?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
netrom: fix locking in nr_find_socket()
[linux/fpc-iii.git] / net / ipv4 / raw.c
blob59d8770055ed08ac6c0bb86717a8be8beaa04228
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 int 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 return 0;
109 }
110 EXPORT_SYMBOL_GPL(raw_hash_sk);
111
112 void raw_unhash_sk(struct sock *sk)
113 {
114 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
115
116 write_lock_bh(&h->lock);
117 if (sk_del_node_init(sk))
118 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
119 write_unlock_bh(&h->lock);
120 }
121 EXPORT_SYMBOL_GPL(raw_unhash_sk);
122
123 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
124 unsigned short num, __be32 raddr, __be32 laddr, int dif)
125 {
126 sk_for_each_from(sk) {
127 struct inet_sock *inet = inet_sk(sk);
128
129 if (net_eq(sock_net(sk), net) && inet->inet_num == num &&
130 !(inet->inet_daddr && inet->inet_daddr != raddr) &&
131 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
132 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
133 goto found; /* gotcha */
134 }
135 sk = NULL;
136 found:
137 return sk;
138 }
139
140 /*
141 * 0 - deliver
142 * 1 - block
143 */
144 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
145 {
146 struct icmphdr _hdr;
147 const struct icmphdr *hdr;
148
149 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
150 sizeof(_hdr), &_hdr);
151 if (!hdr)
152 return 1;
153
154 if (hdr->type < 32) {
155 __u32 data = raw_sk(sk)->filter.data;
156
157 return ((1U << hdr->type) & data) != 0;
158 }
159
160 /* Do not block unknown ICMP types */
161 return 0;
162 }
163
164 /* IP input processing comes here for RAW socket delivery.
165 * Caller owns SKB, so we must make clones.
166 *
167 * RFC 1122: SHOULD pass TOS value up to the transport layer.
168 * -> It does. And not only TOS, but all IP header.
169 */
170 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
171 {
172 struct sock *sk;
173 struct hlist_head *head;
174 int delivered = 0;
175 struct net *net;
176
177 read_lock(&raw_v4_hashinfo.lock);
178 head = &raw_v4_hashinfo.ht[hash];
179 if (hlist_empty(head))
180 goto out;
181
182 net = dev_net(skb->dev);
183 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
184 iph->saddr, iph->daddr,
185 skb->dev->ifindex);
186
187 while (sk) {
188 delivered = 1;
189 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
190 ip_mc_sf_allow(sk, iph->daddr, iph->saddr,
191 skb->dev->ifindex)) {
192 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
193
194 /* Not releasing hash table! */
195 if (clone)
196 raw_rcv(sk, clone);
197 }
198 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
199 iph->saddr, iph->daddr,
200 skb->dev->ifindex);
201 }
202 out:
203 read_unlock(&raw_v4_hashinfo.lock);
204 return delivered;
205 }
206
207 int raw_local_deliver(struct sk_buff *skb, int protocol)
208 {
209 int hash;
210 struct sock *raw_sk;
211
212 hash = protocol & (RAW_HTABLE_SIZE - 1);
213 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
214
215 /* If there maybe a raw socket we must check - if not we
216 * don't care less
217 */
218 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
219 raw_sk = NULL;
220
221 return raw_sk != NULL;
222
223 }
224
225 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
226 {
227 struct inet_sock *inet = inet_sk(sk);
228 const int type = icmp_hdr(skb)->type;
229 const int code = icmp_hdr(skb)->code;
230 int err = 0;
231 int harderr = 0;
232
233 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
234 ipv4_sk_update_pmtu(skb, sk, info);
235 else if (type == ICMP_REDIRECT) {
236 ipv4_sk_redirect(skb, sk);
237 return;
238 }
239
240 /* Report error on raw socket, if:
241 1. User requested ip_recverr.
242 2. Socket is connected (otherwise the error indication
243 is useless without ip_recverr and error is hard.
