?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
[linux-2.6.git] / net / ipv4 / raw.c
blobdd44e0ab600cafe76ac50f74aaff5c6d429ea534
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 * Authors: Ross Biro
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 *
11 * Fixes:
12 * Alan Cox : verify_area() fixed up
13 * Alan Cox : ICMP error handling
14 * Alan Cox : EMSGSIZE if you send too big a packet
15 * Alan Cox : Now uses generic datagrams and shared
16 * skbuff library. No more peek crashes,
17 * no more backlogs
18 * Alan Cox : Checks sk->broadcast.
19 * Alan Cox : Uses skb_free_datagram/skb_copy_datagram
20 * Alan Cox : Raw passes ip options too
21 * Alan Cox : Setsocketopt added
22 * Alan Cox : Fixed error return for broadcasts
23 * Alan Cox : Removed wake_up calls
24 * Alan Cox : Use ttl/tos
25 * Alan Cox : Cleaned up old debugging
26 * Alan Cox : Use new kernel side addresses
27 * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets.
28 * Alan Cox : BSD style RAW socket demultiplexing.
29 * Alan Cox : Beginnings of mrouted support.
30 * Alan Cox : Added IP_HDRINCL option.
31 * Alan Cox : Skip broadcast check if BSDism set.
32 * David S. Miller : New socket lookup architecture.
33 *
34 * This program is free software; you can redistribute it and/or
35 * modify it under the terms of the GNU General Public License
36 * as published by the Free Software Foundation; either version
37 * 2 of the License, or (at your option) any later version.
38 */
39
40 #include <linux/types.h>
41 #include <linux/atomic.h>
42 #include <asm/byteorder.h>
43 #include <asm/current.h>
44 #include <asm/uaccess.h>
45 #include <asm/ioctls.h>
46 #include <linux/stddef.h>
47 #include <linux/slab.h>
48 #include <linux/errno.h>
49 #include <linux/aio.h>
50 #include <linux/kernel.h>
51 #include <linux/export.h>
52 #include <linux/spinlock.h>
53 #include <linux/sockios.h>
54 #include <linux/socket.h>
55 #include <linux/in.h>
56 #include <linux/mroute.h>
57 #include <linux/netdevice.h>
58 #include <linux/in_route.h>
59 #include <linux/route.h>
60 #include <linux/skbuff.h>
61 #include <net/net_namespace.h>
62 #include <net/dst.h>
63 #include <net/sock.h>
64 #include <linux/ip.h>
65 #include <linux/net.h>
66 #include <net/ip.h>
67 #include <net/icmp.h>
68 #include <net/udp.h>
69 #include <net/raw.h>
70 #include <net/snmp.h>
71 #include <net/tcp_states.h>
72 #include <net/inet_common.h>
73 #include <net/checksum.h>
74 #include <net/xfrm.h>
75 #include <linux/rtnetlink.h>
76 #include <linux/proc_fs.h>
77 #include <linux/seq_file.h>
78 #include <linux/netfilter.h>
79 #include <linux/netfilter_ipv4.h>
80 #include <linux/compat.h>
81
82 static struct raw_hashinfo raw_v4_hashinfo = {
83 .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
84 };
85
86 void raw_hash_sk(struct sock *sk)
87 {
88 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
89 struct hlist_head *head;
90
91 head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)];
92
93 write_lock_bh(&h->lock);
94 sk_add_node(sk, head);
95 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
96 write_unlock_bh(&h->lock);
97 }
98 EXPORT_SYMBOL_GPL(raw_hash_sk);
99
100 void raw_unhash_sk(struct sock *sk)
101 {
102 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
103
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
111 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
112 unsigned short num, __be32 raddr, __be32 laddr, int dif)
113 {
114 sk_for_each_from(sk) {
115 struct inet_sock *inet = inet_sk(sk);
116
117 if (net_eq(sock_net(sk), net) && inet->inet_num == num &&
118 !(inet->inet_daddr && inet->inet_daddr != raddr) &&
119 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
120 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
121 goto found; /* gotcha */
122 }
123 sk = NULL;
124 found:
125 return sk;
126 }
127
128 /*
129 * 0 - deliver
130 * 1 - block
131 */
132 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
133 {
134 struct icmphdr _hdr;
135 const struct icmphdr *hdr;
136
137 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
138 sizeof(_hdr), &_hdr);
139 if (!hdr)
140 return 1;
141
142 if (hdr->type < 32) {
143 __u32 data = raw_sk(sk)->filter.data;
144
145 return ((1U << hdr->type) & data) != 0;
146 }
147
148 /* Do not block unknown ICMP types */
149 return 0;
150 }
151
152 /* IP input processing comes here for RAW socket delivery.
