net: Allow netdevices to specify needed head/tailroom
[linux-2.6] / 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  * Version:     $Id: ip_options.c,v 1.21 2001/09/01 00:31:50 davem Exp $
9  *
10  * Authors:     A.N.Kuznetsov
11  *
12  */
13
14 #include <linux/capability.h>
15 #include <linux/module.h>
16 #include <linux/types.h>
17 #include <asm/uaccess.h>
18 #include <linux/skbuff.h>
19 #include <linux/ip.h>
20 #include <linux/icmp.h>
21 #include <linux/netdevice.h>
22 #include <linux/rtnetlink.h>
23 #include <net/sock.h>
24 #include <net/ip.h>
25 #include <net/icmp.h>
26 #include <net/route.h>
27 #include <net/cipso_ipv4.h>
28
29 /*
30  * Write options to IP header, record destination address to
31  * source route option, address of outgoing interface
32  * (we should already know it, so that this  function is allowed be
33  * called only after routing decision) and timestamp,
34  * if we originate this datagram.
35  *
36  * daddr is real destination address, next hop is recorded in IP header.
37  * saddr is address of outgoing interface.
38  */
39
40 void ip_options_build(struct sk_buff * skb, struct ip_options * opt,
41                             __be32 daddr, struct rtable *rt, int is_frag)
42 {
43         unsigned char *iph = skb_network_header(skb);
44
45         memcpy(&(IPCB(skb)->opt), opt, sizeof(struct ip_options));
46         memcpy(iph+sizeof(struct iphdr), opt->__data, opt->optlen);
47         opt = &(IPCB(skb)->opt);
48
49         if (opt->srr)
50                 memcpy(iph+opt->srr+iph[opt->srr+1]-4, &daddr, 4);
51
52         if (!is_frag) {
53                 if (opt->rr_needaddr)
54                         ip_rt_get_source(iph+opt->rr+iph[opt->rr+2]-5, rt);
55                 if (opt->ts_needaddr)
56                         ip_rt_get_source(iph+opt->ts+iph[opt->ts+2]-9, rt);
57                 if (opt->ts_needtime) {
58                         struct timespec tv;
59                         __be32 midtime;
60                         getnstimeofday(&tv);
61                         midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
62                         memcpy(iph+opt->ts+iph[opt->ts+2]-5, &midtime, 4);
63                 }
64                 return;
65         }
66         if (opt->rr) {
67                 memset(iph+opt->rr, IPOPT_NOP, iph[opt->rr+1]);
68                 opt->rr = 0;
69                 opt->rr_needaddr = 0;
70         }
71         if (opt->ts) {
72                 memset(iph+opt->ts, IPOPT_NOP, iph[opt->ts+1]);
73                 opt->ts = 0;
74                 opt->ts_needaddr = opt->ts_needtime = 0;
75         }
76 }
77
78 /*
79  * Provided (sopt, skb) points to received options,
80  * build in dopt compiled option set appropriate for answering.
81  * i.e. invert SRR option, copy anothers,
82  * and grab room in RR/TS options.
83  *
84  * NOTE: dopt cannot point to skb.
