Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * INET An implementation of the TCP/IP protocol suite for the LINUX |
| 3 | * operating system. INET is implemented using the BSD Socket |
| 4 | * interface as the means of communication with the user level. |
| 5 | * |
| 6 | * RAW - implementation of IP "raw" sockets. |
| 7 | * |
| 8 | * Version: $Id: raw.c,v 1.64 2002/02/01 22:01:04 davem Exp $ |
| 9 | * |
Jesper Juhl | 02c30a8 | 2005-05-05 16:16:16 -0700 | [diff] [blame] | 10 | * Authors: Ross Biro |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 11 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
| 12 | * |
| 13 | * Fixes: |
| 14 | * Alan Cox : verify_area() fixed up |
| 15 | * Alan Cox : ICMP error handling |
| 16 | * Alan Cox : EMSGSIZE if you send too big a packet |
| 17 | * Alan Cox : Now uses generic datagrams and shared |
| 18 | * skbuff library. No more peek crashes, |
| 19 | * no more backlogs |
| 20 | * Alan Cox : Checks sk->broadcast. |
| 21 | * Alan Cox : Uses skb_free_datagram/skb_copy_datagram |
| 22 | * Alan Cox : Raw passes ip options too |
| 23 | * Alan Cox : Setsocketopt added |
| 24 | * Alan Cox : Fixed error return for broadcasts |
| 25 | * Alan Cox : Removed wake_up calls |
| 26 | * Alan Cox : Use ttl/tos |
| 27 | * Alan Cox : Cleaned up old debugging |
| 28 | * Alan Cox : Use new kernel side addresses |
| 29 | * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets. |
| 30 | * Alan Cox : BSD style RAW socket demultiplexing. |
| 31 | * Alan Cox : Beginnings of mrouted support. |
| 32 | * Alan Cox : Added IP_HDRINCL option. |
| 33 | * Alan Cox : Skip broadcast check if BSDism set. |
| 34 | * David S. Miller : New socket lookup architecture. |
| 35 | * |
| 36 | * This program is free software; you can redistribute it and/or |
| 37 | * modify it under the terms of the GNU General Public License |
| 38 | * as published by the Free Software Foundation; either version |
| 39 | * 2 of the License, or (at your option) any later version. |
| 40 | */ |
| 41 | |
| 42 | #include <linux/config.h> |
| 43 | #include <asm/atomic.h> |
| 44 | #include <asm/byteorder.h> |
| 45 | #include <asm/current.h> |
| 46 | #include <asm/uaccess.h> |
| 47 | #include <asm/ioctls.h> |
| 48 | #include <linux/types.h> |
| 49 | #include <linux/stddef.h> |
| 50 | #include <linux/slab.h> |
| 51 | #include <linux/errno.h> |
| 52 | #include <linux/aio.h> |
| 53 | #include <linux/kernel.h> |
| 54 | #include <linux/spinlock.h> |
| 55 | #include <linux/sockios.h> |
| 56 | #include <linux/socket.h> |
| 57 | #include <linux/in.h> |
| 58 | #include <linux/mroute.h> |
| 59 | #include <linux/netdevice.h> |
| 60 | #include <linux/in_route.h> |
| 61 | #include <linux/route.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 62 | #include <linux/skbuff.h> |
| 63 | #include <net/dst.h> |
| 64 | #include <net/sock.h> |
| 65 | #include <linux/gfp.h> |
| 66 | #include <linux/ip.h> |
| 67 | #include <linux/net.h> |
| 68 | #include <net/ip.h> |
| 69 | #include <net/icmp.h> |
| 70 | #include <net/udp.h> |
| 71 | #include <net/raw.h> |
| 72 | #include <net/snmp.h> |
Arnaldo Carvalho de Melo | c752f07 | 2005-08-09 20:08:28 -0700 | [diff] [blame] | 73 | #include <net/tcp_states.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 74 | #include <net/inet_common.h> |
| 75 | #include <net/checksum.h> |
| 76 | #include <net/xfrm.h> |
| 77 | #include <linux/rtnetlink.h> |
| 78 | #include <linux/proc_fs.h> |
| 79 | #include <linux/seq_file.h> |
| 80 | #include <linux/netfilter.h> |
| 81 | #include <linux/netfilter_ipv4.h> |
| 82 | |
| 83 | struct hlist_head raw_v4_htable[RAWV4_HTABLE_SIZE]; |
| 84 | DEFINE_RWLOCK(raw_v4_lock); |
| 85 | |
