raw.c 26 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139
  1. // SPDX-License-Identifier: GPL-2.0-or-later
  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. * RAW - implementation of IP "raw" sockets.
  8. *
  9. * Authors: Ross Biro
  10. * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  11. *
  12. * Fixes:
  13. * Alan Cox : verify_area() fixed up
  14. * Alan Cox : ICMP error handling
  15. * Alan Cox : EMSGSIZE if you send too big a packet
  16. * Alan Cox : Now uses generic datagrams and shared
  17. * skbuff library. No more peek crashes,
  18. * no more backlogs
  19. * Alan Cox : Checks sk->broadcast.
  20. * Alan Cox : Uses skb_free_datagram/skb_copy_datagram
  21. * Alan Cox : Raw passes ip options too
  22. * Alan Cox : Setsocketopt added
  23. * Alan Cox : Fixed error return for broadcasts
  24. * Alan Cox : Removed wake_up calls
  25. * Alan Cox : Use ttl/tos
  26. * Alan Cox : Cleaned up old debugging
  27. * Alan Cox : Use new kernel side addresses
  28. * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets.
  29. * Alan Cox : BSD style RAW socket demultiplexing.
  30. * Alan Cox : Beginnings of mrouted support.
  31. * Alan Cox : Added IP_HDRINCL option.
  32. * Alan Cox : Skip broadcast check if BSDism set.
  33. * David S. Miller : New socket lookup architecture.
  34. */
  35. #include <linux/types.h>
  36. #include <linux/atomic.h>
  37. #include <asm/byteorder.h>
  38. #include <asm/current.h>
  39. #include <linux/uaccess.h>
  40. #include <asm/ioctls.h>
  41. #include <linux/stddef.h>
  42. #include <linux/slab.h>
  43. #include <linux/errno.h>
  44. #include <linux/kernel.h>
  45. #include <linux/export.h>
  46. #include <linux/spinlock.h>
  47. #include <linux/sockios.h>
  48. #include <linux/socket.h>
  49. #include <linux/in.h>
  50. #include <linux/mroute.h>
  51. #include <linux/netdevice.h>
  52. #include <linux/in_route.h>
  53. #include <linux/route.h>
  54. #include <linux/skbuff.h>
  55. #include <linux/igmp.h>
  56. #include <net/net_namespace.h>
  57. #include <net/dst.h>
  58. #include <net/sock.h>
  59. #include <linux/ip.h>
  60. #include <linux/net.h>
  61. #include <net/ip.h>
  62. #include <net/icmp.h>
  63. #include <net/udp.h>
  64. #include <net/raw.h>
  65. #include <net/snmp.h>
  66. #include <net/tcp_states.h>
  67. #include <net/inet_common.h>
  68. #include <net/checksum.h>
  69. #include <net/xfrm.h>
  70. #include <linux/rtnetlink.h>
  71. #include <linux/proc_fs.h>
  72. #include <linux/seq_file.h>
  73. #include <linux/netfilter.h>
  74. #include <linux/netfilter_ipv4.h>
  75. #include <linux/compat.h>
  76. #include <linux/uio.h>
  77. struct raw_frag_vec {
  78. struct msghdr *msg;
  79. union {
  80. struct icmphdr icmph;
  81. char c[1];
  82. } hdr;
  83. int hlen;
  84. };
  85. struct raw_hashinfo raw_v4_hashinfo = {
  86. .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
  87. };
  88. EXPORT_SYMBOL_GPL(raw_v4_hashinfo);
  89. int raw_hash_sk(struct sock *sk)
  90. {
  91. struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
  92. struct hlist_head *head;
  93. head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)];
  94. write_lock_bh(&h->lock);
  95. sk_add_node(sk, head);
  96. sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
  97. write_unlock_bh(&h->lock);
  98. return 0;
  99. }
  100. EXPORT_SYMBOL_GPL(raw_hash_sk);
  101. void raw_unhash_sk(struct sock *sk)
  102. {
  103. struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
  104. write_lock_bh(&h->lock);
  105. if (sk_del_node_init(sk))
  106. sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
  107. write_unlock_bh(&h->lock);
  108. }
  109. EXPORT_SYMBOL_GPL(raw_unhash_sk);
  110. struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
  111. unsigned short num, __be32 raddr, __be32 laddr,
  112. int dif, int sdif)
  113. {
  114. sk_for_each_from(sk) {
  115. struct inet_sock *inet = inet_sk(sk);
  116. if (net_eq(sock_net(sk), net) && inet->inet_num == num &&
  117. !(inet->inet_daddr && inet->inet_daddr != raddr) &&
  118. !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
  119. raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif))
  120. goto found; /* gotcha */
  121. }
  122. sk = NULL;
  123. found:
  124. return sk;
  125. }
  126. EXPORT_SYMBOL_GPL(__raw_v4_lookup);
  127. /*
  128. * 0 - deliver
  129. * 1 - block
  130. */
  131. static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
  132. {
  133. struct icmphdr _hdr;
  134. const struct icmphdr *hdr;
  135. hdr = skb_header_pointer(skb, skb_transport_offset(skb),
  136. sizeof(_hdr), &_hdr);
  137. if (!hdr)
  138. return 1;
  139. if (hdr->type < 32) {
  140. __u32 data = raw_sk(sk)->filter.data;
  141. return ((1U << hdr->type) & data) != 0;
  142. }
  143. /* Do not block unknown ICMP types */
  144. return 0;
  145. }
  146. /* IP input processing comes here for RAW socket delivery.