244 */
245 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
246 return;
247
248 switch (type) {
249 default:
250 case ICMP_TIME_EXCEEDED:
251 err = EHOSTUNREACH;
252 break;
253 case ICMP_SOURCE_QUENCH:
254 return;
255 case ICMP_PARAMETERPROB:
256 err = EPROTO;
257 harderr = 1;
258 break;
259 case ICMP_DEST_UNREACH:
260 err = EHOSTUNREACH;
261 if (code > NR_ICMP_UNREACH)
262 break;
263 err = icmp_err_convert[code].errno;
264 harderr = icmp_err_convert[code].fatal;
265 if (code == ICMP_FRAG_NEEDED) {
266 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
267 err = EMSGSIZE;
268 }
269 }
270
271 if (inet->recverr) {
272 const struct iphdr *iph = (const struct iphdr *)skb->data;
273 u8 *payload = skb->data + (iph->ihl << 2);
274
275 if (inet->hdrincl)
276 payload = skb->data;
277 ip_icmp_error(sk, skb, err, 0, info, payload);
278 }
279
280 if (inet->recverr || harderr) {
281 sk->sk_err = err;
282 sk->sk_error_report(sk);
283 }
284 }
285
286 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
287 {
288 int hash;
289 struct sock *raw_sk;
290 const struct iphdr *iph;
291 struct net *net;
292
293 hash = protocol & (RAW_HTABLE_SIZE - 1);
294
295 read_lock(&raw_v4_hashinfo.lock);
296 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
297 if (raw_sk) {
298 iph = (const struct iphdr *)skb->data;
299 net = dev_net(skb->dev);
300
301 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
302 iph->daddr, iph->saddr,
303 skb->dev->ifindex)) != NULL) {
304 raw_err(raw_sk, skb, info);
305 raw_sk = sk_next(raw_sk);
306 iph = (const struct iphdr *)skb->data;
307 }
308 }
309 read_unlock(&raw_v4_hashinfo.lock);
310 }
311
312 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
313 {
314 /* Charge it to the socket. */
315
316 ipv4_pktinfo_prepare(sk, skb);
317 if (sock_queue_rcv_skb(sk, skb) < 0) {
318 kfree_skb(skb);
319 return NET_RX_DROP;
320 }
321
322 return NET_RX_SUCCESS;
323 }
324
325 int raw_rcv(struct sock *sk, struct sk_buff *skb)
326 {
327 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
328 atomic_inc(&sk->sk_drops);
329 kfree_skb(skb);
330 return NET_RX_DROP;
331 }
332 nf_reset(skb);
333
334 skb_push(skb, skb->data - skb_network_header(skb));
335
336 raw_rcv_skb(sk, skb);
337 return 0;
338 }
339
340 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
341 struct msghdr *msg, size_t length,
342 struct rtable **rtp, unsigned int flags,
343 const struct sockcm_cookie *sockc)
344 {
345 struct inet_sock *inet = inet_sk(sk);
346 struct net *net = sock_net(sk);
347 struct iphdr *iph;
348 struct sk_buff *skb;
349 unsigned int iphlen;
350 int err;
351 struct rtable *rt = *rtp;
352 int hlen, tlen;
353
354 if (length > rt->dst.dev->mtu) {
355 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
356 rt->dst.dev->mtu);
357 return -EMSGSIZE;
358 }
359 if (length < sizeof(struct iphdr))
360 return -EINVAL;
361
362 if (flags&MSG_PROBE)
363 goto out;
364
365 hlen = LL_RESERVED_SPACE(rt->dst.dev);
366 tlen = rt->dst.dev->needed_tailroom;
367 skb = sock_alloc_send_skb(sk,
368 length + hlen + tlen + 15,
369 flags & MSG_DONTWAIT, &err);
370 if (!skb)
371 goto error;
372 skb_reserve(skb, hlen);
373
374 skb->priority = sk->sk_priority;
375 skb->mark = sk->sk_mark;
376 skb_dst_set(skb, &rt->dst);
377 *rtp = NULL;
378
379 skb_reset_network_header(skb);
380 iph = ip_hdr(skb);
381 skb_put(skb, length);
382
383 skb->ip_summed = CHECKSUM_NONE;
384
385 sock_tx_timestamp(sk, sockc->tsflags, &skb_shinfo(skb)->tx_flags);
386
387 skb->transport_header = skb->network_header;
388 err = -EFAULT;
389 if (memcpy_from_msg(iph, msg, length))
390 goto error_free;
391
392 iphlen = iph->ihl * 4;
393
394 /*
