ip_options.c 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641
  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. #define pr_fmt(fmt) "IPv4: " fmt
  13. #include <linux/capability.h>
  14. #include <linux/module.h>
  15. #include <linux/slab.h>
  16. #include <linux/types.h>
  17. #include <linux/uaccess.h>
  18. #include <asm/unaligned.h>
  19. #include <linux/skbuff.h>
  20. #include <linux/ip.h>
  21. #include <linux/icmp.h>
  22. #include <linux/netdevice.h>
  23. #include <linux/rtnetlink.h>
  24. #include <net/sock.h>
  25. #include <net/ip.h>
  26. #include <net/icmp.h>
  27. #include <net/route.h>
  28. #include <net/cipso_ipv4.h>
  29. #include <net/ip_fib.h>
  30. /*
  31. * Write options to IP header, record destination address to
  32. * source route option, address of outgoing interface
  33. * (we should already know it, so that this function is allowed be
  34. * called only after routing decision) and timestamp,
  35. * if we originate this datagram.
  36. *
  37. * daddr is real destination address, next hop is recorded in IP header.
  38. * saddr is address of outgoing interface.
  39. */
  40. void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
  41. __be32 daddr, struct rtable *rt)
  42. {
  43. unsigned char *iph = skb_network_header(skb);
  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. if (opt->srr)
  48. memcpy(iph + opt->srr + iph[opt->srr + 1] - 4, &daddr, 4);
  49. if (opt->rr_needaddr)
  50. ip_rt_get_source(iph + opt->rr + iph[opt->rr + 2] - 5, skb, rt);
  51. if (opt->ts_needaddr)
  52. ip_rt_get_source(iph + opt->ts + iph[opt->ts + 2] - 9, skb, rt);
  53. if (opt->ts_needtime) {
  54. __be32 midtime;
  55. midtime = inet_current_timestamp();
  56. memcpy(iph + opt->ts + iph[opt->ts + 2] - 5, &midtime, 4);
  57. }
  58. }
  59. /*
  60. * Provided (sopt, skb) points to received options,
  61. * build in dopt compiled option set appropriate for answering.
  62. * i.e. invert SRR option, copy anothers,
  63. * and grab room in RR/TS options.
  64. *
  65. * NOTE: dopt cannot point to skb.
  66. */
  67. int __ip_options_echo(struct net *net, struct ip_options *dopt,
  68. struct sk_buff *skb, const struct ip_options *sopt)
  69. {
  70. unsigned char *sptr, *dptr;
  71. int soffset, doffset;
  72. int optlen;
  73. memset(dopt, 0, sizeof(struct ip_options));
  74. if (sopt->optlen == 0)
  75. return 0;
  76. sptr = skb_network_header(skb);
  77. dptr = dopt->__data;
  78. if (sopt->rr) {
  79. optlen = sptr[sopt->rr+1];
  80. soffset = sptr[sopt->rr+2];
  81. dopt->rr = dopt->optlen + sizeof(struct iphdr);
  82. memcpy(dptr, sptr+sopt->rr, optlen);
  83. if (sopt->rr_needaddr && soffset <= optlen) {
  84. if (soffset + 3 > optlen)
  85. return -EINVAL;
  86. dptr[2] = soffset + 4;
  87. dopt->rr_needaddr = 1;
  88. }
  89. dptr += optlen;
  90. dopt->optlen += optlen;
  91. }
  92. if (sopt->ts) {
  93. optlen = sptr[sopt->ts+1];
  94. soffset = sptr[sopt->ts+2];
  95. dopt->ts = dopt->optlen + sizeof(struct iphdr);
  96. memcpy(dptr, sptr+sopt->ts, optlen);
  97. if (soffset <= optlen) {
  98. if (sopt->ts_needaddr) {
  99. if (soffset + 3 > optlen)
  100. return -EINVAL;
  101. dopt->ts_needaddr = 1;
  102. soffset += 4;
  103. }
  104. if (sopt->ts_needtime) {
  105. if (soffset + 3 > optlen)
  106. return -EINVAL;
  107. if ((dptr[3]&0xF) != IPOPT_TS_PRESPEC) {
  108. dopt->ts_needtime = 1;
  109. soffset += 4;
  110. } else {
  111. dopt->ts_needtime = 0;
  112. if (soffset + 7 <= optlen) {
  113. __be32 addr;
  114. memcpy(&addr, dptr+soffset-1, 4);
  115. if (inet_addr_type(net, addr) != RTN_UNICAST) {
  116. dopt->ts_needtime = 1;
  117. soffset += 8;
  118. }
  119. }
  120. }
  121. }
  122. dptr[2] = soffset;
  123. }
  124. dptr += optlen;
  125. dopt->optlen += optlen;
  126. }
  127. if (sopt->srr) {
  128. unsigned char *start = sptr+sopt->srr;
  129. __be32 faddr;
  130. optlen = start[1];
  131. soffset = start[2];
  132. doffset = 0;
  133. if (soffset > optlen)
  134. soffset = optlen + 1;
  135. soffset -= 4;
  136. if (soffset > 3) {
  137. memcpy(&faddr, &start[soffset-1], 4);
  138. for (soffset -= 4, doffset = 4; soffset > 3; soffset -= 4, doffset += 4)
  139. memcpy(&dptr[doffset-1], &start[soffset-1], 4);
  140. /*
  141. * RFC1812 requires to fix illegal source routes.
