xref: /illumos-gate/usr/src/boot/libsa/ip.c (revision 9b9d39d2a32ff806d2431dbcc50968ef1e6d46b2)
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 
42 #include <sys/param.h>
43 #include <sys/socket.h>
44 #include <sys/queue.h>
45 
46 #include <string.h>
47 
48 #include <net/if.h>
49 #include <netinet/in.h>
50 #include <netinet/if_ether.h>
51 #include <netinet/in_systm.h>
52 
53 #include <netinet/ip.h>
54 #include <netinet/ip_var.h>
55 #include <netinet/udp.h>
56 #include <netinet/udp_var.h>
57 
58 #include "stand.h"
59 #include "net.h"
60 
61 typedef STAILQ_HEAD(ipqueue, ip_queue) ip_queue_t;
62 struct ip_queue {
63 	void		*ipq_pkt;
64 	struct ip	*ipq_hdr;
65 	STAILQ_ENTRY(ip_queue) ipq_next;
66 };
67 
68 /*
69  * Fragment re-assembly queue.
70  */
71 struct ip_reasm {
72 	struct in_addr	ip_src;
73 	struct in_addr	ip_dst;
74 	uint16_t	ip_id;
75 	uint8_t		ip_proto;
76 	uint8_t		ip_ttl;
77 	size_t		ip_total_size;
78 	ip_queue_t	ip_queue;
79 	void		*ip_pkt;
80 	struct ip	*ip_hdr;
81 	STAILQ_ENTRY(ip_reasm) ip_next;
82 };
83 
84 STAILQ_HEAD(ire_list, ip_reasm) ire_list = STAILQ_HEAD_INITIALIZER(ire_list);
85 
86 /* Caller must leave room for ethernet and ip headers in front!! */
87 ssize_t
88 sendip(struct iodesc *d, void *pkt, size_t len, uint8_t proto)
89 {
90 	ssize_t cc;
91 	struct ip *ip;
92 	u_char *ea;
93 
94 #ifdef NET_DEBUG
95 	if (debug) {
96 		printf("sendip: proto: %x d=%p called.\n", proto, (void *)d);
97 		if (d) {
98 			printf("saddr: %s:%d",
99 			    inet_ntoa(d->myip), ntohs(d->myport));
100 			printf(" daddr: %s:%d\n",
101 			    inet_ntoa(d->destip), ntohs(d->destport));
102 		}
103 	}
104 #endif
105 
106 	ip = (struct ip *)pkt - 1;
107 	len += sizeof(*ip);
108 
109 	bzero(ip, sizeof(*ip));
110 
111 	ip->ip_v = IPVERSION;			/* half-char */
112 	ip->ip_hl = sizeof(*ip) >> 2;		/* half-char */
113 	ip->ip_len = htons(len);
114 	ip->ip_p = proto;			/* char */
115 	ip->ip_ttl = IPDEFTTL;			/* char */
116 	ip->ip_src = d->myip;
117 	ip->ip_dst = d->destip;
118 	ip->ip_sum = in_cksum(ip, sizeof(*ip));	 /* short, but special */
119 
120 	if (ip->ip_dst.s_addr == INADDR_BROADCAST || ip->ip_src.s_addr == 0 ||
121 	    netmask == 0 || SAMENET(ip->ip_src, ip->ip_dst, netmask))
122 		ea = arpwhohas(d, ip->ip_dst);
123 	else
124 		ea = arpwhohas(d, gateip);
125 
126 	cc = sendether(d, ip, len, ea, ETHERTYPE_IP);
127 	if (cc == -1)
128 		return (-1);
129 	if (cc != len)
130 		panic("sendip: bad write (%zd != %zd)", cc, len);
131 	return (cc - sizeof(*ip));
132 }
133 
134 static void
135 ip_reasm_free(struct ip_reasm *ipr)
136 {
137 	struct ip_queue *ipq;
138 