85  */
86
87 int ip_options_echo(struct ip_options * dopt, struct sk_buff * skb)
88 {
89         struct ip_options *sopt;
90         unsigned char *sptr, *dptr;
91         int soffset, doffset;
92         int     optlen;
93         __be32  daddr;
94
95         memset(dopt, 0, sizeof(struct ip_options));
96
97         sopt = &(IPCB(skb)->opt);
98
99         if (sopt->optlen == 0) {
100                 dopt->optlen = 0;
101                 return 0;
102         }
103
104         sptr = skb_network_header(skb);
105         dptr = dopt->__data;
106
107         daddr = skb->rtable->rt_spec_dst;
108
109         if (sopt->rr) {
110                 optlen  = sptr[sopt->rr+1];
111                 soffset = sptr[sopt->rr+2];
112                 dopt->rr = dopt->optlen + sizeof(struct iphdr);
113                 memcpy(dptr, sptr+sopt->rr, optlen);
114                 if (sopt->rr_needaddr && soffset <= optlen) {
115                         if (soffset + 3 > optlen)
116                                 return -EINVAL;
117                         dptr[2] = soffset + 4;
118                         dopt->rr_needaddr = 1;
119                 }
120                 dptr += optlen;
121                 dopt->optlen += optlen;
122         }
123         if (sopt->ts) {
124                 optlen = sptr[sopt->ts+1];
125                 soffset = sptr[sopt->ts+2];
126                 dopt->ts = dopt->optlen + sizeof(struct iphdr);
127                 memcpy(dptr, sptr+sopt->ts, optlen);
128                 if (soffset <= optlen) {
129                         if (sopt->ts_needaddr) {
130                                 if (soffset + 3 > optlen)
131                                         return -EINVAL;
132                                 dopt->ts_needaddr = 1;
133                                 soffset += 4;
134                         }
135                         if (sopt->ts_needtime) {
136                                 if (soffset + 3 > optlen)
137                                         return -EINVAL;
138                                 if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
139                                         dopt->ts_needtime = 1;
140                                         soffset += 4;
141                                 } else {
142                                         dopt->ts_needtime = 0;
143
144                                         if (soffset + 8 <= optlen) {
145                                                 __be32 addr;
146
147                                                 memcpy(&addr, sptr+soffset-1, 4);
148                                                 if (inet_addr_type(dev_net(skb->dst->dev), addr) != RTN_LOCAL) {
149                                                         dopt->ts_needtime = 1;
150                                                         soffset += 8;
151                                                 }
152                                         }
153                                 }
154                         }
155                         dptr[2] = soffset;
156                 }
157                 dptr += optlen;
158                 dopt->optlen += optlen;
159         }
160         if (sopt->srr) {
161                 unsigned char * start = sptr+sopt->srr;
162                 __be32 faddr;
163
164                 optlen  = start[1];
165                 soffset = start[2];
166                 doffset = 0;
167                 if (soffset > optlen)
168                         soffset = optlen + 1;
169                 soffset -= 4;
170                 if (soffset > 3) {
171                         memcpy(&faddr, &start[soffset-1], 4);
172                         for (soffset-=4, doffset=4; soffset > 3; soffset-=4, doffset+=4)
173                                 memcpy(&dptr[doffset-1], &start[soffset-1], 4);
174                         /*
175                          * RFC1812 requires to fix illegal source routes.
176                          */
177                         if (memcmp(&ip_hdr(skb)->saddr,
178                                    &start[soffset + 3], 4) == 0)
179                                 doffset -= 4;
180                 }
181                 if (doffset > 3) {
182                         memcpy(&start[doffset-1], &daddr, 4);
183                         dopt->faddr = faddr;
184                         dptr[0] = start[0];
185                         dptr[1] = doffset+3;
186                         dptr[2] = 4;
187                         dptr += doffset+3;
188                         dopt->srr = dopt->optlen + sizeof(struct iphdr);
189                         dopt->optlen += doffset+3;
190                         dopt->is_strictroute = sopt->is_strictroute;
191                 }
192         }
193         if (sopt->cipso) {
194                 optlen  = sptr[sopt->cipso+1];
195                 dopt->cipso = dopt->optlen+sizeof(struct iphdr);
196                 memcpy(dptr, sptr+sopt->cipso, optlen);
197                 dptr += optlen;
198                 dopt->optlen += optlen;
199         }
200         while (dopt->optlen & 3) {
201                 *dptr++ = IPOPT_END;
202                 dopt->optlen++;
203         }
204         return 0;
205 }
206
207 /*
208  *      Options "fragmenting", just fill options not
209  *      allowed in fragments with NOOPs.
210  *      Simple and stupid 8), but the most efficient way.
211  */
212
213 void ip_options_fragment(struct sk_buff * skb)
214 {
215         unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
216         struct ip_options * opt = &(IPCB(skb)->opt);
217         int  l = opt->optlen;
218         int  optlen;
219
220         while (l > 0) {
221                 switch (*optptr) {
222                 case IPOPT_END:
223                         return;
224                 case IPOPT_NOOP:
225                         l--;
226                         optptr++;
227                         continue;
228                 }
229                 optlen = optptr[1];
230                 if (optlen<2 || optlen>l)
231                   return;
232                 if (!IPOPT_COPIED(*optptr))
233                         memset(optptr, IPOPT_NOOP, optlen);
234                 l -= optlen;
235                 optptr += optlen;
236         }
237         opt->ts = 0;
238         opt->rr = 0;
239         opt->rr_needaddr = 0;
240         opt->ts_needaddr = 0;
241         opt->ts_needtime = 0;
242         return;
243 }
244
245 /*
246  * Verify options and fill pointers in struct options.