  142. */
  143. if (memcmp(&ip_hdr(skb)->saddr,
  144. &start[soffset + 3], 4) == 0)
  145. doffset -= 4;
  146. }
  147. if (doffset > 3) {
  148. dopt->faddr = faddr;
  149. dptr[0] = start[0];
  150. dptr[1] = doffset+3;
  151. dptr[2] = 4;
  152. dptr += doffset+3;
  153. dopt->srr = dopt->optlen + sizeof(struct iphdr);
  154. dopt->optlen += doffset+3;
  155. dopt->is_strictroute = sopt->is_strictroute;
  156. }
  157. }
  158. if (sopt->cipso) {
  159. optlen = sptr[sopt->cipso+1];
  160. dopt->cipso = dopt->optlen+sizeof(struct iphdr);
  161. memcpy(dptr, sptr+sopt->cipso, optlen);
  162. dptr += optlen;
  163. dopt->optlen += optlen;
  164. }
  165. while (dopt->optlen & 3) {
  166. *dptr++ = IPOPT_END;
  167. dopt->optlen++;
  168. }
  169. return 0;
  170. }
  171. /*
  172. * Options "fragmenting", just fill options not
  173. * allowed in fragments with NOOPs.
  174. * Simple and stupid 8), but the most efficient way.
  175. */
  176. void ip_options_fragment(struct sk_buff *skb)
  177. {
  178. unsigned char *optptr = skb_network_header(skb) + sizeof(struct iphdr);
  179. struct ip_options *opt = &(IPCB(skb)->opt);
  180. int l = opt->optlen;
  181. int optlen;
  182. while (l > 0) {
  183. switch (*optptr) {
  184. case IPOPT_END:
  185. return;
  186. case IPOPT_NOOP:
  187. l--;
  188. optptr++;
  189. continue;
  190. }
  191. optlen = optptr[1];
  192. if (optlen < 2 || optlen > l)
  193. return;
  194. if (!IPOPT_COPIED(*optptr))
  195. memset(optptr, IPOPT_NOOP, optlen);
  196. l -= optlen;
  197. optptr += optlen;
  198. }
  199. opt->ts = 0;
  200. opt->rr = 0;
  201. opt->rr_needaddr = 0;
  202. opt->ts_needaddr = 0;
  203. opt->ts_needtime = 0;
  204. }
  205. /* helper used by ip_options_compile() to call fib_compute_spec_dst()
  206. * at most one time.
  207. */
  208. static void spec_dst_fill(__be32 *spec_dst, struct sk_buff *skb)
  209. {
  210. if (*spec_dst == htonl(INADDR_ANY))
  211. *spec_dst = fib_compute_spec_dst(skb);
  212. }
  213. /*
  214. * Verify options and fill pointers in struct options.
  215. * Caller should clear *opt, and set opt->data.
  216. * If opt == NULL, then skb->data should point to IP header.
