xref: /freebsd/stand/libsa/ip.c (revision 8881d206f4e68b564c2c5f50fc717086fc3e827a)
1 /*
2  * Copyright (c) 1992 Regents of the University of California.
3  * All rights reserved.
4  *
5  * This software was developed by the Computer Systems Engineering group
6  * at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and
7  * contributed to Berkeley.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * The send and receive functions were originally implemented in udp.c and
36  * moved here. Also it is likely some more cleanup can be done, especially
37  * once we will implement the support for tcp.
38  */
39 
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include <sys/param.h>
44 #include <sys/socket.h>
45 #include <sys/queue.h>
46 
47 #include <string.h>
48 
49 #include <net/if.h>
50 #include <netinet/in.h>
51 #include <netinet/if_ether.h>
52 #include <netinet/in_systm.h>
53 
54 #include <netinet/ip.h>
55 #include <netinet/ip_var.h>
56 #include <netinet/udp.h>
57 #include <netinet/udp_var.h>
58 
59 #include "stand.h"
60 #include "net.h"
61 
62 typedef STAILQ_HEAD(ipqueue, ip_queue) ip_queue_t;
63 struct ip_queue {
64 	void		*ipq_pkt;
65 	struct ip	*ipq_hdr;
66 	STAILQ_ENTRY(ip_queue) ipq_next;
67 };
68 
69 /*
70  * Fragment re-assembly queue.
71  */
72 struct ip_reasm {
73 	struct in_addr	ip_src;
74 	struct in_addr	ip_dst;
75 	uint16_t	ip_id;
76 	uint8_t		ip_proto;
77 	uint8_t		ip_ttl;
78 	size_t		ip_total_size;
79 	ip_queue_t	ip_queue;
80 	void		*ip_pkt;
81 	struct ip	*ip_hdr;
82 	STAILQ_ENTRY(ip_reasm) ip_next;
83 };
84 
85 STAILQ_HEAD(ire_list, ip_reasm) ire_list = STAILQ_HEAD_INITIALIZER(ire_list);
86 
87 /* Caller must leave room for ethernet and ip headers in front!! */
88 ssize_t
89 sendip(struct iodesc *d, void *pkt, size_t len, uint8_t proto)
90 {
91 	ssize_t cc;
92 	struct ip *ip;
93 	u_char *ea;
94 
95 #ifdef NET_DEBUG
96 	if (debug) {
97 		printf("sendip: proto: %x d=%p called.\n", proto, (void *)d);
98 		if (d) {
99 			printf("saddr: %s:%d",
100 			    inet_ntoa(d->myip), ntohs(d->myport));
101 			printf(" daddr: %s:%d\n",
102 			    inet_ntoa(d->destip), ntohs(d->destport));
103 		}
104 	}
105 #endif
106 
107 	ip = (struct ip *)pkt - 1;
108 	len += sizeof(*ip);
109 
110 	bzero(ip, sizeof(*ip));
111 
112 	ip->ip_v = IPVERSION;			/* half-char */
113 	ip->ip_hl = sizeof(*ip) >> 2;		/* half-char */
114 	ip->ip_len = htons(len);
115 	ip->ip_p = proto;			/* char */
116 	ip->ip_ttl = IPDEFTTL;			/* char */
117 	ip->ip_src = d->myip;
118 	ip->ip_dst = d->destip;
119 	ip->ip_sum = in_cksum(ip, sizeof(*ip));	 /* short, but special */
120 
121 	if (ip->ip_dst.s_addr == INADDR_BROADCAST || ip->ip_src.s_addr == 0 ||
122 	    netmask == 0 || SAMENET(ip->ip_src, ip->ip_dst, netmask))
123 		ea = arpwhohas(d, ip->ip_dst);
124 	else
