projects/linux·blob master · refs

645 lines · 14.6 KiB

1// SPDX-License-Identifier: GPL-2.0
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * The options processing module for ip.c
8 *
9 * Authors: A.N.Kuznetsov
10 *
11 */
12
13#define pr_fmt(fmt) "IPv4: " fmt
14
15#include <linux/capability.h>
16#include <linux/module.h>
17#include <linux/slab.h>
18#include <linux/types.h>
19#include <linux/uaccess.h>
20#include <linux/unaligned.h>
21#include <linux/skbuff.h>
22#include <linux/ip.h>
23#include <linux/icmp.h>
24#include <linux/netdevice.h>
25#include <linux/rtnetlink.h>
26#include <net/sock.h>
27#include <net/ip.h>
28#include <net/icmp.h>
29#include <net/route.h>
30#include <net/cipso_ipv4.h>
31#include <net/ip_fib.h>
32
33/*
34 * Write options to IP header, record destination address to
35 * source route option, address of outgoing interface
36 * (we should already know it, so that this function is allowed be
37 * called only after routing decision) and timestamp,
38 * if we originate this datagram.
39 *
40 * daddr is real destination address, next hop is recorded in IP header.
41 * saddr is address of outgoing interface.
42 */
43
44void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
45 __be32 daddr, struct rtable *rt)
46{
47 unsigned char *iph = skb_network_header(skb);
48
49 memcpy(&(IPCB(skb)->opt), opt, sizeof(struct ip_options));
50 memcpy(iph + sizeof(struct iphdr), opt->__data, opt->optlen);
51 opt = &(IPCB(skb)->opt);
52
53 if (opt->srr)
54 memcpy(iph + opt->srr + iph[opt->srr + 1] - 4, &daddr, 4);
55
56 if (opt->rr_needaddr)
57 ip_rt_get_source(iph + opt->rr + iph[opt->rr + 2] - 5, skb, rt);
58 if (opt->ts_needaddr)
59 ip_rt_get_source(iph + opt->ts + iph[opt->ts + 2] - 9, skb, rt);
60 if (opt->ts_needtime) {
61 __be32 midtime;
62
63 midtime = inet_current_timestamp();
64 memcpy(iph + opt->ts + iph[opt->ts + 2] - 5, &midtime, 4);
65 }
66}
67
68/*
69 * Provided (sopt, skb) points to received options,
70 * build in dopt compiled option set appropriate for answering.
71 * i.e. invert SRR option, copy anothers,
72 * and grab room in RR/TS options.
73 *
74 * NOTE: dopt cannot point to skb.
75 */
76
77int __ip_options_echo(struct net *net, struct ip_options *dopt,
78 struct sk_buff *skb, const struct ip_options *sopt)
79{
80 unsigned char *sptr, *dptr;
81 int soffset, doffset;
82 int optlen;
83
84 memset(dopt, 0, sizeof(struct ip_options));
85
86 if (sopt->optlen == 0)
87 return 0;
88
89 sptr = skb_network_header(skb);
90 dptr = dopt->__data;
91
92 if (sopt->rr) {
93 optlen = sptr[sopt->rr+1];
94 soffset = sptr[sopt->rr+2];
95 dopt->rr = dopt->optlen + sizeof(struct iphdr);
96 memcpy(dptr, sptr+sopt->rr, optlen);
97 if (sopt->rr_needaddr && soffset <= optlen) {
98 if (soffset + 3 > optlen)
99 return -EINVAL;
100 dptr[2] = soffset + 4;
101 dopt->rr_needaddr = 1;
102 }
103 dptr += optlen;
104 dopt->optlen += optlen;
105 }
106 if (sopt->ts) {
107 optlen = sptr[sopt->ts+1];
108 soffset = sptr[sopt->ts+2];
109 dopt->ts = dopt->optlen + sizeof(struct iphdr);
110 memcpy(dptr, sptr+sopt->ts, optlen);
111 if (soffset <= optlen) {