  217. */
  218. int __ip_options_compile(struct net *net,
  219. struct ip_options *opt, struct sk_buff *skb,
  220. __be32 *info)
  221. {
  222. __be32 spec_dst = htonl(INADDR_ANY);
  223. unsigned char *pp_ptr = NULL;
  224. struct rtable *rt = NULL;
  225. unsigned char *optptr;
  226. unsigned char *iph;
  227. int optlen, l;
  228. if (skb) {
  229. rt = skb_rtable(skb);
  230. optptr = (unsigned char *)&(ip_hdr(skb)[1]);
  231. } else
  232. optptr = opt->__data;
  233. iph = optptr - sizeof(struct iphdr);
  234. for (l = opt->optlen; l > 0; ) {
  235. switch (*optptr) {
  236. case IPOPT_END:
  237. for (optptr++, l--; l > 0; optptr++, l--) {
  238. if (*optptr != IPOPT_END) {
  239. *optptr = IPOPT_END;
  240. opt->is_changed = 1;
  241. }
  242. }
  243. goto eol;
  244. case IPOPT_NOOP:
  245. l--;
  246. optptr++;
  247. continue;
  248. }
  249. if (unlikely(l < 2)) {
  250. pp_ptr = optptr;
  251. goto error;
  252. }
  253. optlen = optptr[1];
  254. if (optlen < 2 || optlen > l) {
  255. pp_ptr = optptr;
  256. goto error;
  257. }
  258. switch (*optptr) {
  259. case IPOPT_SSRR:
  260. case IPOPT_LSRR:
  261. if (optlen < 3) {
  262. pp_ptr = optptr + 1;
  263. goto error;
  264. }
  265. if (optptr[2] < 4) {
  266. pp_ptr = optptr + 2;
  267. goto error;
  268. }
  269. /* NB: cf RFC-1812 5.2.4.1 */
  270. if (opt->srr) {
  271. pp_ptr = optptr;
  272. goto error;
  273. }
  274. if (!skb) {
  275. if (optptr[2] != 4 || optlen < 7 || ((optlen-3) & 3)) {
  276. pp_ptr = optptr + 1;
  277. goto error;
  278. }
  279. memcpy(&opt->faddr, &optptr[3], 4);
  280. if (optlen > 7)
  281. memmove(&optptr[3], &optptr[7], optlen-7);
  282. }
  283. opt->is_strictroute = (optptr[0] == IPOPT_SSRR);
  284. opt->srr = optptr - iph;
  285. break;
  286. case IPOPT_RR:
  287. if (opt->rr) {
  288. pp_ptr = optptr;
  289. goto error;
  290. }
  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. if (optptr[2] <= optlen) {
  300. if (optptr[2]+3 > optlen) {
  301. pp_ptr = optptr + 2;
  302. goto error;
  303. }
  304. if (rt) {
  305. spec_dst_fill(&spec_dst, skb);
  306. memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
  307. opt->is_changed = 1;
  308. }
  309. optptr[2] += 4;
  310. opt->rr_needaddr = 1;
  311. }
  312. opt->rr = optptr - iph;
  313. break;
  314. case IPOPT_TIMESTAMP:
  315. if (opt->ts) {
  316. pp_ptr = optptr;
  317. goto error;
  318. }
  319. if (optlen < 4) {
  320. pp_ptr = optptr + 1;
  321. goto error;
  322. }
  323. if (optptr[2] < 5) {
  324. pp_ptr = optptr + 2;
  325. goto error;
  326. }
  327. if (optptr[2] <= optlen) {
  328. unsigned char *timeptr = NULL;
  329. if (optptr[2]+3 > optlen) {
  330. pp_ptr = optptr + 2;
  331. goto error;
  332. }
  333. switch (optptr[3]&0xF) {
  334. case IPOPT_TS_TSONLY:
  335. if (skb)
  336. timeptr = &optptr[optptr[2]-1];
  337. opt->ts_needtime = 1;
  338. optptr[2] += 4;
  339. break;
  340. case IPOPT_TS_TSANDADDR:
  341. if (optptr[2]+7 > optlen) {
  342. pp_ptr = optptr + 2;
  343. goto error;
  344. }
  345. if (rt) {
  346. spec_dst_fill(&spec_dst, skb);
  347. memcpy(&optptr[optptr[2]-1], &spec_dst, 4);
  348. timeptr = &optptr[optptr[2]+3];
  349. }
  350. opt->ts_needaddr = 1;
  351. opt->ts_needtime = 1;
  352. optptr[2] += 8;
  353. break;
  354. case IPOPT_TS_PRESPEC:
  355. if (optptr[2]+7 > optlen) {
  356. pp_ptr = optptr + 2;
  357. goto error;
  358. }
  359. {
  360. __be32 addr;