125 		ea = arpwhohas(d, gateip);
126 
127 	cc = sendether(d, ip, len, ea, ETHERTYPE_IP);
128 	if (cc == -1)
129 		return (-1);
130 	if (cc != len)
131 		panic("sendip: bad write (%zd != %zd)", cc, len);
132 	return (cc - sizeof(*ip));
133 }
134 
135 static void
136 ip_reasm_free(struct ip_reasm *ipr)
137 {
138 	struct ip_queue *ipq;
139 
140 	while ((ipq = STAILQ_FIRST(&ipr->ip_queue)) != NULL) {
141 		STAILQ_REMOVE_HEAD(&ipr->ip_queue, ipq_next);
142 		free(ipq->ipq_pkt);
143 		free(ipq);
144 	}
145 	free(ipr->ip_pkt);
146 	free(ipr);
147 }
148 
149 static int
150 ip_reasm_add(struct ip_reasm *ipr, void *pkt, struct ip *ip)
151 {
152 	struct ip_queue *ipq, *prev, *p;
153 
154 	if ((ipq = calloc(1, sizeof (*ipq))) == NULL)
155 		return (1);
156 
157 	ipq->ipq_pkt = pkt;
158 	ipq->ipq_hdr = ip;
159 
160 	prev = NULL;
161 	STAILQ_FOREACH(p, &ipr->ip_queue, ipq_next) {
162 		if ((ntohs(p->ipq_hdr->ip_off) & IP_OFFMASK) <
163 		    (ntohs(ip->ip_off) & IP_OFFMASK)) {
164 			prev = p;
165 			continue;
166 		}
167 		if (prev == NULL)
168 			break;
169 
170 		STAILQ_INSERT_AFTER(&ipr->ip_queue, prev, ipq, ipq_next);
171 		return (0);
172 	}
173 	STAILQ_INSERT_HEAD(&ipr->ip_queue, ipq, ipq_next);
174 	return (0);
175 }
176 
177 /*
178  * Receive a IP packet and validate it is for us.
179  */
180 static ssize_t
181 readipv4(struct iodesc *d, void **pkt, void **payload, time_t tleft,
182     uint8_t proto)
183 {
184 	ssize_t n;
185 	size_t hlen;
186 	struct ether_header *eh;
187 	struct ip *ip;
188 	struct udphdr *uh;
189 	uint16_t etype;		/* host order */
190 	char *ptr;
191 	struct ip_reasm *ipr;
192 	struct ip_queue *ipq, *last;
193 
194 #ifdef NET_DEBUG
195 	if (debug)
196 		printf("readip: called\n");
197 #endif
198 
199 	ip = NULL;
200 	ptr = NULL;
201 	n = readether(d, (void **)&ptr, (void **)&ip, tleft, &etype);
202 	if (n == -1 || n < sizeof(*ip) + sizeof(*uh)) {
203 		free(ptr);
204 		return (-1);
205 	}
206 
207 	/* Ethernet address checks now in readether() */
208 
209 	/* Need to respond to ARP requests. */
210 	if (etype == ETHERTYPE_ARP) {
211 		struct arphdr *ah = (void *)ip;
212 		if (ah->ar_op == htons(ARPOP_REQUEST)) {
213 			/* Send ARP reply */
214 			arp_reply(d, ah);
215 		}
216 		free(ptr);
217 		errno = EAGAIN;	/* Call me again. */
218 		return (-1);
219 	}
220 
221 	if (etype != ETHERTYPE_IP) {
222 #ifdef NET_DEBUG
223 		if (debug)
224 			printf("readip: not IP. ether_type=%x\n", etype);
225 #endif
226 		free(ptr);
227 		return (-1);
228 	}
229 
230 	/* Check ip header */
231 	if (ip->ip_v != IPVERSION ||	/* half char */
232 	    ip->ip_p != proto) {
233 #ifdef NET_DEBUG
234 		if (debug) {
235 			printf("readip: IP version or proto. ip_v=%d ip_p=%d\n",
236 			    ip->ip_v, ip->ip_p);
237 		}
238 #endif
239 		free(ptr);
