projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ipv4: Handle PMTU in all ICMP error handlers.
[deliverable/linux.git] / net / ipv4 / raw.c
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 struct hlist_node *node;
115
116 sk_for_each_from(sk, node) {
117 struct inet_sock *inet = inet_sk(sk);
118
119 if (net_eq(sock_net(sk), net) && inet->inet_num == num &&
120 !(inet->inet_daddr && inet->inet_daddr != raddr) &&
121 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
122 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
123 goto found; /* gotcha */
124 }
125 sk = NULL;
126 found:
127 return sk;
128 }
129
130 /*
131 * 0 - deliver
132 * 1 - block
133 */
134 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
135 {
136 int type;
137
138 if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
139 return 1;
140
141 type = icmp_hdr(skb)->type;
142 if (type < 32) {
143 __u32 data = raw_sk(sk)->filter.data;
144
145 return ((1 << 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
222 /* Report error on raw socket, if:
223 1. User requested ip_recverr.
224 2. Socket is connected (otherwise the error indication
225 is useless without ip_recverr and error is hard.
226 */
227 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
228 return;
229
230 switch (type) {
231 default:
232 case ICMP_TIME_EXCEEDED:
233 err = EHOSTUNREACH;
234 break;
235 case ICMP_SOURCE_QUENCH:
236 return;
237 case ICMP_PARAMETERPROB:
238 err = EPROTO;
239 harderr = 1;
240 break;
241 case ICMP_DEST_UNREACH:
242 err = EHOSTUNREACH;
243 if (code > NR_ICMP_UNREACH)
244 break;
245 err = icmp_err_convert[code].errno;
246 harderr = icmp_err_convert[code].fatal;
247 if (code == ICMP_FRAG_NEEDED) {
248 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
249 err = EMSGSIZE;
250 }
251 }
252
253 if (inet->recverr) {
254 const struct iphdr *iph = (const struct iphdr *)skb->data;
255 u8 *payload = skb->data + (iph->ihl << 2);
256
257 if (inet->hdrincl)
258 payload = skb->data;
259 ip_icmp_error(sk, skb, err, 0, info, payload);
260 }
261
262 if (inet->recverr || harderr) {
263 sk->sk_err = err;
264 sk->sk_error_report(sk);
265 }
266 }
267
268 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
269 {
270 int hash;
271 struct sock *raw_sk;
272 const struct iphdr *iph;
273 struct net *net;
274
275 hash = protocol & (RAW_HTABLE_SIZE - 1);
276
277 read_lock(&raw_v4_hashinfo.lock);
278 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
279 if (raw_sk != NULL) {
280 iph = (const struct iphdr *)skb->data;
281 net = dev_net(skb->dev);
282
283 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
284 iph->daddr, iph->saddr,
285 skb->dev->ifindex)) != NULL) {
286 raw_err(raw_sk, skb, info);
287 raw_sk = sk_next(raw_sk);
288 iph = (const struct iphdr *)skb->data;
289 }
290 }
291 read_unlock(&raw_v4_hashinfo.lock);
292 }
293
294 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
295 {
296 /* Charge it to the socket. */
297
298 ipv4_pktinfo_prepare(skb);
299 if (sock_queue_rcv_skb(sk, skb) < 0) {
300 kfree_skb(skb);
301 return NET_RX_DROP;
302 }
303
304 return NET_RX_SUCCESS;
305 }
306
307 int raw_rcv(struct sock *sk, struct sk_buff *skb)
308 {
309 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
310 atomic_inc(&sk->sk_drops);
311 kfree_skb(skb);
312 return NET_RX_DROP;
313 }
314 nf_reset(skb);
315
316 skb_push(skb, skb->data - skb_network_header(skb));
317
318 raw_rcv_skb(sk, skb);
319 return 0;
320 }
321
322 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
323 void *from, size_t length,
324 struct rtable **rtp,
325 unsigned int flags)
326 {
327 struct inet_sock *inet = inet_sk(sk);