139 	while ((ipq = STAILQ_FIRST(&ipr->ip_queue)) != NULL) {
140 		STAILQ_REMOVE_HEAD(&ipr->ip_queue, ipq_next);
141 		free(ipq->ipq_pkt);
142 		free(ipq);
143 	}
144 	free(ipr->ip_pkt);
145 	free(ipr);
146 }
147 
148 static int
149 ip_reasm_add(struct ip_reasm *ipr, void *pkt, struct ip *ip)
150 {
151 	struct ip_queue *ipq, *prev, *p;
152 
153 	if ((ipq = calloc(1, sizeof (*ipq))) == NULL)
154 		return (1);
155 
156 	ipq->ipq_pkt = pkt;
157 	ipq->ipq_hdr = ip;
158 
159 	prev = NULL;
160 	STAILQ_FOREACH(p, &ipr->ip_queue, ipq_next) {
161 		if ((ntohs(p->ipq_hdr->ip_off) & IP_OFFMASK) <
162 		    (ntohs(ip->ip_off) & IP_OFFMASK)) {
163 			prev = p;
164 			continue;
165 		}
166 		if (prev == NULL)
167 			break;
168 
169 		STAILQ_INSERT_AFTER(&ipr->ip_queue, prev, ipq, ipq_next);
170 		return (0);
171 	}
172 	STAILQ_INSERT_HEAD(&ipr->ip_queue, ipq, ipq_next);
173 	return (0);
174 }
175 
176 /*
177  * Receive a IP packet and validate it is for us.
178  */
179 static ssize_t
180 readipv4(struct iodesc *d, void **pkt, void **payload, time_t tleft,
181     uint8_t proto)
182 {
183 	ssize_t n;
184 	size_t hlen;
185 	struct ether_header *eh;
186 	struct ip *ip;
187 	struct udphdr *uh;
188 	uint16_t etype;		/* host order */
189 	char *ptr;
190 	struct ip_reasm *ipr;
191 	struct ip_queue *ipq, *last;
192 
193 #ifdef NET_DEBUG
194 	if (debug)
195 		printf("readip: called\n");
196 #endif
197 
198 	ip = NULL;
199 	ptr = NULL;
200 	n = readether(d, (void **)&ptr, (void **)&ip, tleft, &etype);
201 	if (n == -1 || n < sizeof(*ip) + sizeof(*uh)) {
202 		free(ptr);
203 		return (-1);
204 	}
205 
206 	/* Ethernet address checks now in readether() */
207 
208 	/* Need to respond to ARP requests. */
209 	if (etype == ETHERTYPE_ARP) {
210 		struct arphdr *ah = (void *)ip;
211 		if (ah->ar_op == htons(ARPOP_REQUEST)) {
212 			/* Send ARP reply */
213 			arp_reply(d, ah);
214 		}
215 		free(ptr);
216 		errno = EAGAIN;	/* Call me again. */
217 		return (-1);
218 	}
219 
220 	if (etype != ETHERTYPE_IP) {
221 #ifdef NET_DEBUG
222 		if (debug)
223 			printf("readip: not IP. ether_type=%x\n", etype);
224 #endif
225 		free(ptr);
226 		return (-1);
227 	}
228 
229 	/* Check ip header */
230 	if (ip->ip_v != IPVERSION ||	/* half char */
231 	    ip->ip_p != proto) {
232 #ifdef NET_DEBUG
233 		if (debug) {
234 			printf("readip: IP version or proto. ip_v=%d ip_p=%d\n",
235 			    ip->ip_v, ip->ip_p);
236 		}
237 #endif
238 		free(ptr);
239 		return (-1);
240 	}
241 
242 	hlen = ip->ip_hl << 2;
243 	if (hlen < sizeof(*ip) ||
244 	    in_cksum(ip, hlen) != 0) {
245 #ifdef NET_DEBUG
246 		if (debug)