240 		return (-1);
241 	}
242 
243 	hlen = ip->ip_hl << 2;
244 	if (hlen < sizeof(*ip) ||
245 	    in_cksum(ip, hlen) != 0) {
246 #ifdef NET_DEBUG
247 		if (debug)
248 			printf("readip: short hdr or bad cksum.\n");
249 #endif
250 		free(ptr);
251 		return (-1);
252 	}
253 	if (n < ntohs(ip->ip_len)) {
254 #ifdef NET_DEBUG
255 		if (debug)
256 			printf("readip: bad length %d < %d.\n",
257 			       (int)n, ntohs(ip->ip_len));
258 #endif
259 		free(ptr);
260 		return (-1);
261 	}
262 	if (d->myip.s_addr && ip->ip_dst.s_addr != d->myip.s_addr) {
263 #ifdef NET_DEBUG
264 		if (debug) {
265 			printf("readip: bad saddr %s != ", inet_ntoa(d->myip));
266 			printf("%s\n", inet_ntoa(ip->ip_dst));
267 		}
268 #endif
269 		free(ptr);
270 		return (-1);
271 	}
272 
273 	/* Unfragmented packet. */
274 	if ((ntohs(ip->ip_off) & IP_MF) == 0 &&
275 	    (ntohs(ip->ip_off) & IP_OFFMASK) == 0) {
276 		uh = (struct udphdr *)((uintptr_t)ip + sizeof (*ip));
277 		/* If there were ip options, make them go away */
278 		if (hlen != sizeof(*ip)) {
279 			bcopy(((u_char *)ip) + hlen, uh, uh->uh_ulen - hlen);
280 			ip->ip_len = htons(sizeof(*ip));
281 			n -= hlen - sizeof(*ip);
282 		}
283 
284 		n = (n > (ntohs(ip->ip_len) - sizeof(*ip))) ?
285 		    ntohs(ip->ip_len) - sizeof(*ip) : n;
286 		*pkt = ptr;
287 		*payload = (void *)((uintptr_t)ip + sizeof(*ip));
288 		return (n);
289 	}
290 
291 	STAILQ_FOREACH(ipr, &ire_list, ip_next) {
292 		if (ipr->ip_src.s_addr == ip->ip_src.s_addr &&
293 		    ipr->ip_dst.s_addr == ip->ip_dst.s_addr &&
294 		    ipr->ip_id == ip->ip_id &&
295 		    ipr->ip_proto == ip->ip_p)
296 			break;
297 	}
298 
299 	/* Allocate new reassembly entry */
300 	if (ipr == NULL) {
301 		if ((ipr = calloc(1, sizeof (*ipr))) == NULL) {
302 			free(ptr);
303 			return (-1);
304 		}
305 
306 		ipr->ip_src = ip->ip_src;
307 		ipr->ip_dst = ip->ip_dst;
308 		ipr->ip_id = ip->ip_id;
309 		ipr->ip_proto = ip->ip_p;
310 		ipr->ip_ttl = MAXTTL;
311 		STAILQ_INIT(&ipr->ip_queue);
312 		STAILQ_INSERT_TAIL(&ire_list, ipr, ip_next);
313 	}
314 
315 	if (ip_reasm_add(ipr, ptr, ip) != 0) {
316 		STAILQ_REMOVE(&ire_list, ipr, ip_reasm, ip_next);
317 		free(ipr);
318 		free(ptr);
319 		return (-1);
320 	}
321 
322 	if ((ntohs(ip->ip_off) & IP_MF) == 0) {
323 		ipr->ip_total_size = (8 * (ntohs(ip->ip_off) & IP_OFFMASK));
324 		ipr->ip_total_size += n + sizeof (*ip);
325 		ipr->ip_total_size += sizeof (struct ether_header);
326 
327 		ipr->ip_pkt = malloc(ipr->ip_total_size + 2);
328 		if (ipr->ip_pkt == NULL) {
329 			STAILQ_REMOVE(&ire_list, ipr, ip_reasm, ip_next);
330 			ip_reasm_free(ipr);
331 			return (-1);
332 		}
333 	}
334 
335 	/*
336 	 * If we do not have re-assembly buffer ipr->ip_pkt, we are still
337 	 * missing fragments, so just restart the read.