153 * Caller owns SKB, so we must make clones.
154 *
155 * RFC 1122: SHOULD pass TOS value up to the transport layer.
156 * -> It does. And not only TOS, but all IP header.
157 */
158 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
159 {
160 struct sock *sk;
161 struct hlist_head *head;
162 int delivered = 0;
163 struct net *net;
164
165 read_lock(&raw_v4_hashinfo.lock);
166 head = &raw_v4_hashinfo.ht[hash];
167 if (hlist_empty(head))
168 goto out;
169
170 net = dev_net(skb->dev);
171 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
172 iph->saddr, iph->daddr,
173 skb->dev->ifindex);
174
175 while (sk) {
176 delivered = 1;
177 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
178 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
179
180 /* Not releasing hash table! */
181 if (clone)
182 raw_rcv(sk, clone);
183 }
184 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
185 iph->saddr, iph->daddr,
186 skb->dev->ifindex);
187 }
188 out:
189 read_unlock(&raw_v4_hashinfo.lock);
190 return delivered;
191 }
192
193 int raw_local_deliver(struct sk_buff *skb, int protocol)
194 {
195 int hash;
196 struct sock *raw_sk;
197
198 hash = protocol & (RAW_HTABLE_SIZE - 1);
199 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
200
201 /* If there maybe a raw socket we must check - if not we
202 * don't care less
203 */
204 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
205 raw_sk = NULL;
206
207 return raw_sk != NULL;
208
209 }
210
211 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
212 {
213 struct inet_sock *inet = inet_sk(sk);
214 const int type = icmp_hdr(skb)->type;
215 const int code = icmp_hdr(skb)->code;
216 int err = 0;
217 int harderr = 0;
218
219 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
220 ipv4_sk_update_pmtu(skb, sk, info);
221 else if (type == ICMP_REDIRECT)
222 ipv4_sk_redirect(skb, sk);
223
224 /* Report error on raw socket, if:
225 1. User requested ip_recverr.
226 2. Socket is connected (otherwise the error indication
227 is useless without ip_recverr and error is hard.
