xref: /illumos-gate/usr/src/boot/libsa/ip.c (revision 608eb926e14f4ba4736b2d59e891335f1cba9e1e)
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 	uchar_t *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 	void *buf;
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 **)&buf, 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 = buf;
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 	ip = buf;
231 	/* Check ip header */
232 	if (ip->ip_v != IPVERSION ||	/* half char */
233 	    ip->ip_p != proto) {
234 #ifdef NET_DEBUG
235 		if (debug) {
236 			printf("readip: IP version or proto. ip_v=%d ip_p=%d\n",
237 			    ip->ip_v, ip->ip_p);
238 		}
239 #endif
240 		free(ptr);
241 		return (-1);
242 	}
243 
244 	hlen = ip->ip_hl << 2;
245 	if (hlen < sizeof (*ip) ||
246 	    in_cksum(ip, hlen) != 0) {
247 #ifdef NET_DEBUG
248 		if (debug)
249 			printf("readip: short hdr or bad cksum.\n");
250 #endif
251 		free(ptr);
252 		return (-1);
253 	}
254 	if (n < ntohs(ip->ip_len)) {
255 #ifdef NET_DEBUG
256 		if (debug) {
257 			printf("readip: bad length %d < %d.\n",
258 			    (int)n, ntohs(ip->ip_len));
259 		}
260 #endif
261 		free(ptr);
262 		return (-1);
263 	}
264 	if (d->myip.s_addr && ip->ip_dst.s_addr != d->myip.s_addr) {
265 #ifdef NET_DEBUG
266 		if (debug) {
267 			printf("readip: bad saddr %s != ", inet_ntoa(d->myip));
268 			printf("%s\n", inet_ntoa(ip->ip_dst));
269 		}
270 #endif
271 		free(ptr);
272 		return (-1);
273 	}
274 
275 	/* Unfragmented packet. */
276 	if ((ntohs(ip->ip_off) & IP_MF) == 0 &&
277 	    (ntohs(ip->ip_off) & IP_OFFMASK) == 0) {
278 		uh = (struct udphdr *)((uintptr_t)ip + sizeof (*ip));
279 		/* If there were ip options, make them go away */
280 		if (hlen != sizeof (*ip)) {
281 			bcopy(((uchar_t *)ip) + hlen, uh, uh->uh_ulen - hlen);
282 			ip->ip_len = htons(sizeof (*ip));
283 			n -= hlen - sizeof (*ip);
284 		}
285 
286 		n = (n > (ntohs(ip->ip_len) - sizeof (*ip))) ?
287 		    ntohs(ip->ip_len) - sizeof (*ip) : n;
288 		*pkt = ptr;
289 		*payload = (void *)((uintptr_t)ip + sizeof (*ip));
290 		return (n);
291 	}
292 
293 	STAILQ_FOREACH(ipr, &ire_list, ip_next) {
294 		if (ipr->ip_src.s_addr == ip->ip_src.s_addr &&
295 		    ipr->ip_dst.s_addr == ip->ip_dst.s_addr &&
296 		    ipr->ip_id == ip->ip_id &&
297 		    ipr->ip_proto == ip->ip_p)
298 			break;
299 	}
300 
301 	/* Allocate new reassembly entry */
302 	if (ipr == NULL) {
303 		if ((ipr = calloc(1, sizeof (*ipr))) == NULL) {
304 			free(ptr);
305 			return (-1);
306 		}
307 
308 		ipr->ip_src = ip->ip_src;
309 		ipr->ip_dst = ip->ip_dst;
310 		ipr->ip_id = ip->ip_id;
311 		ipr->ip_proto = ip->ip_p;
312 		ipr->ip_ttl = MAXTTL;
313 		STAILQ_INIT(&ipr->ip_queue);
314 		STAILQ_INSERT_TAIL(&ire_list, ipr, ip_next);
315 	}
316 
317 	if (ip_reasm_add(ipr, ptr, ip) != 0) {
318 		STAILQ_REMOVE(&ire_list, ipr, ip_reasm, ip_next);
319 		free(ipr);
320 		free(ptr);
321 		return (-1);
322 	}
323 
324 	if ((ntohs(ip->ip_off) & IP_MF) == 0) {
325 		ipr->ip_total_size = (8 * (ntohs(ip->ip_off) & IP_OFFMASK));
326 		ipr->ip_total_size += n + sizeof (*ip);
327 		ipr->ip_total_size += sizeof (struct ether_header);
328 
329 		ipr->ip_pkt = malloc(ipr->ip_total_size + 2);
330 		if (ipr->ip_pkt == NULL) {
331 			STAILQ_REMOVE(&ire_list, ipr, ip_reasm, ip_next);
332 			ip_reasm_free(ipr);
333 			return (-1);
334 		}
335 	}
336 
337 	/*
338 	 * If we do not have re-assembly buffer ipr->ip_pkt, we are still
339 	 * missing fragments, so just restart the read.
