?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
dm thin metadata: fix __udivdi3 undefined on 32-bit
[linux/fpc-iii.git] / net / ipv4 / raw.c
blob4d3d4291c82fcd4fa1873d93763b1fc9311f51f2
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/kernel.h>
50 #include <linux/export.h>
51 #include <linux/spinlock.h>
52 #include <linux/sockios.h>
53 #include <linux/socket.h>
54 #include <linux/in.h>
55 #include <linux/mroute.h>
56 #include <linux/netdevice.h>
57 #include <linux/in_route.h>
58 #include <linux/route.h>
59 #include <linux/skbuff.h>
60 #include <linux/igmp.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 #include <linux/uio.h>
82
83 struct raw_frag_vec {
84 struct msghdr *msg;
85 union {
86 struct icmphdr icmph;
87 char c[1];
88 } hdr;
89 int hlen;
90 };
91
92 static struct raw_hashinfo raw_v4_hashinfo = {
93 .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
94 };
95
96 void raw_hash_sk(struct sock *sk)
97 {
98 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
99 struct hlist_head *head;
100
101 head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)];
102
103 write_lock_bh(&h->lock);
104 sk_add_node(sk, head);
105 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
106 write_unlock_bh(&h->lock);
107 }
108 EXPORT_SYMBOL_GPL(raw_hash_sk);
109
110 void raw_unhash_sk(struct sock *sk)
111 {
112 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
113
114 write_lock_bh(&h->lock);
115 if (sk_del_node_init(sk))
116 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
117 write_unlock_bh(&h->lock);
118 }
119 EXPORT_SYMBOL_GPL(raw_unhash_sk);
120
121 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
122 unsigned short num, __be32 raddr, __be32 laddr, int dif)
123 {
124 sk_for_each_from(sk) {
125 struct inet_sock *inet = inet_sk(sk);
126
127 if (net_eq(sock_net(sk), net) && inet->inet_num == num &&
128 !(inet->inet_daddr && inet->inet_daddr != raddr) &&
129 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
130 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
131 goto found; /* gotcha */
132 }
133 sk = NULL;
134 found:
135 return sk;
136 }
137
138 /*
139 * 0 - deliver
140 * 1 - block
141 */
142 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
143 {
144 struct icmphdr _hdr;
145 const struct icmphdr *hdr;
146
147 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
148 sizeof(_hdr), &_hdr);
149 if (!hdr)
150 return 1;
151
152 if (hdr->type < 32) {
153 __u32 data = raw_sk(sk)->filter.data;
154
155 return ((1U << hdr->type) & data) != 0;
156 }
157
158 /* Do not block unknown ICMP types */
159 return 0;
160 }
161
162 /* IP input processing comes here for RAW socket delivery.
163 * Caller owns SKB, so we must make clones.
164 *
165 * RFC 1122: SHOULD pass TOS value up to the transport layer.
166 * -> It does. And not only TOS, but all IP header.
167 */
168 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
169 {
170 struct sock *sk;
171 struct hlist_head *head;
172 int delivered = 0;
173 struct net *net;
174
175 read_lock(&raw_v4_hashinfo.lock);
176 head = &raw_v4_hashinfo.ht[hash];
177 if (hlist_empty(head))
178 goto out;
179
180 net = dev_net(skb->dev);
181 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
182 iph->saddr, iph->daddr,
183 skb->dev->ifindex);
184
185 while (sk) {
186 delivered = 1;
187 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
188 ip_mc_sf_allow(sk, iph->daddr, iph->saddr,
189 skb->dev->ifindex)) {
190 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
191
192 /* Not releasing hash table! */
193 if (clone)
194 raw_rcv(sk, clone);
195 }
196 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
197 iph->saddr, iph->daddr,
198 skb->dev->ifindex);
199 }
200 out:
201 read_unlock(&raw_v4_hashinfo.lock);
202 return delivered;
203 }
204
205 int raw_local_deliver(struct sk_buff *skb, int protocol)
206 {
207 int hash;
208 struct sock *raw_sk;
209
210 hash = protocol & (RAW_HTABLE_SIZE - 1);
211 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
212
213 /* If there maybe a raw socket we must check - if not we
214 * don't care less
215 */
216 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
217 raw_sk = NULL;
218
219 return raw_sk != NULL;
220
221 }
222
223 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
224 {
225 struct inet_sock *inet = inet_sk(sk);
226 const int type = icmp_hdr(skb)->type;
227 const int code = icmp_hdr(skb)->code;
228 int err = 0;
229 int harderr = 0;
230
231 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
232 ipv4_sk_update_pmtu(skb, sk, info);
233 else if (type == ICMP_REDIRECT) {
234 ipv4_sk_redirect(skb, sk);
235 return;
236 }
237
238 /* Report error on raw socket, if:
239 1. User requested ip_recverr.