228 */
229 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
230 return;
231
232 switch (type) {
233 default:
234 case ICMP_TIME_EXCEEDED:
235 err = EHOSTUNREACH;
236 break;
237 case ICMP_SOURCE_QUENCH:
238 return;
239 case ICMP_PARAMETERPROB:
240 err = EPROTO;
241 harderr = 1;
242 break;
243 case ICMP_DEST_UNREACH:
244 err = EHOSTUNREACH;
245 if (code > NR_ICMP_UNREACH)
246 break;
247 err = icmp_err_convert[code].errno;
248 harderr = icmp_err_convert[code].fatal;
249 if (code == ICMP_FRAG_NEEDED) {
250 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
251 err = EMSGSIZE;
252 }
253 }
254
255 if (inet->recverr) {
256 const struct iphdr *iph = (const struct iphdr *)skb->data;
257 u8 *payload = skb->data + (iph->ihl << 2);
258
259 if (inet->hdrincl)
260 payload = skb->data;
261 ip_icmp_error(sk, skb, err, 0, info, payload);
262 }
263
264 if (inet->recverr || harderr) {
265 sk->sk_err = err;
266 sk->sk_error_report(sk);
267 }
268 }
269
270 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
271 {
272 int hash;
273 struct sock *raw_sk;
274 const struct iphdr *iph;
275 struct net *net;
276
277 hash = protocol & (RAW_HTABLE_SIZE - 1);
278
279 read_lock(&raw_v4_hashinfo.lock);
280 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
281 if (raw_sk != NULL) {
282 iph = (const struct iphdr *)skb->data;
283 net = dev_net(skb->dev);
284
285 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
286 iph->daddr, iph->saddr,
287 skb->dev->ifindex)) != NULL) {
288 raw_err(raw_sk, skb, info);
289 raw_sk = sk_next(raw_sk);
290 iph = (const struct iphdr *)skb->data;
291 }
292 }
293 read_unlock(&raw_v4_hashinfo.lock);
294 }
295
296 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
297 {
298 /* Charge it to the socket. */
299
300 ipv4_pktinfo_prepare(skb);
301 if (sock_queue_rcv_skb(sk, skb) < 0) {
302 kfree_skb(skb);
303 return NET_RX_DROP;
304 }
305
306 return NET_RX_SUCCESS;
307 }
308
309 int raw_rcv(struct sock *sk, struct sk_buff *skb)
310 {
311 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
312 atomic_inc(&sk->sk_drops);
313 kfree_skb(skb);
314 return NET_RX_DROP;
315 }
316 nf_reset(skb);
317
318 skb_push(skb, skb->data - skb_network_header(skb));
319
320 raw_rcv_skb(sk, skb);
321 return 0;
322 }
323
324 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
325 void *from, size_t length,
326 struct rtable **rtp,
327 unsigned int flags)
328 {
329 struct inet_sock *inet = inet_sk(sk);
330 struct net *net = sock_net(sk);
331 struct iphdr *iph;
332 struct sk_buff *skb;
333 unsigned int iphlen;
334 int err;
335 struct rtable *rt = *rtp;
336 int hlen, tlen;
337
338 if (length > rt->dst.dev->mtu) {
339 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
340 rt->dst.dev->mtu);
341 return -EMSGSIZE;
342 }
343 if (flags&MSG_PROBE)
344 goto out;
345
346 hlen = LL_RESERVED_SPACE(rt->dst.dev);
347 tlen = rt->dst.dev->needed_tailroom;
348 skb = sock_alloc_send_skb(sk,
349 length + hlen + tlen + 15,
350 flags & MSG_DONTWAIT, &err);
351 if (skb == NULL)
352 goto error;
353 skb_reserve(skb, hlen);
354
355 skb->priority = sk->sk_priority;
356 skb->mark = sk->sk_mark;
357 skb_dst_set(skb, &rt->dst);
358 *rtp = NULL;
359
360 skb_reset_network_header(skb);
361 iph = ip_hdr(skb);
362 skb_put(skb, length);
363
364 skb->ip_summed = CHECKSUM_NONE;
365
366 skb->transport_header = skb->network_header;
367 err = -EFAULT;
368 if (memcpy_fromiovecend((void *)iph, from, 0, length))
369 goto error_free;
370
371 iphlen = iph->ihl * 4;
372
373 /*
374 * We don't want to modify the ip header, but we do need to
375 * be sure that it won't cause problems later along the network
376 * stack. Specifically we want to make sure that iph->ihl is a
377 * sane value. If ihl points beyond the length of the buffer passed
378 * in, reject the frame as invalid
379 */
380 err = -EINVAL;
381 if (iphlen > length)
