net: add a noref bit on skb dst
[linux-2.6.git] / net / ipv4 / ip_options.c
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.
5  *
6  *              The options processing module for ip.c
7  *
8  * Authors:     A.N.Kuznetsov
9  *
10  */
11
12 #include <linux/capability.h>
13 #include <linux/module.h>
14 #include <linux/slab.h>
15 #include <linux/types.h>
16 #include <asm/uaccess.h>
17 #include <linux/skbuff.h>
18 #include <linux/ip.h>
19 #include <linux/icmp.h>
20 #include <linux/netdevice.h>
21 #include <linux/rtnetlink.h>
22 #include <net/sock.h>
23 #include <net/ip.h>
24 #include <net/icmp.h>
25 #include <net/route.h>
26 #include <net/cipso_ipv4.h>
27
28 /*
29  * Write options to IP header, record destination address to
30  * source route option, address of outgoing interface
31  * (we should already know it, so that this  function is allowed be
32  * called only after routing decision) and timestamp,
33  * if we originate this datagram.
34  *
35  * daddr is real destination address, next hop is recorded in IP header.
36  * saddr is address of outgoing interface.
37  */
38
39 void ip_options_build(struct sk_buff * skb, struct ip_options * opt,
40                             __be32 daddr, struct rtable *rt, int is_frag)
41 {
42         unsigned char *iph = skb_network_header(skb);
43
44         memcpy(&(IPCB(skb)->opt), opt, sizeof(struct ip_options));
45         memcpy(iph+sizeof(struct iphdr), opt->__data, opt->optlen);
46         opt = &(IPCB(skb)->opt);
47
48         if (opt->srr)
49                 memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4);
50
51         if (!is_frag) {
52                 if (opt->rr_needaddr)
53                         ip_rt_get_source(iph+opt->rr+iph[opt->rr+2]-5, rt);
54                 if (opt->ts_needaddr)
55                         ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, rt);
56                 if (opt->ts_needtime) {
57                         struct timespec tv;
58                         __be32 midtime;
59                         getnstimeofday(&tv);
60                         midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
61                         memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4);
62                 }
63                 return;
64         }
65         if (opt->rr) {
66                 memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]);
67                 opt->rr = 0;
68                 opt->rr_needaddr = 0;
69         }
70         if (opt->ts) {
71                 memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]);
72                 opt->ts = 0;
73                 opt->ts_needaddr = opt->ts_needtime = 0;
74         }
75 }
76
77 /*
78  * Provided (sopt, skb) points to received options,
79  * build in dopt compiled option set appropriate for answering.
80  * i.e. invert SRR option, copy anothers,
81  * and grab room in RR/TS options.
82  *
83  * NOTE: dopt cannot point to skb.
84  */
85
86 int ip_options_echo(struct ip_options * dopt, struct sk_buff * skb)
87 {
88         struct ip_options *sopt;
89         unsigned char *sptr, *dptr;
90         int soffset, doffset;
91         int     optlen;
92         __be32  daddr;
93
94         memset(dopt, 0, sizeof(struct ip_options));
95
96         sopt = &(IPCB(skb)->opt);
97
98         if (sopt->optlen == 0) {
99                 dopt->optlen = 0;
100                 return 0;
101         }
102
103         sptr = skb_network_header(skb);
104         dptr = dopt->__data;
105
106         daddr = skb_rtable(skb)->rt_spec_dst;
107
108         if (sopt->rr) {
109                 optlen  = sptr[sopt->rr+1];
110                 soffset = sptr[sopt->rr+2];
111                 dopt->rr = dopt->optlen + sizeof(struct iphdr);
112                 memcpy(dptr, sptr+sopt->rr, optlen);
113                 if (sopt->rr_needaddr && soffset <= optlen) {
114                         if (soffset + 3 > optlen)
115                                 return -EINVAL;
116                         dptr[2] = soffset + 4;
117                         dopt->rr_needaddr = 1;
118                 }
119                 dptr += optlen;
120                 dopt->optlen += optlen;
121         }
122         if (sopt->ts) {
123                 optlen = sptr[sopt->ts+1];
124                 soffset = sptr[sopt->ts+2];
125                 dopt->ts = dopt->optlen + sizeof(struct iphdr);
126                 memcpy(dptr, sptr+sopt->ts, optlen);
127                 if (soffset <= optlen) {
128                         if (sopt->ts_needaddr) {
129                                 if (soffset + 3 > optlen)
130                                         return -EINVAL;
131                                 dopt->ts_needaddr = 1;
132                                 soffset += 4;
133                         }