| 86 | static void raw_v4_hash(struct sock *sk) |
| 87 | { |
| 88 | struct hlist_head *head = &raw_v4_htable[inet_sk(sk)->num & |
| 89 | (RAWV4_HTABLE_SIZE - 1)]; |
| 90 | |
| 91 | write_lock_bh(&raw_v4_lock); |
| 92 | sk_add_node(sk, head); |
| 93 | sock_prot_inc_use(sk->sk_prot); |
| 94 | write_unlock_bh(&raw_v4_lock); |
| 95 | } |
| 96 | |
| 97 | static void raw_v4_unhash(struct sock *sk) |
| 98 | { |
| 99 | write_lock_bh(&raw_v4_lock); |
| 100 | if (sk_del_node_init(sk)) |
| 101 | sock_prot_dec_use(sk->sk_prot); |
| 102 | write_unlock_bh(&raw_v4_lock); |
| 103 | } |
| 104 | |
| 105 | struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num, |
| 106 | unsigned long raddr, unsigned long laddr, |
| 107 | int dif) |
| 108 | { |
| 109 | struct hlist_node *node; |
| 110 | |
| 111 | sk_for_each_from(sk, node) { |
| 112 | struct inet_sock *inet = inet_sk(sk); |
| 113 | |
| 114 | if (inet->num == num && |
| 115 | !(inet->daddr && inet->daddr != raddr) && |
| 116 | !(inet->rcv_saddr && inet->rcv_saddr != laddr) && |
| 117 | !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)) |
| 118 | goto found; /* gotcha */ |
| 119 | } |
| 120 | sk = NULL; |
| 121 | found: |
| 122 | return sk; |
| 123 | } |
| 124 | |
| 125 | /* |
| 126 | * 0 - deliver |
| 127 | * 1 - block |
| 128 | */ |
| 129 | static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb) |
| 130 | { |
| 131 | int type; |
| 132 | |
| 133 | if (!pskb_may_pull(skb, sizeof(struct icmphdr))) |
| 134 | return 1; |
| 135 | |
| 136 | type = skb->h.icmph->type; |
| 137 | if (type < 32) { |
| 138 | __u32 data = raw_sk(sk)->filter.data; |
| 139 | |
| 140 | return ((1 << type) & data) != 0; |
| 141 | } |
| 142 | |
| 143 | /* Do not block unknown ICMP types */ |
| 144 | return 0; |
| 145 | } |
| 146 | |
| 147 | /* IP input processing comes here for RAW socket delivery. |
| 148 | * Caller owns SKB, so we must make clones. |
| 149 | * |
| 150 | * RFC 1122: SHOULD pass TOS value up to the transport layer. |
| 151 | * -> It does. And not only TOS, but all IP header. |
| 152 | */ |
Patrick McHardy | d13964f | 2005-08-09 19:45:02 -0700 | [diff] [blame] | 153 | int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 154 | { |
| 155 | struct sock *sk; |
| 156 | struct hlist_head *head; |
Patrick McHardy | d13964f | 2005-08-09 19:45:02 -0700 | [diff] [blame] | 157 | int delivered = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 158 | |
| 159 | read_lock(&raw_v4_lock); |
| 160 | head = &raw_v4_htable[hash]; |
| 161 | if (hlist_empty(head)) |
| 162 | goto out; |
| 163 | sk = __raw_v4_lookup(__sk_head(head), iph->protocol, |
| 164 | iph->saddr, iph->daddr, |
| 165 | skb->dev->ifindex); |
| 166 | |
| 167 | while (sk) { |
Patrick McHardy | d13964f | 2005-08-09 19:45:02 -0700 | [diff] [blame] | 168 | delivered = 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 169 | if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) { |
| 170 | struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); |
| 171 | |
| 172 | /* Not releasing hash table! */ |
| 173 | if (clone) |
| 174 | raw_rcv(sk, clone); |
| 175 | } |
| 176 | sk = __raw_v4_lookup(sk_next(sk), iph->protocol, |
| 177 | iph->saddr, iph->daddr, |
| 178 | skb->dev->ifindex); |
| 179 | } |
| 180 | out: |
| 181 | read_unlock(&raw_v4_lock); |
Patrick McHardy | d13964f | 2005-08-09 19:45:02 -0700 | [diff] [blame] | 182 | return delivered; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 183 | } |
| 184 | |
| 185 | void raw_err (struct sock *sk, struct sk_buff *skb, u32 info) |
| 186 | { |
| 187 | struct inet_sock *inet = inet_sk(sk); |
| 188 | int type = skb->h.icmph->type; |
| 189 | int code = skb->h.icmph->code; |