382 goto error_free;
383
384 if (iphlen >= sizeof(*iph)) {
385 if (!iph->saddr)
386 iph->saddr = fl4->saddr;
387 iph->check = 0;
388 iph->tot_len = htons(length);
389 if (!iph->id)
390 ip_select_ident(iph, &rt->dst, NULL);
391
392 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
393 }
394 if (iph->protocol == IPPROTO_ICMP)
395 icmp_out_count(net, ((struct icmphdr *)
396 skb_transport_header(skb))->type);
397
398 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT, skb, NULL,
399 rt->dst.dev, dst_output);
400 if (err > 0)
401 err = net_xmit_errno(err);
402 if (err)
403 goto error;
404 out:
405 return 0;
406
407 error_free:
408 kfree_skb(skb);
409 error:
410 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
411 if (err == -ENOBUFS && !inet->recverr)
412 err = 0;
413 return err;
414 }
415
416 static int raw_probe_proto_opt(struct flowi4 *fl4, struct msghdr *msg)
417 {
418 struct iovec *iov;
419 u8 __user *type = NULL;
420 u8 __user *code = NULL;
421 int probed = 0;
422 unsigned int i;
423
424 if (!msg->msg_iov)
425 return 0;
426
427 for (i = 0; i < msg->msg_iovlen; i++) {
428 iov = &msg->msg_iov[i];
429 if (!iov)
430 continue;
431
432 switch (fl4->flowi4_proto) {
433 case IPPROTO_ICMP:
434 /* check if one-byte field is readable or not. */
435 if (iov->iov_base && iov->iov_len < 1)
436 break;
437
438 if (!type) {
439 type = iov->iov_base;
440 /* check if code field is readable or not. */
441 if (iov->iov_len > 1)
442 code = type + 1;
443 } else if (!code)
444 code = iov->iov_base;
445
446 if (type && code) {
447 if (get_user(fl4->fl4_icmp_type, type) ||
448 get_user(fl4->fl4_icmp_code, code))
449 return -EFAULT;
450 probed = 1;
451 }
452 break;
453 default:
454 probed = 1;
455 break;
456 }
457 if (probed)
458 break;
459 }
460 return 0;
461 }
462
463 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
464 size_t len)
465 {
466 struct inet_sock *inet = inet_sk(sk);
467 struct ipcm_cookie ipc;
468 struct rtable *rt = NULL;
469 struct flowi4 fl4;
470 int free = 0;
471 __be32 daddr;
472 __be32 saddr;
473 u8 tos;
474 int err;
475 struct ip_options_data opt_copy;
476
477 err = -EMSGSIZE;
478 if (len > 0xFFFF)
479 goto out;
480
481 /*
482 * Check the flags.
483 */
484
485 err = -EOPNOTSUPP;
486 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
487 goto out; /* compatibility */
488
489 /*
490 * Get and verify the address.
491 */
492
493 if (msg->msg_namelen) {
494 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
495 err = -EINVAL;
496 if (msg->msg_namelen < sizeof(*usin))
497 goto out;
498 if (usin->sin_family != AF_INET) {
499 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
500 __func__, current->comm);
501 err = -EAFNOSUPPORT;
502 if (usin->sin_family)
503 goto out;
504 }
505 daddr = usin->sin_addr.s_addr;
506 /* ANK: I did not forget to get protocol from port field.
507 * I just do not know, who uses this weirdness.
508 * IP_HDRINCL is much more convenient.
509 */
510 } else {
511 err = -EDESTADDRREQ;
512 if (sk->sk_state != TCP_ESTABLISHED)
513 goto out;
514 daddr = inet->inet_daddr;
515 }
516
517 ipc.addr = inet->inet_saddr;
518 ipc.opt = NULL;
519 ipc.tx_flags = 0;
520 ipc.oif = sk->sk_bound_dev_if;
521
522 if (msg->msg_controllen) {
523 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
524 if (err)
525 goto out;
526 if (ipc.opt)
527 free = 1;
528 }
529
530 saddr = ipc.addr;
531 ipc.addr = daddr;
532
533 if (!ipc.opt) {
534 struct ip_options_rcu *inet_opt;
535
536 rcu_read_lock();
537 inet_opt = rcu_dereference(inet->inet_opt);
538 if (inet_opt) {
539 memcpy(&opt_copy, inet_opt,
540 sizeof(*inet_opt) + inet_opt->opt.optlen);
541 ipc.opt = &opt_copy.opt;
542 }
543 rcu_read_unlock();
544 }
545
546 if (ipc.opt) {
547 err = -EINVAL;
548 /* Linux does not mangle headers on raw sockets,
549 * so that IP options + IP_HDRINCL is non-sense.