  147. * Caller owns SKB, so we must make clones.
  148. *
  149. * RFC 1122: SHOULD pass TOS value up to the transport layer.
  150. * -> It does. And not only TOS, but all IP header.
  151. */
  152. static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
  153. {
  154. int sdif = inet_sdif(skb);
  155. int dif = inet_iif(skb);
  156. struct sock *sk;
  157. struct hlist_head *head;
  158. int delivered = 0;
  159. struct net *net;
  160. read_lock(&raw_v4_hashinfo.lock);
  161. head = &raw_v4_hashinfo.ht[hash];
  162. if (hlist_empty(head))
  163. goto out;
  164. net = dev_net(skb->dev);
  165. sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
  166. iph->saddr, iph->daddr, dif, sdif);
  167. while (sk) {
  168. delivered = 1;
  169. if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
  170. ip_mc_sf_allow(sk, iph->daddr, iph->saddr,
  171. skb->dev->ifindex, sdif)) {
  172. struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
  173. /* Not releasing hash table! */
  174. if (clone)
  175. raw_rcv(sk, clone);
  176. }
  177. sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
  178. iph->saddr, iph->daddr,
  179. dif, sdif);
  180. }
  181. out:
  182. read_unlock(&raw_v4_hashinfo.lock);
  183. return delivered;
  184. }
  185. int raw_local_deliver(struct sk_buff *skb, int protocol)
  186. {
  187. int hash;
  188. struct sock *raw_sk;
  189. hash = protocol & (RAW_HTABLE_SIZE - 1);
  190. raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
  191. /* If there maybe a raw socket we must check - if not we
  192. * don't care less
  193. */
  194. if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
  195. raw_sk = NULL;
  196. return raw_sk != NULL;
  197. }
  198. static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
  199. {
  200. struct inet_sock *inet = inet_sk(sk);
  201. const int type = icmp_hdr(skb)->type;
  202. const int code = icmp_hdr(skb)->code;
  203. int err = 0;
  204. int harderr = 0;
  205. if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
  206. ipv4_sk_update_pmtu(skb, sk, info);
  207. else if (type == ICMP_REDIRECT) {
  208. ipv4_sk_redirect(skb, sk);
  209. return;
  210. }
  211. /* Report error on raw socket, if:
  212. 1. User requested ip_recverr.
  213. 2. Socket is connected (otherwise the error indication
  214. is useless without ip_recverr and error is hard.
