Kyle Swenson | 8d8f654 | 2021-03-15 11:02:55 -0600 | [diff] [blame^] | 1 | /* |
| 2 | * RAW sockets for IPv6 |
| 3 | * Linux INET6 implementation |
| 4 | * |
| 5 | * Authors: |
| 6 | * Pedro Roque <roque@di.fc.ul.pt> |
| 7 | * |
| 8 | * Adapted from linux/net/ipv4/raw.c |
| 9 | * |
| 10 | * Fixes: |
| 11 | * Hideaki YOSHIFUJI : sin6_scope_id support |
| 12 | * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance) |
| 13 | * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data |
| 14 | * |
| 15 | * This program is free software; you can redistribute it and/or |
| 16 | * modify it under the terms of the GNU General Public License |
| 17 | * as published by the Free Software Foundation; either version |
| 18 | * 2 of the License, or (at your option) any later version. |
| 19 | */ |
| 20 | |
| 21 | #include <linux/errno.h> |
| 22 | #include <linux/types.h> |
| 23 | #include <linux/socket.h> |
| 24 | #include <linux/slab.h> |
| 25 | #include <linux/sockios.h> |
| 26 | #include <linux/net.h> |
| 27 | #include <linux/in6.h> |
| 28 | #include <linux/netdevice.h> |
| 29 | #include <linux/if_arp.h> |
| 30 | #include <linux/icmpv6.h> |
| 31 | #include <linux/netfilter.h> |
| 32 | #include <linux/netfilter_ipv6.h> |
| 33 | #include <linux/skbuff.h> |
| 34 | #include <linux/compat.h> |
| 35 | #include <linux/uaccess.h> |
| 36 | #include <asm/ioctls.h> |
| 37 | |
| 38 | #include <net/net_namespace.h> |
| 39 | #include <net/ip.h> |
| 40 | #include <net/sock.h> |
| 41 | #include <net/snmp.h> |
| 42 | |
| 43 | #include <net/ipv6.h> |
| 44 | #include <net/ndisc.h> |
| 45 | #include <net/protocol.h> |
| 46 | #include <net/ip6_route.h> |
| 47 | #include <net/ip6_checksum.h> |
| 48 | #include <net/addrconf.h> |
| 49 | #include <net/transp_v6.h> |
| 50 | #include <net/udp.h> |
| 51 | #include <net/inet_common.h> |
| 52 | #include <net/tcp_states.h> |
| 53 | #if IS_ENABLED(CONFIG_IPV6_MIP6) |
| 54 | #include <net/mip6.h> |
| 55 | #endif |
| 56 | #include <linux/mroute6.h> |
| 57 | |
| 58 | #include <net/raw.h> |
| 59 | #include <net/rawv6.h> |
| 60 | #include <net/xfrm.h> |
| 61 | |
| 62 | #include <linux/proc_fs.h> |
| 63 | #include <linux/seq_file.h> |
| 64 | #include <linux/export.h> |
| 65 | |
| 66 | #define ICMPV6_HDRLEN 4 /* ICMPv6 header, RFC 4443 Section 2.1 */ |
| 67 | |
| 68 | static struct raw_hashinfo raw_v6_hashinfo = { |
| 69 | .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock), |
| 70 | }; |
| 71 | |
| 72 | static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk, |
| 73 | unsigned short num, const struct in6_addr *loc_addr, |
| 74 | const struct in6_addr *rmt_addr, int dif) |
| 75 | { |
| 76 | bool is_multicast = ipv6_addr_is_multicast(loc_addr); |
| 77 | |
| 78 | sk_for_each_from(sk) |
| 79 | if (inet_sk(sk)->inet_num == num) { |
| 80 | |
| 81 | if (!net_eq(sock_net(sk), net)) |
| 82 | continue; |
| 83 | |
| 84 | if (!ipv6_addr_any(&sk->sk_v6_daddr) && |
| 85 | !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) |
| 86 | continue; |
| 87 | |
| 88 | if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) |
| 89 | continue; |
| 90 | |
| 91 | if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) { |
| 92 | if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)) |
| 93 | goto found; |
| 94 | if (is_multicast && |
| 95 | inet6_mc_check(sk, loc_addr, rmt_addr)) |
| 96 | goto found; |
| 97 | continue; |
| 98 | } |
| 99 | goto found; |
| 100 | } |
| 101 | sk = NULL; |
| 102 | found: |
| 103 | return sk; |
| 104 | } |
| 105 | |
| 106 | /* |
| 107 | * 0 - deliver |
| 108 | * 1 - block |
| 109 | */ |
| 110 | static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb) |
| 111 | { |
| 112 | struct icmp6hdr _hdr; |
| 113 | const struct icmp6hdr *hdr; |
| 114 | |
| 115 | /* We require only the four bytes of the ICMPv6 header, not any |
| 116 | * additional bytes of message body in "struct icmp6hdr". |
| 117 | */ |
| 118 | hdr = skb_header_pointer(skb, skb_transport_offset(skb), |
| 119 | ICMPV6_HDRLEN, &_hdr); |
| 120 | if (hdr) { |
| 121 | const __u32 *data = &raw6_sk(sk)->filter.data[0]; |
| 122 | unsigned int type = hdr->icmp6_type; |
| 123 | |
| 124 | return (data[type >> 5] & (1U << (type & 31))) != 0; |
| 125 | } |
| 126 | return 1; |
| 127 | } |
| 128 | |
| 129 | #if IS_ENABLED(CONFIG_IPV6_MIP6) |
| 130 | typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb); |
| 131 | |
| 132 | static mh_filter_t __rcu *mh_filter __read_mostly; |
| 133 | |
| 134 | int rawv6_mh_filter_register(mh_filter_t filter) |
| 135 | { |
| 136 | rcu_assign_pointer(mh_filter, filter); |
| 137 | return 0; |
| 138 | } |
| 139 | EXPORT_SYMBOL(rawv6_mh_filter_register); |
| 140 | |
| 141 | int rawv6_mh_filter_unregister(mh_filter_t filter) |
| 142 | { |
| 143 | RCU_INIT_POINTER(mh_filter, NULL); |
| 144 | synchronize_rcu(); |
| 145 | return 0; |
| 146 | } |
| 147 | EXPORT_SYMBOL(rawv6_mh_filter_unregister); |
| 148 | |
| 149 | #endif |
| 150 | |
| 151 | /* |
| 152 | * demultiplex raw sockets. |
| 153 | * (should consider queueing the skb in the sock receive_queue |
| 154 | * without calling rawv6.c) |
| 155 | * |
| 156 | * Caller owns SKB so we must make clones. |
| 157 | */ |
| 158 | static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr) |
| 159 | { |
| 160 | const struct in6_addr *saddr; |
| 161 | const struct in6_addr *daddr; |