240 2. Socket is connected (otherwise the error indication
241 is useless without ip_recverr and error is hard.
242 */
243 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
244 return;
245
246 switch (type) {
247 default:
248 case ICMP_TIME_EXCEEDED:
249 err = EHOSTUNREACH;
250 break;
251 case ICMP_SOURCE_QUENCH:
252 return;
253 case ICMP_PARAMETERPROB:
254 err = EPROTO;
255 harderr = 1;
256 break;
257 case ICMP_DEST_UNREACH:
258 err = EHOSTUNREACH;
259 if (code > NR_ICMP_UNREACH)
260 break;
261 err = icmp_err_convert[code].errno;
262 harderr = icmp_err_convert[code].fatal;
263 if (code == ICMP_FRAG_NEEDED) {
264 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
265 err = EMSGSIZE;
266 }
267 }
268
269 if (inet->recverr) {
270 const struct iphdr *iph = (const struct iphdr *)skb->data;
271 u8 *payload = skb->data + (iph->ihl << 2);
272
273 if (inet->hdrincl)
274 payload = skb->data;
275 ip_icmp_error(sk, skb, err, 0, info, payload);
276 }
277
278 if (inet->recverr || harderr) {
279 sk->sk_err = err;
280 sk->sk_error_report(sk);
281 }
282 }
283
284 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
285 {
286 int hash;
287 struct sock *raw_sk;
288 const struct iphdr *iph;
289 struct net *net;
290
291 hash = protocol & (RAW_HTABLE_SIZE - 1);
292
293 read_lock(&raw_v4_hashinfo.lock);
294 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
295 if (raw_sk) {
296 iph = (const struct iphdr *)skb->data;
297 net = dev_net(skb->dev);
298
299 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
300 iph->daddr, iph->saddr,
301 skb->dev->ifindex)) != NULL) {
302 raw_err(raw_sk, skb, info);
303 raw_sk = sk_next(raw_sk);
304 iph = (const struct iphdr *)skb->data;
305 }
306 }
307 read_unlock(&raw_v4_hashinfo.lock);
308 }
309
310 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
311 {
312 /* Charge it to the socket. */
313
314 ipv4_pktinfo_prepare(sk, skb);
315 if (sock_queue_rcv_skb(sk, skb) < 0) {
316 kfree_skb(skb);
317 return NET_RX_DROP;
318 }
319
320 return NET_RX_SUCCESS;
321 }
322
323 int raw_rcv(struct sock *sk, struct sk_buff *skb)
324 {
325 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
326 atomic_inc(&sk->sk_drops);
327 kfree_skb(skb);
328 return NET_RX_DROP;
329 }
330 nf_reset(skb);
331
332 skb_push(skb, skb->data - skb_network_header(skb));
333
334 raw_rcv_skb(sk, skb);
335 return 0;
336 }
337
338 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
339 struct msghdr *msg, size_t length,
340 struct rtable **rtp,
341 unsigned int flags)
342 {
343 struct inet_sock *inet = inet_sk(sk);
344 struct net *net = sock_net(sk);
345 struct iphdr *iph;
346 struct sk_buff *skb;
347 unsigned int iphlen;
348 int err;
349 struct rtable *rt = *rtp;
350 int hlen, tlen;
351
352 if (length > rt->dst.dev->mtu) {
353 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
354 rt->dst.dev->mtu);
355 return -EMSGSIZE;
356 }
357 if (length < sizeof(struct iphdr))
358 return -EINVAL;
359
360 if (flags&MSG_PROBE)
361 goto out;
362
363 hlen = LL_RESERVED_SPACE(rt->dst.dev);
364 tlen = rt->dst.dev->needed_tailroom;
365 skb = sock_alloc_send_skb(sk,
366 length + hlen + tlen + 15,
367 flags & MSG_DONTWAIT, &err);
368 if (!skb)
369 goto error;
370 skb_reserve(skb, hlen);
371
372 skb->priority = sk->sk_priority;
373 skb->mark = sk->sk_mark;
374 skb_dst_set(skb, &rt->dst);
375 *rtp = NULL;
376
377 skb_reset_network_header(skb);
378 iph = ip_hdr(skb);