| 190 | int err = 0; |
| 191 | int harderr = 0; |
| 192 | |
| 193 | /* Report error on raw socket, if: |
| 194 | 1. User requested ip_recverr. |
| 195 | 2. Socket is connected (otherwise the error indication |
| 196 | is useless without ip_recverr and error is hard. |
| 197 | */ |
| 198 | if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED) |
| 199 | return; |
| 200 | |
| 201 | switch (type) { |
| 202 | default: |
| 203 | case ICMP_TIME_EXCEEDED: |
| 204 | err = EHOSTUNREACH; |
| 205 | break; |
| 206 | case ICMP_SOURCE_QUENCH: |
| 207 | return; |
| 208 | case ICMP_PARAMETERPROB: |
| 209 | err = EPROTO; |
| 210 | harderr = 1; |
| 211 | break; |
| 212 | case ICMP_DEST_UNREACH: |
| 213 | err = EHOSTUNREACH; |
| 214 | if (code > NR_ICMP_UNREACH) |
| 215 | break; |
| 216 | err = icmp_err_convert[code].errno; |
| 217 | harderr = icmp_err_convert[code].fatal; |
| 218 | if (code == ICMP_FRAG_NEEDED) { |
| 219 | harderr = inet->pmtudisc != IP_PMTUDISC_DONT; |
| 220 | err = EMSGSIZE; |
| 221 | } |
| 222 | } |
| 223 | |
| 224 | if (inet->recverr) { |
| 225 | struct iphdr *iph = (struct iphdr*)skb->data; |
| 226 | u8 *payload = skb->data + (iph->ihl << 2); |
| 227 | |
| 228 | if (inet->hdrincl) |
| 229 | payload = skb->data; |
| 230 | ip_icmp_error(sk, skb, err, 0, info, payload); |
| 231 | } |
| 232 | |
| 233 | if (inet->recverr || harderr) { |
| 234 | sk->sk_err = err; |
| 235 | sk->sk_error_report(sk); |
| 236 | } |
| 237 | } |
| 238 | |
| 239 | static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb) |
| 240 | { |
| 241 | /* Charge it to the socket. */ |
| 242 | |
| 243 | if (sock_queue_rcv_skb(sk, skb) < 0) { |
| 244 | /* FIXME: increment a raw drops counter here */ |
| 245 | kfree_skb(skb); |
| 246 | return NET_RX_DROP; |
| 247 | } |
| 248 | |
| 249 | return NET_RX_SUCCESS; |
| 250 | } |
| 251 | |
| 252 | int raw_rcv(struct sock *sk, struct sk_buff *skb) |
| 253 | { |
| 254 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) { |
| 255 | kfree_skb(skb); |
| 256 | return NET_RX_DROP; |
| 257 | } |
| 258 | |
| 259 | skb_push(skb, skb->data - skb->nh.raw); |
| 260 | |
| 261 | raw_rcv_skb(sk, skb); |
| 262 | return 0; |
| 263 | } |
| 264 | |
Jesper Juhl | f7d7fc0 | 2005-06-18 23:00:34 -0700 | [diff] [blame] | 265 | static int raw_send_hdrinc(struct sock *sk, void *from, size_t length, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 266 | struct rtable *rt, |
| 267 | unsigned int flags) |
| 268 | { |
| 269 | struct inet_sock *inet = inet_sk(sk); |
| 270 | int hh_len; |
| 271 | struct iphdr *iph; |
| 272 | struct sk_buff *skb; |
| 273 | int err; |
| 274 | |
| 275 | if (length > rt->u.dst.dev->mtu) { |
| 276 | ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport, |
| 277 | rt->u.dst.dev->mtu); |
| 278 | return -EMSGSIZE; |
| 279 | } |
| 280 | if (flags&MSG_PROBE) |
| 281 | goto out; |
| 282 | |
| 283 | hh_len = LL_RESERVED_SPACE(rt->u.dst.dev); |
| 284 | |
| 285 | skb = sock_alloc_send_skb(sk, length+hh_len+15, |
| 286 | flags&MSG_DONTWAIT, &err); |
| 287 | if (skb == NULL) |
| 288 | goto error; |
| 289 | skb_reserve(skb, hh_len); |
| 290 | |
| 291 | skb->priority = sk->sk_priority; |
| 292 | skb->dst = dst_clone(&rt->u.dst); |
| 293 | |
| 294 | skb->nh.iph = iph = (struct iphdr *)skb_put(skb, length); |
| 295 | |
| 296 | skb->ip_summed = CHECKSUM_NONE; |
| 297 | |
| 298 | skb->h.raw = skb->nh.raw; |
| 299 | err = memcpy_fromiovecend((void *)iph, from, 0, length); |
| 300 | if (err) |
| 301 | goto error_fault; |
| 302 | |
| 303 | /* We don't modify invalid header */ |
Jesper Juhl | f7d7fc0 | 2005-06-18 23:00:34 -0700 | [diff] [blame] | 304 | if (length >= sizeof(*iph) && iph->ihl * 4U <= length) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 305 | if (!iph->saddr) |