| 162 | struct sock *sk; |
| 163 | bool delivered = false; |
| 164 | __u8 hash; |
| 165 | struct net *net; |
| 166 | |
| 167 | saddr = &ipv6_hdr(skb)->saddr; |
| 168 | daddr = saddr + 1; |
| 169 | |
| 170 | hash = nexthdr & (RAW_HTABLE_SIZE - 1); |
| 171 | |
| 172 | read_lock(&raw_v6_hashinfo.lock); |
| 173 | sk = sk_head(&raw_v6_hashinfo.ht[hash]); |
| 174 | |
| 175 | if (!sk) |
| 176 | goto out; |
| 177 | |
| 178 | net = dev_net(skb->dev); |
| 179 | sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, inet6_iif(skb)); |
| 180 | |
| 181 | while (sk) { |
| 182 | int filtered; |
| 183 | |
| 184 | delivered = true; |
| 185 | switch (nexthdr) { |
| 186 | case IPPROTO_ICMPV6: |
| 187 | filtered = icmpv6_filter(sk, skb); |
| 188 | break; |
| 189 | |
| 190 | #if IS_ENABLED(CONFIG_IPV6_MIP6) |
| 191 | case IPPROTO_MH: |
| 192 | { |
| 193 | /* XXX: To validate MH only once for each packet, |
| 194 | * this is placed here. It should be after checking |
| 195 | * xfrm policy, however it doesn't. The checking xfrm |
| 196 | * policy is placed in rawv6_rcv() because it is |
| 197 | * required for each socket. |
| 198 | */ |
| 199 | mh_filter_t *filter; |
| 200 | |
| 201 | filter = rcu_dereference(mh_filter); |
| 202 | filtered = filter ? (*filter)(sk, skb) : 0; |
| 203 | break; |
| 204 | } |
| 205 | #endif |
| 206 | default: |
| 207 | filtered = 0; |
| 208 | break; |
| 209 | } |
| 210 | |
| 211 | if (filtered < 0) |
| 212 | break; |
| 213 | if (filtered == 0) { |
| 214 | struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); |
| 215 | |
| 216 | /* Not releasing hash table! */ |
| 217 | if (clone) { |
| 218 | nf_reset(clone); |
| 219 | rawv6_rcv(sk, clone); |
| 220 | } |
| 221 | } |
| 222 | sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr, |
| 223 | inet6_iif(skb)); |
| 224 | } |
| 225 | out: |
| 226 | read_unlock(&raw_v6_hashinfo.lock); |
| 227 | return delivered; |
| 228 | } |
| 229 | |
| 230 | bool raw6_local_deliver(struct sk_buff *skb, int nexthdr) |
| 231 | { |
| 232 | struct sock *raw_sk; |
| 233 | |
| 234 | raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]); |
| 235 | if (raw_sk && !ipv6_raw_deliver(skb, nexthdr)) |
| 236 | raw_sk = NULL; |
| 237 | |
| 238 | return raw_sk != NULL; |
| 239 | } |
| 240 | |
| 241 | /* This cleans up af_inet6 a bit. -DaveM */ |
| 242 | static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) |
| 243 | { |
| 244 | struct inet_sock *inet = inet_sk(sk); |
| 245 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 246 | struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr; |
| 247 | __be32 v4addr = 0; |
| 248 | int addr_type; |
| 249 | int err; |
| 250 | |
| 251 | if (addr_len < SIN6_LEN_RFC2133) |
| 252 | return -EINVAL; |
| 253 | |
| 254 | if (addr->sin6_family != AF_INET6) |
| 255 | return -EINVAL; |
| 256 | |
| 257 | addr_type = ipv6_addr_type(&addr->sin6_addr); |
| 258 | |
| 259 | /* Raw sockets are IPv6 only */ |
| 260 | if (addr_type == IPV6_ADDR_MAPPED) |
| 261 | return -EADDRNOTAVAIL; |
| 262 | |
| 263 | lock_sock(sk); |
| 264 | |
| 265 | err = -EINVAL; |
| 266 | if (sk->sk_state != TCP_CLOSE) |
| 267 | goto out; |
| 268 | |
| 269 | rcu_read_lock(); |
| 270 | /* Check if the address belongs to the host. */ |
| 271 | if (addr_type != IPV6_ADDR_ANY) { |
| 272 | struct net_device *dev = NULL; |
| 273 | |
| 274 | if (__ipv6_addr_needs_scope_id(addr_type)) { |
| 275 | if (addr_len >= sizeof(struct sockaddr_in6) && |
| 276 | addr->sin6_scope_id) { |
| 277 | /* Override any existing binding, if another |
| 278 | * one is supplied by user. |
| 279 | */ |
| 280 | sk->sk_bound_dev_if = addr->sin6_scope_id; |
| 281 | } |
| 282 | |
| 283 | /* Binding to link-local address requires an interface */ |
| 284 | if (!sk->sk_bound_dev_if) |
| 285 | goto out_unlock; |
| 286 | |
| 287 | err = -ENODEV; |
| 288 | dev = dev_get_by_index_rcu(sock_net(sk), |
| 289 | sk->sk_bound_dev_if); |
| 290 | if (!dev) |
| 291 | goto out_unlock; |
| 292 | } |
| 293 | |
| 294 | /* ipv4 addr of the socket is invalid. Only the |
| 295 | * unspecified and mapped address have a v4 equivalent. |
| 296 | */ |
| 297 | v4addr = LOOPBACK4_IPV6; |
| 298 | if (!(addr_type & IPV6_ADDR_MULTICAST) && |
| 299 | !sock_net(sk)->ipv6.sysctl.ip_nonlocal_bind) { |
| 300 | err = -EADDRNOTAVAIL; |
| 301 | if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr, |
| 302 | dev, 0)) { |
| 303 | goto out_unlock; |
| 304 | } |
| 305 | } |
| 306 | } |
| 307 | |
| 308 | inet->inet_rcv_saddr = inet->inet_saddr = v4addr; |
| 309 | sk->sk_v6_rcv_saddr = addr->sin6_addr; |
| 310 | if (!(addr_type & IPV6_ADDR_MULTICAST)) |
| 311 | np->saddr = addr->sin6_addr; |
| 312 | err = 0; |
| 313 | out_unlock: |
| 314 | rcu_read_unlock(); |
| 315 | out: |
| 316 | release_sock(sk); |
| 317 | return err; |
| 318 | } |
| 319 | |
| 320 | static void rawv6_err(struct sock *sk, struct sk_buff *skb, |
| 321 | struct inet6_skb_parm *opt, |
| 322 | u8 type, u8 code, int offset, __be32 info) |
| 323 | { |
| 324 | struct inet_sock *inet = inet_sk(sk); |
| 325 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 326 | int err; |
| 327 | int harderr; |
| 328 | |
| 329 | /* Report error on raw socket, if: |
| 330 | 1. User requested recverr. |
| 331 | 2. Socket is connected (otherwise the error indication |
| 332 | is useless without recverr and error is hard. |
| 333 | */ |