379 skb_put(skb, length);
380
381 skb->ip_summed = CHECKSUM_NONE;
382
383 sock_tx_timestamp(sk, &skb_shinfo(skb)->tx_flags);
384
385 skb->transport_header = skb->network_header;
386 err = -EFAULT;
387 if (memcpy_from_msg(iph, msg, length))
388 goto error_free;
389
390 iphlen = iph->ihl * 4;
391
392 /*
393 * We don't want to modify the ip header, but we do need to
394 * be sure that it won't cause problems later along the network
395 * stack. Specifically we want to make sure that iph->ihl is a
396 * sane value. If ihl points beyond the length of the buffer passed
397 * in, reject the frame as invalid
398 */
399 err = -EINVAL;
400 if (iphlen > length)
401 goto error_free;
402
403 if (iphlen >= sizeof(*iph)) {
404 if (!iph->saddr)
405 iph->saddr = fl4->saddr;
406 iph->check = 0;
407 iph->tot_len = htons(length);
408 if (!iph->id)
409 ip_select_ident(net, skb, NULL);
410
411 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
412 skb->transport_header += iphlen;
413 if (iph->protocol == IPPROTO_ICMP &&
414 length >= iphlen + sizeof(struct icmphdr))
415 icmp_out_count(net, ((struct icmphdr *)
416 skb_transport_header(skb))->type);
417 }
418
419 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
420 net, sk, skb, NULL, rt->dst.dev,
421 dst_output);
422 if (err > 0)
423 err = net_xmit_errno(err);
424 if (err)
425 goto error;
426 out:
427 return 0;
428
429 error_free:
430 kfree_skb(skb);
431 error:
432 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
433 if (err == -ENOBUFS && !inet->recverr)
434 err = 0;
435 return err;
436 }
437
438 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
439 {
440 int err;
441
442 if (fl4->flowi4_proto != IPPROTO_ICMP)
443 return 0;
444
445 /* We only need the first two bytes. */
446 rfv->hlen = 2;
447
448 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
449 if (err)
450 return err;
451
452 fl4->fl4_icmp_type = rfv->hdr.icmph.type;
453 fl4->fl4_icmp_code = rfv->hdr.icmph.code;
454
455 return 0;
456 }
457
458 static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
459 struct sk_buff *skb)
460 {
461 struct raw_frag_vec *rfv = from;
462
463 if (offset < rfv->hlen) {
464 int copy = min(rfv->hlen - offset, len);
465
466 if (skb->ip_summed == CHECKSUM_PARTIAL)
467 memcpy(to, rfv->hdr.c + offset, copy);
468 else
469 skb->csum = csum_block_add(
470 skb->csum,
471 csum_partial_copy_nocheck(rfv->hdr.c + offset,
472 to, copy, 0),
473 odd);
474
475 odd = 0;
476 offset += copy;
477 to += copy;
478 len -= copy;
479
480 if (!len)
481 return 0;
482 }
483
484 offset -= rfv->hlen;
485
486 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
487 }
488
489 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
490 {
491 struct inet_sock *inet = inet_sk(sk);
492 struct net *net = sock_net(sk);
493 struct ipcm_cookie ipc;
494 struct rtable *rt = NULL;
495 struct flowi4 fl4;
496 int free = 0;
497 __be32 daddr;
498 __be32 saddr;
499 u8 tos;
500 int err;
501 struct ip_options_data opt_copy;
502 struct raw_frag_vec rfv;
503 int hdrincl;
504
505 err = -EMSGSIZE;
506 if (len > 0xFFFF)
507 goto out;
508
509 /* hdrincl should be READ_ONCE(inet->hdrincl)
510 * but READ_ONCE() doesn't work with bit fields
511 */
512 hdrincl = inet->hdrincl;
513 /*
514 * Check the flags.
515 */
516
517 err = -EOPNOTSUPP;
518 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
519 goto out; /* compatibility */
520
521 /*
522 * Get and verify the address.
