xref: /illumos-gate/usr/src/boot/libsa/ip.c (revision 4db51db7ef7ac07e35da1ac4db463882c1899e83)
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 #include <stdbool.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 	uchar_t *ea;
94 
95 #ifdef NET_DEBUG
96 	if (debug) {
97 		printf("sendip: proto: %x d=%p called.\n", proto, (void *)d);
98 		printf("saddr: %s:%d daddr: %s:%d\n",
99 		    inet_ntoa(d->myip), ntohs(d->myport),
100 		    inet_ntoa(d->destip), ntohs(d->destport));
101 	}
102 #endif
103 
104 	ip = (struct ip *)pkt - 1;
105 	len += sizeof (*ip);
106 
107 	bzero(ip, sizeof (*ip));
108 
109 	ip->ip_v = IPVERSION;			/* half-char */
110 	ip->ip_hl = sizeof (*ip) >> 2;		/* half-char */
111 	ip->ip_len = htons(len);
112 	ip->ip_p = proto;			/* char */
113 	ip->ip_ttl = IPDEFTTL;			/* char */
114 	ip->ip_src = d->myip;
115 	ip->ip_dst = d->destip;
116 	ip->ip_sum = in_cksum(ip, sizeof (*ip)); /* short, but special */
117 
118 	if (ip->ip_dst.s_addr == INADDR_BROADCAST || ip->ip_src.s_addr == 0 ||
119 	    netmask == 0 || SAMENET(ip->ip_src, ip->ip_dst, netmask))
120 		ea = arpwhohas(d, ip->ip_dst);
121 	else
122 		ea = arpwhohas(d, gateip);
123 
124 	cc = sendether(d, ip, len, ea, ETHERTYPE_IP);
125 	if (cc == -1)
126 		return (-1);
127 	if (cc != len)
128 		panic("sendip: bad write (%zd != %zd)", cc, len);
129 	return (cc - sizeof (*ip));
130 }
131 
132 static void
133 ip_reasm_free(struct ip_reasm *ipr)
134 {
135 	struct ip_queue *ipq;
136 
137 	while ((ipq = STAILQ_FIRST(&ipr->ip_queue)) != NULL) {
138 		STAILQ_REMOVE_HEAD(&ipr->ip_queue, ipq_next);
139 		free(ipq->ipq_pkt);
140 		free(ipq);
141 	}
142 	free(ipr->ip_pkt);
143 	free(ipr);
144 }
145 
146 static bool
147 ip_reasm_add(struct ip_reasm *ipr, void *pkt, struct ip *ip)
148 {
149 	struct ip_queue *ipq, *p;
150 	uint16_t off_q, off_ip;
151 
152 	if ((ipq = calloc(1, sizeof (*ipq))) == NULL)
153 		return (false);
154 
155 	ipq->ipq_pkt = pkt;
156 	ipq->ipq_hdr = ip;
157 
158 	STAILQ_FOREACH(p, &ipr->ip_queue, ipq_next) {
159 		off_q = ntohs(p->ipq_hdr->ip_off) & IP_OFFMASK;
160 		off_ip = ntohs(ip->ip_off) & IP_OFFMASK;
161 
162 		if (off_q == off_ip) {	/* duplicate */
163 			free(pkt);
164 			free(ipq);
165 			return (true);
166 		}
167 
168 		if (off_ip < off_q) {
169 			/*
170 			 * Everything in queue has larger offset,
171 			 * drop out of loop and insert to HEAD.
172 			 */
173 			break;
174 		}
175 
176 		/*
177 		 * p in queue is smaller than ip, check if we need to put
178 		 * ip after p or after p->next.