| 334 | if (!np->recverr && sk->sk_state != TCP_ESTABLISHED) |
| 335 | return; |
| 336 | |
| 337 | harderr = icmpv6_err_convert(type, code, &err); |
| 338 | if (type == ICMPV6_PKT_TOOBIG) { |
| 339 | ip6_sk_update_pmtu(skb, sk, info); |
| 340 | harderr = (np->pmtudisc == IPV6_PMTUDISC_DO); |
| 341 | } |
| 342 | if (type == NDISC_REDIRECT) { |
| 343 | ip6_sk_redirect(skb, sk); |
| 344 | return; |
| 345 | } |
| 346 | if (np->recverr) { |
| 347 | u8 *payload = skb->data; |
| 348 | if (!inet->hdrincl) |
| 349 | payload += offset; |
| 350 | ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload); |
| 351 | } |
| 352 | |
| 353 | if (np->recverr || harderr) { |
| 354 | sk->sk_err = err; |
| 355 | sk->sk_error_report(sk); |
| 356 | } |
| 357 | } |
| 358 | |
| 359 | void raw6_icmp_error(struct sk_buff *skb, int nexthdr, |
| 360 | u8 type, u8 code, int inner_offset, __be32 info) |
| 361 | { |
| 362 | struct sock *sk; |
| 363 | int hash; |
| 364 | const struct in6_addr *saddr, *daddr; |
| 365 | struct net *net; |
| 366 | |
| 367 | hash = nexthdr & (RAW_HTABLE_SIZE - 1); |
| 368 | |
| 369 | read_lock(&raw_v6_hashinfo.lock); |
| 370 | sk = sk_head(&raw_v6_hashinfo.ht[hash]); |
| 371 | if (sk) { |
| 372 | /* Note: ipv6_hdr(skb) != skb->data */ |
| 373 | const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data; |
| 374 | saddr = &ip6h->saddr; |
| 375 | daddr = &ip6h->daddr; |
| 376 | net = dev_net(skb->dev); |
| 377 | |
| 378 | while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr, |
| 379 | inet6_iif(skb)))) { |
| 380 | rawv6_err(sk, skb, NULL, type, code, |
| 381 | inner_offset, info); |
| 382 | sk = sk_next(sk); |
| 383 | } |
| 384 | } |
| 385 | read_unlock(&raw_v6_hashinfo.lock); |
| 386 | } |
| 387 | |
| 388 | static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb) |
| 389 | { |
| 390 | if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) && |
| 391 | skb_checksum_complete(skb)) { |
| 392 | atomic_inc(&sk->sk_drops); |
| 393 | kfree_skb(skb); |
| 394 | return NET_RX_DROP; |
| 395 | } |
| 396 | |
| 397 | /* Charge it to the socket. */ |
| 398 | skb_dst_drop(skb); |
| 399 | if (sock_queue_rcv_skb(sk, skb) < 0) { |
| 400 | kfree_skb(skb); |
| 401 | return NET_RX_DROP; |
| 402 | } |
| 403 | |
| 404 | return 0; |
| 405 | } |
| 406 | |
| 407 | /* |
| 408 | * This is next to useless... |
| 409 | * if we demultiplex in network layer we don't need the extra call |
| 410 | * just to queue the skb... |
| 411 | * maybe we could have the network decide upon a hint if it |
| 412 | * should call raw_rcv for demultiplexing |
| 413 | */ |
| 414 | int rawv6_rcv(struct sock *sk, struct sk_buff *skb) |
| 415 | { |
| 416 | struct inet_sock *inet = inet_sk(sk); |
| 417 | struct raw6_sock *rp = raw6_sk(sk); |
| 418 | |
| 419 | if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) { |
| 420 | atomic_inc(&sk->sk_drops); |
| 421 | kfree_skb(skb); |
| 422 | return NET_RX_DROP; |
| 423 | } |
| 424 | |
| 425 | if (!rp->checksum) |
| 426 | skb->ip_summed = CHECKSUM_UNNECESSARY; |
| 427 | |
| 428 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
| 429 | skb_postpull_rcsum(skb, skb_network_header(skb), |
| 430 | skb_network_header_len(skb)); |
| 431 | if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr, |
| 432 | &ipv6_hdr(skb)->daddr, |
| 433 | skb->len, inet->inet_num, skb->csum)) |
| 434 | skb->ip_summed = CHECKSUM_UNNECESSARY; |
| 435 | } |
| 436 | if (!skb_csum_unnecessary(skb)) |
| 437 | skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr, |
| 438 | &ipv6_hdr(skb)->daddr, |
| 439 | skb->len, |
| 440 | inet->inet_num, 0)); |
| 441 | |
| 442 | if (inet->hdrincl) { |
| 443 | if (skb_checksum_complete(skb)) { |
| 444 | atomic_inc(&sk->sk_drops); |
| 445 | kfree_skb(skb); |
| 446 | return NET_RX_DROP; |
| 447 | } |
| 448 | } |
| 449 | |
| 450 | rawv6_rcv_skb(sk, skb); |
| 451 | return 0; |
| 452 | } |
| 453 | |
| 454 | |
| 455 | /* |
| 456 | * This should be easy, if there is something there |
| 457 | * we return it, otherwise we block. |
| 458 | */ |
| 459 | |
| 460 | static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, |
| 461 | int noblock, int flags, int *addr_len) |
| 462 | { |
| 463 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 464 | DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
| 465 | struct sk_buff *skb; |
| 466 | size_t copied; |
| 467 | int err; |
| 468 | |
| 469 | if (flags & MSG_OOB) |
| 470 | return -EOPNOTSUPP; |
| 471 | |
| 472 | if (flags & MSG_ERRQUEUE) |
| 473 | return ipv6_recv_error(sk, msg, len, addr_len); |
| 474 | |
| 475 | if (np->rxpmtu && np->rxopt.bits.rxpmtu) |
| 476 | return ipv6_recv_rxpmtu(sk, msg, len, addr_len); |
| 477 | |
| 478 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
| 479 | if (!skb) |
| 480 | goto out; |
| 481 | |
| 482 | copied = skb->len; |
| 483 | if (copied > len) { |
| 484 | copied = len; |
| 485 | msg->msg_flags |= MSG_TRUNC; |
| 486 | } |
| 487 | |
| 488 | if (skb_csum_unnecessary(skb)) { |
| 489 | err = skb_copy_datagram_msg(skb, 0, msg, copied); |
| 490 | } else if (msg->msg_flags&MSG_TRUNC) { |
| 491 | if (__skb_checksum_complete(skb)) |
| 492 | goto csum_copy_err; |
| 493 | err = skb_copy_datagram_msg(skb, 0, msg, copied); |
| 494 | } else { |
| 495 | err = skb_copy_and_csum_datagram_msg(skb, 0, msg); |
| 496 | if (err == -EINVAL) |