395 * We don't want to modify the ip header, but we do need to
396 * be sure that it won't cause problems later along the network
397 * stack. Specifically we want to make sure that iph->ihl is a
398 * sane value. If ihl points beyond the length of the buffer passed
399 * in, reject the frame as invalid
400 */
401 err = -EINVAL;
402 if (iphlen > length)
403 goto error_free;
404
405 if (iphlen >= sizeof(*iph)) {
406 if (!iph->saddr)
407 iph->saddr = fl4->saddr;
408 iph->check = 0;
409 iph->tot_len = htons(length);
410 if (!iph->id)
411 ip_select_ident(net, skb, NULL);
412
413 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
414 skb->transport_header += iphlen;
415 if (iph->protocol == IPPROTO_ICMP &&
416 length >= iphlen + sizeof(struct icmphdr))
417 icmp_out_count(net, ((struct icmphdr *)
418 skb_transport_header(skb))->type);
419 }
420
421 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
422 net, sk, skb, NULL, rt->dst.dev,
423 dst_output);
424 if (err > 0)
425 err = net_xmit_errno(err);
426 if (err)
427 goto error;
428 out:
429 return 0;
430
431 error_free:
432 kfree_skb(skb);
433 error:
434 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
435 if (err == -ENOBUFS && !inet->recverr)
436 err = 0;
437 return err;
438 }
439
440 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
441 {
442 int err;
443
444 if (fl4->flowi4_proto != IPPROTO_ICMP)
445 return 0;
446
447 /* We only need the first two bytes. */
448 rfv->hlen = 2;
449
450 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
451 if (err)
452 return err;
453
454 fl4->fl4_icmp_type = rfv->hdr.icmph.type;
455 fl4->fl4_icmp_code = rfv->hdr.icmph.code;
456
457 return 0;
458 }
459
460 static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
461 struct sk_buff *skb)
462 {
463 struct raw_frag_vec *rfv = from;
464
465 if (offset < rfv->hlen) {
466 int copy = min(rfv->hlen - offset, len);
467
468 if (skb->ip_summed == CHECKSUM_PARTIAL)
469 memcpy(to, rfv->hdr.c + offset, copy);
470 else
471 skb->csum = csum_block_add(
472 skb->csum,
473 csum_partial_copy_nocheck(rfv->hdr.c + offset,
474 to, copy, 0),
475 odd);
476
477 odd = 0;
478 offset += copy;
479 to += copy;
480 len -= copy;
481
482 if (!len)
483 return 0;
484 }
485
486 offset -= rfv->hlen;
487
488 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
489 }
490
491 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
492 {
493 struct inet_sock *inet = inet_sk(sk);
494 struct net *net = sock_net(sk);
495 struct ipcm_cookie ipc;
496 struct rtable *rt = NULL;
497 struct flowi4 fl4;
498 int free = 0;
499 __be32 daddr;
500 __be32 saddr;
501 u8 tos;
502 int err;
503 struct ip_options_data opt_copy;
504 struct raw_frag_vec rfv;
505 int hdrincl;
506
507 err = -EMSGSIZE;
508 if (len > 0xFFFF)
509 goto out;
510
511 /* hdrincl should be READ_ONCE(inet->hdrincl)
512 * but READ_ONCE() doesn't work with bit fields
513 */
514 hdrincl = inet->hdrincl;
515 /*
516 * Check the flags.
517 */
518
519 err = -EOPNOTSUPP;
520 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
521 goto out; /* compatibility */
522
523 /*
524 * Get and verify the address.
525 */
526
527 if (msg->msg_namelen) {
528 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
529 err = -EINVAL;
530 if (msg->msg_namelen < sizeof(*usin))
531 goto out;
532 if (usin->sin_family != AF_INET) {
533 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
534 __func__, current->comm);
535 err = -EAFNOSUPPORT;
536 if (usin->sin_family)
537 goto out;
538 }
539 daddr = usin->sin_addr.s_addr;
540 /* ANK: I did not forget to get protocol from port field.
541 * I just do not know, who uses this weirdness.
542 * IP_HDRINCL is much more convenient.