| 306 | iph->saddr = rt->rt_src; |
| 307 | iph->check = 0; |
| 308 | iph->tot_len = htons(length); |
| 309 | if (!iph->id) |
| 310 | ip_select_ident(iph, &rt->u.dst, NULL); |
| 311 | |
| 312 | iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl); |
| 313 | } |
| 314 | |
| 315 | err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev, |
| 316 | dst_output); |
| 317 | if (err > 0) |
| 318 | err = inet->recverr ? net_xmit_errno(err) : 0; |
| 319 | if (err) |
| 320 | goto error; |
| 321 | out: |
| 322 | return 0; |
| 323 | |
| 324 | error_fault: |
| 325 | err = -EFAULT; |
| 326 | kfree_skb(skb); |
| 327 | error: |
| 328 | IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS); |
| 329 | return err; |
| 330 | } |
| 331 | |
| 332 | static void raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg) |
| 333 | { |
| 334 | struct iovec *iov; |
| 335 | u8 __user *type = NULL; |
| 336 | u8 __user *code = NULL; |
| 337 | int probed = 0; |
Jesper Juhl | 93765d8 | 2005-06-18 23:00:15 -0700 | [diff] [blame] | 338 | unsigned int i; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 339 | |
| 340 | if (!msg->msg_iov) |
| 341 | return; |
| 342 | |
| 343 | for (i = 0; i < msg->msg_iovlen; i++) { |
| 344 | iov = &msg->msg_iov[i]; |
| 345 | if (!iov) |
| 346 | continue; |
| 347 | |
| 348 | switch (fl->proto) { |
| 349 | case IPPROTO_ICMP: |
| 350 | /* check if one-byte field is readable or not. */ |
| 351 | if (iov->iov_base && iov->iov_len < 1) |
| 352 | break; |
| 353 | |
| 354 | if (!type) { |
| 355 | type = iov->iov_base; |
| 356 | /* check if code field is readable or not. */ |
| 357 | if (iov->iov_len > 1) |
| 358 | code = type + 1; |
| 359 | } else if (!code) |
| 360 | code = iov->iov_base; |
| 361 | |
| 362 | if (type && code) { |
| 363 | get_user(fl->fl_icmp_type, type); |
Mark J Cox | 6d1cfe3 | 2005-09-19 17:55:30 -0700 | [diff] [blame] | 364 | get_user(fl->fl_icmp_code, code); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 365 | probed = 1; |
| 366 | } |
| 367 | break; |
| 368 | default: |
| 369 | probed = 1; |
| 370 | break; |
| 371 | } |
| 372 | if (probed) |
| 373 | break; |
| 374 | } |
| 375 | } |
| 376 | |
| 377 | static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, |
| 378 | size_t len) |
| 379 | { |
| 380 | struct inet_sock *inet = inet_sk(sk); |
| 381 | struct ipcm_cookie ipc; |
| 382 | struct rtable *rt = NULL; |
| 383 | int free = 0; |
| 384 | u32 daddr; |
| 385 | u32 saddr; |
| 386 | u8 tos; |
| 387 | int err; |
| 388 | |
| 389 | err = -EMSGSIZE; |
Jesper Juhl | 926d4b8 | 2005-06-18 23:00:00 -0700 | [diff] [blame] | 390 | if (len > 0xFFFF) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 391 | goto out; |
| 392 | |
| 393 | /* |
| 394 | * Check the flags. |
| 395 | */ |
| 396 | |
| 397 | err = -EOPNOTSUPP; |
| 398 | if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */ |
| 399 | goto out; /* compatibility */ |
| 400 | |
| 401 | /* |
| 402 | * Get and verify the address. |
| 403 | */ |
| 404 | |
| 405 | if (msg->msg_namelen) { |
| 406 | struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name; |
| 407 | err = -EINVAL; |
| 408 | if (msg->msg_namelen < sizeof(*usin)) |
| 409 | goto out; |
| 410 | if (usin->sin_family != AF_INET) { |
| 411 | static int complained; |
| 412 | if (!complained++) |
| 413 | printk(KERN_INFO "%s forgot to set AF_INET in " |
| 414 | "raw sendmsg. Fix it!\n", |
| 415 | current->comm); |
| 416 | err = -EAFNOSUPPORT; |
| 417 | if (usin->sin_family) |
| 418 | goto out; |
| 419 | } |
| 420 | daddr = usin->sin_addr.s_addr; |
| 421 | /* ANK: I did not forget to get protocol from port field. |