| 497 | goto csum_copy_err; |
| 498 | } |
| 499 | if (err) |
| 500 | goto out_free; |
| 501 | |
| 502 | /* Copy the address. */ |
| 503 | if (sin6) { |
| 504 | sin6->sin6_family = AF_INET6; |
| 505 | sin6->sin6_port = 0; |
| 506 | sin6->sin6_addr = ipv6_hdr(skb)->saddr; |
| 507 | sin6->sin6_flowinfo = 0; |
| 508 | sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr, |
| 509 | inet6_iif(skb)); |
| 510 | *addr_len = sizeof(*sin6); |
| 511 | } |
| 512 | |
| 513 | sock_recv_ts_and_drops(msg, sk, skb); |
| 514 | |
| 515 | if (np->rxopt.all) |
| 516 | ip6_datagram_recv_ctl(sk, msg, skb); |
| 517 | |
| 518 | err = copied; |
| 519 | if (flags & MSG_TRUNC) |
| 520 | err = skb->len; |
| 521 | |
| 522 | out_free: |
| 523 | skb_free_datagram(sk, skb); |
| 524 | out: |
| 525 | return err; |
| 526 | |
| 527 | csum_copy_err: |
| 528 | skb_kill_datagram(sk, skb, flags); |
| 529 | |
| 530 | /* Error for blocking case is chosen to masquerade |
| 531 | as some normal condition. |
| 532 | */ |
| 533 | err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH; |
| 534 | goto out; |
| 535 | } |
| 536 | |
| 537 | static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, |
| 538 | struct raw6_sock *rp) |
| 539 | { |
| 540 | struct sk_buff *skb; |
| 541 | int err = 0; |
| 542 | int offset; |
| 543 | int len; |
| 544 | int total_len; |
| 545 | __wsum tmp_csum; |
| 546 | __sum16 csum; |
| 547 | |
| 548 | if (!rp->checksum) |
| 549 | goto send; |
| 550 | |
| 551 | skb = skb_peek(&sk->sk_write_queue); |
| 552 | if (!skb) |
| 553 | goto out; |
| 554 | |
| 555 | offset = rp->offset; |
| 556 | total_len = inet_sk(sk)->cork.base.length; |
| 557 | if (offset >= total_len - 1) { |
| 558 | err = -EINVAL; |
| 559 | ip6_flush_pending_frames(sk); |
| 560 | goto out; |
| 561 | } |
| 562 | |
| 563 | /* should be check HW csum miyazawa */ |
| 564 | if (skb_queue_len(&sk->sk_write_queue) == 1) { |
| 565 | /* |
| 566 | * Only one fragment on the socket. |
| 567 | */ |
| 568 | tmp_csum = skb->csum; |
| 569 | } else { |
| 570 | struct sk_buff *csum_skb = NULL; |
| 571 | tmp_csum = 0; |
| 572 | |
| 573 | skb_queue_walk(&sk->sk_write_queue, skb) { |
| 574 | tmp_csum = csum_add(tmp_csum, skb->csum); |
| 575 | |
| 576 | if (csum_skb) |
| 577 | continue; |
| 578 | |
| 579 | len = skb->len - skb_transport_offset(skb); |
| 580 | if (offset >= len) { |
| 581 | offset -= len; |
| 582 | continue; |
| 583 | } |
| 584 | |
| 585 | csum_skb = skb; |
| 586 | } |
| 587 | |
| 588 | skb = csum_skb; |
| 589 | } |
| 590 | |
| 591 | offset += skb_transport_offset(skb); |
| 592 | err = skb_copy_bits(skb, offset, &csum, 2); |
| 593 | if (err < 0) { |
| 594 | ip6_flush_pending_frames(sk); |
| 595 | goto out; |
| 596 | } |
| 597 | |
| 598 | /* in case cksum was not initialized */ |
| 599 | if (unlikely(csum)) |
| 600 | tmp_csum = csum_sub(tmp_csum, csum_unfold(csum)); |
| 601 | |
| 602 | csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr, |
| 603 | total_len, fl6->flowi6_proto, tmp_csum); |
| 604 | |
| 605 | if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP) |
| 606 | csum = CSUM_MANGLED_0; |
| 607 | |
| 608 | BUG_ON(skb_store_bits(skb, offset, &csum, 2)); |
| 609 | |
| 610 | send: |
| 611 | err = ip6_push_pending_frames(sk); |
| 612 | out: |
| 613 | return err; |
| 614 | } |
| 615 | |
| 616 | static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length, |
| 617 | struct flowi6 *fl6, struct dst_entry **dstp, |
| 618 | unsigned int flags) |
| 619 | { |
| 620 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 621 | struct net *net = sock_net(sk); |
| 622 | struct ipv6hdr *iph; |
| 623 | struct sk_buff *skb; |
| 624 | int err; |
| 625 | struct rt6_info *rt = (struct rt6_info *)*dstp; |
| 626 | int hlen = LL_RESERVED_SPACE(rt->dst.dev); |
| 627 | int tlen = rt->dst.dev->needed_tailroom; |
| 628 | |
| 629 | if (length > rt->dst.dev->mtu) { |
| 630 | ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu); |
| 631 | return -EMSGSIZE; |
| 632 | } |
| 633 | if (length < sizeof(struct ipv6hdr)) |
| 634 | return -EINVAL; |
| 635 | if (flags&MSG_PROBE) |
| 636 | goto out; |
| 637 | |
| 638 | skb = sock_alloc_send_skb(sk, |
| 639 | length + hlen + tlen + 15, |
| 640 | flags & MSG_DONTWAIT, &err); |
| 641 | if (!skb) |
| 642 | goto error; |
| 643 | skb_reserve(skb, hlen); |
| 644 | |
| 645 | skb->protocol = htons(ETH_P_IPV6); |
| 646 | skb->priority = sk->sk_priority; |
| 647 | skb->mark = sk->sk_mark; |
| 648 | skb_dst_set(skb, &rt->dst); |
| 649 | *dstp = NULL; |
| 650 | |
| 651 | skb_put(skb, length); |
| 652 | skb_reset_network_header(skb); |
| 653 | iph = ipv6_hdr(skb); |
| 654 | |
| 655 | skb->ip_summed = CHECKSUM_NONE; |
| 656 | |
| 657 | skb->transport_header = skb->network_header; |
| 658 | err = memcpy_from_msg(iph, msg, length); |
| 659 | if (err) |
| 660 | goto error_fault; |
| 661 | |
| 662 | IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len); |
| 663 | err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb, |
| 664 | NULL, rt->dst.dev, dst_output); |
| 665 | if (err > 0) |
| 666 | err = net_xmit_errno(err); |
| 667 | if (err) |
| 668 | goto error; |
| 669 | out: |
| 670 | return 0; |
| 671 | |
| 672 | error_fault: |
| 673 | err = -EFAULT; |
| 674 | kfree_skb(skb); |
| 675 | error: |
| 676 | IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS); |
| 677 | if (err == -ENOBUFS && !np->recverr) |