112 if (sopt->ts_needaddr) {
113 if (soffset + 3 > optlen)
114 return -EINVAL;
115 dopt->ts_needaddr = 1;
116 soffset += 4;
117 }
118 if (sopt->ts_needtime) {
119 if (soffset + 3 > optlen)
120 return -EINVAL;
121 if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
122 dopt->ts_needtime = 1;
123 soffset += 4;
124 } else {
125 dopt->ts_needtime = 0;
126
127 if (soffset + 7 <= optlen) {
128 __be32 addr;
129
130 memcpy(&addr, dptr+soffset-1, 4);
131 if (inet_addr_type(net, addr) != RTN_UNICAST) {
132 dopt->ts_needtime = 1;
133 soffset += 8;
134 }
135 }
136 }
137 }
138 dptr[2] = soffset;
139 }
140 dptr += optlen;
141 dopt->optlen += optlen;
142 }
143 if (sopt->srr) {
144 unsigned char *start = sptr+sopt->srr;
145 __be32 faddr;
146
147 optlen = start[1];
148 soffset = start[2];
149 doffset = 0;
150 if (soffset > optlen)
151 soffset = optlen + 1;
152 soffset -= 4;
153 if (soffset > 3) {
154 memcpy(&faddr, &start[soffset-1], 4);
155 for (soffset -= 4, doffset = 4; soffset > 3; soffset -= 4, doffset += 4)
156 memcpy(&dptr[doffset-1], &start[soffset-1], 4);
157 /*
158 * RFC1812 requires to fix illegal source routes.
159 */
160 if (memcmp(&ip_hdr(skb)->saddr,
161 &start[soffset + 3], 4) == 0)
162 doffset -= 4;
163 }
164 if (doffset > 3) {
165 dopt->faddr = faddr;
166 dptr[0] = start[0];
167 dptr[1] = doffset+3;
168 dptr[2] = 4;
169 dptr += doffset+3;
170 dopt->srr = dopt->optlen + sizeof(struct iphdr);
171 dopt->optlen += doffset+3;
172 dopt->is_strictroute = sopt->is_strictroute;
173 }
174 }
175 if (sopt->cipso) {
176 optlen = sptr[sopt->cipso+1];
177 dopt->cipso = dopt->optlen+sizeof(struct iphdr);
178 memcpy(dptr, sptr+sopt->cipso, optlen);
179 dptr += optlen;
180 dopt->optlen += optlen;
181 }
182 while (dopt->optlen & 3) {
183 *dptr++ = IPOPT_END;
184 dopt->optlen++;
185 }
186 return 0;
187}
188
189/*
190 * Options "fragmenting", just fill options not
191 * allowed in fragments with NOOPs.
192 * Simple and stupid 8), but the most efficient way.
193 */
194
195void ip_options_fragment(struct sk_buff *skb)
196{
197 unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
198 struct ip_options *opt = &(IPCB(skb)->opt);
199 int l = opt->optlen;
200 int optlen;
201
202 while (l > 0) {
203 switch (*optptr) {
204 case IPOPT_END:
205 return;
206 case IPOPT_NOOP:
207 l--;
208 optptr++;
209 continue;
210 }
211 optlen = optptr[1];
212 if (optlen < 2 || optlen > l)
213 return;
214 if (!IPOPT_COPIED(*optptr))
215 memset(optptr, IPOPT_NOOP, optlen);
216 l -= optlen;
217 optptr += optlen;
218 }
219 opt->ts = 0;
220 opt->rr = 0;
221 opt->rr_needaddr = 0;
222 opt->ts_needaddr = 0;
223 opt->ts_needtime = 0;
224}
225
226/* helper used by ip_options_compile() to call fib_compute_spec_dst()
227 * at most one time.
228 */
229static void spec_dst_fill(__be32 *spec_dst, struct sk_buff *skb)
230{
231 if (*spec_dst == htonl(INADDR_ANY))
232 *spec_dst = fib_compute_spec_dst(skb);
233}
234
235/*
236 * Verify options and fill pointers in struct options.
237 * Caller should clear *opt, and set opt->data.
238 * If opt == NULL, then skb->data should point to IP header.