247  * Caller should clear *opt, and set opt->data.
248  * If opt == NULL, then skb->data should point to IP header.
249  */
250
251 int ip_options_compile(struct net *net,
252                        struct ip_options * opt, struct sk_buff * skb)
253 {
254         int l;
255         unsigned char * iph;
256         unsigned char * optptr;
257         int optlen;
258         unsigned char * pp_ptr = NULL;
259         struct rtable *rt = NULL;
260
261         if (skb != NULL) {
262                 rt = skb->rtable;
263                 optptr = (unsigned char *)&(ip_hdr(skb)[1]);
264         } else
265                 optptr = opt->__data;
266         iph = optptr - sizeof(struct iphdr);
267
268         for (l = opt->optlen; l > 0; ) {
269                 switch (*optptr) {
270                       case IPOPT_END:
271                         for (optptr++, l--; l>0; optptr++, l--) {
272                                 if (*optptr != IPOPT_END) {
273                                         *optptr = IPOPT_END;
274                                         opt->is_changed = 1;
275                                 }
276                         }
277                         goto eol;
278                       case IPOPT_NOOP:
279                         l--;
280                         optptr++;
281                         continue;
282                 }
283                 optlen = optptr[1];
284                 if (optlen<2 || optlen>l) {
285                         pp_ptr = optptr;
286                         goto error;
287                 }
288                 switch (*optptr) {
289                       case IPOPT_SSRR:
290                       case IPOPT_LSRR:
291                         if (optlen < 3) {
292                                 pp_ptr = optptr + 1;
293                                 goto error;
294                         }
295                         if (optptr[2] < 4) {
296                                 pp_ptr = optptr + 2;
297                                 goto error;
298                         }
299                         /* NB: cf RFC-1812 5.2.4.1 */
300                         if (opt->srr) {
301                                 pp_ptr = optptr;
302                                 goto error;
303                         }
304                         if (!skb) {
305                                 if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) {
306                                         pp_ptr = optptr + 1;
307                                         goto error;
308                                 }
309                                 memcpy(&opt->faddr, &optptr[3], 4);
310                                 if (optlen > 7)
311                                         memmove(&optptr[3], &optptr[7], optlen-7);
312                         }
313                         opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
314                         opt->srr = optptr - iph;
315                         break;
316                       case IPOPT_RR:
317                         if (opt->rr) {
318                                 pp_ptr = optptr;
319                                 goto error;
320                         }
321                         if (optlen < 3) {
322                                 pp_ptr = optptr + 1;
323                                 goto error;
324                         }
325                         if (optptr[2] < 4) {
326                                 pp_ptr = optptr + 2;
327                                 goto error;
328                         }
329                         if (optptr[2] <= optlen) {
330                                 if (optptr[2]+3 > optlen) {
331                                         pp_ptr = optptr + 2;
332                                         goto error;
333                                 }
334                                 if (skb) {
335                                         memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
336                                         opt->is_changed = 1;
337                                 }
338                                 optptr[2] += 4;
339                                 opt->rr_needaddr = 1;
340                         }
341                         opt->rr = optptr - iph;
342                         break;
343                       case IPOPT_TIMESTAMP:
344                         if (opt->ts) {
345                                 pp_ptr = optptr;
346                                 goto error;
347                         }
348                         if (optlen < 4) {
349                                 pp_ptr = optptr + 1;
350                                 goto error;
351                         }
352                         if (optptr[2] < 5) {
353                                 pp_ptr = optptr + 2;
354                                 goto error;
355                         }
356                         if (optptr[2] <= optlen) {
357                                 __be32 *timeptr = NULL;
358                                 if (optptr[2]+3 > optptr[1]) {
359                                         pp_ptr = optptr + 2;
360                                         goto error;
361                                 }
362                                 switch (optptr[3]&0xF) {
363                                       case IPOPT_TS_TSONLY:
364                                         opt->ts = optptr - iph;
365                                         if (skb)
366                                                 timeptr = (__be32*)&optptr[optptr[2]-1];