| 678 | err = 0; |
| 679 | return err; |
| 680 | } |
| 681 | |
| 682 | struct raw6_frag_vec { |
| 683 | struct msghdr *msg; |
| 684 | int hlen; |
| 685 | char c[4]; |
| 686 | }; |
| 687 | |
| 688 | static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6) |
| 689 | { |
| 690 | int err = 0; |
| 691 | switch (fl6->flowi6_proto) { |
| 692 | case IPPROTO_ICMPV6: |
| 693 | rfv->hlen = 2; |
| 694 | err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen); |
| 695 | if (!err) { |
| 696 | fl6->fl6_icmp_type = rfv->c[0]; |
| 697 | fl6->fl6_icmp_code = rfv->c[1]; |
| 698 | } |
| 699 | break; |
| 700 | case IPPROTO_MH: |
| 701 | rfv->hlen = 4; |
| 702 | err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen); |
| 703 | if (!err) |
| 704 | fl6->fl6_mh_type = rfv->c[2]; |
| 705 | } |
| 706 | return err; |
| 707 | } |
| 708 | |
| 709 | static int raw6_getfrag(void *from, char *to, int offset, int len, int odd, |
| 710 | struct sk_buff *skb) |
| 711 | { |
| 712 | struct raw6_frag_vec *rfv = from; |
| 713 | |
| 714 | if (offset < rfv->hlen) { |
| 715 | int copy = min(rfv->hlen - offset, len); |
| 716 | |
| 717 | if (skb->ip_summed == CHECKSUM_PARTIAL) |
| 718 | memcpy(to, rfv->c + offset, copy); |
| 719 | else |
| 720 | skb->csum = csum_block_add( |
| 721 | skb->csum, |
| 722 | csum_partial_copy_nocheck(rfv->c + offset, |
| 723 | to, copy, 0), |
| 724 | odd); |
| 725 | |
| 726 | odd = 0; |
| 727 | offset += copy; |
| 728 | to += copy; |
| 729 | len -= copy; |
| 730 | |
| 731 | if (!len) |
| 732 | return 0; |
| 733 | } |
| 734 | |
| 735 | offset -= rfv->hlen; |
| 736 | |
| 737 | return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb); |
| 738 | } |
| 739 | |
| 740 | static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) |
| 741 | { |
| 742 | struct ipv6_txoptions *opt_to_free = NULL; |
| 743 | struct ipv6_txoptions opt_space; |
| 744 | DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); |
| 745 | struct in6_addr *daddr, *final_p, final; |
| 746 | struct inet_sock *inet = inet_sk(sk); |
| 747 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 748 | struct raw6_sock *rp = raw6_sk(sk); |
| 749 | struct ipv6_txoptions *opt = NULL; |
| 750 | struct ip6_flowlabel *flowlabel = NULL; |
| 751 | struct dst_entry *dst = NULL; |
| 752 | struct raw6_frag_vec rfv; |
| 753 | struct flowi6 fl6; |
| 754 | int addr_len = msg->msg_namelen; |
| 755 | int hlimit = -1; |
| 756 | int tclass = -1; |
| 757 | int dontfrag = -1; |
| 758 | u16 proto; |
| 759 | int err; |
| 760 | |
| 761 | /* Rough check on arithmetic overflow, |
| 762 | better check is made in ip6_append_data(). |
| 763 | */ |
| 764 | if (len > INT_MAX) |
| 765 | return -EMSGSIZE; |
| 766 | |
| 767 | /* Mirror BSD error message compatibility */ |
| 768 | if (msg->msg_flags & MSG_OOB) |
| 769 | return -EOPNOTSUPP; |
| 770 | |
| 771 | /* |
| 772 | * Get and verify the address. |
| 773 | */ |
| 774 | memset(&fl6, 0, sizeof(fl6)); |
| 775 | |
| 776 | fl6.flowi6_mark = sk->sk_mark; |
| 777 | |
| 778 | if (sin6) { |
| 779 | if (addr_len < SIN6_LEN_RFC2133) |
| 780 | return -EINVAL; |
| 781 | |
| 782 | if (sin6->sin6_family && sin6->sin6_family != AF_INET6) |
| 783 | return -EAFNOSUPPORT; |
| 784 | |
| 785 | /* port is the proto value [0..255] carried in nexthdr */ |
| 786 | proto = ntohs(sin6->sin6_port); |
| 787 | |
| 788 | if (!proto) |
| 789 | proto = inet->inet_num; |
| 790 | else if (proto != inet->inet_num) |
| 791 | return -EINVAL; |
| 792 | |
| 793 | if (proto > 255) |
| 794 | return -EINVAL; |
| 795 | |
| 796 | daddr = &sin6->sin6_addr; |
| 797 | if (np->sndflow) { |
| 798 | fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK; |
| 799 | if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { |
| 800 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); |
| 801 | if (!flowlabel) |
| 802 | return -EINVAL; |
| 803 | } |
| 804 | } |
| 805 | |
| 806 | /* |
| 807 | * Otherwise it will be difficult to maintain |
| 808 | * sk->sk_dst_cache. |
| 809 | */ |
| 810 | if (sk->sk_state == TCP_ESTABLISHED && |
| 811 | ipv6_addr_equal(daddr, &sk->sk_v6_daddr)) |
| 812 | daddr = &sk->sk_v6_daddr; |
| 813 | |
| 814 | if (addr_len >= sizeof(struct sockaddr_in6) && |
| 815 | sin6->sin6_scope_id && |
| 816 | __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr))) |
| 817 | fl6.flowi6_oif = sin6->sin6_scope_id; |
| 818 | } else { |
| 819 | if (sk->sk_state != TCP_ESTABLISHED) |
| 820 | return -EDESTADDRREQ; |
| 821 | |
| 822 | proto = inet->inet_num; |
| 823 | daddr = &sk->sk_v6_daddr; |
| 824 | fl6.flowlabel = np->flow_label; |
| 825 | } |
| 826 | |
| 827 | if (fl6.flowi6_oif == 0) |
| 828 | fl6.flowi6_oif = sk->sk_bound_dev_if; |
| 829 | |
| 830 | if (msg->msg_controllen) { |
| 831 | opt = &opt_space; |
| 832 | memset(opt, 0, sizeof(struct ipv6_txoptions)); |
| 833 | opt->tot_len = sizeof(struct ipv6_txoptions); |
| 834 | |
| 835 | err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt, |
| 836 | &hlimit, &tclass, &dontfrag); |
| 837 | if (err < 0) { |
| 838 | fl6_sock_release(flowlabel); |
| 839 | return err; |
| 840 | } |
| 841 | if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) { |
| 842 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); |
| 843 | if (!flowlabel) |
| 844 | return -EINVAL; |
| 845 | } |
| 846 | if (!(opt->opt_nflen|opt->opt_flen)) |