523 */
524
525 if (msg->msg_namelen) {
526 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
527 err = -EINVAL;
528 if (msg->msg_namelen < sizeof(*usin))
529 goto out;
530 if (usin->sin_family != AF_INET) {
531 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
532 __func__, current->comm);
533 err = -EAFNOSUPPORT;
534 if (usin->sin_family)
535 goto out;
536 }
537 daddr = usin->sin_addr.s_addr;
538 /* ANK: I did not forget to get protocol from port field.
539 * I just do not know, who uses this weirdness.
540 * IP_HDRINCL is much more convenient.
541 */
542 } else {
543 err = -EDESTADDRREQ;
544 if (sk->sk_state != TCP_ESTABLISHED)
545 goto out;
546 daddr = inet->inet_daddr;
547 }
548
549 ipc.addr = inet->inet_saddr;
550 ipc.opt = NULL;
551 ipc.tx_flags = 0;
552 ipc.ttl = 0;
553 ipc.tos = -1;
554 ipc.oif = sk->sk_bound_dev_if;
555
556 if (msg->msg_controllen) {
557 err = ip_cmsg_send(net, msg, &ipc, false);
558 if (unlikely(err)) {
559 kfree(ipc.opt);
560 goto out;
561 }
562 if (ipc.opt)
563 free = 1;
564 }
565
566 saddr = ipc.addr;
567 ipc.addr = daddr;
568
569 if (!ipc.opt) {
570 struct ip_options_rcu *inet_opt;
571
572 rcu_read_lock();
573 inet_opt = rcu_dereference(inet->inet_opt);
574 if (inet_opt) {
575 memcpy(&opt_copy, inet_opt,
576 sizeof(*inet_opt) + inet_opt->opt.optlen);
577 ipc.opt = &opt_copy.opt;
578 }
579 rcu_read_unlock();
580 }
581
582 if (ipc.opt) {
583 err = -EINVAL;
584 /* Linux does not mangle headers on raw sockets,
585 * so that IP options + IP_HDRINCL is non-sense.
586 */
587 if (hdrincl)
588 goto done;
589 if (ipc.opt->opt.srr) {
590 if (!daddr)
591 goto done;
592 daddr = ipc.opt->opt.faddr;
593 }
594 }
595 tos = get_rtconn_flags(&ipc, sk);
596 if (msg->msg_flags & MSG_DONTROUTE)
597 tos |= RTO_ONLINK;
598
599 if (ipv4_is_multicast(daddr)) {
600 if (!ipc.oif)
601 ipc.oif = inet->mc_index;
602 if (!saddr)
603 saddr = inet->mc_addr;
604 } else if (!ipc.oif)
605 ipc.oif = inet->uc_index;
606
607 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
608 RT_SCOPE_UNIVERSE,
609 hdrincl ? IPPROTO_RAW : sk->sk_protocol,
610 inet_sk_flowi_flags(sk) |
611 (hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
612 daddr, saddr, 0, 0);
613
614 if (!saddr && ipc.oif) {
615 err = l3mdev_get_saddr(net, ipc.oif, &fl4);
616 if (err < 0)
617 goto done;
618 }
619
620 if (!hdrincl) {
621 rfv.msg = msg;
622 rfv.hlen = 0;
623
624 err = raw_probe_proto_opt(&rfv, &fl4);
625 if (err)
626 goto done;
627 }
628
629 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
630 rt = ip_route_output_flow(net, &fl4, sk);
631 if (IS_ERR(rt)) {
632 err = PTR_ERR(rt);
633 rt = NULL;
634 goto done;
635 }
636
637 err = -EACCES;
638 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
639 goto done;
640
641 if (msg->msg_flags & MSG_CONFIRM)
642 goto do_confirm;
643 back_from_confirm:
644
645 if (hdrincl)
646 err = raw_send_hdrinc(sk, &fl4, msg, len,
647 &rt, msg->msg_flags);
648
649 else {
650 sock_tx_timestamp(sk, &ipc.tx_flags);
651
652 if (!ipc.addr)
653 ipc.addr = fl4.daddr;
654 lock_sock(sk);
655 err = ip_append_data(sk, &fl4, raw_getfrag,
656 &rfv, len, 0,
657 &ipc, &rt, msg->msg_flags);
658 if (err)
659 ip_flush_pending_frames(sk);
660 else if (!(msg->msg_flags & MSG_MORE)) {
661 err = ip_push_pending_frames(sk, &fl4);
662 if (err == -ENOBUFS && !inet->recverr)
663 err = 0;
664 }
665 release_sock(sk);
666 }
667 done:
668 if (free)
669 kfree(ipc.opt);
670 ip_rt_put(rt);
671
672 out:
673 if (err < 0)
674 return err;
675 return len;
676
677 do_confirm:
678 dst_confirm(&rt->dst);
679 if (!(msg->msg_flags & MSG_PROBE) || len)
680 goto back_from_confirm;
681 err = 0;
682 goto done;
683 }
684
685 static void raw_close(struct sock *sk, long timeout)
686 {
687 /*
688 * Raw sockets may have direct kernel references. Kill them.