239 */
240
241int __ip_options_compile(struct net *net,
242 struct ip_options *opt, struct sk_buff *skb,
243 __be32 *info)
244{
245 __be32 spec_dst = htonl(INADDR_ANY);
246 unsigned char *pp_ptr = NULL;
247 struct rtable *rt = NULL;
248 unsigned char *optptr;
249 unsigned char *iph;
250 int optlen, l;
251
252 if (skb) {
253 rt = skb_rtable(skb);
254 optptr = (unsigned char *)&(ip_hdr(skb)[1]);
255 } else
256 optptr = opt->__data;
257 iph = optptr - sizeof(struct iphdr);
258
259 for (l = opt->optlen; l > 0; ) {
260 switch (*optptr) {
261 case IPOPT_END:
262 for (optptr++, l--; l > 0; optptr++, l--) {
263 if (*optptr != IPOPT_END) {
264 *optptr = IPOPT_END;
265 opt->is_changed = 1;
266 }
267 }
268 goto eol;
269 case IPOPT_NOOP:
270 l--;
271 optptr++;
272 continue;
273 }
274 if (unlikely(l < 2)) {
275 pp_ptr = optptr;
276 goto error;
277 }
278 optlen = optptr[1];
279 if (optlen < 2 || optlen > l) {
280 pp_ptr = optptr;
281 goto error;
282 }
283 switch (*optptr) {
284 case IPOPT_SSRR:
285 case IPOPT_LSRR:
286 if (optlen < 3) {
287 pp_ptr = optptr + 1;
288 goto error;
289 }
290 if (optptr[2] < 4) {
291 pp_ptr = optptr + 2;
292 goto error;
293 }
294 /* NB: cf RFC-1812 5.2.4.1 */
295 if (opt->srr) {
296 pp_ptr = optptr;
297 goto error;
298 }
299 if (!skb) {
300 if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) {
301 pp_ptr = optptr + 1;
302 goto error;
303 }
304 memcpy(&opt->faddr, &optptr[3], 4);
305 if (optlen > 7)
306 memmove(&optptr[3], &optptr[7], optlen-7);
307 }
308 opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
309 opt->srr = optptr - iph;
310 break;
311 case IPOPT_RR:
312 if (opt->rr) {
313 pp_ptr = optptr;
314 goto error;
315 }
316 if (optlen < 3) {
317 pp_ptr = optptr + 1;
318 goto error;
319 }
320 if (optptr[2] < 4) {
321 pp_ptr = optptr + 2;
322 goto error;
323 }
324 if (optptr[2] <= optlen) {
325 if (optptr[2]+3 > optlen) {
326 pp_ptr = optptr + 2;
327 goto error;
328 }
329 if (rt) {
330 spec_dst_fill(&spec_dst, skb);
331 memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
332 opt->is_changed = 1;
333 }
334 optptr[2] += 4;
335 opt->rr_needaddr = 1;
336 }
337 opt->rr = optptr - iph;
338 break;
339 case IPOPT_TIMESTAMP:
340 if (opt->ts) {
341 pp_ptr = optptr;
342 goto error;
343 }
344 if (optlen < 4) {
345 pp_ptr = optptr + 1;
346 goto error;
347 }
348 if (optptr[2] < 5) {
349 pp_ptr = optptr + 2;
350 goto error;
351 }
352 if (optptr[2] <= optlen) {
353 unsigned char *timeptr = NULL;
354 if (optptr[2]+3 > optlen) {
355 pp_ptr = optptr + 2;
356 goto error;
357 }
358 switch (optptr[3]&0xF) {
359 case IPOPT_TS_TSONLY:
360 if (skb)
361 timeptr = &optptr[optptr[2]-1];
362 opt->ts_needtime = 1;
363 optptr[2] += 4;
364 break;
365 case IPOPT_TS_TSANDADDR:
366 if (optptr[2]+7 > optlen) {
367 pp_ptr = optptr + 2;
368 goto error;
369 }
370 if (rt) {
371 spec_dst_fill(&spec_dst, skb);
372 memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
373 timeptr = &optptr[optptr[2]+3];
374 }
375 opt->ts_needaddr = 1;
376 opt->ts_needtime = 1;
377 optptr[2] += 8;
378 break;
379 case IPOPT_TS_PRESPEC:
380 if (optptr[2]+7 > optlen) {
381 pp_ptr = optptr + 2;
382 goto error;
383 }
384 {
385 __be32 addr;