338 	 */
339 	if (ipr->ip_pkt == NULL) {
340 		errno = EAGAIN;
341 		return (-1);
342 	}
343 
344 	/*
345 	 * Walk the packet list in reassembly queue, if we got all the
346 	 * fragments, build the packet.
347 	 */
348 	n = 0;
349 	last = NULL;
350 	STAILQ_FOREACH(ipq, &ipr->ip_queue, ipq_next) {
351 		if ((ntohs(ipq->ipq_hdr->ip_off) & IP_OFFMASK) != n / 8) {
352 			STAILQ_REMOVE(&ire_list, ipr, ip_reasm, ip_next);
353 			ip_reasm_free(ipr);
354 			return (-1);
355 		}
356 
357 		n += ntohs(ipq->ipq_hdr->ip_len) - (ipq->ipq_hdr->ip_hl << 2);
358 		last = ipq;
359 	}
360 	if ((ntohs(last->ipq_hdr->ip_off) & IP_MF) != 0) {
361 		errno = EAGAIN;
362 		return (-1);
363 	}
364 
365 	ipq = STAILQ_FIRST(&ipr->ip_queue);
366 	/* Fabricate ethernet header */
367 	eh = (struct ether_header *)((uintptr_t)ipr->ip_pkt + 2);
368 	bcopy((void *)((uintptr_t)ipq->ipq_pkt + 2), eh, sizeof (*eh));
369 
370 	/* Fabricate IP header */
371 	ipr->ip_hdr = (struct ip *)((uintptr_t)eh + sizeof (*eh));
372 	bcopy(ipq->ipq_hdr, ipr->ip_hdr, sizeof (*ipr->ip_hdr));
373 	ipr->ip_hdr->ip_hl = sizeof (*ipr->ip_hdr) >> 2;
374 	ipr->ip_hdr->ip_len = htons(n);
375 	ipr->ip_hdr->ip_sum = 0;
376 	ipr->ip_hdr->ip_sum = in_cksum(ipr->ip_hdr, sizeof (*ipr->ip_hdr));
377 
378 	n = 0;
379 	ptr = (char *)((uintptr_t)ipr->ip_hdr + sizeof (*ipr->ip_hdr));
380 	STAILQ_FOREACH(ipq, &ipr->ip_queue, ipq_next) {
381 		char *data;
382 		size_t len;
383 
384 		hlen = ipq->ipq_hdr->ip_hl << 2;
385 		len = ntohs(ipq->ipq_hdr->ip_len) - hlen;
386 		data = (char *)((uintptr_t)ipq->ipq_hdr + hlen);
387 
388 		bcopy(data, ptr + n, len);
389 		n += len;
390 	}
391 
392 	*pkt = ipr->ip_pkt;
393 	ipr->ip_pkt = NULL;	/* Avoid free from ip_reasm_free() */
394 	*payload = ptr;
395 
396 	/* Clean up the reassembly list */
397 	while ((ipr = STAILQ_FIRST(&ire_list)) != NULL) {
398 		STAILQ_REMOVE_HEAD(&ire_list, ip_next);
399 		ip_reasm_free(ipr);
400 	}
401 	return (n);
402 }
403 
404 /*
405  * Receive a IP packet.
406  */
407 ssize_t
408 readip(struct iodesc *d, void **pkt, void **payload, time_t tleft,
409     uint8_t proto)
410 {
411 	time_t t;
412 	ssize_t ret = -1;
413 
414 	t = getsecs();
415 	while ((getsecs() - t) < tleft) {
416 		errno = 0;
417 		ret = readipv4(d, pkt, payload, tleft, proto);
418 		if (ret >= 0)
419 			return (ret);
420 		/* Bubble up the error if it wasn't successful */
421 		if (errno != EAGAIN)
422 			return (-1);
423 	}
424 	/* We've exhausted tleft; timeout */
425 	errno = ETIMEDOUT;
426 	return (-1);
427 }
428