367                                         opt->ts_needtime = 1;
368                                         optptr[2] += 4;
369                                         break;
370                                       case IPOPT_TS_TSANDADDR:
371                                         if (optptr[2]+7 > optptr[1]) {
372                                                 pp_ptr = optptr + 2;
373                                                 goto error;
374                                         }
375                                         opt->ts = optptr - iph;
376                                         if (skb) {
377                                                 memcpy(&optptr[optptr[2]-1], &rt->rt_spec_dst, 4);
378                                                 timeptr = (__be32*)&optptr[optptr[2]+3];
379                                         }
380                                         opt->ts_needaddr = 1;
381                                         opt->ts_needtime = 1;
382                                         optptr[2] += 8;
383                                         break;
384                                       case IPOPT_TS_PRESPEC:
385                                         if (optptr[2]+7 > optptr[1]) {
386                                                 pp_ptr = optptr + 2;
387                                                 goto error;
388                                         }
389                                         opt->ts = optptr - iph;
390                                         {
391                                                 __be32 addr;
392                                                 memcpy(&addr, &optptr[optptr[2]-1], 4);
393                                                 if (inet_addr_type(net, addr) == RTN_UNICAST)
394                                                         break;
395                                                 if (skb)
396                                                         timeptr = (__be32*)&optptr[optptr[2]+3];
397                                         }
398                                         opt->ts_needtime = 1;
399                                         optptr[2] += 8;
400                                         break;
401                                       default:
402                                         if (!skb && !capable(CAP_NET_RAW)) {
403                                                 pp_ptr = optptr + 3;
404                                                 goto error;
405                                         }
406                                         break;
407                                 }
408                                 if (timeptr) {
409                                         struct timespec tv;
410                                         __be32  midtime;
411                                         getnstimeofday(&tv);
412                                         midtime = htonl((tv.tv_sec % 86400) * MSEC_PER_SEC + tv.tv_nsec / NSEC_PER_MSEC);
413                                         memcpy(timeptr, &midtime, sizeof(__be32));
414                                         opt->is_changed = 1;
415                                 }
416                         } else {
417                                 unsigned overflow = optptr[3]>>4;
418                                 if (overflow == 15) {
419                                         pp_ptr = optptr + 3;
420                                         goto error;
421                                 }
422                                 opt->ts = optptr - iph;
423                                 if (skb) {
424                                         optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
425                                         opt->is_changed = 1;
426                                 }
427                         }
428                         break;
429                       case IPOPT_RA:
430                         if (optlen < 4) {
431                                 pp_ptr = optptr + 1;
432                                 goto error;
433                         }
434                         if (optptr[2] == 0 && optptr[3] == 0)
435                                 opt->router_alert = optptr - iph;
436                         break;
437                       case IPOPT_CIPSO:
438                         if ((!skb && !capable(CAP_NET_RAW)) || opt->cipso) {
439                                 pp_ptr = optptr;
440                                 goto error;
441                         }
442                         opt->cipso = optptr - iph;
443                         if (cipso_v4_validate(&optptr)) {
444                                 pp_ptr = optptr;
445                                 goto error;
446                         }
447                         break;
448                       case IPOPT_SEC:
449                       case IPOPT_SID:
450                       default:
451                         if (!skb && !capable(CAP_NET_RAW)) {
452                                 pp_ptr = optptr;
453                                 goto error;
454                         }
455                         break;
456                 }
457                 l -= optlen;
458                 optptr += optlen;
459         }
460
461 eol:
462         if (!pp_ptr)
463                 return 0;
464
465 error:
466         if (skb) {
467                 icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((pp_ptr-iph)<<24));
468         }
469         return -EINVAL;
470 }
471
472
473 /*
474  *      Undo all the changes done by ip_options_compile().