247 			printf("readip: short hdr or bad cksum.\n");
248 #endif
249 		free(ptr);
250 		return (-1);
251 	}
252 	if (n < ntohs(ip->ip_len)) {
253 #ifdef NET_DEBUG
254 		if (debug)
255 			printf("readip: bad length %d < %d.\n",
256 			       (int)n, ntohs(ip->ip_len));
257 #endif
258 		free(ptr);
259 		return (-1);
260 	}
261 	if (d->myip.s_addr && ip->ip_dst.s_addr != d->myip.s_addr) {
262 #ifdef NET_DEBUG
263 		if (debug) {
264 			printf("readip: bad saddr %s != ", inet_ntoa(d->myip));
265 			printf("%s\n", inet_ntoa(ip->ip_dst));
266 		}
267 #endif
268 		free(ptr);
269 		return (-1);
270 	}
271 
272 	/* Unfragmented packet. */
273 	if ((ntohs(ip->ip_off) & IP_MF) == 0 &&
274 	    (ntohs(ip->ip_off) & IP_OFFMASK) == 0) {
275 		uh = (struct udphdr *)((uintptr_t)ip + sizeof (*ip));
276 		/* If there were ip options, make them go away */
277 		if (hlen != sizeof(*ip)) {
278 			bcopy(((u_char *)ip) + hlen, uh, uh->uh_ulen - hlen);
279 			ip->ip_len = htons(sizeof(*ip));
280 			n -= hlen - sizeof(*ip);
281 		}
282 
283 		n = (n > (ntohs(ip->ip_len) - sizeof(*ip))) ?
284 		    ntohs(ip->ip_len) - sizeof(*ip) : n;
285 		*pkt = ptr;
286 		*payload = (void *)((uintptr_t)ip + sizeof(*ip));
287 		return (n);
288 	}
289 
290 	STAILQ_FOREACH(ipr, &ire_list, ip_next) {
291 		if (ipr->ip_src.s_addr == ip->ip_src.s_addr &&
292 		    ipr->ip_dst.s_addr == ip->ip_dst.s_addr &&
293 		    ipr->ip_id == ip->ip_id &&
294 		    ipr->ip_proto == ip->ip_p)
295 			break;
296 	}
297 
298 	/* Allocate new reassembly entry */
299 	if (ipr == NULL) {
300 		if ((ipr = calloc(1, sizeof (*ipr))) == NULL) {
301 			free(ptr);
302 			return (-1);
303 		}
304 
305 		ipr->ip_src = ip->ip_src;
306 		ipr->ip_dst = ip->ip_dst;
307 		ipr->ip_id = ip->ip_id;
308 		ipr->ip_proto = ip->ip_p;
309 		ipr->ip_ttl = MAXTTL;
310 		STAILQ_INIT(&ipr->ip_queue);
311 		STAILQ_INSERT_TAIL(&ire_list, ipr, ip_next);
312 	}
313 
314 	if (ip_reasm_add(ipr, ptr, ip) != 0) {
315 		STAILQ_REMOVE(&ire_list, ipr, ip_reasm, ip_next);
316 		free(ipr);
317 		free(ptr);
318 		return (-1);
319 	}
320 
321 	if ((ntohs(ip->ip_off) & IP_MF) == 0) {
322 		ipr->ip_total_size = (8 * (ntohs(ip->ip_off) & IP_OFFMASK));
323 		ipr->ip_total_size += n + sizeof (*ip);
324 		ipr->ip_total_size += sizeof (struct ether_header);
325 
326 		ipr->ip_pkt = malloc(ipr->ip_total_size + 2);
327 		if (ipr->ip_pkt == NULL) {
328 			STAILQ_REMOVE(&ire_list, ipr, ip_reasm, ip_next);
329 			ip_reasm_free(ipr);
330 			return (-1);
331 		}
332 	}
333 
334 	/*
335 	 * If we do not have re-assembly buffer ipr->ip_pkt, we are still
336 	 * missing fragments, so just restart the read.