386 memcpy(&addr, &optptr[optptr[2]-1], 4);
387 if (inet_addr_type(net, addr) == RTN_UNICAST)
388 break;
389 if (skb)
390 timeptr = &optptr[optptr[2]+3];
391 }
392 opt->ts_needtime = 1;
393 optptr[2] += 8;
394 break;
395 default:
396 if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
397 pp_ptr = optptr + 3;
398 goto error;
399 }
400 break;
401 }
402 if (timeptr) {
403 __be32 midtime;
404
405 midtime = inet_current_timestamp();
406 memcpy(timeptr, &midtime, 4);
407 opt->is_changed = 1;
408 }
409 } else if ((optptr[3]&0xF) != IPOPT_TS_PRESPEC) {
410 unsigned int overflow = optptr[3]>>4;
411 if (overflow == 15) {
412 pp_ptr = optptr + 3;
413 goto error;
414 }
415 if (skb) {
416 optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
417 opt->is_changed = 1;
418 }
419 }
420 opt->ts = optptr - iph;
421 break;
422 case IPOPT_RA:
423 if (optlen < 4) {
424 pp_ptr = optptr + 1;
425 goto error;
426 }
427 if (optptr[2] == 0 && optptr[3] == 0)
428 opt->router_alert = optptr - iph;
429 break;
430 case IPOPT_CIPSO:
431 if ((!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) || opt->cipso) {
432 pp_ptr = optptr;
433 goto error;
434 }
435 opt->cipso = optptr - iph;
436 if (cipso_v4_validate(skb, &optptr)) {
437 pp_ptr = optptr;
438 goto error;
439 }
440 break;
441 case IPOPT_SEC:
442 case IPOPT_SID:
443 default:
444 if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
445 pp_ptr = optptr;
446 goto error;
447 }
448 break;
449 }
450 l -= optlen;
451 optptr += optlen;
452 }
453
454eol:
455 if (!pp_ptr)
456 return 0;
457
458error:
459 if (info)
460 *info = htonl((pp_ptr-iph)<<24);
461 return -EINVAL;
462}
463EXPORT_SYMBOL(__ip_options_compile);
464
465int ip_options_compile(struct net *net,
466 struct ip_options *opt, struct sk_buff *skb)
467{
468 int ret;
469 __be32 info;
470
471 ret = __ip_options_compile(net, opt, skb, &info);
472 if (ret != 0 && skb)
473 icmp_send(skb, ICMP_PARAMETERPROB, 0, info);
474 return ret;
475}
476EXPORT_SYMBOL(ip_options_compile);
477
478/*
479 * Undo all the changes done by ip_options_compile().
480 */
481
482void ip_options_undo(struct ip_options *opt)
483{
484 if (opt->srr) {
485 unsigned char *optptr = opt->__data + opt->srr - sizeof(struct iphdr);
486
487 memmove(optptr + 7, optptr + 3, optptr[1] - 7);
488 memcpy(optptr + 3, &opt->faddr, 4);
489 }
490 if (opt->rr_needaddr) {
491 unsigned char *optptr = opt->__data + opt->rr - sizeof(struct iphdr);
492
493 optptr[2] -= 4;
494 memset(&optptr[optptr[2] - 1], 0, 4);
495 }
496 if (opt->ts) {
497 unsigned char *optptr = opt->__data + opt->ts - sizeof(struct iphdr);
498
499 if (opt->ts_needtime) {
500 optptr[2] -= 4;
501 memset(&optptr[optptr[2] - 1], 0, 4);
502 if ((optptr[3] & 0xF) == IPOPT_TS_PRESPEC)
503 optptr[2] -= 4;
504 }
505 if (opt->ts_needaddr) {
506 optptr[2] -= 4;
507 memset(&optptr[optptr[2] - 1], 0, 4);
508 }
509 }
510}
511
512int ip_options_get(struct net *net, struct ip_options_rcu **optp,
513 sockptr_t data, int optlen)
514{
515 struct ip_options_rcu *opt;
516
517 opt = kzalloc(sizeof(struct ip_options_rcu) + ((optlen + 3) & ~3),
518 GFP_KERNEL);
519 if (!opt)
520 return -ENOMEM;
521 if (optlen && copy_from_sockptr(opt->opt.__data, data, optlen)) {