475  */
476
477 void ip_options_undo(struct ip_options * opt)
478 {
479         if (opt->srr) {
480                 unsigned  char * optptr = opt->__data+opt->srr-sizeof(struct  iphdr);
481                 memmove(optptr+7, optptr+3, optptr[1]-7);
482                 memcpy(optptr+3, &opt->faddr, 4);
483         }
484         if (opt->rr_needaddr) {
485                 unsigned  char * optptr = opt->__data+opt->rr-sizeof(struct  iphdr);
486                 optptr[2] -= 4;
487                 memset(&optptr[optptr[2]-1], 0, 4);
488         }
489         if (opt->ts) {
490                 unsigned  char * optptr = opt->__data+opt->ts-sizeof(struct  iphdr);
491                 if (opt->ts_needtime) {
492                         optptr[2] -= 4;
493                         memset(&optptr[optptr[2]-1], 0, 4);
494                         if ((optptr[3]&0xF) == IPOPT_TS_PRESPEC)
495                                 optptr[2] -= 4;
496                 }
497                 if (opt->ts_needaddr) {
498                         optptr[2] -= 4;
499                         memset(&optptr[optptr[2]-1], 0, 4);
500                 }
501         }
502 }
503
504 static struct ip_options *ip_options_get_alloc(const int optlen)
505 {
506         return kzalloc(sizeof(struct ip_options) + ((optlen + 3) & ~3),
507                        GFP_KERNEL);
508 }
509
510 static int ip_options_get_finish(struct net *net, struct ip_options **optp,
511                                  struct ip_options *opt, int optlen)
512 {
513         while (optlen & 3)
514                 opt->__data[optlen++] = IPOPT_END;
515         opt->optlen = optlen;
516         if (optlen && ip_options_compile(net, opt, NULL)) {
517                 kfree(opt);
518                 return -EINVAL;
519         }
520         kfree(*optp);
521         *optp = opt;
522         return 0;
523 }
524
525 int ip_options_get_from_user(struct net *net, struct ip_options **optp,
526                              unsigned char __user *data, int optlen)
527 {
528         struct ip_options *opt = ip_options_get_alloc(optlen);
529
530         if (!opt)
531                 return -ENOMEM;
532         if (optlen && copy_from_user(opt->__data, data, optlen)) {
533                 kfree(opt);
534                 return -EFAULT;
535         }
536         return ip_options_get_finish(net, optp, opt, optlen);
537 }
538
539 int ip_options_get(struct net *net, struct ip_options **optp,
540                    unsigned char *data, int optlen)
541 {
542         struct ip_options *opt = ip_options_get_alloc(optlen);
543
544         if (!opt)
545                 return -ENOMEM;
546         if (optlen)
547                 memcpy(opt->__data, data, optlen);
548         return ip_options_get_finish(net, optp, opt, optlen);
549 }
550
551 void ip_forward_options(struct sk_buff *skb)
552 {
553         struct   ip_options * opt       = &(IPCB(skb)->opt);
554         unsigned char * optptr;
555         struct rtable *rt = skb->rtable;
556         unsigned char *raw = skb_network_header(skb);
557
558         if (opt->rr_needaddr) {
559                 optptr = (unsigned char *)raw + opt->rr;
560                 ip_rt_get_source(&optptr[optptr[2]-5], rt);
561                 opt->is_changed = 1;
562         }
563         if (opt->srr_is_hit) {
564                 int srrptr, srrspace;
565
566                 optptr = raw + opt->srr;
567
568                 for ( srrptr=optptr[2], srrspace = optptr[1];
569                      srrptr <= srrspace;
570                      srrptr += 4
571                      ) {
572                         if (srrptr + 3 > srrspace)
573                                 break;
574                         if (memcmp(&rt->rt_dst, &optptr[srrptr-1], 4) == 0)
575                                 break;
576                 }
577                 if (srrptr + 3 <= srrspace) {
578                         opt->is_changed = 1;
579                         ip_rt_get_source(&optptr[srrptr-1], rt);
580                         ip_hdr(skb)->daddr = rt->rt_dst;
581                         optptr[2] = srrptr+4;
582                 } else if (net_ratelimit())
583                         printk(KERN_CRIT "ip_forward(): Argh! Destination lost!\n");
584                 if (opt->ts_needaddr) {
585                         optptr = raw + opt->ts;
586                         ip_rt_get_source(&optptr[optptr[2]-9], rt);
587                         opt->is_changed = 1;
588                 }
589         }
590         if (opt->is_changed) {
591                 opt->is_changed = 0;
592                 ip_send_check(ip_hdr(skb));
593         }
594 }
595
596 int ip_options_rcv_srr(struct sk_buff *skb)
597 {
598         struct ip_options *opt = &(IPCB(skb)->opt);
599         int srrspace, srrptr;
600         __be32 nexthop;
601         struct iphdr *iph = ip_hdr(skb);
602         unsigned char *optptr = skb_network_header(skb) + opt->srr;
603         struct rtable *rt = skb->rtable;
604         struct rtable *rt2;
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                 rt = skb->rtable;
629                 skb->rtable = NULL;
630                 err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, skb->dev);
631                 rt2 = skb->rtable;
632                 if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
633                         ip_rt_put(rt2);
634                         skb->rtable = rt;
635                         return -EINVAL;
636                 }
637                 ip_rt_put(rt);
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 }