179 		 */
180 		struct ip_queue *next = STAILQ_NEXT(p, ipq_next);
181 		if (next == NULL) {
182 			/* insert after p */
183 			STAILQ_INSERT_AFTER(&ipr->ip_queue, p, ipq, ipq_next);
184 			return (true);
185 		}
186 
187 		off_q = ntohs(next->ipq_hdr->ip_off) & IP_OFFMASK;
188 		if (off_ip < off_q) {
189 			/* next fragment offset is larger, insert after p. */
190 			STAILQ_INSERT_AFTER(&ipr->ip_queue, p, ipq, ipq_next);
191 			return (true);
192 		}
193 		/* next fragment offset is smaller, loop */
194 	}
195 	STAILQ_INSERT_HEAD(&ipr->ip_queue, ipq, ipq_next);
196 	return (true);
197 }
198 
199 /*
200  * Receive a IP packet and validate it is for us.
201  */
202 static ssize_t
203 readipv4(struct iodesc *d, void **pkt, void **payload, ssize_t n)
204 {
205 	struct ip *ip = *payload;
206 	size_t hlen;
207 	struct ether_header *eh;
208 	struct udphdr *uh;
209 	char *ptr = *pkt;
210 	struct ip_reasm *ipr;
211 	struct ip_queue *ipq, *last;
212 	bool morefrag, isfrag;
213 	uint16_t fragoffset;
214 
215 	if (n < sizeof (*ip)) {
216 		free(ptr);
217 		errno = EAGAIN;	/* Call me again. */
218 		return (-1);
219 	}
220 
221 	hlen = ip->ip_hl << 2;
222 	if (hlen < sizeof (*ip) ||
223 	    in_cksum(ip, hlen) != 0) {
224 #ifdef NET_DEBUG
225 		if (debug)
226 			printf("%s: short hdr or bad cksum.\n", __func__);
227 #endif
228 		free(ptr);
229 		errno = EAGAIN;	/* Call me again. */
230 		return (-1);
231 	}
232 
233 	if (n < ntohs(ip->ip_len)) {
234 #ifdef NET_DEBUG
235 		if (debug) {
236 			printf("readip: bad length %dz < %d.\n",
237 			    n, ntohs(ip->ip_len));
238 		}
239 #endif
240 		free(ptr);
241 		errno = EAGAIN;	/* Call me again. */
242 		return (-1);
243 	}
244 
245 	fragoffset = (ntohs(ip->ip_off) & IP_OFFMASK) * 8;
246 	morefrag = (ntohs(ip->ip_off) & IP_MF) == 0 ? false : true;
247 	isfrag = morefrag || fragoffset != 0;
248 
249 	uh = (struct udphdr *)((uintptr_t)ip + sizeof (*ip));
250 
251 	if (d->myip.s_addr && ip->ip_dst.s_addr != d->myip.s_addr) {
252 #ifdef NET_DEBUG
253 		if (debug) {
254 			printf("%s: not for us: saddr %s (%d) != %s (%d)\n",
255 			    __func__, inet_ntoa(d->myip), ntohs(d->myport),
256 			    inet_ntoa(ip->ip_dst), ntohs(uh->uh_dport));
257 		}
258 #endif
259 		free(ptr);
260 		errno = EAGAIN;	/* Call me again. */
261 		return (-1);
262 	}
263 
264 	/* Unfragmented packet. */
265 	if (!isfrag) {
266 #ifdef NET_DEBUG
267 		if (debug) {
268 			printf("%s: unfragmented saddr %s:%d -> ",
269 			    __func__,
270 			    inet_ntoa(ip->ip_src), ntohs(uh->uh_sport));
271 			printf("%s:%d\n",
272 			    inet_ntoa(ip->ip_dst), ntohs(uh->uh_dport));
273 		}
274 #endif
275 		/* If there were ip options, make them go away */
276 		if (hlen != sizeof (*ip)) {
277 			bcopy(((uchar_t *)ip) + hlen, uh,
278 			    ntohs(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 #ifdef NET_DEBUG
313 		if (debug) {
314 			printf("%s: new reassembly ID=%d %s -> ",
315 			    __func__, ntohs(ip->ip_id), inet_ntoa(ip->ip_src));
316 			printf("%s\n", inet_ntoa(ip->ip_dst));
317 		}
318 #endif
319 	}
320 
321 	/*
322 	 * NOTE: with ip_reasm_add() ptr will be stored in reassembly
323 	 * queue and we can not free it without destroying the queue.