689 */
690 ip_ra_control(sk, 0, NULL);
691
692 sk_common_release(sk);
693 }
694
695 static void raw_destroy(struct sock *sk)
696 {
697 lock_sock(sk);
698 ip_flush_pending_frames(sk);
699 release_sock(sk);
700 }
701
702 /* This gets rid of all the nasties in af_inet. -DaveM */
703 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
704 {
705 struct inet_sock *inet = inet_sk(sk);
706 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
707 int ret = -EINVAL;
708 int chk_addr_ret;
709
710 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
711 goto out;
712 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
713 ret = -EADDRNOTAVAIL;
714 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
715 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
716 goto out;
717 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
718 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
719 inet->inet_saddr = 0; /* Use device */
720 sk_dst_reset(sk);
721 ret = 0;
722 out: return ret;
723 }
724
725 /*
726 * This should be easy, if there is something there
727 * we return it, otherwise we block.
728 */
729
730 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
731 int noblock, int flags, int *addr_len)
732 {
733 struct inet_sock *inet = inet_sk(sk);
734 size_t copied = 0;
735 int err = -EOPNOTSUPP;
736 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
737 struct sk_buff *skb;
738
739 if (flags & MSG_OOB)
740 goto out;
741
742 if (flags & MSG_ERRQUEUE) {
743 err = ip_recv_error(sk, msg, len, addr_len);
744 goto out;
745 }
746
747 skb = skb_recv_datagram(sk, flags, noblock, &err);
748 if (!skb)
749 goto out;
750
751 copied = skb->len;
752 if (len < copied) {
753 msg->msg_flags |= MSG_TRUNC;
754 copied = len;
755 }
756
757 err = skb_copy_datagram_msg(skb, 0, msg, copied);
758 if (err)
759 goto done;
760
761 sock_recv_ts_and_drops(msg, sk, skb);
762
763 /* Copy the address. */
764 if (sin) {
765 sin->sin_family = AF_INET;
766 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
767 sin->sin_port = 0;
768 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
769 *addr_len = sizeof(*sin);
770 }
771 if (inet->cmsg_flags)
772 ip_cmsg_recv(msg, skb);
773 if (flags & MSG_TRUNC)
774 copied = skb->len;
775 done:
776 skb_free_datagram(sk, skb);
777 out:
778 if (err)
779 return err;
780 return copied;
781 }
782
783 static int raw_init(struct sock *sk)
784 {
785 struct raw_sock *rp = raw_sk(sk);
786
787 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
788 memset(&rp->filter, 0, sizeof(rp->filter));
789 return 0;
790 }
791
792 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
793 {
794 if (optlen > sizeof(struct icmp_filter))
795 optlen = sizeof(struct icmp_filter);
796 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
797 return -EFAULT;
798 return 0;
799 }
800
801 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
802 {
803 int len, ret = -EFAULT;
804
805 if (get_user(len, optlen))
806 goto out;
807 ret = -EINVAL;
808 if (len < 0)
809 goto out;
810 if (len > sizeof(struct icmp_filter))
811 len = sizeof(struct icmp_filter);
812 ret = -EFAULT;
813 if (put_user(len, optlen) ||
814 copy_to_user(optval, &raw_sk(sk)->filter, len))
815 goto out;
816 ret = 0;
817 out: return ret;
818 }
819
820 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
821 char __user *optval, unsigned int optlen)
822 {
823 if (optname == ICMP_FILTER) {
824 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
825 return -EOPNOTSUPP;
826 else
827 return raw_seticmpfilter(sk, optval, optlen);
828 }
829 return -ENOPROTOOPT;
830 }
831
832 static int raw_setsockopt(struct sock *sk, int level, int optname,
833 char __user *optval, unsigned int optlen)
834 {
835 if (level != SOL_RAW)
836 return ip_setsockopt(sk, level, optname, optval, optlen);