  215. */
  216. if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
  217. return;
  218. switch (type) {
  219. default:
  220. case ICMP_TIME_EXCEEDED:
  221. err = EHOSTUNREACH;
  222. break;
  223. case ICMP_SOURCE_QUENCH:
  224. return;
  225. case ICMP_PARAMETERPROB:
  226. err = EPROTO;
  227. harderr = 1;
  228. break;
  229. case ICMP_DEST_UNREACH:
  230. err = EHOSTUNREACH;
  231. if (code > NR_ICMP_UNREACH)
  232. break;
  233. if (code == ICMP_FRAG_NEEDED) {
  234. harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
  235. err = EMSGSIZE;
  236. } else {
  237. err = icmp_err_convert[code].errno;
  238. harderr = icmp_err_convert[code].fatal;
  239. }
  240. }
  241. if (inet->recverr) {
  242. const struct iphdr *iph = (const struct iphdr *)skb->data;
  243. u8 *payload = skb->data + (iph->ihl << 2);
  244. if (inet->hdrincl)
  245. payload = skb->data;
  246. ip_icmp_error(sk, skb, err, 0, info, payload);
  247. }
  248. if (inet->recverr || harderr) {
  249. sk->sk_err = err;
  250. sk->sk_error_report(sk);
  251. }
  252. }
  253. void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
  254. {
  255. int hash;
  256. struct sock *raw_sk;
  257. const struct iphdr *iph;
  258. struct net *net;
  259. hash = protocol & (RAW_HTABLE_SIZE - 1);
  260. read_lock(&raw_v4_hashinfo.lock);
  261. raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
  262. if (raw_sk) {
  263. int dif = skb->dev->ifindex;
  264. int sdif = inet_sdif(skb);
  265. iph = (const struct iphdr *)skb->data;
  266. net = dev_net(skb->dev);
  267. while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
  268. iph->daddr, iph->saddr,
  269. dif, sdif)) != NULL) {
  270. raw_err(raw_sk, skb, info);
  271. raw_sk = sk_next(raw_sk);
  272. iph = (const struct iphdr *)skb->data;
  273. }
  274. }
  275. read_unlock(&raw_v4_hashinfo.lock);
  276. }
  277. static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
  278. {
  279. /* Charge it to the socket. */
  280. ipv4_pktinfo_prepare(sk, skb);
  281. if (sock_queue_rcv_skb(sk, skb) < 0) {
  282. kfree_skb(skb);
  283. return NET_RX_DROP;
  284. }
  285. return NET_RX_SUCCESS;
  286. }
  287. int raw_rcv(struct sock *sk, struct sk_buff *skb)
  288. {
  289. if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
  290. atomic_inc(&sk->sk_drops);
  291. kfree_skb(skb);
  292. return NET_RX_DROP;
  293. }
  294. nf_reset_ct(skb);
  295. skb_push(skb, skb->data - skb_network_header(skb));
  296. raw_rcv_skb(sk, skb);
  297. return 0;
  298. }
  299. static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
  300. struct msghdr *msg, size_t length,
  301. struct rtable **rtp, unsigned int flags,
  302. const struct sockcm_cookie *sockc)
  303. {
  304. struct inet_sock *inet = inet_sk(sk);
  305. struct net *net = sock_net(sk);
  306. struct iphdr *iph;
  307. struct sk_buff *skb;
  308. unsigned int iphlen;
  309. int err;
  310. struct rtable *rt = *rtp;
  311. int hlen, tlen;
  312. if (length > rt->dst.dev->mtu) {
  313. ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
  314. rt->dst.dev->mtu);
  315. return -EMSGSIZE;
  316. }
  317. if (length < sizeof(struct iphdr))
  318. return -EINVAL;
  319. if (flags&MSG_PROBE)
  320. goto out;
  321. hlen = LL_RESERVED_SPACE(rt->dst.dev);
  322. tlen = rt->dst.dev->needed_tailroom;
  323. skb = sock_alloc_send_skb(sk,
  324. length + hlen + tlen + 15,
  325. flags & MSG_DONTWAIT, &err);
  326. if (!skb)
  327. goto error;
  328. skb_reserve(skb, hlen);
  329. skb->priority = sk->sk_priority;
  330. skb->mark = sockc->mark;
  331. skb->tstamp = sockc->transmit_time;
  332. skb_dst_set(skb, &rt->dst);
  333. *rtp = NULL;
  334. skb_reset_network_header(skb);
  335. iph = ip_hdr(skb);
  336. skb_put(skb, length);
  337. skb->ip_summed = CHECKSUM_NONE;
  338. skb_setup_tx_timestamp(skb, sockc->tsflags);
  339. if (flags & MSG_CONFIRM)
  340. skb_set_dst_pending_confirm(skb, 1);
  341. skb->transport_header = skb->network_header;
  342. err = -EFAULT;
  343. if (memcpy_from_msg(iph, msg, length))
  344. goto error_free;
  345. iphlen = iph->ihl * 4;
  346. /*
  347. * We don't want to modify the ip header, but we do need to
  348. * be sure that it won't cause problems later along the network
  349. * stack. Specifically we want to make sure that iph->ihl is a
  350. * sane value. If ihl points beyond the length of the buffer passed
  351. * in, reject the frame as invalid
  352. */
  353. err = -EINVAL;
  354. if (iphlen > length)
  355. goto error_free;
  356. if (iphlen >= sizeof(*iph)) {