543 */
544 } else {
545 err = -EDESTADDRREQ;
546 if (sk->sk_state != TCP_ESTABLISHED)
547 goto out;
548 daddr = inet->inet_daddr;
549 }
550
551 ipc.sockc.tsflags = sk->sk_tsflags;
552 ipc.addr = inet->inet_saddr;
553 ipc.opt = NULL;
554 ipc.tx_flags = 0;
555 ipc.ttl = 0;
556 ipc.tos = -1;
557 ipc.oif = sk->sk_bound_dev_if;
558
559 if (msg->msg_controllen) {
560 err = ip_cmsg_send(sk, msg, &ipc, false);
561 if (unlikely(err)) {
562 kfree(ipc.opt);
563 goto out;
564 }
565 if (ipc.opt)
566 free = 1;
567 }
568
569 saddr = ipc.addr;
570 ipc.addr = daddr;
571
572 if (!ipc.opt) {
573 struct ip_options_rcu *inet_opt;
574
575 rcu_read_lock();
576 inet_opt = rcu_dereference(inet->inet_opt);
577 if (inet_opt) {
578 memcpy(&opt_copy, inet_opt,
579 sizeof(*inet_opt) + inet_opt->opt.optlen);
580 ipc.opt = &opt_copy.opt;
581 }
582 rcu_read_unlock();
583 }
584
585 if (ipc.opt) {
586 err = -EINVAL;
587 /* Linux does not mangle headers on raw sockets,
588 * so that IP options + IP_HDRINCL is non-sense.
589 */
590 if (hdrincl)
591 goto done;
592 if (ipc.opt->opt.srr) {
593 if (!daddr)
594 goto done;
595 daddr = ipc.opt->opt.faddr;
596 }
597 }
598 tos = get_rtconn_flags(&ipc, sk);
599 if (msg->msg_flags & MSG_DONTROUTE)
600 tos |= RTO_ONLINK;
601
602 if (ipv4_is_multicast(daddr)) {
603 if (!ipc.oif)
604 ipc.oif = inet->mc_index;
605 if (!saddr)
606 saddr = inet->mc_addr;
607 } else if (!ipc.oif)
608 ipc.oif = inet->uc_index;
609
610 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
611 RT_SCOPE_UNIVERSE,
612 hdrincl ? IPPROTO_RAW : sk->sk_protocol,
613 inet_sk_flowi_flags(sk) |
614 (hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
615 daddr, saddr, 0, 0);
616
617 if (!hdrincl) {
618 rfv.msg = msg;
619 rfv.hlen = 0;
620
621 err = raw_probe_proto_opt(&rfv, &fl4);
622 if (err)
623 goto done;
624 }
625
626 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
627 rt = ip_route_output_flow(net, &fl4, sk);
628 if (IS_ERR(rt)) {
629 err = PTR_ERR(rt);
630 rt = NULL;
631 goto done;
632 }
633
634 err = -EACCES;
635 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
636 goto done;
637
638 if (msg->msg_flags & MSG_CONFIRM)
639 goto do_confirm;
640 back_from_confirm:
641
642 if (hdrincl)
643 err = raw_send_hdrinc(sk, &fl4, msg, len,
644 &rt, msg->msg_flags, &ipc.sockc);
645
646 else {
647 sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
648
649 if (!ipc.addr)
650 ipc.addr = fl4.daddr;
651 lock_sock(sk);
652 err = ip_append_data(sk, &fl4, raw_getfrag,
653 &rfv, len, 0,
654 &ipc, &rt, msg->msg_flags);
655 if (err)
656 ip_flush_pending_frames(sk);
657 else if (!(msg->msg_flags & MSG_MORE)) {
658 err = ip_push_pending_frames(sk, &fl4);
659 if (err == -ENOBUFS && !inet->recverr)
660 err = 0;
661 }
662 release_sock(sk);
663 }
664 done:
665 if (free)
666 kfree(ipc.opt);
667 ip_rt_put(rt);
668
669 out:
670 if (err < 0)
671 return err;
672 return len;
673
674 do_confirm:
675 dst_confirm(&rt->dst);
676 if (!(msg->msg_flags & MSG_PROBE) || len)
677 goto back_from_confirm;
678 err = 0;
679 goto done;
680 }
681
682 static void raw_close(struct sock *sk, long timeout)
683 {
684 /*
685 * Raw sockets may have direct kernel references. Kill them.