| 422 | * I just do not know, who uses this weirdness. |
| 423 | * IP_HDRINCL is much more convenient. |
| 424 | */ |
| 425 | } else { |
| 426 | err = -EDESTADDRREQ; |
| 427 | if (sk->sk_state != TCP_ESTABLISHED) |
| 428 | goto out; |
| 429 | daddr = inet->daddr; |
| 430 | } |
| 431 | |
| 432 | ipc.addr = inet->saddr; |
| 433 | ipc.opt = NULL; |
| 434 | ipc.oif = sk->sk_bound_dev_if; |
| 435 | |
| 436 | if (msg->msg_controllen) { |
| 437 | err = ip_cmsg_send(msg, &ipc); |
| 438 | if (err) |
| 439 | goto out; |
| 440 | if (ipc.opt) |
| 441 | free = 1; |
| 442 | } |
| 443 | |
| 444 | saddr = ipc.addr; |
| 445 | ipc.addr = daddr; |
| 446 | |
| 447 | if (!ipc.opt) |
| 448 | ipc.opt = inet->opt; |
| 449 | |
| 450 | if (ipc.opt) { |
| 451 | err = -EINVAL; |
| 452 | /* Linux does not mangle headers on raw sockets, |
| 453 | * so that IP options + IP_HDRINCL is non-sense. |
| 454 | */ |
| 455 | if (inet->hdrincl) |
| 456 | goto done; |
| 457 | if (ipc.opt->srr) { |
| 458 | if (!daddr) |
| 459 | goto done; |
| 460 | daddr = ipc.opt->faddr; |
| 461 | } |
| 462 | } |
| 463 | tos = RT_CONN_FLAGS(sk); |
| 464 | if (msg->msg_flags & MSG_DONTROUTE) |
| 465 | tos |= RTO_ONLINK; |
| 466 | |
| 467 | if (MULTICAST(daddr)) { |
| 468 | if (!ipc.oif) |
| 469 | ipc.oif = inet->mc_index; |
| 470 | if (!saddr) |
| 471 | saddr = inet->mc_addr; |
| 472 | } |
| 473 | |
| 474 | { |
| 475 | struct flowi fl = { .oif = ipc.oif, |
| 476 | .nl_u = { .ip4_u = |
| 477 | { .daddr = daddr, |
| 478 | .saddr = saddr, |
| 479 | .tos = tos } }, |
| 480 | .proto = inet->hdrincl ? IPPROTO_RAW : |
| 481 | sk->sk_protocol, |
| 482 | }; |
| 483 | if (!inet->hdrincl) |
| 484 | raw_probe_proto_opt(&fl, msg); |
| 485 | |
| 486 | err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT)); |
| 487 | } |
| 488 | if (err) |
| 489 | goto done; |
| 490 | |
| 491 | err = -EACCES; |
| 492 | if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST)) |
| 493 | goto done; |
| 494 | |
| 495 | if (msg->msg_flags & MSG_CONFIRM) |
| 496 | goto do_confirm; |
| 497 | back_from_confirm: |
| 498 | |
| 499 | if (inet->hdrincl) |
| 500 | err = raw_send_hdrinc(sk, msg->msg_iov, len, |
| 501 | rt, msg->msg_flags); |
| 502 | |
| 503 | else { |
| 504 | if (!ipc.addr) |
| 505 | ipc.addr = rt->rt_dst; |
| 506 | lock_sock(sk); |
| 507 | err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0, |
| 508 | &ipc, rt, msg->msg_flags); |
| 509 | if (err) |
| 510 | ip_flush_pending_frames(sk); |
| 511 | else if (!(msg->msg_flags & MSG_MORE)) |
| 512 | err = ip_push_pending_frames(sk); |
| 513 | release_sock(sk); |
| 514 | } |
| 515 | done: |
| 516 | if (free) |
| 517 | kfree(ipc.opt); |
| 518 | ip_rt_put(rt); |
| 519 | |
Jesper Juhl | 5418c69 | 2005-06-18 22:59:45 -0700 | [diff] [blame] | 520 | out: |
| 521 | if (err < 0) |
| 522 | return err; |
| 523 | return len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 524 | |
| 525 | do_confirm: |
| 526 | dst_confirm(&rt->u.dst); |
| 527 | if (!(msg->msg_flags & MSG_PROBE) || len) |
| 528 | goto back_from_confirm; |
| 529 | err = 0; |
| 530 | goto done; |
| 531 | } |
| 532 | |
| 533 | static void raw_close(struct sock *sk, long timeout) |
| 534 | { |
| 535 | /* |
| 536 | * Raw sockets may have direct kernel refereneces. Kill them. |
| 537 | */ |
| 538 | ip_ra_control(sk, 0, NULL); |
| 539 | |
| 540 | sk_common_release(sk); |
| 541 | } |
| 542 | |
| 543 | /* This gets rid of all the nasties in af_inet. -DaveM */ |
| 544 | static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) |
| 545 | { |
| 546 | struct inet_sock *inet = inet_sk(sk); |
| 547 | struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; |
| 548 | int ret = -EINVAL; |