328 struct net *net = sock_net(sk);
329 struct iphdr *iph;
330 struct sk_buff *skb;
331 unsigned int iphlen;
332 int err;
333 struct rtable *rt = *rtp;
334 int hlen, tlen;
335
336 if (length > rt->dst.dev->mtu) {
337 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
338 rt->dst.dev->mtu);
339 return -EMSGSIZE;
340 }
341 if (flags&MSG_PROBE)
342 goto out;
343
344 hlen = LL_RESERVED_SPACE(rt->dst.dev);
345 tlen = rt->dst.dev->needed_tailroom;
346 skb = sock_alloc_send_skb(sk,
347 length + hlen + tlen + 15,
348 flags & MSG_DONTWAIT, &err);
349 if (skb == NULL)
350 goto error;
351 skb_reserve(skb, hlen);
352
353 skb->priority = sk->sk_priority;
354 skb->mark = sk->sk_mark;
355 skb_dst_set(skb, &rt->dst);
356 *rtp = NULL;
357
358 skb_reset_network_header(skb);
359 iph = ip_hdr(skb);
360 skb_put(skb, length);
361
362 skb->ip_summed = CHECKSUM_NONE;
363
364 skb->transport_header = skb->network_header;
365 err = -EFAULT;
366 if (memcpy_fromiovecend((void *)iph, from, 0, length))
367 goto error_free;
368
369 iphlen = iph->ihl * 4;
370
371 /*
372 * We don't want to modify the ip header, but we do need to
373 * be sure that it won't cause problems later along the network
374 * stack. Specifically we want to make sure that iph->ihl is a
375 * sane value. If ihl points beyond the length of the buffer passed
376 * in, reject the frame as invalid
377 */
378 err = -EINVAL;
379 if (iphlen > length)
380 goto error_free;
381
382 if (iphlen >= sizeof(*iph)) {
383 if (!iph->saddr)
384 iph->saddr = fl4->saddr;
385 iph->check = 0;
386 iph->tot_len = htons(length);
387 if (!iph->id)
388 ip_select_ident(iph, &rt->dst, NULL);
389
390 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
391 }
392 if (iph->protocol == IPPROTO_ICMP)
393 icmp_out_count(net, ((struct icmphdr *)
394 skb_transport_header(skb))->type);
395
396 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT, skb, NULL,
397 rt->dst.dev, dst_output);
398 if (err > 0)
399 err = net_xmit_errno(err);
400 if (err)
401 goto error;
402 out:
403 return 0;
404
405 error_free:
406 kfree_skb(skb);
407 error:
408 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
409 if (err == -ENOBUFS && !inet->recverr)
410 err = 0;
411 return err;
412 }
413
414 static int raw_probe_proto_opt(struct flowi4 *fl4, struct msghdr *msg)
415 {
416 struct iovec *iov;
417 u8 __user *type = NULL;
418 u8 __user *code = NULL;
419 int probed = 0;
420 unsigned int i;
421
422 if (!msg->msg_iov)
423 return 0;
424
425 for (i = 0; i < msg->msg_iovlen; i++) {
426 iov = &msg->msg_iov[i];
427 if (!iov)
428 continue;
429
430 switch (fl4->flowi4_proto) {
431 case IPPROTO_ICMP:
432 /* check if one-byte field is readable or not. */
433 if (iov->iov_base && iov->iov_len < 1)
434 break;
435
436 if (!type) {
437 type = iov->iov_base;
438 /* check if code field is readable or not. */
439 if (iov->iov_len > 1)
440 code = type + 1;
441 } else if (!code)
442 code = iov->iov_base;
443
444 if (type && code) {
445 if (get_user(fl4->fl4_icmp_type, type) ||
446 get_user(fl4->fl4_icmp_code, code))
447 return -EFAULT;
448 probed = 1;
449 }
450 break;
451 default:
452 probed = 1;
453 break;
454 }
455 if (probed)
456 break;
457 }
458 return 0;
459 }
460
461 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
462 size_t len)
463 {
464 struct inet_sock *inet = inet_sk(sk);
465 struct ipcm_cookie ipc;
466 struct rtable *rt = NULL;
467 struct flowi4 fl4;
468 int free = 0;
469 __be32 daddr;
470 __be32 saddr;
471 u8 tos;
472 int err;
473 struct ip_options_data opt_copy;
474
475 err = -EMSGSIZE;
476 if (len > 0xFFFF)
477 goto out;
478
479 /*
480 * Check the flags.