550 */
551 if (inet->hdrincl)
552 goto done;
553 if (ipc.opt->opt.srr) {
554 if (!daddr)
555 goto done;
556 daddr = ipc.opt->opt.faddr;
557 }
558 }
559 tos = RT_CONN_FLAGS(sk);
560 if (msg->msg_flags & MSG_DONTROUTE)
561 tos |= RTO_ONLINK;
562
563 if (ipv4_is_multicast(daddr)) {
564 if (!ipc.oif)
565 ipc.oif = inet->mc_index;
566 if (!saddr)
567 saddr = inet->mc_addr;
568 } else if (!ipc.oif)
569 ipc.oif = inet->uc_index;
570
571 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
572 RT_SCOPE_UNIVERSE,
573 inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
574 inet_sk_flowi_flags(sk) | FLOWI_FLAG_CAN_SLEEP,
575 daddr, saddr, 0, 0);
576
577 if (!inet->hdrincl) {
578 err = raw_probe_proto_opt(&fl4, msg);
579 if (err)
580 goto done;
581 }
582
583 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
584 rt = ip_route_output_flow(sock_net(sk), &fl4, sk);
585 if (IS_ERR(rt)) {
586 err = PTR_ERR(rt);
587 rt = NULL;
588 goto done;
589 }
590
591 err = -EACCES;
592 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
593 goto done;
594
595 if (msg->msg_flags & MSG_CONFIRM)
596 goto do_confirm;
597 back_from_confirm:
598
599 if (inet->hdrincl)
600 err = raw_send_hdrinc(sk, &fl4, msg->msg_iov, len,
601 &rt, msg->msg_flags);
602
603 else {
604 if (!ipc.addr)
605 ipc.addr = fl4.daddr;
606 lock_sock(sk);
607 err = ip_append_data(sk, &fl4, ip_generic_getfrag,
608 msg->msg_iov, len, 0,
609 &ipc, &rt, msg->msg_flags);
610 if (err)
611 ip_flush_pending_frames(sk);
612 else if (!(msg->msg_flags & MSG_MORE)) {
613 err = ip_push_pending_frames(sk, &fl4);
614 if (err == -ENOBUFS && !inet->recverr)
615 err = 0;
616 }
617 release_sock(sk);
618 }
619 done:
620 if (free)
621 kfree(ipc.opt);
622 ip_rt_put(rt);
623
624 out:
625 if (err < 0)
626 return err;
627 return len;
628
629 do_confirm:
630 dst_confirm(&rt->dst);
631 if (!(msg->msg_flags & MSG_PROBE) || len)
632 goto back_from_confirm;
633 err = 0;
634 goto done;
635 }
636
637 static void raw_close(struct sock *sk, long timeout)
638 {
639 /*
640 * Raw sockets may have direct kernel references. Kill them.
641 */
642 ip_ra_control(sk, 0, NULL);
643
644 sk_common_release(sk);
645 }
646
647 static void raw_destroy(struct sock *sk)
648 {
649 lock_sock(sk);
650 ip_flush_pending_frames(sk);
651 release_sock(sk);
652 }
653
654 /* This gets rid of all the nasties in af_inet. -DaveM */
655 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
656 {
657 struct inet_sock *inet = inet_sk(sk);
658 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
659 int ret = -EINVAL;
660 int chk_addr_ret;
661
662 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
663 goto out;
664 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
665 ret = -EADDRNOTAVAIL;
666 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
667 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
668 goto out;
669 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
670 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
671 inet->inet_saddr = 0; /* Use device */
672 sk_dst_reset(sk);
673 ret = 0;
674 out: return ret;
675 }
676
677 /*
678 * This should be easy, if there is something there
679 * we return it, otherwise we block.