| 847 | opt = NULL; |
| 848 | } |
| 849 | if (!opt) { |
| 850 | opt = txopt_get(np); |
| 851 | opt_to_free = opt; |
| 852 | } |
| 853 | if (flowlabel) |
| 854 | opt = fl6_merge_options(&opt_space, flowlabel, opt); |
| 855 | opt = ipv6_fixup_options(&opt_space, opt); |
| 856 | |
| 857 | fl6.flowi6_proto = proto; |
| 858 | rfv.msg = msg; |
| 859 | rfv.hlen = 0; |
| 860 | err = rawv6_probe_proto_opt(&rfv, &fl6); |
| 861 | if (err) |
| 862 | goto out; |
| 863 | |
| 864 | if (!ipv6_addr_any(daddr)) |
| 865 | fl6.daddr = *daddr; |
| 866 | else |
| 867 | fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */ |
| 868 | if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr)) |
| 869 | fl6.saddr = np->saddr; |
| 870 | |
| 871 | final_p = fl6_update_dst(&fl6, opt, &final); |
| 872 | |
| 873 | if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) |
| 874 | fl6.flowi6_oif = np->mcast_oif; |
| 875 | else if (!fl6.flowi6_oif) |
| 876 | fl6.flowi6_oif = np->ucast_oif; |
| 877 | security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); |
| 878 | |
| 879 | if (inet->hdrincl) |
| 880 | fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH; |
| 881 | |
| 882 | dst = ip6_dst_lookup_flow(sk, &fl6, final_p); |
| 883 | if (IS_ERR(dst)) { |
| 884 | err = PTR_ERR(dst); |
| 885 | goto out; |
| 886 | } |
| 887 | if (hlimit < 0) |
| 888 | hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst); |
| 889 | |
| 890 | if (tclass < 0) |
| 891 | tclass = np->tclass; |
| 892 | |
| 893 | if (dontfrag < 0) |
| 894 | dontfrag = np->dontfrag; |
| 895 | |
| 896 | if (msg->msg_flags&MSG_CONFIRM) |
| 897 | goto do_confirm; |
| 898 | |
| 899 | back_from_confirm: |
| 900 | if (inet->hdrincl) |
| 901 | err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst, msg->msg_flags); |
| 902 | else { |
| 903 | lock_sock(sk); |
| 904 | err = ip6_append_data(sk, raw6_getfrag, &rfv, |
| 905 | len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info *)dst, |
| 906 | msg->msg_flags, dontfrag); |
| 907 | |
| 908 | if (err) |
| 909 | ip6_flush_pending_frames(sk); |
| 910 | else if (!(msg->msg_flags & MSG_MORE)) |
| 911 | err = rawv6_push_pending_frames(sk, &fl6, rp); |
| 912 | release_sock(sk); |
| 913 | } |
| 914 | done: |
| 915 | dst_release(dst); |
| 916 | out: |
| 917 | fl6_sock_release(flowlabel); |
| 918 | txopt_put(opt_to_free); |
| 919 | return err < 0 ? err : len; |
| 920 | do_confirm: |
| 921 | dst_confirm(dst); |
| 922 | if (!(msg->msg_flags & MSG_PROBE) || len) |
| 923 | goto back_from_confirm; |
| 924 | err = 0; |
| 925 | goto done; |
| 926 | } |
| 927 | |
| 928 | static int rawv6_seticmpfilter(struct sock *sk, int level, int optname, |
| 929 | char __user *optval, int optlen) |
| 930 | { |
| 931 | switch (optname) { |
| 932 | case ICMPV6_FILTER: |
| 933 | if (optlen > sizeof(struct icmp6_filter)) |
| 934 | optlen = sizeof(struct icmp6_filter); |
| 935 | if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen)) |
| 936 | return -EFAULT; |
| 937 | return 0; |
| 938 | default: |
| 939 | return -ENOPROTOOPT; |
| 940 | } |
| 941 | |
| 942 | return 0; |
| 943 | } |
| 944 | |
| 945 | static int rawv6_geticmpfilter(struct sock *sk, int level, int optname, |
| 946 | char __user *optval, int __user *optlen) |
| 947 | { |
| 948 | int len; |
| 949 | |
| 950 | switch (optname) { |
| 951 | case ICMPV6_FILTER: |
| 952 | if (get_user(len, optlen)) |
| 953 | return -EFAULT; |
| 954 | if (len < 0) |
| 955 | return -EINVAL; |
| 956 | if (len > sizeof(struct icmp6_filter)) |
| 957 | len = sizeof(struct icmp6_filter); |
| 958 | if (put_user(len, optlen)) |
| 959 | return -EFAULT; |
| 960 | if (copy_to_user(optval, &raw6_sk(sk)->filter, len)) |
| 961 | return -EFAULT; |
| 962 | return 0; |
| 963 | default: |
| 964 | return -ENOPROTOOPT; |
| 965 | } |
| 966 | |
| 967 | return 0; |
| 968 | } |
| 969 | |
| 970 | |
| 971 | static int do_rawv6_setsockopt(struct sock *sk, int level, int optname, |
| 972 | char __user *optval, unsigned int optlen) |
| 973 | { |
| 974 | struct raw6_sock *rp = raw6_sk(sk); |
| 975 | int val; |
| 976 | |
| 977 | if (get_user(val, (int __user *)optval)) |
| 978 | return -EFAULT; |
| 979 | |
| 980 | switch (optname) { |
| 981 | case IPV6_CHECKSUM: |
| 982 | if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 && |
| 983 | level == IPPROTO_IPV6) { |
| 984 | /* |
| 985 | * RFC3542 tells that IPV6_CHECKSUM socket |
| 986 | * option in the IPPROTO_IPV6 level is not |
| 987 | * allowed on ICMPv6 sockets. |
| 988 | * If you want to set it, use IPPROTO_RAW |
| 989 | * level IPV6_CHECKSUM socket option |
| 990 | * (Linux extension). |
| 991 | */ |
| 992 | return -EINVAL; |
| 993 | } |
| 994 | |
| 995 | /* You may get strange result with a positive odd offset; |
| 996 | RFC2292bis agrees with me. */ |
| 997 | if (val > 0 && (val&1)) |
| 998 | return -EINVAL; |
| 999 | if (val < 0) { |
| 1000 | rp->checksum = 0; |
| 1001 | } else { |
| 1002 | rp->checksum = 1; |
| 1003 | rp->offset = val; |
| 1004 | } |
| 1005 | |
| 1006 | return 0; |
| 1007 | |
| 1008 | default: |
| 1009 | return -ENOPROTOOPT; |
| 1010 | } |
| 1011 | } |
| 1012 | |
| 1013 | static int rawv6_setsockopt(struct sock *sk, int level, int optname, |
| 1014 | char __user *optval, unsigned int optlen) |
| 1015 | { |
| 1016 | switch (level) { |
| 1017 | case SOL_RAW: |
| 1018 | break; |
| 1019 | |
