gro: Allow tunnel stacking in the case of FOU/GUE
[linux/fpc-iii.git] / net / ipv4 / ip_options.c
blob4d158ff1def1a8fae218b268c7e66224eed95adf
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * The options processing module for ip.c
8 * Authors: A.N.Kuznetsov
12 #define pr_fmt(fmt) "IPv4: " fmt
14 #include <linux/capability.h>
15 #include <linux/module.h>
16 #include <linux/slab.h>
17 #include <linux/types.h>
18 #include <asm/uaccess.h>
19 #include <asm/unaligned.h>
20 #include <linux/skbuff.h>
21 #include <linux/ip.h>
22 #include <linux/icmp.h>
23 #include <linux/netdevice.h>
24 #include <linux/rtnetlink.h>
25 #include <net/sock.h>
26 #include <net/ip.h>
27 #include <net/icmp.h>
28 #include <net/route.h>
29 #include <net/cipso_ipv4.h>
30 #include <net/ip_fib.h>
33 * Write options to IP header, record destination address to
34 * source route option, address of outgoing interface
35 * (we should already know it, so that this function is allowed be
36 * called only after routing decision) and timestamp,
37 * if we originate this datagram.
39 * daddr is real destination address, next hop is recorded in IP header.
40 * saddr is address of outgoing interface.
43 void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
44 __be32 daddr, struct rtable *rt, int is_frag)
46 unsigned char *iph = skb_network_header(skb);
48 memcpy(&(IPCB(skb)->opt), opt, sizeof(struct ip_options));
49 memcpy(iph+sizeof(struct iphdr), opt->__data, opt->optlen);
50 opt = &(IPCB(skb)->opt);
52 if (opt->srr)
53 memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4);
55 if (!is_frag) {
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;
63 midtime = inet_current_timestamp();
64 memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4);
66 return;
68 if (opt->rr) {
69 memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]);
70 opt->rr = 0;
71 opt->rr_needaddr = 0;
73 if (opt->ts) {
74 memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]);
75 opt->ts = 0;
76 opt->ts_needaddr = opt->ts_needtime = 0;
81 * Provided (sopt, skb) points to received options,
82 * build in dopt compiled option set appropriate for answering.
83 * i.e. invert SRR option, copy anothers,
84 * and grab room in RR/TS options.
86 * NOTE: dopt cannot point to skb.
89 int __ip_options_echo(struct ip_options *dopt, struct sk_buff *skb,
90 const struct ip_options *sopt)
92 unsigned char *sptr, *dptr;
93 int soffset, doffset;
94 int optlen;
96 memset(dopt, 0, sizeof(struct ip_options));
98 if (sopt->optlen == 0)
99 return 0;
101 sptr = skb_network_header(skb);
102 dptr = dopt->__data;
104 if (sopt->rr) {
105 optlen = sptr[sopt->rr+1];
106 soffset = sptr[sopt->rr+2];
107 dopt->rr = dopt->optlen + sizeof(struct iphdr);
108 memcpy(dptr, sptr+sopt->rr, optlen);
109 if (sopt->rr_needaddr && soffset <= optlen) {
110 if (soffset + 3 > optlen)
111 return -EINVAL;
112 dptr[2] = soffset + 4;
113 dopt->rr_needaddr = 1;
115 dptr += optlen;
116 dopt->optlen += optlen;
118 if (sopt->ts) {
119 optlen = sptr[sopt->ts+1];
120 soffset = sptr[sopt->ts+2];
121 dopt->ts = dopt->optlen + sizeof(struct iphdr);