680 */
681
682 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
683 size_t len, int noblock, int flags, int *addr_len)
684 {
685 struct inet_sock *inet = inet_sk(sk);
686 size_t copied = 0;
687 int err = -EOPNOTSUPP;
688 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
689 struct sk_buff *skb;
690
691 if (flags & MSG_OOB)
692 goto out;
693
694 if (addr_len)
695 *addr_len = sizeof(*sin);
696
697 if (flags & MSG_ERRQUEUE) {
698 err = ip_recv_error(sk, msg, len);
699 goto out;
700 }
701
702 skb = skb_recv_datagram(sk, flags, noblock, &err);
703 if (!skb)
704 goto out;
705
706 copied = skb->len;
707 if (len < copied) {
708 msg->msg_flags |= MSG_TRUNC;
709 copied = len;
710 }
711
712 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
713 if (err)
714 goto done;
715
716 sock_recv_ts_and_drops(msg, sk, skb);
717
718 /* Copy the address. */
719 if (sin) {
720 sin->sin_family = AF_INET;
721 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
722 sin->sin_port = 0;
723 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
724 }
725 if (inet->cmsg_flags)
726 ip_cmsg_recv(msg, skb);
727 if (flags & MSG_TRUNC)
728 copied = skb->len;
729 done:
730 skb_free_datagram(sk, skb);
731 out:
732 if (err)
733 return err;
734 return copied;
735 }
736
737 static int raw_init(struct sock *sk)
738 {
739 struct raw_sock *rp = raw_sk(sk);
740
741 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
742 memset(&rp->filter, 0, sizeof(rp->filter));
743 return 0;
744 }
745
746 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
747 {
748 if (optlen > sizeof(struct icmp_filter))
749 optlen = sizeof(struct icmp_filter);
750 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
751 return -EFAULT;
752 return 0;
753 }
754
755 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
756 {
757 int len, ret = -EFAULT;
758
759 if (get_user(len, optlen))
760 goto out;
761 ret = -EINVAL;
762 if (len < 0)
763 goto out;
764 if (len > sizeof(struct icmp_filter))
765 len = sizeof(struct icmp_filter);
766 ret = -EFAULT;
767 if (put_user(len, optlen) ||
768 copy_to_user(optval, &raw_sk(sk)->filter, len))
769 goto out;
770 ret = 0;
771 out: return ret;
772 }
773
774 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
775 char __user *optval, unsigned int optlen)
776 {
777 if (optname == ICMP_FILTER) {
778 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
779 return -EOPNOTSUPP;
780 else
781 return raw_seticmpfilter(sk, optval, optlen);
782 }
783 return -ENOPROTOOPT;
784 }
785
786 static int raw_setsockopt(struct sock *sk, int level, int optname,
787 char __user *optval, unsigned int optlen)
788 {
789 if (level != SOL_RAW)
790 return ip_setsockopt(sk, level, optname, optval, optlen);
791 return do_raw_setsockopt(sk, level, optname, optval, optlen);
792 }
793
794 #ifdef CONFIG_COMPAT
795 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
796 char __user *optval, unsigned int optlen)
797 {
798 if (level != SOL_RAW)
799 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
800 return do_raw_setsockopt(sk, level, optname, optval, optlen);
801 }
802 #endif
803
804 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
805 char __user *optval, int __user *optlen)
806 {
807 if (optname == ICMP_FILTER) {
808 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
809 return -EOPNOTSUPP;
810 else
811 return raw_geticmpfilter(sk, optval, optlen);
812 }