522 kfree(opt);
523 return -EFAULT;
524 }
525
526 while (optlen & 3)
527 opt->opt.__data[optlen++] = IPOPT_END;
528 opt->opt.optlen = optlen;
529 if (optlen && ip_options_compile(net, &opt->opt, NULL)) {
530 kfree(opt);
531 return -EINVAL;
532 }
533 if (opt->opt.srr && !ns_capable(net->user_ns, CAP_NET_RAW)) {
534 kfree(opt);
535 return -EPERM;
536 }
537 kfree(*optp);
538 *optp = opt;
539 return 0;
540}
541
542void ip_forward_options(struct sk_buff *skb)
543{
544 struct ip_options *opt = &(IPCB(skb)->opt);
545 unsigned char *optptr;
546 struct rtable *rt = skb_rtable(skb);
547 unsigned char *raw = skb_network_header(skb);
548
549 if (opt->rr_needaddr) {
550 optptr = (unsigned char *)raw + opt->rr;
551 ip_rt_get_source(&optptr[optptr[2]-5], skb, rt);
552 opt->is_changed = 1;
553 }
554 if (opt->srr_is_hit) {
555 int srrptr, srrspace;
556
557 optptr = raw + opt->srr;
558
559 for ( srrptr = optptr[2], srrspace = optptr[1];
560 srrptr <= srrspace;
561 srrptr += 4
562 ) {
563 if (srrptr + 3 > srrspace)
564 break;
565 if (memcmp(&opt->nexthop, &optptr[srrptr-1], 4) == 0)
566 break;
567 }
568 if (srrptr + 3 <= srrspace) {
569 opt->is_changed = 1;
570 ip_hdr(skb)->daddr = opt->nexthop;
571 ip_rt_get_source(&optptr[srrptr-1], skb, rt);
572 optptr[2] = srrptr+4;
573 } else {
574 net_crit_ratelimited("%s(): Argh! Destination lost!\n",
575 __func__);
576 }
577 if (opt->ts_needaddr) {
578 optptr = raw + opt->ts;
579 ip_rt_get_source(&optptr[optptr[2]-9], skb, rt);
580 opt->is_changed = 1;
581 }
582 }
583 if (opt->is_changed) {
584 opt->is_changed = 0;
585 ip_send_check(ip_hdr(skb));
586 }
587}
588
589int ip_options_rcv_srr(struct sk_buff *skb, struct net_device *dev)
590{
591 struct ip_options *opt = &(IPCB(skb)->opt);
592 int srrspace, srrptr;
593 __be32 nexthop;
594 struct iphdr *iph = ip_hdr(skb);
595 unsigned char *optptr = skb_network_header(skb) + opt->srr;
596 struct rtable *rt = skb_rtable(skb);
597 struct rtable *rt2;
598 unsigned long orefdst;
599 int err;
600
601 if (!rt)
602 return 0;
603
604 if (skb->pkt_type != PACKET_HOST)
605 return -EINVAL;
606 if (rt->rt_type == RTN_UNICAST) {
607 if (!opt->is_strictroute)
608 return 0;
609 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
610 return -EINVAL;
611 }
612 if (rt->rt_type != RTN_LOCAL)
613 return -EINVAL;
614
615 for (srrptr = optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
616 if (srrptr + 3 > srrspace) {
617 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
618 return -EINVAL;
619 }
620 memcpy(&nexthop, &optptr[srrptr-1], 4);
621
622 orefdst = skb_dstref_steal(skb);
623 err = ip_route_input(skb, nexthop, iph->saddr, ip4h_dscp(iph),
624 dev) ? -EINVAL : 0;
625 rt2 = skb_rtable(skb);
626 if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
627 skb_dst_drop(skb);
628 skb_dstref_restore(skb, orefdst);
629 return -EINVAL;
630 }
631 refdst_drop(orefdst);
632 if (rt2->rt_type != RTN_LOCAL)
633 break;
634 /* Superfast 8) loopback forward */
635 iph->daddr = nexthop;
636 opt->is_changed = 1;
637 }
638 if (srrptr <= srrspace) {
639 opt->srr_is_hit = 1;
640 opt->nexthop = nexthop;
641 opt->is_changed = 1;
642 }
643 return 0;
644}
645EXPORT_SYMBOL(ip_options_rcv_srr);