  357. if (!iph->saddr)
  358. iph->saddr = fl4->saddr;
  359. iph->check = 0;
  360. iph->tot_len = htons(length);
  361. if (!iph->id)
  362. ip_select_ident(net, skb, NULL);
  363. iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
  364. skb->transport_header += iphlen;
  365. if (iph->protocol == IPPROTO_ICMP &&
  366. length >= iphlen + sizeof(struct icmphdr))
  367. icmp_out_count(net, ((struct icmphdr *)
  368. skb_transport_header(skb))->type);
  369. }
  370. err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
  371. net, sk, skb, NULL, rt->dst.dev,
  372. dst_output);
  373. if (err > 0)
  374. err = net_xmit_errno(err);
  375. if (err)
  376. goto error;
  377. out:
  378. return 0;
  379. error_free:
  380. kfree_skb(skb);
  381. error:
  382. IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
  383. if (err == -ENOBUFS && !inet->recverr)
  384. err = 0;
  385. return err;
  386. }
  387. static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
  388. {
  389. int err;
  390. if (fl4->flowi4_proto != IPPROTO_ICMP)
  391. return 0;
  392. /* We only need the first two bytes. */
  393. rfv->hlen = 2;
  394. err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
  395. if (err)
  396. return err;
  397. fl4->fl4_icmp_type = rfv->hdr.icmph.type;
  398. fl4->fl4_icmp_code = rfv->hdr.icmph.code;
  399. return 0;
  400. }
  401. static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
  402. struct sk_buff *skb)
  403. {
  404. struct raw_frag_vec *rfv = from;
  405. if (offset < rfv->hlen) {
  406. int copy = min(rfv->hlen - offset, len);
  407. if (skb->ip_summed == CHECKSUM_PARTIAL)
  408. memcpy(to, rfv->hdr.c + offset, copy);
  409. else
  410. skb->csum = csum_block_add(
  411. skb->csum,
  412. csum_partial_copy_nocheck(rfv->hdr.c + offset,
  413. to, copy),
  414. odd);
  415. odd = 0;
  416. offset += copy;
  417. to += copy;
  418. len -= copy;
  419. if (!len)
  420. return 0;
  421. }
  422. offset -= rfv->hlen;
  423. return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
  424. }
  425. static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
  426. {
  427. struct inet_sock *inet = inet_sk(sk);
  428. struct net *net = sock_net(sk);
  429. struct ipcm_cookie ipc;
  430. struct rtable *rt = NULL;
  431. struct flowi4 fl4;
  432. int free = 0;
  433. __be32 daddr;
  434. __be32 saddr;
  435. u8 tos;
  436. int err;
  437. struct ip_options_data opt_copy;
  438. struct raw_frag_vec rfv;
  439. int hdrincl;
  440. err = -EMSGSIZE;
  441. if (len > 0xFFFF)
  442. goto out;
  443. /* hdrincl should be READ_ONCE(inet->hdrincl)
  444. * but READ_ONCE() doesn't work with bit fields.
  445. * Doing this indirectly yields the same result.
  446. */
  447. hdrincl = inet->hdrincl;
  448. hdrincl = READ_ONCE(hdrincl);
  449. /*
  450. * Check the flags.
  451. */
  452. err = -EOPNOTSUPP;
  453. if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
  454. goto out; /* compatibility */
  455. /*
  456. * Get and verify the address.
  457. */
  458. if (msg->msg_namelen) {
  459. DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
  460. err = -EINVAL;
  461. if (msg->msg_namelen < sizeof(*usin))
  462. goto out;
  463. if (usin->sin_family != AF_INET) {
  464. pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
  465. __func__, current->comm);
  466. err = -EAFNOSUPPORT;
  467. if (usin->sin_family)
  468. goto out;
  469. }
  470. daddr = usin->sin_addr.s_addr;
  471. /* ANK: I did not forget to get protocol from port field.
  472. * I just do not know, who uses this weirdness.
  473. * IP_HDRINCL is much more convenient.
  474. */
  475. } else {
  476. err = -EDESTADDRREQ;
  477. if (sk->sk_state != TCP_ESTABLISHED)
  478. goto out;
  479. daddr = inet->inet_daddr;
  480. }
  481. ipcm_init_sk(&ipc, inet);
  482. if (msg->msg_controllen) {
  483. err = ip_cmsg_send(sk, msg, &ipc, false);
  484. if (unlikely(err)) {
  485. kfree(ipc.opt);
  486. goto out;
  487. }
  488. if (ipc.opt)
  489. free = 1;
  490. }
  491. saddr = ipc.addr;
  492. ipc.addr = daddr;
  493. if (!ipc.opt) {
  494. struct ip_options_rcu *inet_opt;
  495. rcu_read_lock();
  496. inet_opt = rcu_dereference(inet->inet_opt);
  497. if (inet_opt) {
  498. memcpy(&opt_copy, inet_opt,
  499. sizeof(*inet_opt) + inet_opt->opt.optlen);
  500. ipc.opt = &opt_copy.opt;
  501. }
  502. rcu_read_unlock();
  503. }
  504. if (ipc.opt) {
  505. err = -EINVAL;
  506. /* Linux does not mangle headers on raw sockets,
  507. * so that IP options + IP_HDRINCL is non-sense.