| 549 | int chk_addr_ret; |
| 550 | |
| 551 | if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in)) |
| 552 | goto out; |
| 553 | chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr); |
| 554 | ret = -EADDRNOTAVAIL; |
| 555 | if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL && |
| 556 | chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST) |
| 557 | goto out; |
| 558 | inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr; |
| 559 | if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) |
| 560 | inet->saddr = 0; /* Use device */ |
| 561 | sk_dst_reset(sk); |
| 562 | ret = 0; |
| 563 | out: return ret; |
| 564 | } |
| 565 | |
| 566 | /* |
| 567 | * This should be easy, if there is something there |
| 568 | * we return it, otherwise we block. |
| 569 | */ |
| 570 | |
| 571 | static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, |
| 572 | size_t len, int noblock, int flags, int *addr_len) |
| 573 | { |
| 574 | struct inet_sock *inet = inet_sk(sk); |
| 575 | size_t copied = 0; |
| 576 | int err = -EOPNOTSUPP; |
| 577 | struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; |
| 578 | struct sk_buff *skb; |
| 579 | |
| 580 | if (flags & MSG_OOB) |
| 581 | goto out; |
| 582 | |
| 583 | if (addr_len) |
| 584 | *addr_len = sizeof(*sin); |
| 585 | |
| 586 | if (flags & MSG_ERRQUEUE) { |
| 587 | err = ip_recv_error(sk, msg, len); |
| 588 | goto out; |
| 589 | } |
| 590 | |
| 591 | skb = skb_recv_datagram(sk, flags, noblock, &err); |
| 592 | if (!skb) |
| 593 | goto out; |
| 594 | |
| 595 | copied = skb->len; |
| 596 | if (len < copied) { |
| 597 | msg->msg_flags |= MSG_TRUNC; |
| 598 | copied = len; |
| 599 | } |
| 600 | |
| 601 | err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); |
| 602 | if (err) |
| 603 | goto done; |
| 604 | |
| 605 | sock_recv_timestamp(msg, sk, skb); |
| 606 | |
| 607 | /* Copy the address. */ |
| 608 | if (sin) { |
| 609 | sin->sin_family = AF_INET; |
| 610 | sin->sin_addr.s_addr = skb->nh.iph->saddr; |
| 611 | memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); |
| 612 | } |
| 613 | if (inet->cmsg_flags) |
| 614 | ip_cmsg_recv(msg, skb); |
| 615 | if (flags & MSG_TRUNC) |
| 616 | copied = skb->len; |
| 617 | done: |
| 618 | skb_free_datagram(sk, skb); |
Jesper Juhl | 5418c69 | 2005-06-18 22:59:45 -0700 | [diff] [blame] | 619 | out: |
| 620 | if (err) |
| 621 | return err; |
| 622 | return copied; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 623 | } |
| 624 | |
| 625 | static int raw_init(struct sock *sk) |
| 626 | { |
| 627 | struct raw_sock *rp = raw_sk(sk); |
| 628 | |
| 629 | if (inet_sk(sk)->num == IPPROTO_ICMP) |
| 630 | memset(&rp->filter, 0, sizeof(rp->filter)); |
| 631 | return 0; |
| 632 | } |
| 633 | |
| 634 | static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen) |
| 635 | { |
| 636 | if (optlen > sizeof(struct icmp_filter)) |
| 637 | optlen = sizeof(struct icmp_filter); |
| 638 | if (copy_from_user(&raw_sk(sk)->filter, optval, optlen)) |
| 639 | return -EFAULT; |
| 640 | return 0; |
| 641 | } |
| 642 | |
| 643 | static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen) |
| 644 | { |
| 645 | int len, ret = -EFAULT; |
| 646 | |
| 647 | if (get_user(len, optlen)) |
| 648 | goto out; |
| 649 | ret = -EINVAL; |
| 650 | if (len < 0) |
| 651 | goto out; |
| 652 | if (len > sizeof(struct icmp_filter)) |
| 653 | len = sizeof(struct icmp_filter); |
| 654 | ret = -EFAULT; |
| 655 | if (put_user(len, optlen) || |
| 656 | copy_to_user(optval, &raw_sk(sk)->filter, len)) |
| 657 | goto out; |
| 658 | ret = 0; |
| 659 | out: return ret; |
| 660 | } |
| 661 | |
| 662 | static int raw_setsockopt(struct sock *sk, int level, int optname, |