813 return -ENOPROTOOPT;
814 }
815
816 static int raw_getsockopt(struct sock *sk, int level, int optname,
817 char __user *optval, int __user *optlen)
818 {
819 if (level != SOL_RAW)
820 return ip_getsockopt(sk, level, optname, optval, optlen);
821 return do_raw_getsockopt(sk, level, optname, optval, optlen);
822 }
823
824 #ifdef CONFIG_COMPAT
825 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
826 char __user *optval, int __user *optlen)
827 {
828 if (level != SOL_RAW)
829 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
830 return do_raw_getsockopt(sk, level, optname, optval, optlen);
831 }
832 #endif
833
834 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
835 {
836 switch (cmd) {
837 case SIOCOUTQ: {
838 int amount = sk_wmem_alloc_get(sk);
839
840 return put_user(amount, (int __user *)arg);
841 }
842 case SIOCINQ: {
843 struct sk_buff *skb;
844 int amount = 0;
845
846 spin_lock_bh(&sk->sk_receive_queue.lock);
847 skb = skb_peek(&sk->sk_receive_queue);
848 if (skb != NULL)
849 amount = skb->len;
850 spin_unlock_bh(&sk->sk_receive_queue.lock);
851 return put_user(amount, (int __user *)arg);
852 }
853
854 default:
855 #ifdef CONFIG_IP_MROUTE
856 return ipmr_ioctl(sk, cmd, (void __user *)arg);
857 #else
858 return -ENOIOCTLCMD;
859 #endif
860 }
861 }
862
863 #ifdef CONFIG_COMPAT
864 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
865 {
866 switch (cmd) {
867 case SIOCOUTQ:
868 case SIOCINQ:
869 return -ENOIOCTLCMD;
870 default:
871 #ifdef CONFIG_IP_MROUTE
872 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
873 #else
874 return -ENOIOCTLCMD;
875 #endif
876 }
877 }
878 #endif
879
880 struct proto raw_prot = {
881 .name = "RAW",
882 .owner = THIS_MODULE,
883 .close = raw_close,
884 .destroy = raw_destroy,
885 .connect = ip4_datagram_connect,
886 .disconnect = udp_disconnect,
887 .ioctl = raw_ioctl,
888 .init = raw_init,
889 .setsockopt = raw_setsockopt,
890 .getsockopt = raw_getsockopt,
891 .sendmsg = raw_sendmsg,
892 .recvmsg = raw_recvmsg,
893 .bind = raw_bind,
894 .backlog_rcv = raw_rcv_skb,
895 .release_cb = ip4_datagram_release_cb,
896 .hash = raw_hash_sk,
897 .unhash = raw_unhash_sk,
898 .obj_size = sizeof(struct raw_sock),
899 .h.raw_hash = &raw_v4_hashinfo,
900 #ifdef CONFIG_COMPAT
901 .compat_setsockopt = compat_raw_setsockopt,
902 .compat_getsockopt = compat_raw_getsockopt,
903 .compat_ioctl = compat_raw_ioctl,
904 #endif
905 };
906
907 #ifdef CONFIG_PROC_FS
908 static struct sock *raw_get_first(struct seq_file *seq)
909 {
910 struct sock *sk;
911 struct raw_iter_state *state = raw_seq_private(seq);
912
913 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
914 ++state->bucket) {
915 sk_for_each(sk, &state->h->ht[state->bucket])
916 if (sock_net(sk) == seq_file_net(seq))
917 goto found;
918 }
919 sk = NULL;
920 found:
921 return sk;
922 }
923
924 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
925 {
926 struct raw_iter_state *state = raw_seq_private(seq);
927
928 do {
929 sk = sk_next(sk);
930 try_again:
931 ;
932 } while (sk && sock_net(sk) != seq_file_net(seq));
933
934 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
935 sk = sk_head(&state->h->ht[state->bucket]);
936 goto try_again;
937 }
938 return sk;
939 }
940
941 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
942 {
943 struct sock *sk = raw_get_first(seq);
944
945 if (sk)
946 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