481 */
482
483 err = -EOPNOTSUPP;
484 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
485 goto out; /* compatibility */
486
487 /*
488 * Get and verify the address.
489 */
490
491 if (msg->msg_namelen) {
492 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
493 err = -EINVAL;
494 if (msg->msg_namelen < sizeof(*usin))
495 goto out;
496 if (usin->sin_family != AF_INET) {
497 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
498 __func__, current->comm);
499 err = -EAFNOSUPPORT;
500 if (usin->sin_family)
501 goto out;
502 }
503 daddr = usin->sin_addr.s_addr;
504 /* ANK: I did not forget to get protocol from port field.
505 * I just do not know, who uses this weirdness.
506 * IP_HDRINCL is much more convenient.
507 */
508 } else {
509 err = -EDESTADDRREQ;
510 if (sk->sk_state != TCP_ESTABLISHED)
511 goto out;
512 daddr = inet->inet_daddr;
513 }
514
515 ipc.addr = inet->inet_saddr;
516 ipc.opt = NULL;
517 ipc.tx_flags = 0;
518 ipc.oif = sk->sk_bound_dev_if;
519
520 if (msg->msg_controllen) {
521 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
522 if (err)
523 goto out;
524 if (ipc.opt)
525 free = 1;
526 }
527
528 saddr = ipc.addr;
529 ipc.addr = daddr;
530
531 if (!ipc.opt) {
532 struct ip_options_rcu *inet_opt;
533
534 rcu_read_lock();
535 inet_opt = rcu_dereference(inet->inet_opt);
536 if (inet_opt) {
537 memcpy(&opt_copy, inet_opt,
538 sizeof(*inet_opt) + inet_opt->opt.optlen);
539 ipc.opt = &opt_copy.opt;
540 }
541 rcu_read_unlock();
542 }
543
544 if (ipc.opt) {
545 err = -EINVAL;
546 /* Linux does not mangle headers on raw sockets,
547 * so that IP options + IP_HDRINCL is non-sense.
548 */
549 if (inet->hdrincl)
550 goto done;
551 if (ipc.opt->opt.srr) {
552 if (!daddr)
553 goto done;
554 daddr = ipc.opt->opt.faddr;
555 }
556 }
557 tos = RT_CONN_FLAGS(sk);
558 if (msg->msg_flags & MSG_DONTROUTE)
559 tos |= RTO_ONLINK;
560
561 if (ipv4_is_multicast(daddr)) {
562 if (!ipc.oif)
563 ipc.oif = inet->mc_index;
564 if (!saddr)
565 saddr = inet->mc_addr;
566 } else if (!ipc.oif)
567 ipc.oif = inet->uc_index;
568
569 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
570 RT_SCOPE_UNIVERSE,
571 inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
572 inet_sk_flowi_flags(sk) | FLOWI_FLAG_CAN_SLEEP,
573 daddr, saddr, 0, 0);
574
575 if (!inet->hdrincl) {
576 err = raw_probe_proto_opt(&fl4, msg);
577 if (err)
578 goto done;
579 }
580
581 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
582 rt = ip_route_output_flow(sock_net(sk), &fl4, sk);
583 if (IS_ERR(rt)) {
584 err = PTR_ERR(rt);
585 rt = NULL;
586 goto done;
587 }
588
589 err = -EACCES;
590 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
591 goto done;
592
593 if (msg->msg_flags & MSG_CONFIRM)
594 goto do_confirm;
595 back_from_confirm:
596
597 if (inet->hdrincl)
598 err = raw_send_hdrinc(sk, &fl4, msg->msg_iov, len,
599 &rt, msg->msg_flags);
600
601 else {
602 if (!ipc.addr)
603 ipc.addr = fl4.daddr;
604 lock_sock(sk);
605 err = ip_append_data(sk, &fl4, ip_generic_getfrag,
606 msg->msg_iov, len, 0,
607 &ipc, &rt, msg->msg_flags);
608 if (err)
609 ip_flush_pending_frames(sk);
610 else if (!(msg->msg_flags & MSG_MORE)) {
611 err = ip_push_pending_frames(sk, &fl4);
612 if (err == -ENOBUFS && !inet->recverr)
613 err = 0;
614 }
615 release_sock(sk);
616 }
617 done:
618 if (free)
619 kfree(ipc.opt);
620 ip_rt_put(rt);
621
622 out:
623 if (err < 0)
624 return err;
625 return len;
626
627 do_confirm:
628 dst_confirm(&rt->dst);
629 if (!(msg->msg_flags & MSG_PROBE) || len)
630 goto back_from_confirm;
631 err = 0;
632 goto done;
633 }
634
635 static void raw_close(struct sock *sk, long timeout)
636 {
637 /*
638 * Raw sockets may have direct kernel references. Kill them.