122 memcpy(dptr, sptr+sopt->ts, optlen);
123 if (soffset <= optlen) {
124 if (sopt->ts_needaddr) {
125 if (soffset + 3 > optlen)
126 return -EINVAL;
127 dopt->ts_needaddr = 1;
128 soffset += 4;
130 if (sopt->ts_needtime) {
131 if (soffset + 3 > optlen)
132 return -EINVAL;
133 if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
134 dopt->ts_needtime = 1;
135 soffset += 4;
136 } else {
137 dopt->ts_needtime = 0;
139 if (soffset + 7 <= optlen) {
140 __be32 addr;
142 memcpy(&addr, dptr+soffset-1, 4);
143 if (inet_addr_type(dev_net(skb_dst(skb)->dev), addr) != RTN_UNICAST) {
144 dopt->ts_needtime = 1;
145 soffset += 8;
150 dptr[2] = soffset;
152 dptr += optlen;
153 dopt->optlen += optlen;
155 if (sopt->srr) {
156 unsigned char *start = sptr+sopt->srr;
157 __be32 faddr;
159 optlen = start[1];
160 soffset = start[2];
161 doffset = 0;
162 if (soffset > optlen)
163 soffset = optlen + 1;
164 soffset -= 4;
165 if (soffset > 3) {
166 memcpy(&faddr, &start[soffset-1], 4);
167 for (soffset -= 4, doffset = 4; soffset > 3; soffset -= 4, doffset += 4)
168 memcpy(&dptr[doffset-1], &start[soffset-1], 4);
170 * RFC1812 requires to fix illegal source routes.
172 if (memcmp(&ip_hdr(skb)->saddr,
173 &start[soffset + 3], 4) == 0)
174 doffset -= 4;
176 if (doffset > 3) {
177 __be32 daddr = fib_compute_spec_dst(skb);
179 memcpy(&start[doffset-1], &daddr, 4);
180 dopt->faddr = faddr;
181 dptr[0] = start[0];
182 dptr[1] = doffset+3;
183 dptr[2] = 4;
184 dptr += doffset+3;
185 dopt->srr = dopt->optlen + sizeof(struct iphdr);
186 dopt->optlen += doffset+3;
187 dopt->is_strictroute = sopt->is_strictroute;
190 if (sopt->cipso) {
191 optlen = sptr[sopt->cipso+1];
192 dopt->cipso = dopt->optlen+sizeof(struct iphdr);
193 memcpy(dptr, sptr+sopt->cipso, optlen);
194 dptr += optlen;
195 dopt->optlen += optlen;
197 while (dopt->optlen & 3) {
198 *dptr++ = IPOPT_END;
199 dopt->optlen++;
201 return 0;
205 * Options "fragmenting", just fill options not
206 * allowed in fragments with NOOPs.
207 * Simple and stupid 8), but the most efficient way.
210 void ip_options_fragment(struct sk_buff *skb)
212 unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
213 struct ip_options *opt = &(IPCB(skb)->opt);
214 int l = opt->optlen;
215 int optlen;
217 while (l > 0) {
218 switch (*optptr) {
219 case IPOPT_END:
220 return;
221 case IPOPT_NOOP:
222 l--;
223 optptr++;
224 continue;
226 optlen = optptr[1];
227 if (optlen < 2 || optlen > l)
228 return;
229 if (!IPOPT_COPIED(*optptr))
230 memset(optptr, IPOPT_NOOP, optlen);
231 l -= optlen;
232 optptr += optlen;
234 opt->ts = 0;
235 opt->rr = 0;
236 opt->rr_needaddr = 0;
237 opt->ts_needaddr = 0;
238 opt->ts_needtime = 0;
241 /* helper used by ip_options_compile() to call fib_compute_spec_dst()
242 * at most one time.
244 static void spec_dst_fill(__be32 *spec_dst, struct sk_buff *skb)
246 if (*spec_dst == htonl(INADDR_ANY))
247 *spec_dst = fib_compute_spec_dst(skb);
251 * Verify options and fill pointers in struct options.
252 * Caller should clear *opt, and set opt->data.
253 * If opt == NULL, then skb->data should point to IP header.