837 return do_raw_setsockopt(sk, level, optname, optval, optlen);
838 }
839
840 #ifdef CONFIG_COMPAT
841 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
842 char __user *optval, unsigned int optlen)
843 {
844 if (level != SOL_RAW)
845 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
846 return do_raw_setsockopt(sk, level, optname, optval, optlen);
847 }
848 #endif
849
850 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
851 char __user *optval, int __user *optlen)
852 {
853 if (optname == ICMP_FILTER) {
854 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
855 return -EOPNOTSUPP;
856 else
857 return raw_geticmpfilter(sk, optval, optlen);
858 }
859 return -ENOPROTOOPT;
860 }
861
862 static int raw_getsockopt(struct sock *sk, int level, int optname,
863 char __user *optval, int __user *optlen)
864 {
865 if (level != SOL_RAW)
866 return ip_getsockopt(sk, level, optname, optval, optlen);
867 return do_raw_getsockopt(sk, level, optname, optval, optlen);
868 }
869
870 #ifdef CONFIG_COMPAT
871 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
872 char __user *optval, int __user *optlen)
873 {
874 if (level != SOL_RAW)
875 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
876 return do_raw_getsockopt(sk, level, optname, optval, optlen);
877 }
878 #endif
879
880 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
881 {
882 switch (cmd) {
883 case SIOCOUTQ: {
884 int amount = sk_wmem_alloc_get(sk);
885
886 return put_user(amount, (int __user *)arg);
887 }
888 case SIOCINQ: {
889 struct sk_buff *skb;
890 int amount = 0;
891
892 spin_lock_bh(&sk->sk_receive_queue.lock);
893 skb = skb_peek(&sk->sk_receive_queue);
894 if (skb)
895 amount = skb->len;
896 spin_unlock_bh(&sk->sk_receive_queue.lock);
897 return put_user(amount, (int __user *)arg);
898 }
899
900 default:
901 #ifdef CONFIG_IP_MROUTE
902 return ipmr_ioctl(sk, cmd, (void __user *)arg);
903 #else
904 return -ENOIOCTLCMD;
905 #endif
906 }
907 }
908
909 #ifdef CONFIG_COMPAT
910 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
911 {
912 switch (cmd) {
913 case SIOCOUTQ:
914 case SIOCINQ:
915 return -ENOIOCTLCMD;
916 default:
917 #ifdef CONFIG_IP_MROUTE
918 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
919 #else
920 return -ENOIOCTLCMD;
921 #endif
922 }
923 }
924 #endif
925
926 struct proto raw_prot = {
927 .name = "RAW",
928 .owner = THIS_MODULE,
929 .close = raw_close,
930 .destroy = raw_destroy,
931 .connect = ip4_datagram_connect,
932 .disconnect = udp_disconnect,
933 .ioctl = raw_ioctl,
934 .init = raw_init,
935 .setsockopt = raw_setsockopt,
936 .getsockopt = raw_getsockopt,
937 .sendmsg = raw_sendmsg,
938 .recvmsg = raw_recvmsg,
939 .bind = raw_bind,
940 .backlog_rcv = raw_rcv_skb,
941 .release_cb = ip4_datagram_release_cb,
942 .hash = raw_hash_sk,
943 .unhash = raw_unhash_sk,
944 .obj_size = sizeof(struct raw_sock),
945 .h.raw_hash = &raw_v4_hashinfo,
946 #ifdef CONFIG_COMPAT
947 .compat_setsockopt = compat_raw_setsockopt,
948 .compat_getsockopt = compat_raw_getsockopt,
949 .compat_ioctl = compat_raw_ioctl,
950 #endif
951 };
952
953 #ifdef CONFIG_PROC_FS
954 static struct sock *raw_get_first(struct seq_file *seq)
955 {
956 struct sock *sk;
957 struct raw_iter_state *state = raw_seq_private(seq);
958
959 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
960 ++state->bucket) {
961 sk_for_each(sk, &state->h->ht[state->bucket])
962 if (sock_net(sk) == seq_file_net(seq))
963 goto found;
964 }
965 sk = NULL;
966 found:
967 return sk;
968 }
969
970 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
971 {
972 struct raw_iter_state *state = raw_seq_private(seq);
973
974 do {
975 sk = sk_next(sk);
976 try_again:
977 ;
978 } while (sk && sock_net(sk) != seq_file_net(seq));
979