| 1020 | case SOL_ICMPV6: |
| 1021 | if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) |
| 1022 | return -EOPNOTSUPP; |
| 1023 | return rawv6_seticmpfilter(sk, level, optname, optval, optlen); |
| 1024 | case SOL_IPV6: |
| 1025 | if (optname == IPV6_CHECKSUM) |
| 1026 | break; |
| 1027 | default: |
| 1028 | return ipv6_setsockopt(sk, level, optname, optval, optlen); |
| 1029 | } |
| 1030 | |
| 1031 | return do_rawv6_setsockopt(sk, level, optname, optval, optlen); |
| 1032 | } |
| 1033 | |
| 1034 | #ifdef CONFIG_COMPAT |
| 1035 | static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname, |
| 1036 | char __user *optval, unsigned int optlen) |
| 1037 | { |
| 1038 | switch (level) { |
| 1039 | case SOL_RAW: |
| 1040 | break; |
| 1041 | case SOL_ICMPV6: |
| 1042 | if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) |
| 1043 | return -EOPNOTSUPP; |
| 1044 | return rawv6_seticmpfilter(sk, level, optname, optval, optlen); |
| 1045 | case SOL_IPV6: |
| 1046 | if (optname == IPV6_CHECKSUM) |
| 1047 | break; |
| 1048 | default: |
| 1049 | return compat_ipv6_setsockopt(sk, level, optname, |
| 1050 | optval, optlen); |
| 1051 | } |
| 1052 | return do_rawv6_setsockopt(sk, level, optname, optval, optlen); |
| 1053 | } |
| 1054 | #endif |
| 1055 | |
| 1056 | static int do_rawv6_getsockopt(struct sock *sk, int level, int optname, |
| 1057 | char __user *optval, int __user *optlen) |
| 1058 | { |
| 1059 | struct raw6_sock *rp = raw6_sk(sk); |
| 1060 | int val, len; |
| 1061 | |
| 1062 | if (get_user(len, optlen)) |
| 1063 | return -EFAULT; |
| 1064 | |
| 1065 | switch (optname) { |
| 1066 | case IPV6_CHECKSUM: |
| 1067 | /* |
| 1068 | * We allow getsockopt() for IPPROTO_IPV6-level |
| 1069 | * IPV6_CHECKSUM socket option on ICMPv6 sockets |
| 1070 | * since RFC3542 is silent about it. |
| 1071 | */ |
| 1072 | if (rp->checksum == 0) |
| 1073 | val = -1; |
| 1074 | else |
| 1075 | val = rp->offset; |
| 1076 | break; |
| 1077 | |
| 1078 | default: |
| 1079 | return -ENOPROTOOPT; |
| 1080 | } |
| 1081 | |
| 1082 | len = min_t(unsigned int, sizeof(int), len); |
| 1083 | |
| 1084 | if (put_user(len, optlen)) |
| 1085 | return -EFAULT; |
| 1086 | if (copy_to_user(optval, &val, len)) |
| 1087 | return -EFAULT; |
| 1088 | return 0; |
| 1089 | } |
| 1090 | |
| 1091 | static int rawv6_getsockopt(struct sock *sk, int level, int optname, |
| 1092 | char __user *optval, int __user *optlen) |
| 1093 | { |
| 1094 | switch (level) { |
| 1095 | case SOL_RAW: |
| 1096 | break; |
| 1097 | |
| 1098 | case SOL_ICMPV6: |
| 1099 | if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) |
| 1100 | return -EOPNOTSUPP; |
| 1101 | return rawv6_geticmpfilter(sk, level, optname, optval, optlen); |
| 1102 | case SOL_IPV6: |
| 1103 | if (optname == IPV6_CHECKSUM) |
| 1104 | break; |
| 1105 | default: |
| 1106 | return ipv6_getsockopt(sk, level, optname, optval, optlen); |
| 1107 | } |
| 1108 | |
| 1109 | return do_rawv6_getsockopt(sk, level, optname, optval, optlen); |
| 1110 | } |
| 1111 | |
| 1112 | #ifdef CONFIG_COMPAT |
| 1113 | static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname, |
| 1114 | char __user *optval, int __user *optlen) |
| 1115 | { |
| 1116 | switch (level) { |
| 1117 | case SOL_RAW: |
| 1118 | break; |
| 1119 | case SOL_ICMPV6: |
| 1120 | if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6) |
| 1121 | return -EOPNOTSUPP; |
| 1122 | return rawv6_geticmpfilter(sk, level, optname, optval, optlen); |
| 1123 | case SOL_IPV6: |
| 1124 | if (optname == IPV6_CHECKSUM) |
| 1125 | break; |
| 1126 | default: |
| 1127 | return compat_ipv6_getsockopt(sk, level, optname, |
| 1128 | optval, optlen); |
| 1129 | } |
| 1130 | return do_rawv6_getsockopt(sk, level, optname, optval, optlen); |
| 1131 | } |
| 1132 | #endif |
| 1133 | |
| 1134 | static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg) |
| 1135 | { |
| 1136 | switch (cmd) { |
| 1137 | case SIOCOUTQ: { |
| 1138 | int amount = sk_wmem_alloc_get(sk); |
| 1139 | |
| 1140 | return put_user(amount, (int __user *)arg); |
| 1141 | } |
| 1142 | case SIOCINQ: { |
| 1143 | struct sk_buff *skb; |
| 1144 | int amount = 0; |
| 1145 | |
| 1146 | spin_lock_bh(&sk->sk_receive_queue.lock); |
| 1147 | skb = skb_peek(&sk->sk_receive_queue); |
| 1148 | if (skb) |
| 1149 | amount = skb->len; |
| 1150 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
| 1151 | return put_user(amount, (int __user *)arg); |
| 1152 | } |
| 1153 | |
| 1154 | default: |
| 1155 | #ifdef CONFIG_IPV6_MROUTE |
| 1156 | return ip6mr_ioctl(sk, cmd, (void __user *)arg); |
| 1157 | #else |
| 1158 | return -ENOIOCTLCMD; |
| 1159 | #endif |
| 1160 | } |
| 1161 | } |
| 1162 | |
| 1163 | #ifdef CONFIG_COMPAT |
| 1164 | static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg) |
| 1165 | { |
| 1166 | switch (cmd) { |
| 1167 | case SIOCOUTQ: |
| 1168 | case SIOCINQ: |
| 1169 | return -ENOIOCTLCMD; |
| 1170 | default: |
| 1171 | #ifdef CONFIG_IPV6_MROUTE |
| 1172 | return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg)); |
| 1173 | #else |
| 1174 | return -ENOIOCTLCMD; |
| 1175 | #endif |
| 1176 | } |
| 1177 | } |
| 1178 | #endif |
| 1179 | |
| 1180 | static void rawv6_close(struct sock *sk, long timeout) |
| 1181 | { |
| 1182 | if (inet_sk(sk)->inet_num == IPPROTO_RAW) |
| 1183 | ip6_ra_control(sk, -1); |
| 1184 | ip6mr_sk_done(sk); |