| 663 | char __user *optval, int optlen) |
| 664 | { |
| 665 | if (level != SOL_RAW) |
| 666 | return ip_setsockopt(sk, level, optname, optval, optlen); |
| 667 | |
| 668 | if (optname == ICMP_FILTER) { |
| 669 | if (inet_sk(sk)->num != IPPROTO_ICMP) |
| 670 | return -EOPNOTSUPP; |
| 671 | else |
| 672 | return raw_seticmpfilter(sk, optval, optlen); |
| 673 | } |
| 674 | return -ENOPROTOOPT; |
| 675 | } |
| 676 | |
| 677 | static int raw_getsockopt(struct sock *sk, int level, int optname, |
| 678 | char __user *optval, int __user *optlen) |
| 679 | { |
| 680 | if (level != SOL_RAW) |
| 681 | return ip_getsockopt(sk, level, optname, optval, optlen); |
| 682 | |
| 683 | if (optname == ICMP_FILTER) { |
| 684 | if (inet_sk(sk)->num != IPPROTO_ICMP) |
| 685 | return -EOPNOTSUPP; |
| 686 | else |
| 687 | return raw_geticmpfilter(sk, optval, optlen); |
| 688 | } |
| 689 | return -ENOPROTOOPT; |
| 690 | } |
| 691 | |
| 692 | static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg) |
| 693 | { |
| 694 | switch (cmd) { |
| 695 | case SIOCOUTQ: { |
| 696 | int amount = atomic_read(&sk->sk_wmem_alloc); |
| 697 | return put_user(amount, (int __user *)arg); |
| 698 | } |
| 699 | case SIOCINQ: { |
| 700 | struct sk_buff *skb; |
| 701 | int amount = 0; |
| 702 | |
Herbert Xu | e0f9f85 | 2005-06-18 22:56:18 -0700 | [diff] [blame] | 703 | spin_lock_bh(&sk->sk_receive_queue.lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 704 | skb = skb_peek(&sk->sk_receive_queue); |
| 705 | if (skb != NULL) |
| 706 | amount = skb->len; |
Herbert Xu | e0f9f85 | 2005-06-18 22:56:18 -0700 | [diff] [blame] | 707 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 708 | return put_user(amount, (int __user *)arg); |
| 709 | } |
| 710 | |
| 711 | default: |
| 712 | #ifdef CONFIG_IP_MROUTE |
| 713 | return ipmr_ioctl(sk, cmd, (void __user *)arg); |
| 714 | #else |
| 715 | return -ENOIOCTLCMD; |
| 716 | #endif |
| 717 | } |
| 718 | } |
| 719 | |
| 720 | struct proto raw_prot = { |
| 721 | .name = "RAW", |
| 722 | .owner = THIS_MODULE, |
| 723 | .close = raw_close, |
| 724 | .connect = ip4_datagram_connect, |
| 725 | .disconnect = udp_disconnect, |
| 726 | .ioctl = raw_ioctl, |
| 727 | .init = raw_init, |
| 728 | .setsockopt = raw_setsockopt, |
| 729 | .getsockopt = raw_getsockopt, |
| 730 | .sendmsg = raw_sendmsg, |
| 731 | .recvmsg = raw_recvmsg, |
| 732 | .bind = raw_bind, |
| 733 | .backlog_rcv = raw_rcv_skb, |
| 734 | .hash = raw_v4_hash, |
| 735 | .unhash = raw_v4_unhash, |
| 736 | .obj_size = sizeof(struct raw_sock), |
| 737 | }; |
| 738 | |
| 739 | #ifdef CONFIG_PROC_FS |
| 740 | struct raw_iter_state { |
| 741 | int bucket; |
| 742 | }; |
| 743 | |
| 744 | #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private) |
| 745 | |
| 746 | static struct sock *raw_get_first(struct seq_file *seq) |
| 747 | { |
| 748 | struct sock *sk; |
| 749 | struct raw_iter_state* state = raw_seq_private(seq); |
| 750 | |
| 751 | for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) { |
| 752 | struct hlist_node *node; |
| 753 | |
| 754 | sk_for_each(sk, node, &raw_v4_htable[state->bucket]) |
| 755 | if (sk->sk_family == PF_INET) |
| 756 | goto found; |
| 757 | } |
| 758 | sk = NULL; |
| 759 | found: |
| 760 | return sk; |
| 761 | } |
| 762 | |
| 763 | static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk) |
| 764 | { |
| 765 | struct raw_iter_state* state = raw_seq_private(seq); |
| 766 | |
| 767 | do { |
| 768 | sk = sk_next(sk); |
| 769 | try_again: |
| 770 | ; |
| 771 | } while (sk && sk->sk_family != PF_INET); |
| 772 | |
| 773 | if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) { |
| 774 | sk = sk_head(&raw_v4_htable[state->bucket]); |