639 */
640 ip_ra_control(sk, 0, NULL);
641
642 sk_common_release(sk);
643 }
644
645 static void raw_destroy(struct sock *sk)
646 {
647 lock_sock(sk);
648 ip_flush_pending_frames(sk);
649 release_sock(sk);
650 }
651
652 /* This gets rid of all the nasties in af_inet. -DaveM */
653 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
654 {
655 struct inet_sock *inet = inet_sk(sk);
656 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
657 int ret = -EINVAL;
658 int chk_addr_ret;
659
660 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
661 goto out;
662 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
663 ret = -EADDRNOTAVAIL;
664 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
665 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
666 goto out;
667 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
668 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
669 inet->inet_saddr = 0; /* Use device */
670 sk_dst_reset(sk);
671 ret = 0;
672 out: return ret;
673 }
674
675 /*
676 * This should be easy, if there is something there
677 * we return it, otherwise we block.
678 */
679
680 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
681 size_t len, int noblock, int flags, int *addr_len)
682 {
683 struct inet_sock *inet = inet_sk(sk);
684 size_t copied = 0;
685 int err = -EOPNOTSUPP;
686 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
687 struct sk_buff *skb;
688
689 if (flags & MSG_OOB)
690 goto out;
691
692 if (addr_len)
693 *addr_len = sizeof(*sin);
694
695 if (flags & MSG_ERRQUEUE) {
696 err = ip_recv_error(sk, msg, len);
697 goto out;
698 }
699
700 skb = skb_recv_datagram(sk, flags, noblock, &err);
701 if (!skb)
702 goto out;
703
704 copied = skb->len;
705 if (len < copied) {
706 msg->msg_flags |= MSG_TRUNC;
707 copied = len;
708 }
709
710 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
711 if (err)
712 goto done;
713
714 sock_recv_ts_and_drops(msg, sk, skb);
715
716 /* Copy the address. */
717 if (sin) {
718 sin->sin_family = AF_INET;
719 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
720 sin->sin_port = 0;
721 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
722 }
723 if (inet->cmsg_flags)
724 ip_cmsg_recv(msg, skb);
725 if (flags & MSG_TRUNC)
726 copied = skb->len;
727 done:
728 skb_free_datagram(sk, skb);
729 out:
730 if (err)
731 return err;
732 return copied;
733 }
734
735 static int raw_init(struct sock *sk)
736 {
737 struct raw_sock *rp = raw_sk(sk);
738
739 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
740 memset(&rp->filter, 0, sizeof(rp->filter));
741 return 0;
742 }
743
744 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
745 {
746 if (optlen > sizeof(struct icmp_filter))
747 optlen = sizeof(struct icmp_filter);
748 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
749 return -EFAULT;
750 return 0;
751 }
752
753 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
754 {
755 int len, ret = -EFAULT;
756
757 if (get_user(len, optlen))
758 goto out;
759 ret = -EINVAL;
760 if (len < 0)
761 goto out;
762 if (len > sizeof(struct icmp_filter))
763 len = sizeof(struct icmp_filter);
764 ret = -EFAULT;
765 if (put_user(len, optlen) ||
766 copy_to_user(optval, &raw_sk(sk)->filter, len))
767 goto out;
768 ret = 0;
769 out: return ret;
770 }
771
772 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
773 char __user *optval, unsigned int optlen)
774 {
775 if (optname == ICMP_FILTER) {
776 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
777 return -EOPNOTSUPP;
778 else
779 return raw_seticmpfilter(sk, optval, optlen);
780 }
781 return -ENOPROTOOPT;
782 }
783
784 static int raw_setsockopt(struct sock *sk, int level, int optname,
785 char __user *optval, unsigned int optlen)
786 {
787 if (level != SOL_RAW)