  508. */
  509. if (hdrincl)
  510. goto done;
  511. if (ipc.opt->opt.srr) {
  512. if (!daddr)
  513. goto done;
  514. daddr = ipc.opt->opt.faddr;
  515. }
  516. }
  517. tos = get_rtconn_flags(&ipc, sk);
  518. if (msg->msg_flags & MSG_DONTROUTE)
  519. tos |= RTO_ONLINK;
  520. if (ipv4_is_multicast(daddr)) {
  521. if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
  522. ipc.oif = inet->mc_index;
  523. if (!saddr)
  524. saddr = inet->mc_addr;
  525. } else if (!ipc.oif) {
  526. ipc.oif = inet->uc_index;
  527. } else if (ipv4_is_lbcast(daddr) && inet->uc_index) {
  528. /* oif is set, packet is to local broadcast
  529. * and uc_index is set. oif is most likely set
  530. * by sk_bound_dev_if. If uc_index != oif check if the
  531. * oif is an L3 master and uc_index is an L3 slave.
  532. * If so, we want to allow the send using the uc_index.
  533. */
  534. if (ipc.oif != inet->uc_index &&
  535. ipc.oif == l3mdev_master_ifindex_by_index(sock_net(sk),
  536. inet->uc_index)) {
  537. ipc.oif = inet->uc_index;
  538. }
  539. }
  540. flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
  541. RT_SCOPE_UNIVERSE,
  542. hdrincl ? IPPROTO_RAW : sk->sk_protocol,
  543. inet_sk_flowi_flags(sk) |
  544. (hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
  545. daddr, saddr, 0, 0, sk->sk_uid);
  546. if (!hdrincl) {
  547. rfv.msg = msg;
  548. rfv.hlen = 0;
  549. err = raw_probe_proto_opt(&rfv, &fl4);
  550. if (err)
  551. goto done;
  552. }
  553. security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
  554. rt = ip_route_output_flow(net, &fl4, sk);
  555. if (IS_ERR(rt)) {
  556. err = PTR_ERR(rt);
  557. rt = NULL;
  558. goto done;
  559. }
  560. err = -EACCES;
  561. if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
  562. goto done;
  563. if (msg->msg_flags & MSG_CONFIRM)
  564. goto do_confirm;
  565. back_from_confirm:
  566. if (hdrincl)
  567. err = raw_send_hdrinc(sk, &fl4, msg, len,
  568. &rt, msg->msg_flags, &ipc.sockc);
  569. else {
  570. if (!ipc.addr)
  571. ipc.addr = fl4.daddr;
  572. lock_sock(sk);
  573. err = ip_append_data(sk, &fl4, raw_getfrag,
  574. &rfv, len, 0,
  575. &ipc, &rt, msg->msg_flags);
  576. if (err)
  577. ip_flush_pending_frames(sk);
  578. else if (!(msg->msg_flags & MSG_MORE)) {
  579. err = ip_push_pending_frames(sk, &fl4);
  580. if (err == -ENOBUFS && !inet->recverr)
  581. err = 0;
  582. }
  583. release_sock(sk);
  584. }
  585. done:
  586. if (free)
  587. kfree(ipc.opt);
  588. ip_rt_put(rt);
  589. out:
  590. if (err < 0)
  591. return err;
  592. return len;
  593. do_confirm:
  594. if (msg->msg_flags & MSG_PROBE)
  595. dst_confirm_neigh(&rt->dst, &fl4.daddr);
  596. if (!(msg->msg_flags & MSG_PROBE) || len)
  597. goto back_from_confirm;
  598. err = 0;
  599. goto done;
  600. }
  601. static void raw_close(struct sock *sk, long timeout)
  602. {
  603. /*
  604. * Raw sockets may have direct kernel references. Kill them.