  361. memcpy(&addr, &optptr[optptr[2]-1], 4);
  362. if (inet_addr_type(net, addr) == RTN_UNICAST)
  363. break;
  364. if (skb)
  365. timeptr = &optptr[optptr[2]+3];
  366. }
  367. opt->ts_needtime = 1;
  368. optptr[2] += 8;
  369. break;
  370. default:
  371. if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
  372. pp_ptr = optptr + 3;
  373. goto error;
  374. }
  375. break;
  376. }
  377. if (timeptr) {
  378. __be32 midtime;
  379. midtime = inet_current_timestamp();
  380. memcpy(timeptr, &midtime, 4);
  381. opt->is_changed = 1;
  382. }
  383. } else if ((optptr[3]&0xF) != IPOPT_TS_PRESPEC) {
  384. unsigned int overflow = optptr[3]>>4;
  385. if (overflow == 15) {
  386. pp_ptr = optptr + 3;
  387. goto error;
  388. }
  389. if (skb) {
  390. optptr[3] = (optptr[3]&0xF)|((overflow+1)<<4);
  391. opt->is_changed = 1;
  392. }
  393. }
  394. opt->ts = optptr - iph;
  395. break;
  396. case IPOPT_RA:
  397. if (optlen < 4) {
  398. pp_ptr = optptr + 1;
  399. goto error;
  400. }
  401. if (optptr[2] == 0 && optptr[3] == 0)
  402. opt->router_alert = optptr - iph;
  403. break;
  404. case IPOPT_CIPSO:
  405. if ((!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) || opt->cipso) {
  406. pp_ptr = optptr;
  407. goto error;
  408. }
  409. opt->cipso = optptr - iph;
  410. if (cipso_v4_validate(skb, &optptr)) {
  411. pp_ptr = optptr;
  412. goto error;
  413. }
  414. break;
  415. case IPOPT_SEC:
  416. case IPOPT_SID:
  417. default:
  418. if (!skb && !ns_capable(net->user_ns, CAP_NET_RAW)) {
  419. pp_ptr = optptr;
  420. goto error;
  421. }
  422. break;
  423. }
  424. l -= optlen;
  425. optptr += optlen;
  426. }
  427. eol:
  428. if (!pp_ptr)
  429. return 0;
  430. error:
  431. if (info)
  432. *info = htonl((pp_ptr-iph)<<24);
  433. return -EINVAL;
  434. }
  435. EXPORT_SYMBOL(__ip_options_compile);
  436. int ip_options_compile(struct net *net,
  437. struct ip_options *opt, struct sk_buff *skb)
  438. {
  439. int ret;
  440. __be32 info;
  441. ret = __ip_options_compile(net, opt, skb, &info);
  442. if (ret != 0 && skb)
  443. icmp_send(skb, ICMP_PARAMETERPROB, 0, info);
  444. return ret;
  445. }
  446. EXPORT_SYMBOL(ip_options_compile);
  447. /*
  448. * Undo all the changes done by ip_options_compile().
  449. */
  450. void ip_options_undo(struct ip_options *opt)
  451. {
  452. if (opt->srr) {
  453. unsigned char *optptr = opt->__data + opt->srr - sizeof(struct iphdr);
  454. memmove(optptr + 7, optptr + 3, optptr[1] - 7);
  455. memcpy(optptr + 3, &opt->faddr, 4);
  456. }
  457. if (opt->rr_needaddr) {
  458. unsigned char *optptr = opt->__data + opt->rr - sizeof(struct iphdr);
  459. optptr[2] -= 4;
  460. memset(&optptr[optptr[2] - 1], 0, 4);
  461. }
  462. if (opt->ts) {
  463. unsigned char *optptr = opt->__data + opt->ts - sizeof(struct iphdr);
  464. if (opt->ts_needtime) {
  465. optptr[2] -= 4;
  466. memset(&optptr[optptr[2] - 1], 0, 4);
  467. if ((optptr[3] & 0xF) == IPOPT_TS_PRESPEC)
  468. optptr[2] -= 4;
  469. }
  470. if (opt->ts_needaddr) {
  471. optptr[2] -= 4;
  472. memset(&optptr[optptr[2] - 1], 0, 4);
  473. }
  474. }
  475. }
  476. int ip_options_get(struct net *net, struct ip_options_rcu **optp,
  477. sockptr_t data, int optlen)
  478. {
  479. struct ip_options_rcu *opt;
  480. opt = kzalloc(sizeof(struct ip_options_rcu) + ((optlen + 3) & ~3),
  481. GFP_KERNEL);
  482. if (!opt)
  483. return -ENOMEM;