134                         if (sopt->ts_needtime) {
135                                 if (soffset + 3 > optlen)
136                                         return -EINVAL;
137                                 if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
138                                         dopt->ts_needtime = 1;
139                                         soffset += 4;
140                                 } else {
141                                         dopt->ts_needtime = 0;
142
143                                         if (soffset + 8 <= optlen) {
144                                                 __be32 addr;
145
146                                                 memcpy(&addr, sptr+soffset-1, 4);
147                                                 if (inet_addr_type(dev_net(skb_dst(skb)->dev), addr) != RTN_LOCAL) {
148                                                         dopt->ts_needtime = 1;
149                                                         soffset += 8;
150                                                 }
151                                         }
152                                 }
153                         }
154                         dptr[2] = soffset;
155                 }
156                 dptr += optlen;
157                 dopt->optlen += optlen;
158         }
159         if (sopt->srr) {
160                 unsigned char * start = sptr+sopt->srr;
161                 __be32 faddr;
162
163                 optlen  = start[1];
164                 soffset = start[2];
165                 doffset = 0;
166                 if (soffset > optlen)
167                         soffset = optlen + 1;
168                 soffset -= 4;
169                 if (soffset > 3) {
170                         memcpy(&faddr, &start[soffset-1], 4);
171                         for (soffset-=4, doffset=4; soffset > 3; soffset-=4, doffset+=4)
172                                 memcpy(&dptr[doffset-1], &start[soffset-1], 4);
173                         /*
174                          * RFC1812 requires to fix illegal source routes.
175                          */
176                         if (memcmp(&ip_hdr(skb)->saddr,
177                                    &start[soffset + 3], 4) == 0)
178                                 doffset -= 4;
179                 }
180                 if (doffset > 3) {
181                         memcpy(&start[doffset-1], &daddr, 4);
182                         dopt->faddr = faddr;
183                         dptr[0] = start[0];
184                         dptr[1] = doffset+3;
185                         dptr[2] = 4;
186                         dptr += doffset+3;
187                         dopt->srr = dopt->optlen + sizeof(struct iphdr);
188                         dopt->optlen += doffset+3;
189                         dopt->is_strictroute = sopt->is_strictroute;
190                 }
191         }
192         if (sopt->cipso) {
193                 optlen  = sptr[sopt->cipso+1];
194                 dopt->cipso = dopt->optlen+sizeof(struct iphdr);
195                 memcpy(dptr, sptr+sopt->cipso, optlen);
196                 dptr += optlen;
197                 dopt->optlen += optlen;
198         }
199         while (dopt->optlen & 3) {
200                 *dptr++ = IPOPT_END;
201                 dopt->optlen++;
202         }
203         return 0;
204 }
205
206 /*
207  *      Options "fragmenting", just fill options not
208  *      allowed in fragments with NOOPs.
209  *      Simple and stupid 8), but the most efficient way.
210  */
211
212 void ip_options_fragment(struct sk_buff * skb)
213 {
214         unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
215         struct ip_options * opt = &(IPCB(skb)->opt);
216         int  l = opt->optlen;
217         int  optlen;
218
219         while (l > 0) {
220                 switch (*optptr) {
221                 case IPOPT_END:
222                         return;
223                 case IPOPT_NOOP:
224                         l--;
225                         optptr++;
226                         continue;
227                 }
228                 optlen = optptr[1];
229                 if (optlen<2 || optlen>l)
230                   return;
231                 if (!IPOPT_COPIED(*optptr))
232                         memset(optptr, IPOPT_NOOP, optlen);
233                 l -= optlen;
234                 optptr += optlen;
235         }
236         opt->ts = 0;
237         opt->rr = 0;
238         opt->rr_needaddr = 0;
239         opt->ts_needaddr = 0;
240         opt->ts_needtime = 0;
241         return;
242 }
243
244 /*
245  * Verify options and fill pointers in struct options.
246  * Caller should clear *opt, and set opt->data.
247  * If opt == NULL, then skb->data should point to IP header.