256 int ip_options_compile(struct net *net,
257 struct ip_options *opt, struct sk_buff *skb)
259 __be32 spec_dst = htonl(INADDR_ANY);
260 unsigned char *pp_ptr = NULL;
261 struct rtable *rt = NULL;
262 unsigned char *optptr;
263 unsigned char *iph;
264 int optlen, l;
266 if (skb) {
267 rt = skb_rtable(skb);
268 optptr = (unsigned char *)&(ip_hdr(skb)[1]);
269 } else
270 optptr = opt->__data;
271 iph = optptr - sizeof(struct iphdr);
273 for (l = opt->optlen; l > 0; ) {
274 switch (*optptr) {
275 case IPOPT_END:
276 for (optptr++, l--; l > 0; optptr++, l--) {
277 if (*optptr != IPOPT_END) {
278 *optptr = IPOPT_END;
279 opt->is_changed = 1;
282 goto eol;
283 case IPOPT_NOOP:
284 l--;
285 optptr++;
286 continue;
288 if (unlikely(l < 2)) {
289 pp_ptr = optptr;
290 goto error;
292 optlen = optptr[1];
293 if (optlen < 2 || optlen > l) {
294 pp_ptr = optptr;
295 goto error;
297 switch (*optptr) {
298 case IPOPT_SSRR:
299 case IPOPT_LSRR:
300 if (optlen < 3) {
301 pp_ptr = optptr + 1;
302 goto error;
304 if (optptr[2] < 4) {
305 pp_ptr = optptr + 2;
306 goto error;
308 /* NB: cf RFC-1812 5.2.4.1 */
309 if (opt->srr) {
310 pp_ptr = optptr;
311 goto error;
313 if (!skb) {
314 if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) {
315 pp_ptr = optptr + 1;
316 goto error;
318 memcpy(&opt->faddr, &optptr[3], 4);
319 if (optlen > 7)
320 memmove(&optptr[3], &optptr[7], optlen-7);
322 opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
323 opt->srr = optptr - iph;
324 break;
325 case IPOPT_RR:
326 if (opt->rr) {
327 pp_ptr = optptr;
328 goto error;
330 if (optlen < 3) {
331 pp_ptr = optptr + 1;
332 goto error;
334 if (optptr[2] < 4) {
335 pp_ptr = optptr + 2;
336 goto error;
338 if (optptr[2] <= optlen) {
339 if (optptr[2]+3 > optlen) {
340 pp_ptr = optptr + 2;
341 goto error;
343 if (rt) {
344 spec_dst_fill(&spec_dst, skb);
345 memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
346 opt->is_changed = 1;
348 optptr[2] += 4;
349 opt->rr_needaddr = 1;
351 opt->rr = optptr - iph;
352 break;
353 case IPOPT_TIMESTAMP:
354 if (opt->ts) {
355 pp_ptr = optptr;
356 goto error;
358 if (optlen < 4) {
359 pp_ptr = optptr + 1;
360 goto error;
362 if (optptr[2] < 5) {
363 pp_ptr = optptr + 2;
364 goto error;
366 if (optptr[2] <= optlen) {
367 unsigned char *timeptr = NULL;
368 if (optptr[2]+3 > optlen) {
369 pp_ptr = optptr + 2;
370 goto error;
372 switch (optptr[3]&0xF) {
373 case IPOPT_TS_TSONLY:
374 if (skb)
375 timeptr = &optptr[optptr[2]-1];
376 opt->ts_needtime = 1;
377 optptr[2] += 4;
378 break;
379 case IPOPT_TS_TSANDADDR:
380 if (optptr[2]+7 > optlen) {
381 pp_ptr = optptr + 2;
382 goto error;
384 if (rt) {
385 spec_dst_fill(&spec_dst, skb);
386 memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
387 timeptr = &optptr[optptr[2]+3];
389 opt->ts_needaddr = 1;
390 opt->ts_needtime = 1;
391 optptr[2] += 8;
392 break;
393 case IPOPT_TS_PRESPEC:
394 if (optptr[2]+7 > optlen) {
395 pp_ptr = optptr + 2;
396 goto error;
399 __be32 addr;
400 memcpy(&addr, &optptr[optptr[2]-1], 4);
401 if (inet_addr_type(net, addr) == RTN_UNICAST)