  484. if (optlen && copy_from_sockptr(opt->opt.__data, data, optlen)) {
  485. kfree(opt);
  486. return -EFAULT;
  487. }
  488. while (optlen & 3)
  489. opt->opt.__data[optlen++] = IPOPT_END;
  490. opt->opt.optlen = optlen;
  491. if (optlen && ip_options_compile(net, &opt->opt, NULL)) {
  492. kfree(opt);
  493. return -EINVAL;
  494. }
  495. kfree(*optp);
  496. *optp = opt;
  497. return 0;
  498. }
  499. void ip_forward_options(struct sk_buff *skb)
  500. {
  501. struct ip_options *opt = &(IPCB(skb)->opt);
  502. unsigned char *optptr;
  503. struct rtable *rt = skb_rtable(skb);
  504. unsigned char *raw = skb_network_header(skb);
  505. if (opt->rr_needaddr) {
  506. optptr = (unsigned char *)raw + opt->rr;
  507. ip_rt_get_source(&optptr[optptr[2]-5], skb, rt);
  508. opt->is_changed = 1;
  509. }
  510. if (opt->srr_is_hit) {
  511. int srrptr, srrspace;
  512. optptr = raw + opt->srr;
  513. for ( srrptr = optptr[2], srrspace = optptr[1];
  514. srrptr <= srrspace;
  515. srrptr += 4
  516. ) {
  517. if (srrptr + 3 > srrspace)
  518. break;
  519. if (memcmp(&opt->nexthop, &optptr[srrptr-1], 4) == 0)
  520. break;
  521. }
  522. if (srrptr + 3 <= srrspace) {
  523. opt->is_changed = 1;
  524. ip_hdr(skb)->daddr = opt->nexthop;
  525. ip_rt_get_source(&optptr[srrptr-1], skb, rt);
  526. optptr[2] = srrptr+4;
  527. } else {
  528. net_crit_ratelimited("%s(): Argh! Destination lost!\n",
  529. __func__);
  530. }
  531. if (opt->ts_needaddr) {
  532. optptr = raw + opt->ts;
  533. ip_rt_get_source(&optptr[optptr[2]-9], skb, rt);
  534. opt->is_changed = 1;
  535. }
  536. }
  537. if (opt->is_changed) {
  538. opt->is_changed = 0;
  539. ip_send_check(ip_hdr(skb));
  540. }
  541. }
  542. int ip_options_rcv_srr(struct sk_buff *skb, struct net_device *dev)
  543. {
  544. struct ip_options *opt = &(IPCB(skb)->opt);
  545. int srrspace, srrptr;
  546. __be32 nexthop;
  547. struct iphdr *iph = ip_hdr(skb);
  548. unsigned char *optptr = skb_network_header(skb) + opt->srr;
  549. struct rtable *rt = skb_rtable(skb);
  550. struct rtable *rt2;
  551. unsigned long orefdst;
  552. int err;
  553. if (!rt)
  554. return 0;
  555. if (skb->pkt_type != PACKET_HOST)
  556. return -EINVAL;
  557. if (rt->rt_type == RTN_UNICAST) {
  558. if (!opt->is_strictroute)
  559. return 0;
  560. icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl(16<<24));
  561. return -EINVAL;
  562. }
  563. if (rt->rt_type != RTN_LOCAL)
  564. return -EINVAL;
  565. for (srrptr = optptr[2], srrspace = optptr[1]; srrptr <= srrspace; srrptr += 4) {
  566. if (srrptr + 3 > srrspace) {
  567. icmp_send(skb, ICMP_PARAMETERPROB, 0, htonl((opt->srr+2)<<24));
  568. return -EINVAL;
  569. }
  570. memcpy(&nexthop, &optptr[srrptr-1], 4);
  571. orefdst = skb->_skb_refdst;
  572. skb_dst_set(skb, NULL);
  573. err = ip_route_input(skb, nexthop, iph->saddr, iph->tos, dev);
  574. rt2 = skb_rtable(skb);
  575. if (err || (rt2->rt_type != RTN_UNICAST && rt2->rt_type != RTN_LOCAL)) {
  576. skb_dst_drop(skb);
  577. skb->_skb_refdst = orefdst;
  578. return -EINVAL;
  579. }
  580. refdst_drop(orefdst);
  581. if (rt2->rt_type != RTN_LOCAL)
  582. break;
  583. /* Superfast 8) loopback forward */
  584. iph->daddr = nexthop;
  585. opt->is_changed = 1;
  586. }
  587. if (srrptr <= srrspace) {
  588. opt->srr_is_hit = 1;
  589. opt->nexthop = nexthop;
  590. opt->is_changed = 1;
  591. }
  592. return 0;
  593. }
  594. EXPORT_SYMBOL(ip_options_rcv_srr);