337 	 */
338 	if (ipr->ip_pkt == NULL) {
339 		errno = EAGAIN;
340 		return (-1);
341 	}
342 
343 	/*
344 	 * Walk the packet list in reassembly queue, if we got all the
345 	 * fragments, build the packet.
346 	 */
347 	n = 0;
348 	last = NULL;
349 	STAILQ_FOREACH(ipq, &ipr->ip_queue, ipq_next) {
350 		if ((ntohs(ipq->ipq_hdr->ip_off) & IP_OFFMASK) != n / 8) {
351 			STAILQ_REMOVE(&ire_list, ipr, ip_reasm, ip_next);
352 			ip_reasm_free(ipr);
353 			return (-1);
354 		}
355 
356 		n += ntohs(ipq->ipq_hdr->ip_len) - (ipq->ipq_hdr->ip_hl << 2);
357 		last = ipq;
358 	}
359 	if ((ntohs(last->ipq_hdr->ip_off) & IP_MF) != 0) {
360 		errno = EAGAIN;
361 		return (-1);
362 	}
363 
364 	ipq = STAILQ_FIRST(&ipr->ip_queue);
365 	/* Fabricate ethernet header */
366 	eh = (struct ether_header *)((uintptr_t)ipr->ip_pkt + 2);
367 	bcopy((void *)((uintptr_t)ipq->ipq_pkt + 2), eh, sizeof (*eh));
368 
369 	/* Fabricate IP header */
370 	ipr->ip_hdr = (struct ip *)((uintptr_t)eh + sizeof (*eh));
371 	bcopy(ipq->ipq_hdr, ipr->ip_hdr, sizeof (*ipr->ip_hdr));
372 	ipr->ip_hdr->ip_hl = sizeof (*ipr->ip_hdr) >> 2;
373 	ipr->ip_hdr->ip_len = htons(n);
374 	ipr->ip_hdr->ip_sum = 0;
375 	ipr->ip_hdr->ip_sum = in_cksum(ipr->ip_hdr, sizeof (*ipr->ip_hdr));
376 
377 	n = 0;
378 	ptr = (char *)((uintptr_t)ipr->ip_hdr + sizeof (*ipr->ip_hdr));
379 	STAILQ_FOREACH(ipq, &ipr->ip_queue, ipq_next) {
380 		char *data;
381 		size_t len;
382 
383 		hlen = ipq->ipq_hdr->ip_hl << 2;
384 		len = ntohs(ipq->ipq_hdr->ip_len) - hlen;
385 		data = (char *)((uintptr_t)ipq->ipq_hdr + hlen);
386 
387 		bcopy(data, ptr + n, len);
388 		n += len;
389 	}
390 
391 	*pkt = ipr->ip_pkt;
392 	ipr->ip_pkt = NULL;	/* Avoid free from ip_reasm_free() */
393 	*payload = ptr;
394 
395 	/* Clean up the reassembly list */
396 	while ((ipr = STAILQ_FIRST(&ire_list)) != NULL) {
397 		STAILQ_REMOVE_HEAD(&ire_list, ip_next);
398 		ip_reasm_free(ipr);
399 	}
400 	return (n);
401 }
402 
403 /*
404  * Receive a IP packet.
405  */
406 ssize_t
407 readip(struct iodesc *d, void **pkt, void **payload, time_t tleft,
408     uint8_t proto)
409 {
410 	time_t t;
411 	ssize_t ret = -1;
412 
413 	t = getsecs();
414 	while ((getsecs() - t) < tleft) {
415 		errno = 0;
416 		ret = readipv4(d, pkt, payload, tleft, proto);
417 		if (ret >= 0)
418 			return (ret);
419 		/* Bubble up the error if it wasn't successful */
420 		if (errno != EAGAIN)
421 			return (-1);
422 	}
423 	/* We've exhausted tleft; timeout */
424 	errno = ETIMEDOUT;
425 	return (-1);
426 }
427