  605. */
  606. ip_ra_control(sk, 0, NULL);
  607. sk_common_release(sk);
  608. }
  609. static void raw_destroy(struct sock *sk)
  610. {
  611. lock_sock(sk);
  612. ip_flush_pending_frames(sk);
  613. release_sock(sk);
  614. }
  615. /* This gets rid of all the nasties in af_inet. -DaveM */
  616. static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
  617. {
  618. struct inet_sock *inet = inet_sk(sk);
  619. struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
  620. u32 tb_id = RT_TABLE_LOCAL;
  621. int ret = -EINVAL;
  622. int chk_addr_ret;
  623. lock_sock(sk);
  624. if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
  625. goto out;
  626. if (sk->sk_bound_dev_if)
  627. tb_id = l3mdev_fib_table_by_index(sock_net(sk),
  628. sk->sk_bound_dev_if) ? : tb_id;
  629. chk_addr_ret = inet_addr_type_table(sock_net(sk), addr->sin_addr.s_addr,
  630. tb_id);
  631. ret = -EADDRNOTAVAIL;
  632. if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
  633. chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
  634. goto out;
  635. inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
  636. if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
  637. inet->inet_saddr = 0; /* Use device */
  638. sk_dst_reset(sk);
  639. ret = 0;
  640. out:
  641. release_sock(sk);
  642. return ret;
  643. }
  644. /*
  645. * This should be easy, if there is something there
  646. * we return it, otherwise we block.
  647. */
  648. static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
  649. int noblock, int flags, int *addr_len)
  650. {
  651. struct inet_sock *inet = inet_sk(sk);
  652. size_t copied = 0;
  653. int err = -EOPNOTSUPP;
  654. DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
  655. struct sk_buff *skb;
  656. if (flags & MSG_OOB)
  657. goto out;
  658. if (flags & MSG_ERRQUEUE) {
  659. err = ip_recv_error(sk, msg, len, addr_len);
  660. goto out;
  661. }
  662. skb = skb_recv_datagram(sk, flags, noblock, &err);
  663. if (!skb)
  664. goto out;
  665. copied = skb->len;
  666. if (len < copied) {
  667. msg->msg_flags |= MSG_TRUNC;
  668. copied = len;
  669. }
  670. err = skb_copy_datagram_msg(skb, 0, msg, copied);
  671. if (err)
  672. goto done;
  673. sock_recv_ts_and_drops(msg, sk, skb);
  674. /* Copy the address. */
  675. if (sin) {
  676. sin->sin_family = AF_INET;
  677. sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
  678. sin->sin_port = 0;
  679. memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
  680. *addr_len = sizeof(*sin);
  681. }
  682. if (inet->cmsg_flags)
  683. ip_cmsg_recv(msg, skb);
  684. if (flags & MSG_TRUNC)
  685. copied = skb->len;
  686. done:
  687. skb_free_datagram(sk, skb);
  688. out:
  689. if (err)
  690. return err;
  691. return copied;
  692. }
  693. static int raw_sk_init(struct sock *sk)
  694. {
  695. struct raw_sock *rp = raw_sk(sk);
  696. if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
  697. memset(&rp->filter, 0, sizeof(rp->filter));
  698. return 0;
  699. }
  700. static int raw_seticmpfilter(struct sock *sk, sockptr_t optval, int optlen)
  701. {
  702. if (optlen > sizeof(struct icmp_filter))
  703. optlen = sizeof(struct icmp_filter);
  704. if (copy_from_sockptr(&raw_sk(sk)->filter, optval, optlen))
  705. return -EFAULT;
  706. return 0;
  707. }
  708. static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
  709. {
  710. int len, ret = -EFAULT;
  711. if (get_user(len, optlen))
  712. goto out;
  713. ret = -EINVAL;
  714. if (len < 0)
  715. goto out;
  716. if (len > sizeof(struct icmp_filter))
  717. len = sizeof(struct icmp_filter);
  718. ret = -EFAULT;
  719. if (put_user(len, optlen) ||
  720. copy_to_user(optval, &raw_sk(sk)->filter, len))
  721. goto out;
  722. ret = 0;
  723. out: return ret;
  724. }
  725. static int do_raw_setsockopt(struct sock *sk, int level, int optname,
  726. sockptr_t optval, unsigned int optlen)
  727. {
  728. if (optname == ICMP_FILTER) {
  729. if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