402 break;
403 if (skb)
404 timeptr = &optptr[optptr[2]+3];
406 opt->ts_needtime = 1;
407 optptr[2] += 8;
408 break;
409 default:
410 if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
411 pp_ptr = optptr + 3;
412 goto error;
414 break;
416 if (timeptr) {
417 __be32 midtime;
419 midtime = inet_current_timestamp();
420 memcpy(timeptr, &midtime, 4);
421 opt->is_changed = 1;
423 } else if ((optptr[3]&0xF) != IPOPT_TS_PRESPEC) {
424 unsigned int overflow = optptr[3]>>4;
425 if (overflow == 15) {
426 pp_ptr = optptr + 3;
427 goto error;
429 if (skb) {
430 optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
431 opt->is_changed = 1;
434 opt->ts = optptr - iph;
435 break;
436 case IPOPT_RA:
437 if (optlen < 4) {
438 pp_ptr = optptr + 1;
439 goto error;
441 if (optptr[2] == 0 && optptr[3] == 0)
442 opt->router_alert = optptr - iph;
443 break;
444 case IPOPT_CIPSO:
445 if ((!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) || opt->cipso) {
446 pp_ptr = optptr;
447 goto error;
449 opt->cipso = optptr - iph;
450 if (cipso_v4_validate(skb, &optptr)) {
451 pp_ptr = optptr;
452 goto error;
454 break;
455 case IPOPT_SEC:
456 case IPOPT_SID:
457 default:
458 if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
459 pp_ptr = optptr;
460 goto error;
462 break;
464 l -= optlen;
465 optptr += optlen;
468 eol:
469 if (!pp_ptr)
470 return 0;
472 error:
473 if (skb) {
474 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
476 return -EINVAL;
478 EXPORT_SYMBOL(ip_options_compile);
481 * Undo all the changes done by ip_options_compile().
484 void ip_options_undo(struct ip_options *opt)
486 if (opt->srr) {
487 unsigned char *optptr = opt->__data+opt->srr-sizeof(struct iphdr);
488 memmove(optptr+7, optptr+3, optptr[1]-7);
489 memcpy(optptr+3, &opt->faddr, 4);
491 if (opt->rr_needaddr) {
492 unsigned char *optptr = opt->__data+opt->rr-sizeof(struct iphdr);
493 optptr[2] -= 4;
494 memset(&optptr[optptr[2]-1], 0, 4);
496 if (opt->ts) {
497 unsigned char *optptr = opt->__data+opt->ts-sizeof(struct iphdr);
498 if (opt->ts_needtime) {
499 optptr[2] -= 4;
500 memset(&optptr[optptr[2]-1], 0, 4);
501 if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
502 optptr[2] -= 4;
504 if (opt->ts_needaddr) {
505 optptr[2] -= 4;
506 memset(&optptr[optptr[2]-1], 0, 4);
511 static struct ip_options_rcu *ip_options_get_alloc(const int optlen)
513 return kzalloc(sizeof(struct ip_options_rcu) + ((optlen + 3) & ~3),
514 GFP_KERNEL);
517 static int ip_options_get_finish(struct net *net, struct ip_options_rcu **optp,
518 struct ip_options_rcu *opt, int optlen)
520 while (optlen & 3)
521 opt->opt.__data[optlen++] = IPOPT_END;
522 opt->opt.optlen = optlen;
523 if (optlen && ip_options_compile(net, &opt->opt, NULL)) {
524 kfree(opt);
525 return -EINVAL;
527 kfree(*optp);
528 *optp = opt;
529 return 0;
532 int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp,
533 unsigned char __user *data, int optlen)
535 struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
537 if (!opt)
538 return -ENOMEM;