248  */
249
250 int ip_options_compile(struct net *net,
251                        struct ip_options * opt, struct sk_buff * skb)
252 {
253         int l;
254         unsigned char * iph;
255         unsigned char * optptr;
256         int optlen;
257         unsigned char * pp_ptr = NULL;
258         struct rtable *rt = NULL;
259
260         if (skb != NULL) {
261                 rt = skb_rtable(skb);
262                 optptr = (unsigned char *)&(ip_hdr(skb)[1]);
263         } else
264                 optptr = opt->__data;
265         iph = optptr - sizeof(struct iphdr);
266
267         for (l = opt->optlen; l > 0; ) {
268                 switch (*optptr) {
269                       case IPOPT_END:
270                         for (optptr++, l--; l>0; optptr++, l--) {
271                                 if (*optptr != IPOPT_END) {
272                                         *optptr = IPOPT_END;
273                                         opt->is_changed = 1;
274                                 }
275                         }
276                         goto eol;
277                       case IPOPT_NOOP:
278                         l--;
279                         optptr++;
280                         continue;
281                 }
282                 optlen = optptr[1];
283                 if (optlen<2 || optlen>l) {
284                         pp_ptr = optptr;
285                         goto error;
286                 }
287                 switch (*optptr) {
288                       case IPOPT_SSRR:
289                       case IPOPT_LSRR:
290                         if (optlen < 3) {
291                                 pp_ptr = optptr + 1;
292                                 goto error;
293                         }
294                         if (optptr[2] < 4) {
295                                 pp_ptr = optptr + 2;
296                                 goto error;
297                         }
298                         /* NB: cf RFC-1812 5.2.4.1 */
299                         if (opt->srr) {
300                                 pp_ptr = optptr;
301                                 goto error;
302                         }
303                         if (!skb) {
304                                 if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) {
305                                         pp_ptr = optptr + 1;
306                                         goto error;
307                                 }
308                                 memcpy(&opt->faddr, &optptr[3], 4);
309                                 if (optlen > 7)
310                                         memmove(&optptr[3], &optptr[7], optlen-7);
311                         }
312                         opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
313                         opt->srr = optptr - iph;
314                         break;
315                       case IPOPT_RR:
316                         if (opt->rr) {
317                                 pp_ptr = optptr;
318                                 goto error;
319                         }
320                         if (optlen < 3) {
321                                 pp_ptr = optptr + 1;
322                                 goto error;
323                         }
324                         if (optptr[2] < 4) {
325                                 pp_ptr = optptr + 2;
326                                 goto error;
327                         }
328                         if (optptr[2] <= optlen) {
329                                 if (optptr[2]+3 > optlen) {
330                                         pp_ptr = optptr + 2;
331                                         goto error;
332                                 }
333                                 if (skb) {
334                                         memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
335                                         opt->is_changed = 1;
336                                 }
337                                 optptr[2] += 4;
338                                 opt->rr_needaddr = 1;
339                         }
340                         opt->rr = optptr - iph;
341                         break;
342                       case IPOPT_TIMESTAMP:
343                         if (opt->ts) {
344                                 pp_ptr = optptr;
345                                 goto error;
346                         }
347                         if (optlen < 4) {
348                                 pp_ptr = optptr + 1;
349                                 goto error;
350                         }
351                         if (optptr[2] < 5) {
352                                 pp_ptr = optptr + 2;
353                                 goto error;
354                         }
355                         if (optptr[2] <= optlen) {
356                                 __be32 *timeptr = NULL;
357                                 if (optptr[2]+3 > optptr[1]) {
358                                         pp_ptr = optptr + 2;
359                                         goto error;
360                                 }
361                                 switch (optptr[3]&0xF) {
362                                       case IPOPT_TS_TSONLY:
363                                         opt->ts = optptr - iph;
364                                         if (skb)
365                                                 timeptr = (__be32*)&optptr[optptr[2]-1];
366                                         opt->ts_needtime = 1;
367                                         optptr[2] += 4;
368                                         break;
369                                       case IPOPT_TS_TSANDADDR:
370                                         if (optptr[2]+7 > optptr[1]) {
371                                                 pp_ptr = optptr + 2;
372                                                 goto error;
373                                         }