| 775 | goto try_again; |
| 776 | } |
| 777 | return sk; |
| 778 | } |
| 779 | |
| 780 | static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos) |
| 781 | { |
| 782 | struct sock *sk = raw_get_first(seq); |
| 783 | |
| 784 | if (sk) |
| 785 | while (pos && (sk = raw_get_next(seq, sk)) != NULL) |
| 786 | --pos; |
| 787 | return pos ? NULL : sk; |
| 788 | } |
| 789 | |
| 790 | static void *raw_seq_start(struct seq_file *seq, loff_t *pos) |
| 791 | { |
| 792 | read_lock(&raw_v4_lock); |
| 793 | return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; |
| 794 | } |
| 795 | |
| 796 | static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos) |
| 797 | { |
| 798 | struct sock *sk; |
| 799 | |
| 800 | if (v == SEQ_START_TOKEN) |
| 801 | sk = raw_get_first(seq); |
| 802 | else |
| 803 | sk = raw_get_next(seq, v); |
| 804 | ++*pos; |
| 805 | return sk; |
| 806 | } |
| 807 | |
| 808 | static void raw_seq_stop(struct seq_file *seq, void *v) |
| 809 | { |
| 810 | read_unlock(&raw_v4_lock); |
| 811 | } |
| 812 | |
| 813 | static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i) |
| 814 | { |
| 815 | struct inet_sock *inet = inet_sk(sp); |
| 816 | unsigned int dest = inet->daddr, |
| 817 | src = inet->rcv_saddr; |
| 818 | __u16 destp = 0, |
| 819 | srcp = inet->num; |
| 820 | |
| 821 | sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X" |
| 822 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p", |
| 823 | i, src, srcp, dest, destp, sp->sk_state, |
| 824 | atomic_read(&sp->sk_wmem_alloc), |
| 825 | atomic_read(&sp->sk_rmem_alloc), |
| 826 | 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp), |
| 827 | atomic_read(&sp->sk_refcnt), sp); |
| 828 | return tmpbuf; |
| 829 | } |
| 830 | |
| 831 | static int raw_seq_show(struct seq_file *seq, void *v) |
| 832 | { |
| 833 | char tmpbuf[129]; |
| 834 | |
| 835 | if (v == SEQ_START_TOKEN) |
| 836 | seq_printf(seq, "%-127s\n", |
| 837 | " sl local_address rem_address st tx_queue " |
| 838 | "rx_queue tr tm->when retrnsmt uid timeout " |
| 839 | "inode"); |
| 840 | else { |
| 841 | struct raw_iter_state *state = raw_seq_private(seq); |
| 842 | |
| 843 | seq_printf(seq, "%-127s\n", |
| 844 | get_raw_sock(v, tmpbuf, state->bucket)); |
| 845 | } |
| 846 | return 0; |
| 847 | } |
| 848 | |
| 849 | static struct seq_operations raw_seq_ops = { |
| 850 | .start = raw_seq_start, |
| 851 | .next = raw_seq_next, |
| 852 | .stop = raw_seq_stop, |
| 853 | .show = raw_seq_show, |
| 854 | }; |
| 855 | |
| 856 | static int raw_seq_open(struct inode *inode, struct file *file) |
| 857 | { |
| 858 | struct seq_file *seq; |
| 859 | int rc = -ENOMEM; |
| 860 | struct raw_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL); |
| 861 | |
| 862 | if (!s) |
| 863 | goto out; |
| 864 | rc = seq_open(file, &raw_seq_ops); |
| 865 | if (rc) |
| 866 | goto out_kfree; |
| 867 | |
| 868 | seq = file->private_data; |
| 869 | seq->private = s; |
| 870 | memset(s, 0, sizeof(*s)); |
| 871 | out: |
| 872 | return rc; |
| 873 | out_kfree: |
| 874 | kfree(s); |
| 875 | goto out; |
| 876 | } |
| 877 | |
| 878 | static struct file_operations raw_seq_fops = { |
| 879 | .owner = THIS_MODULE, |
| 880 | .open = raw_seq_open, |
| 881 | .read = seq_read, |
| 882 | .llseek = seq_lseek, |
| 883 | .release = seq_release_private, |
| 884 | }; |
| 885 | |
| 886 | int __init raw_proc_init(void) |
| 887 | { |
| 888 | if (!proc_net_fops_create("raw", S_IRUGO, &raw_seq_fops)) |
| 889 | return -ENOMEM; |
| 890 | return 0; |
| 891 | } |
| 892 | |
| 893 | void __init raw_proc_exit(void) |
| 894 | { |
| 895 | proc_net_remove("raw"); |
| 896 | } |
| 897 | #endif /* CONFIG_PROC_FS */ |