324 	 */
325 	if (!ip_reasm_add(ipr, ptr, ip)) {
326 		STAILQ_REMOVE(&ire_list, ipr, ip_reasm, ip_next);
327 		free(ipr);
328 		free(ptr);
329 		return (-1);
330 	}
331 
332 	/*
333 	 * Walk the packet list in reassembly queue, if we got all the
334 	 * fragments, build the packet.
335 	 */
336 	n = 0;
337 	last = NULL;
338 	STAILQ_FOREACH(ipq, &ipr->ip_queue, ipq_next) {
339 		fragoffset = (ntohs(ipq->ipq_hdr->ip_off) & IP_OFFMASK) * 8;
340 		if (fragoffset != n) {
341 #ifdef NET_DEBUG
342 			if (debug) {
343 				printf("%s: need more fragments %d %s -> ",
344 				    __func__, ntohs(ipq->ipq_hdr->ip_id),
345 				    inet_ntoa(ipq->ipq_hdr->ip_src));
346 				printf("%s offset=%d MF=%d\n",
347 				    inet_ntoa(ipq->ipq_hdr->ip_dst),
348 				    fragoffset,
349 				    (ntohs(ipq->ipq_hdr->ip_off) & IP_MF) != 0);
350 			}
351 #endif
352 			errno = EAGAIN;
353 			return (-1);
354 		}
355 
356 		n += ntohs(ipq->ipq_hdr->ip_len) - (ipq->ipq_hdr->ip_hl << 2);
357 		last = ipq;
358 	}
359 
360 	/* complete queue has last packet with MF 0 */
361 	if ((ntohs(last->ipq_hdr->ip_off) & IP_MF) != 0) {
362 #ifdef NET_DEBUG
363 		if (debug) {
364 			printf("%s: need more fragments %d %s -> ",
365 			    __func__, ntohs(last->ipq_hdr->ip_id),
366 			    inet_ntoa(last->ipq_hdr->ip_src));
367 			printf("%s offset=%d MF=%d\n",
368 			    inet_ntoa(last->ipq_hdr->ip_dst),
369 			    (ntohs(last->ipq_hdr->ip_off) & IP_OFFMASK) * 8,
370 			    (ntohs(last->ipq_hdr->ip_off) & IP_MF) != 0);
371 		}
372 #endif
373 		errno = EAGAIN;
374 		return (-1);
375 	}
376 
377 	ipr->ip_total_size = n + sizeof (*ip) + sizeof (struct ether_header);
378 	ipr->ip_pkt = malloc(ipr->ip_total_size + 2);
379 	if (ipr->ip_pkt == NULL) {
380 		STAILQ_REMOVE(&ire_list, ipr, ip_reasm, ip_next);
381 		ip_reasm_free(ipr);
382 		return (-1);
383 	}
384 
385 	ipq = STAILQ_FIRST(&ipr->ip_queue);
386 	/* Fabricate ethernet header */
387 	eh = (struct ether_header *)((uintptr_t)ipr->ip_pkt + 2);
388 	bcopy((void *)((uintptr_t)ipq->ipq_pkt + 2), eh, sizeof (*eh));
389 
390 	/* Fabricate IP header */
391 	ipr->ip_hdr = (struct ip *)((uintptr_t)eh + sizeof (*eh));
392 	bcopy(ipq->ipq_hdr, ipr->ip_hdr, sizeof (*ipr->ip_hdr));
393 	ipr->ip_hdr->ip_hl = sizeof (*ipr->ip_hdr) >> 2;
394 	ipr->ip_hdr->ip_len = htons(n);
395 	ipr->ip_hdr->ip_sum = 0;
396 	ipr->ip_hdr->ip_sum = in_cksum(ipr->ip_hdr, sizeof (*ipr->ip_hdr));
397 
398 	n = 0;
399 	ptr = (char *)((uintptr_t)ipr->ip_hdr + sizeof (*ipr->ip_hdr));
400 	STAILQ_FOREACH(ipq, &ipr->ip_queue, ipq_next) {
401 		char *data;
402 		size_t len;
403 