  730. return -EOPNOTSUPP;
  731. else
  732. return raw_seticmpfilter(sk, optval, optlen);
  733. }
  734. return -ENOPROTOOPT;
  735. }
  736. static int raw_setsockopt(struct sock *sk, int level, int optname,
  737. sockptr_t optval, unsigned int optlen)
  738. {
  739. if (level != SOL_RAW)
  740. return ip_setsockopt(sk, level, optname, optval, optlen);
  741. return do_raw_setsockopt(sk, level, optname, optval, optlen);
  742. }
  743. static int do_raw_getsockopt(struct sock *sk, int level, int optname,
  744. char __user *optval, int __user *optlen)
  745. {
  746. if (optname == ICMP_FILTER) {
  747. if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
  748. return -EOPNOTSUPP;
  749. else
  750. return raw_geticmpfilter(sk, optval, optlen);
  751. }
  752. return -ENOPROTOOPT;
  753. }
  754. static int raw_getsockopt(struct sock *sk, int level, int optname,
  755. char __user *optval, int __user *optlen)
  756. {
  757. if (level != SOL_RAW)
  758. return ip_getsockopt(sk, level, optname, optval, optlen);
  759. return do_raw_getsockopt(sk, level, optname, optval, optlen);
  760. }
  761. static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
  762. {
  763. switch (cmd) {
  764. case SIOCOUTQ: {
  765. int amount = sk_wmem_alloc_get(sk);
  766. return put_user(amount, (int __user *)arg);
  767. }
  768. case SIOCINQ: {
  769. struct sk_buff *skb;
  770. int amount = 0;
  771. spin_lock_bh(&sk->sk_receive_queue.lock);
  772. skb = skb_peek(&sk->sk_receive_queue);
  773. if (skb)
  774. amount = skb->len;
  775. spin_unlock_bh(&sk->sk_receive_queue.lock);
  776. return put_user(amount, (int __user *)arg);
  777. }
  778. default:
  779. #ifdef CONFIG_IP_MROUTE
  780. return ipmr_ioctl(sk, cmd, (void __user *)arg);
  781. #else
  782. return -ENOIOCTLCMD;
  783. #endif
  784. }
  785. }
  786. #ifdef CONFIG_COMPAT
  787. static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
  788. {
  789. switch (cmd) {
  790. case SIOCOUTQ:
  791. case SIOCINQ:
  792. return -ENOIOCTLCMD;
  793. default:
  794. #ifdef CONFIG_IP_MROUTE
  795. return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
  796. #else
  797. return -ENOIOCTLCMD;
  798. #endif
  799. }
  800. }
  801. #endif
  802. int raw_abort(struct sock *sk, int err)
  803. {
  804. lock_sock(sk);
  805. sk->sk_err = err;
  806. sk->sk_error_report(sk);
  807. __udp_disconnect(sk, 0);
  808. release_sock(sk);
  809. return 0;
  810. }
  811. EXPORT_SYMBOL_GPL(raw_abort);
  812. struct proto raw_prot = {
  813. .name = "RAW",
  814. .owner = THIS_MODULE,
  815. .close = raw_close,
  816. .destroy = raw_destroy,
  817. .connect = ip4_datagram_connect,
  818. .disconnect = __udp_disconnect,
  819. .ioctl = raw_ioctl,
  820. .init = raw_sk_init,
  821. .setsockopt = raw_setsockopt,
  822. .getsockopt = raw_getsockopt,
  823. .sendmsg = raw_sendmsg,
  824. .recvmsg = raw_recvmsg,
  825. .bind = raw_bind,
  826. .backlog_rcv = raw_rcv_skb,
  827. .release_cb = ip4_datagram_release_cb,
  828. .hash = raw_hash_sk,
  829. .unhash = raw_unhash_sk,
  830. .obj_size = sizeof(struct raw_sock),
  831. .useroffset = offsetof(struct raw_sock, filter),
  832. .usersize = sizeof_field(struct raw_sock, filter),
  833. .h.raw_hash = &raw_v4_hashinfo,
  834. #ifdef CONFIG_COMPAT
  835. .compat_ioctl = compat_raw_ioctl,
  836. #endif
  837. .diag_destroy = raw_abort,
  838. };
  839. #ifdef CONFIG_PROC_FS
  840. static struct sock *raw_get_first(struct seq_file *seq)
  841. {
  842. struct sock *sk;
  843. struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file));
  844. struct raw_iter_state *state = raw_seq_private(seq);
  845. for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
  846. ++state->bucket) {
  847. sk_for_each(sk, &h->ht[state->bucket])
  848. if (sock_net(sk) == seq_file_net(seq))
  849. goto found;
  850. }
  851. sk = NULL;
  852. found:
  853. return sk;
  854. }
  855. static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
  856. {
  857. struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file));