340 	 */
341 	if (ipr->ip_pkt == NULL) {
342 		errno = EAGAIN;
343 		return (-1);
344 	}
345 
346 	/*
347 	 * Walk the packet list in reassembly queue, if we got all the
348 	 * fragments, build the packet.
349 	 */
350 	n = 0;
351 	last = NULL;
352 	STAILQ_FOREACH(ipq, &ipr->ip_queue, ipq_next) {
353 		if ((ntohs(ipq->ipq_hdr->ip_off) & IP_OFFMASK) != n / 8) {
354 			STAILQ_REMOVE(&ire_list, ipr, ip_reasm, ip_next);
355 			ip_reasm_free(ipr);
356 			return (-1);
357 		}
358 
359 		n += ntohs(ipq->ipq_hdr->ip_len) - (ipq->ipq_hdr->ip_hl << 2);
360 		last = ipq;
361 	}
362 	if ((ntohs(last->ipq_hdr->ip_off) & IP_MF) != 0) {
363 		errno = EAGAIN;
364 		return (-1);
365 	}
366 
367 	ipq = STAILQ_FIRST(&ipr->ip_queue);
368 	/* Fabricate ethernet header */
369 	eh = (struct ether_header *)((uintptr_t)ipr->ip_pkt + 2);
370 	bcopy((void *)((uintptr_t)ipq->ipq_pkt + 2), eh, sizeof (*eh));
371 
372 	/* Fabricate IP header */
373 	ipr->ip_hdr = (struct ip *)((uintptr_t)eh + sizeof (*eh));
374 	bcopy(ipq->ipq_hdr, ipr->ip_hdr, sizeof (*ipr->ip_hdr));
375 	ipr->ip_hdr->ip_hl = sizeof (*ipr->ip_hdr) >> 2;
376 	ipr->ip_hdr->ip_len = htons(n);
377 	ipr->ip_hdr->ip_sum = 0;
378 	ipr->ip_hdr->ip_sum = in_cksum(ipr->ip_hdr, sizeof (*ipr->ip_hdr));
379 
380 	n = 0;
381 	ptr = (char *)((uintptr_t)ipr->ip_hdr + sizeof (*ipr->ip_hdr));
382 	STAILQ_FOREACH(ipq, &ipr->ip_queue, ipq_next) {
383 		char *data;
384 		size_t len;
385 
386 		hlen = ipq->ipq_hdr->ip_hl << 2;
387 		len = ntohs(ipq->ipq_hdr->ip_len) - hlen;
388 		data = (char *)((uintptr_t)ipq->ipq_hdr + hlen);
389 
390 		bcopy(data, ptr + n, len);
391 		n += len;
392 	}
393 
394 	*pkt = ipr->ip_pkt;
395 	ipr->ip_pkt = NULL;	/* Avoid free from ip_reasm_free() */
396 	*payload = ptr;
397 
398 	/* Clean up the reassembly list */
399 	while ((ipr = STAILQ_FIRST(&ire_list)) != NULL) {
400 		STAILQ_REMOVE_HEAD(&ire_list, ip_next);
401 		ip_reasm_free(ipr);
402 	}
403 	return (n);
404 }
405 
406 /*
407  * Receive a IP packet.
408  */
409 ssize_t
410 readip(struct iodesc *d, void **pkt, void **payload, time_t tleft,
411     uint8_t proto)
412 {
413 	time_t t;
414 	ssize_t ret = -1;
415 
416 	t = getsecs();
417 	while ((getsecs() - t) < tleft) {
418 		errno = 0;
419 		ret = readipv4(d, pkt, payload, tleft, proto);
420 		if (ret >= 0)
421 			return (ret);
422 		/* Bubble up the error if it wasn't successful */
423 		if (errno != EAGAIN)
424 			return (-1);
425 	}
426 	/* We've exhausted tleft; timeout */
427 	errno = ETIMEDOUT;
428 	return (-1);
429 }
430