if this is not the first fragment,
148 	 * attempt reassembly; if it succeeds, proceed.
149 	 */
150 	if (((struct ipasfrag *)ip)->ipf_mff || ip->ip_off) {
151 		ip = ip_reass((struct ipasfrag *)ip, fp);
152 		if (ip == 0)
153 			return;
154 		hlen = ip->ip_hl << 2;
155 		m = dtom(ip);
156 	} else
157 		if (fp)
158 			(void) ip_freef(fp);
159 	(*protosw[ip_protox[ip->ip_p]].pr_input)(m);
160 }
161 
162 /*
163  * Take incoming datagram fragment and try to
164  * reassamble it into whole datagram.  If a chain for
165  * reassembly of this datagram already exists, then it
166  * is given as fp; otherwise have to make a chain.
167  */
168 struct ip *
169 ip_reass(ip, fp)
170 	register struct ipasfrag *ip;
171 	register struct ipq *fp;
172 {
173 	register struct mbuf *m = dtom(ip);
174 	register struct ipasfrag *q;
175 	struct mbuf *t;
176 	int hlen = ip->ip_hl << 2;
177 	int i, next;
178 
179 	/*
180 	 * Presence of header sizes in mbufs
181 	 * would confuse code below.
182 	 */
183 	m->m_off += hlen;
184 	m->m_len -= hlen;
185 
186 	/*
187 	 * If first fragment to arrive, create a reassembly queue.
188 	 */
189 	if (fp == 0) {
190 		if ((t = m_get(1)) == NULL)
191 			goto dropfrag;
192 		t->m_off = MMINOFF;
193 		fp = mtod(t, struct ipq *);
194 		insque(fp, &ipq);
195 		fp->ipq_ttl = IPFRAGTTL;
196 		fp->ipq_p = ip->ip_p;
197 		fp->ipq_id = ip->ip_id;
198 		fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp;
199 		fp->ipq_src = ((struct ip *)ip)->ip_src;
200 		fp->ipq_dst = ((struct ip *)ip)->ip_dst;
201 	}
202 
203 	/*
204 	 * Find a segment which begins after this one does.
205 	 */
206 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next)
207 		if (q->ip_off > ip->ip_off)
208 			break;
209 
210 	/*
211 	 * If there is a preceding segment, it may provide some of
212 	 * our data already.  If so, drop the data from the incoming
213 	 * segment.  If it provides all of our data, drop us.
214 	 */
215 	if (q->ipf_prev != (struct ipasfrag *)fp) {
216 		i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off;
217 		if (i > 0) {
218 			if (i >= ip->ip_len)
219 				goto dropfrag;
220 			m_adj(dtom(ip), i);
221 			ip->ip_off += i;
222 			ip->ip_len -= i;
223 		}
224 	}
225 
226 	/*
227 	 * While we overlap succeeding segments trim them or,
228 	 * if they are completely covered, dequeue them.
229 	 */
230 	while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) {
231 		i = (ip->ip_off + ip->ip_len) - q->ip_off;
232 		if (i < q->ip_len) {
233 			q->ip_len -= i;
234 			m_adj(dtom(q), i);
235 			break;
236 		}
237 		q = q->ipf_next;
238 		m_freem(dtom(q->ipf_prev));
239 		ip_deq(q->ipf_prev);
240 	}
241 
242 	/*
243 	 * Stick new segment in its place;
244 	 * check for complete reassembly.
245 	 */
246 	ip_enq(ip, q->ipf_prev);
247 	next = 0;
248 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) {
249 		if (q->ip_off != next)
250 			return (0);
251 		next += q->ip_len;
252 	}
253 	if (q->ipf_prev->ipf_mff)
254 		return (0);
255 
256 	/*
257 	 * Reassembly is complete; concatenate fragments.
258 	 */
259 	q = fp->ipq_next;
260 	m = dtom(q);
261 	t = m->m_next;
262 	m->m_next = 0;
263 	m_cat(m, t);
264 	while ((q = q->ipf_next) != (struct ipasfrag *)fp)
265 		m_cat(m, dtom(q));
266 
267 	/*
268 	 * Create header for new ip packet by
269 	 * modifying header of first packet;
270 	 * dequeue and discard fragment reassembly header.
271 	 * Make header visible.
272 	 */
273 	ip = fp->ipq_next;
274 	ip->ip_len = next;
275 	((struct ip *)ip)->ip_src = fp->ipq_src;
276 	((struct ip *)ip)->ip_dst = fp->ipq_dst;
277 	remque(fp);
278 	m_free(dtom(fp));
279 	m = dtom(ip);
280 	m->m_len += sizeof (struct ipasfrag);
281 	m->m_off -= sizeof (struct ipasfrag);
282 	return ((struct ip *)ip);
283 
284 dropfrag:
285 	m_freem(m);
286 	return (0);
287 }
288 
289 /*
290  * Free a fragment reassembly header and all
291  * associated datagrams.