  858. struct raw_iter_state *state = raw_seq_private(seq);
  859. do {
  860. sk = sk_next(sk);
  861. try_again:
  862. ;
  863. } while (sk && sock_net(sk) != seq_file_net(seq));
  864. if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
  865. sk = sk_head(&h->ht[state->bucket]);
  866. goto try_again;
  867. }
  868. return sk;
  869. }
  870. static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
  871. {
  872. struct sock *sk = raw_get_first(seq);
  873. if (sk)
  874. while (pos && (sk = raw_get_next(seq, sk)) != NULL)
  875. --pos;
  876. return pos ? NULL : sk;
  877. }
  878. void *raw_seq_start(struct seq_file *seq, loff_t *pos)
  879. __acquires(&h->lock)
  880. {
  881. struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file));
  882. read_lock(&h->lock);
  883. return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
  884. }
  885. EXPORT_SYMBOL_GPL(raw_seq_start);
  886. void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
  887. {
  888. struct sock *sk;
  889. if (v == SEQ_START_TOKEN)
  890. sk = raw_get_first(seq);
  891. else
  892. sk = raw_get_next(seq, v);
  893. ++*pos;
  894. return sk;
  895. }
  896. EXPORT_SYMBOL_GPL(raw_seq_next);
  897. void raw_seq_stop(struct seq_file *seq, void *v)
  898. __releases(&h->lock)
  899. {
  900. struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file));
  901. read_unlock(&h->lock);
  902. }
  903. EXPORT_SYMBOL_GPL(raw_seq_stop);
  904. static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
  905. {
  906. struct inet_sock *inet = inet_sk(sp);
  907. __be32 dest = inet->inet_daddr,
  908. src = inet->inet_rcv_saddr;
  909. __u16 destp = 0,
  910. srcp = inet->inet_num;
  911. seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
  912. " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u\n",
  913. i, src, srcp, dest, destp, sp->sk_state,
  914. sk_wmem_alloc_get(sp),
  915. sk_rmem_alloc_get(sp),
  916. 0, 0L, 0,
  917. from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
  918. 0, sock_i_ino(sp),
  919. refcount_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
  920. }
  921. static int raw_seq_show(struct seq_file *seq, void *v)
  922. {
  923. if (v == SEQ_START_TOKEN)
  924. seq_printf(seq, " sl local_address rem_address st tx_queue "
  925. "rx_queue tr tm->when retrnsmt uid timeout "
  926. "inode ref pointer drops\n");
  927. else
  928. raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
  929. return 0;
  930. }
  931. static const struct seq_operations raw_seq_ops = {
  932. .start = raw_seq_start,
  933. .next = raw_seq_next,
  934. .stop = raw_seq_stop,
  935. .show = raw_seq_show,
  936. };
  937. static __net_init int raw_init_net(struct net *net)
  938. {
  939. if (!proc_create_net_data("raw", 0444, net->proc_net, &raw_seq_ops,
  940. sizeof(struct raw_iter_state), &raw_v4_hashinfo))
  941. return -ENOMEM;
  942. return 0;
  943. }
  944. static __net_exit void raw_exit_net(struct net *net)
  945. {
  946. remove_proc_entry("raw", net->proc_net);
  947. }
  948. static __net_initdata struct pernet_operations raw_net_ops = {
  949. .init = raw_init_net,
  950. .exit = raw_exit_net,
  951. };
  952. int __init raw_proc_init(void)
  953. {
  954. return register_pernet_subsys(&raw_net_ops);
  955. }
  956. void __init raw_proc_exit(void)
  957. {
  958. unregister_pernet_subsys(&raw_net_ops);
  959. }
  960. #endif /* CONFIG_PROC_FS */
  961. static void raw_sysctl_init_net(struct net *net)
  962. {
  963. #ifdef CONFIG_NET_L3_MASTER_DEV
  964. net->ipv4.sysctl_raw_l3mdev_accept = 1;
  965. #endif
  966. }
  967. static int __net_init raw_sysctl_init(struct net *net)
  968. {
  969. raw_sysctl_init_net(net);
  970. return 0;
  971. }
  972. static struct pernet_operations __net_initdata raw_sysctl_ops = {
  973. .init = raw_sysctl_init,
  974. };
  975. void __init raw_init(void)
  976. {
  977. raw_sysctl_init_net(&init_net);
  978. if (register_pernet_subsys(&raw_sysctl_ops))
  979. panic("RAW: failed to init sysctl parameters.\n");
  980. }