980 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
981 sk = sk_head(&state->h->ht[state->bucket]);
982 goto try_again;
983 }
984 return sk;
985 }
986
987 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
988 {
989 struct sock *sk = raw_get_first(seq);
990
991 if (sk)
992 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
993 --pos;
994 return pos ? NULL : sk;
995 }
996
997 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
998 {
999 struct raw_iter_state *state = raw_seq_private(seq);
1000
1001 read_lock(&state->h->lock);
1002 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1003 }
1004 EXPORT_SYMBOL_GPL(raw_seq_start);
1005
1006 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1007 {
1008 struct sock *sk;
1009
1010 if (v == SEQ_START_TOKEN)
1011 sk = raw_get_first(seq);
1012 else
1013 sk = raw_get_next(seq, v);
1014 ++*pos;
1015 return sk;
1016 }
1017 EXPORT_SYMBOL_GPL(raw_seq_next);
1018
1019 void raw_seq_stop(struct seq_file *seq, void *v)
1020 {
1021 struct raw_iter_state *state = raw_seq_private(seq);
1022
1023 read_unlock(&state->h->lock);
1024 }
1025 EXPORT_SYMBOL_GPL(raw_seq_stop);
1026
1027 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1028 {
1029 struct inet_sock *inet = inet_sk(sp);
1030 __be32 dest = inet->inet_daddr,
1031 src = inet->inet_rcv_saddr;
1032 __u16 destp = 0,
1033 srcp = inet->inet_num;
1034
1035 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1036 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1037 i, src, srcp, dest, destp, sp->sk_state,
1038 sk_wmem_alloc_get(sp),
1039 sk_rmem_alloc_get(sp),
1040 0, 0L, 0,
1041 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1042 0, sock_i_ino(sp),
1043 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1044 }
1045
1046 static int raw_seq_show(struct seq_file *seq, void *v)
1047 {
1048 if (v == SEQ_START_TOKEN)
1049 seq_printf(seq, " sl local_address rem_address st tx_queue "
1050 "rx_queue tr tm->when retrnsmt uid timeout "
1051 "inode ref pointer drops\n");
1052 else
1053 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1054 return 0;
1055 }
1056
1057 static const struct seq_operations raw_seq_ops = {
1058 .start = raw_seq_start,
1059 .next = raw_seq_next,
1060 .stop = raw_seq_stop,
1061 .show = raw_seq_show,
1062 };
1063
1064 int raw_seq_open(struct inode *ino, struct file *file,
1065 struct raw_hashinfo *h, const struct seq_operations *ops)
1066 {
1067 int err;
1068 struct raw_iter_state *i;
1069
1070 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1071 if (err < 0)
1072 return err;
1073
1074 i = raw_seq_private((struct seq_file *)file->private_data);
1075 i->h = h;
1076 return 0;
1077 }
1078 EXPORT_SYMBOL_GPL(raw_seq_open);
1079
1080 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1081 {
1082 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1083 }
1084
1085 static const struct file_operations raw_seq_fops = {
1086 .owner = THIS_MODULE,
1087 .open = raw_v4_seq_open,
1088 .read = seq_read,
1089 .llseek = seq_lseek,
1090 .release = seq_release_net,
1091 };
1092
1093 static __net_init int raw_init_net(struct net *net)
1094 {
1095 if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops))
1096 return -ENOMEM;
1097
1098 return 0;
1099 }
1100
1101 static __net_exit void raw_exit_net(struct net *net)
1102 {
1103 remove_proc_entry("raw", net->proc_net);
1104 }
1105
1106 static __net_initdata struct pernet_operations raw_net_ops = {
1107 .init = raw_init_net,
1108 .exit = raw_exit_net,
1109 };
1110
1111 int __init raw_proc_init(void)
1112 {
1113 return register_pernet_subsys(&raw_net_ops);
1114 }
1115
1116 void __init raw_proc_exit(void)
1117 {
1118 unregister_pernet_subsys(&raw_net_ops);
1119 }
1120 #endif /* CONFIG_PROC_FS */
Linux with added FPC-III support