374                                         opt->ts = optptr - iph;
375                                         if (skb) {
376                                                 memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
377                                                 timeptr = (__be32*)&optptr[optptr[2]+3];
378                                         }
379                                         opt->ts_needaddr = 1;
380                                         opt->ts_needtime = 1;
381                                         optptr[2] += 8;
382                                         break;
383                                       case IPOPT_TS_PRESPEC:
384                                         if (optptr[2]+7 > optptr[1]) {
385                                                 pp_ptr = optptr + 2;
386                                                 goto error;
387                                         }
388                                         opt->ts = optptr - iph;
389                                         {
390                                                 __be32 addr;
391                                                 memcpy(&addr, &optptr[optptr[2]-1], 4);
392                                                 if (inet_addr_type(net, addr) == RTN_UNICAST)
393                                                         break;
394                                                 if (skb)
395                                                         timeptr = (__be32*)&optptr[optptr[2]+3];
396                                         }
397                                         opt->ts_needtime = 1;
398                                         optptr[2] += 8;
399                                         break;
400                                       default:
401                                         if (!skb && !capable(CAP_NET_RAW)) {
402                                                 pp_ptr = optptr + 3;
403                                                 goto error;
404                                         }
405                                         break;
406                                 }
407                                 if (timeptr) {
408                                         struct timespec tv;
409                                         __be32  midtime;
410                                         getnstimeofday(&tv);
411                                         midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
412                                         memcpy(timeptr, &midtime, sizeof(__be32));
413                                         opt->is_changed = 1;
414                                 }
415                         } else {
416                                 unsigned overflow = optptr[3]>>4;
417                                 if (overflow == 15) {
418                                         pp_ptr = optptr + 3;
419                                         goto error;
420                                 }
421                                 opt->ts = optptr - iph;
422                                 if (skb) {
423                                         optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
424                                         opt->is_changed = 1;
425                                 }
426                         }
427                         break;
428                       case IPOPT_RA:
429                         if (optlen < 4) {
430                                 pp_ptr = optptr + 1;
431                                 goto error;
432                         }
433                         if (optptr[2] == 0 && optptr[3] == 0)
434                                 opt->router_alert = optptr - iph;
435                         break;
436                       case IPOPT_CIPSO:
437                         if ((!skb && !capable(CAP_NET_RAW)) || opt->cipso) {
438                                 pp_ptr = optptr;
439                                 goto error;
440                         }
441                         opt->cipso = optptr - iph;
442                         if (cipso_v4_validate(skb, &optptr)) {
443                                 pp_ptr = optptr;
444                                 goto error;
445                         }
446                         break;
447                       case IPOPT_SEC:
448                       case IPOPT_SID:
449                       default:
450                         if (!skb && !capable(CAP_NET_RAW)) {
451                                 pp_ptr = optptr;
452                                 goto error;
453                         }
454                         break;
455                 }
456                 l -= optlen;
457                 optptr += optlen;
458         }
459
460 eol:
461         if (!pp_ptr)
462                 return 0;
463
464 error:
465         if (skb) {
466                 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
467         }
468         return -EINVAL;
469 }
470
471
472 /*
473  *      Undo all the changes done by ip_options_compile().
474  */
475
476 void ip_options_undo(struct ip_options * opt)
477 {
478         if (opt->srr) {
479                 unsigned  char * optptr = opt->__data+opt->srr-sizeof(struct  iphdr);
480                 memmove(optptr+7, optptr+3, optptr[1]-7);
481                 memcpy(optptr+3, &opt->faddr, 4);
482         }
483         if (opt->rr_needaddr) {
484                 unsigned  char * optptr = opt->__data+opt->rr-sizeof(struct  iphdr);
485                 optptr[2] -= 4;
486                 memset(&optptr[optptr[2]-1], 0, 4);
487         }
488         if (opt->ts) {
489                 unsigned  char * optptr = opt->__data+opt->ts-sizeof(struct  iphdr);
490                 if (opt->ts_needtime) {
491                         optptr[2] -= 4;
492                         memset(&optptr[optptr[2]-1], 0, 4);
493                         if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