539 if (optlen && copy_from_user(opt->opt.__data, data, optlen)) {
540 kfree(opt);
541 return -EFAULT;
543 return ip_options_get_finish(net, optp, opt, optlen);
546 int ip_options_get(struct net *net, struct ip_options_rcu **optp,
547 unsigned char *data, int optlen)
549 struct ip_options_rcu *opt = ip_options_get_alloc(optlen);
551 if (!opt)
552 return -ENOMEM;
553 if (optlen)
554 memcpy(opt->opt.__data, data, optlen);
555 return ip_options_get_finish(net, optp, opt, optlen);
558 void ip_forward_options(struct sk_buff *skb)
560 struct ip_options *opt = &(IPCB(skb)->opt);
561 unsigned char *optptr;
562 struct rtable *rt = skb_rtable(skb);
563 unsigned char *raw = skb_network_header(skb);
565 if (opt->rr_needaddr) {
566 optptr = (unsigned char *)raw + opt->rr;
567 ip_rt_get_source(&optptr[optptr[2]-5], skb, rt);
568 opt->is_changed = 1;
570 if (opt->srr_is_hit) {
571 int srrptr, srrspace;
573 optptr = raw + opt->srr;
575 for ( srrptr = optptr[2], srrspace = optptr[1];
576 srrptr <= srrspace;
577 srrptr += 4
579 if (srrptr + 3 > srrspace)
580 break;
581 if (memcmp(&opt->nexthop, &optptr[srrptr-1], 4) == 0)
582 break;
584 if (srrptr + 3 <= srrspace) {
585 opt->is_changed = 1;
586 ip_hdr(skb)->daddr = opt->nexthop;
587 ip_rt_get_source(&optptr[srrptr-1], skb, rt);
588 optptr[2] = srrptr+4;
589 } else {
590 net_crit_ratelimited("%s(): Argh! Destination lost!\n",
591 __func__);
593 if (opt->ts_needaddr) {
594 optptr = raw + opt->ts;
595 ip_rt_get_source(&optptr[optptr[2]-9], skb, rt);
596 opt->is_changed = 1;
599 if (opt->is_changed) {
600 opt->is_changed = 0;
601 ip_send_check(ip_hdr(skb));
605 int ip_options_rcv_srr(struct sk_buff *skb)
607 struct ip_options *opt = &(IPCB(skb)->opt);
608 int srrspace, srrptr;
609 __be32 nexthop;
610 struct iphdr *iph = ip_hdr(skb);
611 unsigned char *optptr = skb_network_header(skb) + opt->srr;
612 struct rtable *rt = skb_rtable(skb);
613 struct rtable *rt2;
614 unsigned long orefdst;
615 int err;
617 if (!rt)
618 return 0;
620 if (skb->pkt_type != PACKET_HOST)
621 return -EINVAL;
622 if (rt->rt_type == RTN_UNICAST) {
623 if (!opt->is_strictroute)
624 return 0;
625 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
626 return -EINVAL;
628 if (rt->rt_type != RTN_LOCAL)
629 return -EINVAL;
631 for (srrptr = optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
632 if (srrptr + 3 > srrspace) {
633 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
634 return -EINVAL;
636 memcpy(&nexthop, &optptr[srrptr-1], 4);
638 orefdst = skb->_skb_refdst;
639 skb_dst_set(skb, NULL);
640 err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
641 rt2 = skb_rtable(skb);
642 if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
643 skb_dst_drop(skb);
644 skb->_skb_refdst = orefdst;
645 return -EINVAL;
647 refdst_drop(orefdst);
648 if (rt2->rt_type != RTN_LOCAL)
649 break;
650 /* Superfast 8) loopback forward */
651 iph->daddr = nexthop;
652 opt->is_changed = 1;
654 if (srrptr <= srrspace) {
655 opt->srr_is_hit = 1;
656 opt->nexthop = nexthop;
657 opt->is_changed = 1;
659 return 0;
661 EXPORT_SYMBOL(ip_options_rcv_srr);