| 1185 | sk_common_release(sk); |
| 1186 | } |
| 1187 | |
| 1188 | static void raw6_destroy(struct sock *sk) |
| 1189 | { |
| 1190 | lock_sock(sk); |
| 1191 | ip6_flush_pending_frames(sk); |
| 1192 | release_sock(sk); |
| 1193 | |
| 1194 | inet6_destroy_sock(sk); |
| 1195 | } |
| 1196 | |
| 1197 | static int rawv6_init_sk(struct sock *sk) |
| 1198 | { |
| 1199 | struct raw6_sock *rp = raw6_sk(sk); |
| 1200 | |
| 1201 | switch (inet_sk(sk)->inet_num) { |
| 1202 | case IPPROTO_ICMPV6: |
| 1203 | rp->checksum = 1; |
| 1204 | rp->offset = 2; |
| 1205 | break; |
| 1206 | case IPPROTO_MH: |
| 1207 | rp->checksum = 1; |
| 1208 | rp->offset = 4; |
| 1209 | break; |
| 1210 | default: |
| 1211 | break; |
| 1212 | } |
| 1213 | return 0; |
| 1214 | } |
| 1215 | |
| 1216 | struct proto rawv6_prot = { |
| 1217 | .name = "RAWv6", |
| 1218 | .owner = THIS_MODULE, |
| 1219 | .close = rawv6_close, |
| 1220 | .destroy = raw6_destroy, |
| 1221 | .connect = ip6_datagram_connect_v6_only, |
| 1222 | .disconnect = udp_disconnect, |
| 1223 | .ioctl = rawv6_ioctl, |
| 1224 | .init = rawv6_init_sk, |
| 1225 | .setsockopt = rawv6_setsockopt, |
| 1226 | .getsockopt = rawv6_getsockopt, |
| 1227 | .sendmsg = rawv6_sendmsg, |
| 1228 | .recvmsg = rawv6_recvmsg, |
| 1229 | .bind = rawv6_bind, |
| 1230 | .backlog_rcv = rawv6_rcv_skb, |
| 1231 | .hash = raw_hash_sk, |
| 1232 | .unhash = raw_unhash_sk, |
| 1233 | .obj_size = sizeof(struct raw6_sock), |
| 1234 | .h.raw_hash = &raw_v6_hashinfo, |
| 1235 | #ifdef CONFIG_COMPAT |
| 1236 | .compat_setsockopt = compat_rawv6_setsockopt, |
| 1237 | .compat_getsockopt = compat_rawv6_getsockopt, |
| 1238 | .compat_ioctl = compat_rawv6_ioctl, |
| 1239 | #endif |
| 1240 | }; |
| 1241 | |
| 1242 | #ifdef CONFIG_PROC_FS |
| 1243 | static int raw6_seq_show(struct seq_file *seq, void *v) |
| 1244 | { |
| 1245 | if (v == SEQ_START_TOKEN) { |
| 1246 | seq_puts(seq, IPV6_SEQ_DGRAM_HEADER); |
| 1247 | } else { |
| 1248 | struct sock *sp = v; |
| 1249 | __u16 srcp = inet_sk(sp)->inet_num; |
| 1250 | ip6_dgram_sock_seq_show(seq, v, srcp, 0, |
| 1251 | raw_seq_private(seq)->bucket); |
| 1252 | } |
| 1253 | return 0; |
| 1254 | } |
| 1255 | |
| 1256 | static const struct seq_operations raw6_seq_ops = { |
| 1257 | .start = raw_seq_start, |
| 1258 | .next = raw_seq_next, |
| 1259 | .stop = raw_seq_stop, |
| 1260 | .show = raw6_seq_show, |
| 1261 | }; |
| 1262 | |
| 1263 | static int raw6_seq_open(struct inode *inode, struct file *file) |
| 1264 | { |
| 1265 | return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops); |
| 1266 | } |
| 1267 | |
| 1268 | static const struct file_operations raw6_seq_fops = { |
| 1269 | .owner = THIS_MODULE, |
| 1270 | .open = raw6_seq_open, |
| 1271 | .read = seq_read, |
| 1272 | .llseek = seq_lseek, |
| 1273 | .release = seq_release_net, |
| 1274 | }; |
| 1275 | |
| 1276 | static int __net_init raw6_init_net(struct net *net) |
| 1277 | { |
| 1278 | if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops)) |
| 1279 | return -ENOMEM; |
| 1280 | |
| 1281 | return 0; |
| 1282 | } |
| 1283 | |
| 1284 | static void __net_exit raw6_exit_net(struct net *net) |
| 1285 | { |
| 1286 | remove_proc_entry("raw6", net->proc_net); |
| 1287 | } |
| 1288 | |
| 1289 | static struct pernet_operations raw6_net_ops = { |
| 1290 | .init = raw6_init_net, |
| 1291 | .exit = raw6_exit_net, |
| 1292 | }; |
| 1293 | |
| 1294 | int __init raw6_proc_init(void) |
| 1295 | { |
| 1296 | return register_pernet_subsys(&raw6_net_ops); |
| 1297 | } |
| 1298 | |
| 1299 | void raw6_proc_exit(void) |
| 1300 | { |
| 1301 | unregister_pernet_subsys(&raw6_net_ops); |
| 1302 | } |
| 1303 | #endif /* CONFIG_PROC_FS */ |
| 1304 | |
| 1305 | /* Same as inet6_dgram_ops, sans udp_poll. */ |
| 1306 | const struct proto_ops inet6_sockraw_ops = { |
| 1307 | .family = PF_INET6, |
| 1308 | .owner = THIS_MODULE, |
| 1309 | .release = inet6_release, |
| 1310 | .bind = inet6_bind, |
| 1311 | .connect = inet_dgram_connect, /* ok */ |
| 1312 | .socketpair = sock_no_socketpair, /* a do nothing */ |
| 1313 | .accept = sock_no_accept, /* a do nothing */ |
| 1314 | .getname = inet6_getname, |
| 1315 | .poll = datagram_poll, /* ok */ |
| 1316 | .ioctl = inet6_ioctl, /* must change */ |
| 1317 | .listen = sock_no_listen, /* ok */ |
| 1318 | .shutdown = inet_shutdown, /* ok */ |
| 1319 | .setsockopt = sock_common_setsockopt, /* ok */ |
| 1320 | .getsockopt = sock_common_getsockopt, /* ok */ |
| 1321 | .sendmsg = inet_sendmsg, /* ok */ |
| 1322 | .recvmsg = sock_common_recvmsg, /* ok */ |
| 1323 | .mmap = sock_no_mmap, |
| 1324 | .sendpage = sock_no_sendpage, |
| 1325 | #ifdef CONFIG_COMPAT |
| 1326 | .compat_setsockopt = compat_sock_common_setsockopt, |
| 1327 | .compat_getsockopt = compat_sock_common_getsockopt, |
| 1328 | #endif |
| 1329 | }; |
| 1330 | |
| 1331 | static struct inet_protosw rawv6_protosw = { |
| 1332 | .type = SOCK_RAW, |
| 1333 | .protocol = IPPROTO_IP, /* wild card */ |
| 1334 | .prot = &rawv6_prot, |
| 1335 | .ops = &inet6_sockraw_ops, |
| 1336 | .flags = INET_PROTOSW_REUSE, |
| 1337 | }; |
| 1338 | |
| 1339 | int __init rawv6_init(void) |
| 1340 | { |
| 1341 | return inet6_register_protosw(&rawv6_protosw); |
| 1342 | } |
| 1343 | |
| 1344 | void rawv6_exit(void) |
| 1345 | { |
| 1346 | inet6_unregister_protosw(&rawv6_protosw); |
| 1347 | } |