686 */
687 ip_ra_control(sk, 0, NULL);
688
689 sk_common_release(sk);
690 }
691
692 static void raw_destroy(struct sock *sk)
693 {
694 lock_sock(sk);
695 ip_flush_pending_frames(sk);
696 release_sock(sk);
697 }
698
699 /* This gets rid of all the nasties in af_inet. -DaveM */
700 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
701 {
702 struct inet_sock *inet = inet_sk(sk);
703 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
704 int ret = -EINVAL;
705 int chk_addr_ret;
706
707 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
708 goto out;
709 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
710 ret = -EADDRNOTAVAIL;
711 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
712 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
713 goto out;
714 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
715 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
716 inet->inet_saddr = 0; /* Use device */
717 sk_dst_reset(sk);
718 ret = 0;
719 out: return ret;
720 }
721
722 /*
723 * This should be easy, if there is something there
724 * we return it, otherwise we block.
725 */
726
727 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
728 int noblock, int flags, int *addr_len)
729 {
730 struct inet_sock *inet = inet_sk(sk);
731 size_t copied = 0;
732 int err = -EOPNOTSUPP;
733 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
734 struct sk_buff *skb;
735
736 if (flags & MSG_OOB)
737 goto out;
738
739 if (flags & MSG_ERRQUEUE) {
740 err = ip_recv_error(sk, msg, len, addr_len);
741 goto out;
742 }
743
744 skb = skb_recv_datagram(sk, flags, noblock, &err);
745 if (!skb)
746 goto out;
747
748 copied = skb->len;
749 if (len < copied) {
750 msg->msg_flags |= MSG_TRUNC;
751 copied = len;
752 }
753
754 err = skb_copy_datagram_msg(skb, 0, msg, copied);
755 if (err)
756 goto done;
757
758 sock_recv_ts_and_drops(msg, sk, skb);
759
760 /* Copy the address. */
761 if (sin) {
762 sin->sin_family = AF_INET;
763 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
764 sin->sin_port = 0;
765 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
766 *addr_len = sizeof(*sin);
767 }
768 if (inet->cmsg_flags)
769 ip_cmsg_recv(msg, skb);
770 if (flags & MSG_TRUNC)
771 copied = skb->len;
772 done:
773 skb_free_datagram(sk, skb);
774 out:
775 if (err)
776 return err;
777 return copied;
778 }
779
780 static int raw_init(struct sock *sk)
781 {
782 struct raw_sock *rp = raw_sk(sk);
783
784 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
785 memset(&rp->filter, 0, sizeof(rp->filter));
786 return 0;
787 }
788
789 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
790 {
791 if (optlen > sizeof(struct icmp_filter))
792 optlen = sizeof(struct icmp_filter);
793 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
794 return -EFAULT;
795 return 0;
796 }
797
798 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
799 {
800 int len, ret = -EFAULT;
801
802 if (get_user(len, optlen))
803 goto out;
804 ret = -EINVAL;
805 if (len < 0)
806 goto out;
807 if (len > sizeof(struct icmp_filter))
808 len = sizeof(struct icmp_filter);
809 ret = -EFAULT;
810 if (put_user(len, optlen) ||
811 copy_to_user(optval, &raw_sk(sk)->filter, len))
812 goto out;
813 ret = 0;
814 out: return ret;
815 }
816
817 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
818 char __user *optval, unsigned int optlen)
819 {
820 if (optname == ICMP_FILTER) {
821 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
822 return -EOPNOTSUPP;
823 else
824 return raw_seticmpfilter(sk, optval, optlen);
825 }
826 return -ENOPROTOOPT;
827 }
828
829 static int raw_setsockopt(struct sock *sk, int level, int optname,
830 char __user *optval, unsigned int optlen)
831 {
832 if (level != SOL_RAW)
833 return ip_setsockopt(sk, level, optname, optval, optlen);
834 return do_raw_setsockopt(sk, level, optname, optval, optlen);
835 }
836
837 #ifdef CONFIG_COMPAT
838 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
839 char __user *optval, unsigned int optlen)
840 {
841 if (level != SOL_RAW)
842 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
843 return do_raw_setsockopt(sk, level, optname, optval, optlen);
844 }
845 #endif
846
847 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