947 --pos;
948 return pos ? NULL : sk;
949 }
950
951 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
952 {
953 struct raw_iter_state *state = raw_seq_private(seq);
954
955 read_lock(&state->h->lock);
956 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
957 }
958 EXPORT_SYMBOL_GPL(raw_seq_start);
959
960 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
961 {
962 struct sock *sk;
963
964 if (v == SEQ_START_TOKEN)
965 sk = raw_get_first(seq);
966 else
967 sk = raw_get_next(seq, v);
968 ++*pos;
969 return sk;
970 }
971 EXPORT_SYMBOL_GPL(raw_seq_next);
972
973 void raw_seq_stop(struct seq_file *seq, void *v)
974 {
975 struct raw_iter_state *state = raw_seq_private(seq);
976
977 read_unlock(&state->h->lock);
978 }
979 EXPORT_SYMBOL_GPL(raw_seq_stop);
980
981 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
982 {
983 struct inet_sock *inet = inet_sk(sp);
984 __be32 dest = inet->inet_daddr,
985 src = inet->inet_rcv_saddr;
986 __u16 destp = 0,
987 srcp = inet->inet_num;
988
989 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
990 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
991 i, src, srcp, dest, destp, sp->sk_state,
992 sk_wmem_alloc_get(sp),
993 sk_rmem_alloc_get(sp),
994 0, 0L, 0,
995 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
996 0, sock_i_ino(sp),
997 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
998 }
999
1000 static int raw_seq_show(struct seq_file *seq, void *v)
1001 {
1002 if (v == SEQ_START_TOKEN)
1003 seq_printf(seq, " sl local_address rem_address st tx_queue "
1004 "rx_queue tr tm->when retrnsmt uid timeout "
1005 "inode ref pointer drops\n");
1006 else
1007 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1008 return 0;
1009 }
1010
1011 static const struct seq_operations raw_seq_ops = {
1012 .start = raw_seq_start,
1013 .next = raw_seq_next,
1014 .stop = raw_seq_stop,
1015 .show = raw_seq_show,
1016 };
1017
1018 int raw_seq_open(struct inode *ino, struct file *file,
1019 struct raw_hashinfo *h, const struct seq_operations *ops)
1020 {
1021 int err;
1022 struct raw_iter_state *i;
1023
1024 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1025 if (err < 0)
1026 return err;
1027
1028 i = raw_seq_private((struct seq_file *)file->private_data);
1029 i->h = h;
1030 return 0;
1031 }
1032 EXPORT_SYMBOL_GPL(raw_seq_open);
1033
1034 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1035 {
1036 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1037 }
1038
1039 static const struct file_operations raw_seq_fops = {
1040 .owner = THIS_MODULE,
1041 .open = raw_v4_seq_open,
1042 .read = seq_read,
1043 .llseek = seq_lseek,
1044 .release = seq_release_net,
1045 };
1046
1047 static __net_init int raw_init_net(struct net *net)
1048 {
1049 if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops))
1050 return -ENOMEM;
1051
1052 return 0;
1053 }
1054
1055 static __net_exit void raw_exit_net(struct net *net)
1056 {
1057 remove_proc_entry("raw", net->proc_net);
1058 }
1059
1060 static __net_initdata struct pernet_operations raw_net_ops = {
1061 .init = raw_init_net,
1062 .exit = raw_exit_net,
1063 };
1064
1065 int __init raw_proc_init(void)
1066 {
1067 return register_pernet_subsys(&raw_net_ops);
1068 }
1069
1070 void __init raw_proc_exit(void)
1071 {
1072 unregister_pernet_subsys(&raw_net_ops);
1073 }
1074 #endif /* CONFIG_PROC_FS */
Linus' kernel tree