494                                 optptr[2] -= 4;
495                 }
496                 if (opt->ts_needaddr) {
497                         optptr[2] -= 4;
498                         memset(&optptr[optptr[2]-1], 0, 4);
499                 }
500         }
501 }
502
503 static struct ip_options *ip_options_get_alloc(const int optlen)
504 {
505         return kzalloc(sizeof(struct ip_options) + ((optlen + 3) & ~3),
506                        GFP_KERNEL);
507 }
508
509 static int ip_options_get_finish(struct net *net, struct ip_options **optp,
510                                  struct ip_options *opt, int optlen)
511 {
512         while (optlen & 3)
513                 opt->__data[optlen++] = IPOPT_END;
514         opt->optlen = optlen;
515         if (optlen && ip_options_compile(net, opt, NULL)) {
516                 kfree(opt);
517                 return -EINVAL;
518         }
519         kfree(*optp);
520         *optp = opt;
521         return 0;
522 }
523
524 int ip_options_get_from_user(struct net *net, struct ip_options **optp,
525                              unsigned char __user *data, int optlen)
526 {
527         struct ip_options *opt = ip_options_get_alloc(optlen);
528
529         if (!opt)
530                 return -ENOMEM;
531         if (optlen && copy_from_user(opt->__data, data, optlen)) {
532                 kfree(opt);
533                 return -EFAULT;
534         }
535         return ip_options_get_finish(net, optp, opt, optlen);
536 }
537
538 int ip_options_get(struct net *net, struct ip_options **optp,
539                    unsigned char *data, int optlen)
540 {
541         struct ip_options *opt = ip_options_get_alloc(optlen);
542
543         if (!opt)
544                 return -ENOMEM;
545         if (optlen)
546                 memcpy(opt->__data, data, optlen);
547         return ip_options_get_finish(net, optp, opt, optlen);
548 }
549
550 void ip_forward_options(struct sk_buff *skb)
551 {
552         struct   ip_options * opt       = &(IPCB(skb)->opt);
553         unsigned char * optptr;
554         struct rtable *rt = skb_rtable(skb);
555         unsigned char *raw = skb_network_header(skb);
556
557         if (opt->rr_needaddr) {
558                 optptr = (unsigned char *)raw + opt->rr;
559                 ip_rt_get_source(&optptr[optptr[2]-5], rt);
560                 opt->is_changed = 1;
561         }
562         if (opt->srr_is_hit) {
563                 int srrptr, srrspace;
564
565                 optptr = raw + opt->srr;
566
567                 for ( srrptr=optptr[2], srrspace = optptr[1];
568                      srrptr <= srrspace;
569                      srrptr += 4
570                      ) {
571                         if (srrptr + 3 > srrspace)
572                                 break;
573                         if (memcmp(&rt->rt_dst, &optptr[srrptr-1], 4) == 0)
574                                 break;
575                 }
576                 if (srrptr + 3 <= srrspace) {
577                         opt->is_changed = 1;
578                         ip_rt_get_source(&optptr[srrptr-1], rt);
579                         ip_hdr(skb)->daddr = rt->rt_dst;
580                         optptr[2] = srrptr+4;
581                 } else if (net_ratelimit())
582                         printk(KERN_CRIT "ip_forward(): Argh! Destination lost!\n");
583                 if (opt->ts_needaddr) {
584                         optptr = raw + opt->ts;
585                         ip_rt_get_source(&optptr[optptr[2]-9], rt);
586                         opt->is_changed = 1;
587                 }
588         }
589         if (opt->is_changed) {
590                 opt->is_changed = 0;
591                 ip_send_check(ip_hdr(skb));
592         }
593 }
594
595 int ip_options_rcv_srr(struct sk_buff *skb)
596 {
597         struct ip_options *opt = &(IPCB(skb)->opt);
598         int srrspace, srrptr;
599         __be32 nexthop;
600         struct iphdr *iph = ip_hdr(skb);
601         unsigned char *optptr = skb_network_header(skb) + opt->srr;
602         struct rtable *rt = skb_rtable(skb);
603         struct rtable *rt2;
604         unsigned long orefdst;
605         int err;
606
607         if (!opt->srr)
608                 return 0;
609
610         if (skb->pkt_type != PACKET_HOST)
611                 return -EINVAL;
612         if (rt->rt_type == RTN_UNICAST) {
613                 if (!opt->is_strictroute)
614                         return 0;
615                 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
616                 return -EINVAL;
617         }
618         if (rt->rt_type != RTN_LOCAL)
619                 return -EINVAL;
620
621         for (srrptr=optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
622                 if (srrptr + 3 > srrspace) {
623                         icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
624                         return -EINVAL;
625                 }
626                 memcpy(&nexthop, &optptr[srrptr-1], 4);
627
628                 orefdst = skb->_skb_refdst;
629                 skb_dst_set(skb, NULL);
630                 err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
631                 rt2 = skb_rtable(skb);
632                 if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
633                         skb_dst_drop(skb);
634                         skb->_skb_refdst = orefdst;
635                         return -EINVAL;
636                 }
637                 refdst_drop(orefdst);
638                 if (rt2->rt_type != RTN_LOCAL)
639                         break;
640                 /* Superfast 8) loopback forward */
641                 memcpy(&iph->daddr, &optptr[srrptr-1], 4);
642                 opt->is_changed = 1;
643         }
644         if (srrptr <= srrspace) {
645                 opt->srr_is_hit = 1;
646                 opt->is_changed = 1;
647         }
648         return 0;
649 }