404 		hlen = ipq->ipq_hdr->ip_hl << 2;
405 		len = ntohs(ipq->ipq_hdr->ip_len) - hlen;
406 		data = (char *)((uintptr_t)ipq->ipq_hdr + hlen);
407 
408 		bcopy(data, ptr + n, len);
409 		n += len;
410 	}
411 
412 	*pkt = ipr->ip_pkt;
413 	ipr->ip_pkt = NULL;	/* Avoid free from ip_reasm_free() */
414 	*payload = ptr;
415 
416 	/* Clean up the reassembly list */
417 	while ((ipr = STAILQ_FIRST(&ire_list)) != NULL) {
418 		STAILQ_REMOVE_HEAD(&ire_list, ip_next);
419 		ip_reasm_free(ipr);
420 	}
421 #ifdef NET_DEBUG
422 	if (debug) {
423 		printf("%s: completed fragments ID=%d %s -> %s\n",
424 		    __func__, ntohs(ip->ip_id), inet_ntoa(ip->ip_src),
425 		    inet_ntoa(ip->ip_dst));
426 	}
427 #endif
428 	return (n);
429 }
430 
431 /*
432  * Receive a IP packet.
433  */
434 ssize_t
435 readip(struct iodesc *d, void **pkt, void **payload, time_t tleft,
436     uint8_t proto)
437 {
438 	time_t t;
439 	ssize_t ret = -1;
440 
441 	t = getsecs();
442 	while ((getsecs() - t) < tleft) {
443 		ssize_t n;
444 		uint16_t etype;	/* host order */
445 		void *ptr = NULL;
446 		void *data = NULL;
447 
448 		errno = 0;
449 		n = readether(d, &ptr, &data, tleft, &etype);
450 		if (n == -1) {
451 			free(ptr);
452 			continue;
453 		}
454 		/* Ethernet address checks are done in readether() */
455 
456 		/* Need to respond to ARP requests. */
457 		if (etype == ETHERTYPE_ARP) {
458 			struct arphdr *ah = data;
459 
460 #ifdef NET_DEBUG
461 			if (debug)
462 				printf("%s: ARP request\n", __func__);
463 #endif
464 
465 			if (ah->ar_op == htons(ARPOP_REQUEST)) {
466 				/* Send ARP reply */
467 				arp_reply(d, ah);
468 			}
469 			free(ptr);
470 			continue;	/* Get next packet */
471 		}
472 
473 		if (etype == ETHERTYPE_IP) {
474 			struct ip *ip = data;
475 
476 			if (ip->ip_v == IPVERSION &&	/* half char */
477 			    ip->ip_p == proto) {
478 				errno = 0;
479 				ret = readipv4(d, &ptr, &data, n);
480 				if (ret >= 0) {
481 					*pkt = ptr;
482 					*payload = data;
483 					return (ret);
484 				}
485 
486 				/*
487 				 * Bubble up the error if it wasn't successful
488 				 */
489 				if (errno != EAGAIN)
490 					return (-1);
491 				continue;
492 			}
493 #ifdef NET_DEBUG
494 			if (debug) {
495 				printf("%s: IP version or proto. "
496 				    "ip_v=%d ip_p=%d\n",
497 				    __func__, ip->ip_v, ip->ip_p);
498 			}
499 #endif
500 			free(ptr);
501 			continue;
502 		}
503 		free(ptr);
504 	}
505 	/* We've exhausted tleft; timeout */
506 	errno = ETIMEDOUT;
507 #ifdef NET_DEBUG
508 	if (debug) {
509 		printf("%s: timeout\n", __func__);
510 	}
511 #endif
512 	return (-1);
513 }
514