788 return ip_setsockopt(sk, level, optname, optval, optlen);
789 return do_raw_setsockopt(sk, level, optname, optval, optlen);
790 }
791
792 #ifdef CONFIG_COMPAT
793 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
794 char __user *optval, unsigned int optlen)
795 {
796 if (level != SOL_RAW)
797 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
798 return do_raw_setsockopt(sk, level, optname, optval, optlen);
799 }
800 #endif
801
802 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
803 char __user *optval, int __user *optlen)
804 {
805 if (optname == ICMP_FILTER) {
806 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
807 return -EOPNOTSUPP;
808 else
809 return raw_geticmpfilter(sk, optval, optlen);
810 }
811 return -ENOPROTOOPT;
812 }
813
814 static int raw_getsockopt(struct sock *sk, int level, int optname,
815 char __user *optval, int __user *optlen)
816 {
817 if (level != SOL_RAW)
818 return ip_getsockopt(sk, level, optname, optval, optlen);
819 return do_raw_getsockopt(sk, level, optname, optval, optlen);
820 }
821
822 #ifdef CONFIG_COMPAT
823 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
824 char __user *optval, int __user *optlen)
825 {
826 if (level != SOL_RAW)
827 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
828 return do_raw_getsockopt(sk, level, optname, optval, optlen);
829 }
830 #endif
831
832 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
833 {
834 switch (cmd) {
835 case SIOCOUTQ: {
836 int amount = sk_wmem_alloc_get(sk);
837
838 return put_user(amount, (int __user *)arg);
839 }
840 case SIOCINQ: {
841 struct sk_buff *skb;
842 int amount = 0;
843
844 spin_lock_bh(&sk->sk_receive_queue.lock);
845 skb = skb_peek(&sk->sk_receive_queue);
846 if (skb != NULL)
847 amount = skb->len;
848 spin_unlock_bh(&sk->sk_receive_queue.lock);
849 return put_user(amount, (int __user *)arg);
850 }
851
852 default:
853 #ifdef CONFIG_IP_MROUTE
854 return ipmr_ioctl(sk, cmd, (void __user *)arg);
855 #else
856 return -ENOIOCTLCMD;
857 #endif
858 }
859 }
860
861 #ifdef CONFIG_COMPAT
862 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
863 {
864 switch (cmd) {
865 case SIOCOUTQ:
866 case SIOCINQ:
867 return -ENOIOCTLCMD;
868 default:
869 #ifdef CONFIG_IP_MROUTE
870 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
871 #else
872 return -ENOIOCTLCMD;
873 #endif
874 }
875 }
876 #endif
877
878 struct proto raw_prot = {
879 .name = "RAW",
880 .owner = THIS_MODULE,
881 .close = raw_close,
882 .destroy = raw_destroy,
883 .connect = ip4_datagram_connect,
884 .disconnect = udp_disconnect,
885 .ioctl = raw_ioctl,
886 .init = raw_init,
887 .setsockopt = raw_setsockopt,
888 .getsockopt = raw_getsockopt,
889 .sendmsg = raw_sendmsg,
890 .recvmsg = raw_recvmsg,
891 .bind = raw_bind,
892 .backlog_rcv = raw_rcv_skb,
893 .hash = raw_hash_sk,
894 .unhash = raw_unhash_sk,
895 .obj_size = sizeof(struct raw_sock),
896 .h.raw_hash = &raw_v4_hashinfo,
897 #ifdef CONFIG_COMPAT
898 .compat_setsockopt = compat_raw_setsockopt,
899 .compat_getsockopt = compat_raw_getsockopt,
900 .compat_ioctl = compat_raw_ioctl,
901 #endif
902 };
903
904 #ifdef CONFIG_PROC_FS
905 static struct sock *raw_get_first(struct seq_file *seq)
906 {
907 struct sock *sk;
908 struct raw_iter_state *state = raw_seq_private(seq);
909
910 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
911 ++state->bucket) {
912 struct hlist_node *node;
913
914 sk_for_each(sk, node, &state->h->ht[state->bucket])
915 if (sock_net(sk) == seq_file_net(seq))
916 goto found;
917 }
918 sk = NULL;
919 found:
920 return sk;
921 }
922
923 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
924 {
925 struct raw_iter_state *state = raw_seq_private(seq);
926
927 do {
928 sk = sk_next(sk);
929 try_again:
930 ;