292  */
293 struct ipq *
294 ip_freef(fp)
295 	struct ipq *fp;
296 {
297 	register struct ipasfrag *q;
298 	struct mbuf *m;
299 
300 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next)
301 		m_freem(dtom(q));
302 	m = dtom(fp);
303 	fp = fp->next;
304 	remque(fp->prev);
305 	m_free(m);
306 	return (fp);
307 }
308 
309 /*
310  * Put an ip fragment on a reassembly chain.
311  * Like insque, but pointers in middle of structure.
312  */
313 ip_enq(p, prev)
314 	register struct ipasfrag *p, *prev;
315 {
316 COUNT(IP_ENQ);
317 
318 	p->ipf_prev = prev;
319 	p->ipf_next = prev->ipf_next;
320 	prev->ipf_next->ipf_prev = p;
321 	prev->ipf_next = p;
322 }
323 
324 /*
325  * To ip_enq as remque is to insque.
326  */
327 ip_deq(p)
328 	register struct ipasfrag *p;
329 {
330 COUNT(IP_DEQ);
331 
332 	p->ipf_prev->ipf_next = p->ipf_next;
333 	p->ipf_next->ipf_prev = p->ipf_prev;
334 }
335 
336 /*
337  * IP timer processing;
338  * if a timer expires on a reassembly
339  * queue, discard it.
340  */
341 ip_slowtimo()
342 {
343 	register struct ipq *fp;
344 	int s = splnet();
345 COUNT(IP_SLOWTIMO);
346 
347 	for (fp = ipq.next; fp != &ipq; )
348 		if (--fp->ipq_ttl == 0)
349 			fp = ip_freef(fp);
350 		else
351 			fp = fp->next;
352 	splx(s);
353 }
354 
355 ip_drain()
356 {
357 
358 }
359 /*
360  * Do option processing on a datagram,
361  * possibly discarding it if bad options
362  * are encountered.
363  */
364 ip_dooptions(ip)
365 	struct ip *ip;
366 {
367 	register u_char *cp;
368 	int opt, optlen, cnt, s;
369 	struct ip_addr *sp;
370 	register struct ip_timestamp *ipt;
371 	int x;
372 
373 	cp = (u_char *)(ip + 1);
374 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
375 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
376 		opt = cp[0];
377 		if (opt == IPOPT_EOL)
378 			break;
379 		if (opt == IPOPT_NOP)
380 			optlen = 1;
381 		else
382 			optlen = cp[1];
383 		switch (opt) {
384 
385 		default:
386 			break;
387 
388 		case IPOPT_LSRR:
389 		case IPOPT_SSRR:
390 			if (cp[2] < 4 || cp[2] > optlen - (sizeof (long) - 1))
391 				break;
392 			sp = (struct ip_addr *)(cp + cp[2]);
393 			if (n_lhost.s_addr == *(u_long *)sp) {
394 				if (opt == IPOPT_SSRR) {
395 					/* MAKE SURE *SP DIRECTLY ACCESSIBLE */
396 				}
397 				ip->ip_dst = *sp;
398 				*sp = n_lhost;
399 				cp[2] += 4;
400 			}
401 			break;
402 
403 		case IPOPT_TS:
404 			ipt = (struct ip_timestamp *)cp;
405 			if (ipt->ipt_len < 5)
406 				goto bad;
407 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (long)) {
408 				if (++ipt->ipt_oflw == 0)
409 					goto bad;
410 				break;
411 			}
412 			sp = (struct ip_addr *)(cp+cp[2]);
413 			switch (ipt->ipt_flg) {
414 
415 			case IPOPT_TS_TSONLY:
416 				break;
417 
418 			case IPOPT_TS_TSANDADDR:
419 				if (ipt->ipt_ptr + 8 > ipt->ipt_len)
420 					goto bad;
421 				*(struct ip_addr *)sp++ = n_lhost;
422 				break;
423 
424 			case IPOPT_TS_PRESPEC:
425 				if (*(u_long *)sp != n_lhost.s_addr)
426 					break;
427 				if (ipt->ipt_ptr + 8 > ipt->ipt_len)
428 					goto bad;
429 				ipt->ipt_ptr += 4;
430 				break;
431 
432 			default:
433 				goto bad;
434 			}
435 			*(n_time *)sp = ip_time();
436 			ipt->ipt_ptr += 4;
437 		}
438 	}
439 	return (0);
440 bad:
441 	/* SHOULD FORCE ICMP MESSAGE */
442 	return (-1);
443 }
444 
445 /*
446  * Strip out IP options, e.g. before passing
447  * to higher level protocol in the kernel.
448  */
449 ip_stripoptions(ip)
450 	struct ip *ip;
451 {
452 	register int i;
453 	register struct mbuf *m;
454 	char *op;
455 	int olen;
456 COUNT(IP_OPT);
457 
458 	olen = (ip->ip_hl<<2) - sizeof (struct ip);
459 	op = (caddr_t)ip + olen;
460 	m = dtom(++ip);
461 	i = m->m_len - (sizeof (struct ip) + olen);
462 	bcopy((caddr_t)ip+olen, (caddr_t)ip, i);
463 	m->m_len -= i;
464 }
465