931 } while (sk && sock_net(sk) != seq_file_net(seq));
932
933 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
934 sk = sk_head(&state->h->ht[state->bucket]);
935 goto try_again;
936 }
937 return sk;
938 }
939
940 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
941 {
942 struct sock *sk = raw_get_first(seq);
943
944 if (sk)
945 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
946 --pos;
947 return pos ? NULL : sk;
948 }
949
950 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
951 {
952 struct raw_iter_state *state = raw_seq_private(seq);
953
954 read_lock(&state->h->lock);
955 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
956 }
957 EXPORT_SYMBOL_GPL(raw_seq_start);
958
959 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
960 {
961 struct sock *sk;
962
963 if (v == SEQ_START_TOKEN)
964 sk = raw_get_first(seq);
965 else
966 sk = raw_get_next(seq, v);
967 ++*pos;
968 return sk;
969 }
970 EXPORT_SYMBOL_GPL(raw_seq_next);
971
972 void raw_seq_stop(struct seq_file *seq, void *v)
973 {
974 struct raw_iter_state *state = raw_seq_private(seq);
975
976 read_unlock(&state->h->lock);
977 }
978 EXPORT_SYMBOL_GPL(raw_seq_stop);
979
980 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
981 {
982 struct inet_sock *inet = inet_sk(sp);
983 __be32 dest = inet->inet_daddr,
984 src = inet->inet_rcv_saddr;
985 __u16 destp = 0,
986 srcp = inet->inet_num;
987
988 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
989 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
990 i, src, srcp, dest, destp, sp->sk_state,
991 sk_wmem_alloc_get(sp),
992 sk_rmem_alloc_get(sp),
993 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
994 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
995 }
996
997 static int raw_seq_show(struct seq_file *seq, void *v)
998 {
999 if (v == SEQ_START_TOKEN)
1000 seq_printf(seq, " sl local_address rem_address st tx_queue "
1001 "rx_queue tr tm->when retrnsmt uid timeout "
1002 "inode ref pointer drops\n");
1003 else
1004 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1005 return 0;
1006 }
1007
1008 static const struct seq_operations raw_seq_ops = {
1009 .start = raw_seq_start,
1010 .next = raw_seq_next,
1011 .stop = raw_seq_stop,
1012 .show = raw_seq_show,
1013 };
1014
1015 int raw_seq_open(struct inode *ino, struct file *file,
1016 struct raw_hashinfo *h, const struct seq_operations *ops)
1017 {
1018 int err;
1019 struct raw_iter_state *i;
1020
1021 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1022 if (err < 0)
1023 return err;
1024
1025 i = raw_seq_private((struct seq_file *)file->private_data);
1026 i->h = h;
1027 return 0;
1028 }
1029 EXPORT_SYMBOL_GPL(raw_seq_open);
1030
1031 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1032 {
1033 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1034 }
1035
1036 static const struct file_operations raw_seq_fops = {
1037 .owner = THIS_MODULE,
1038 .open = raw_v4_seq_open,
1039 .read = seq_read,
1040 .llseek = seq_lseek,
1041 .release = seq_release_net,
1042 };
1043
1044 static __net_init int raw_init_net(struct net *net)
1045 {
1046 if (!proc_net_fops_create(net, "raw", S_IRUGO, &raw_seq_fops))
1047 return -ENOMEM;
1048
1049 return 0;
1050 }
1051
1052 static __net_exit void raw_exit_net(struct net *net)
1053 {
1054 proc_net_remove(net, "raw");
1055 }
1056
1057 static __net_initdata struct pernet_operations raw_net_ops = {
1058 .init = raw_init_net,
1059 .exit = raw_exit_net,
1060 };
1061
1062 int __init raw_proc_init(void)
1063 {
1064 return register_pernet_subsys(&raw_net_ops);
1065 }
1066
1067 void __init raw_proc_exit(void)
1068 {
1069 unregister_pernet_subsys(&raw_net_ops);
1070 }
1071 #endif /* CONFIG_PROC_FS */
Linux kernel - Deliverables
This page took 0.054824 seconds and 5 git commands to generate.