848 char __user *optval, int __user *optlen)
849 {
850 if (optname == ICMP_FILTER) {
851 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
852 return -EOPNOTSUPP;
853 else
854 return raw_geticmpfilter(sk, optval, optlen);
855 }
856 return -ENOPROTOOPT;
857 }
858
859 static int raw_getsockopt(struct sock *sk, int level, int optname,
860 char __user *optval, int __user *optlen)
861 {
862 if (level != SOL_RAW)
863 return ip_getsockopt(sk, level, optname, optval, optlen);
864 return do_raw_getsockopt(sk, level, optname, optval, optlen);
865 }
866
867 #ifdef CONFIG_COMPAT
868 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
869 char __user *optval, int __user *optlen)
870 {
871 if (level != SOL_RAW)
872 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
873 return do_raw_getsockopt(sk, level, optname, optval, optlen);
874 }
875 #endif
876
877 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
878 {
879 switch (cmd) {
880 case SIOCOUTQ: {
881 int amount = sk_wmem_alloc_get(sk);
882
883 return put_user(amount, (int __user *)arg);
884 }
885 case SIOCINQ: {
886 struct sk_buff *skb;
887 int amount = 0;
888
889 spin_lock_bh(&sk->sk_receive_queue.lock);
890 skb = skb_peek(&sk->sk_receive_queue);
891 if (skb)
892 amount = skb->len;
893 spin_unlock_bh(&sk->sk_receive_queue.lock);
894 return put_user(amount, (int __user *)arg);
895 }
896
897 default:
898 #ifdef CONFIG_IP_MROUTE
899 return ipmr_ioctl(sk, cmd, (void __user *)arg);
900 #else
901 return -ENOIOCTLCMD;
902 #endif
903 }
904 }
905
906 #ifdef CONFIG_COMPAT
907 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
908 {
909 switch (cmd) {
910 case SIOCOUTQ:
911 case SIOCINQ:
912 return -ENOIOCTLCMD;
913 default:
914 #ifdef CONFIG_IP_MROUTE
915 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
916 #else
917 return -ENOIOCTLCMD;
918 #endif
919 }
920 }
921 #endif
922
923 struct proto raw_prot = {
924 .name = "RAW",
925 .owner = THIS_MODULE,
926 .close = raw_close,
927 .destroy = raw_destroy,
928 .connect = ip4_datagram_connect,
929 .disconnect = __udp_disconnect,
930 .ioctl = raw_ioctl,
931 .init = raw_init,
932 .setsockopt = raw_setsockopt,
933 .getsockopt = raw_getsockopt,
934 .sendmsg = raw_sendmsg,
935 .recvmsg = raw_recvmsg,
936 .bind = raw_bind,
937 .backlog_rcv = raw_rcv_skb,
938 .release_cb = ip4_datagram_release_cb,
939 .hash = raw_hash_sk,
940 .unhash = raw_unhash_sk,
941 .obj_size = sizeof(struct raw_sock),
942 .h.raw_hash = &raw_v4_hashinfo,
943 #ifdef CONFIG_COMPAT
944 .compat_setsockopt = compat_raw_setsockopt,
945 .compat_getsockopt = compat_raw_getsockopt,
946 .compat_ioctl = compat_raw_ioctl,
947 #endif
948 };
949
950 #ifdef CONFIG_PROC_FS
951 static struct sock *raw_get_first(struct seq_file *seq)
952 {
953 struct sock *sk;
954 struct raw_iter_state *state = raw_seq_private(seq);
955
956 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
957 ++state->bucket) {
958 sk_for_each(sk, &state->h->ht[state->bucket])
959 if (sock_net(sk) == seq_file_net(seq))
960 goto found;
961 }
962 sk = NULL;
963 found:
964 return sk;
965 }
966
967 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
968 {
969 struct raw_iter_state *state = raw_seq_private(seq);
970
971 do {
972 sk = sk_next(sk);
973 try_again:
974 ;
975 } while (sk && sock_net(sk) != seq_file_net(seq));
976
977 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
978 sk = sk_head(&state->h->ht[state->bucket]);
979 goto try_again;
980 }
981 return sk;
982 }
983
984 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
985 {
986 struct sock *sk = raw_get_first(seq);
987
988 if (sk)
989 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
990 --pos;
991 return pos ? NULL : sk;
992 }
993
994 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
995 {
996 struct raw_iter_state *state = raw_seq_private(seq);
997
998 read_lock(&state->h->lock);
999 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1000 }
1001 EXPORT_SYMBOL_GPL(raw_seq_start);
1002
1003 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1004 {
1005 struct sock *sk;
1006
1007 if (v == SEQ_START_TOKEN)
1008 sk = raw_get_first(seq);
1009 else
1010 sk = raw_get_next(seq, v);
1011 ++*pos;
1012 return sk;
1013 }
1014 EXPORT_SYMBOL_GPL(raw_seq_next);
1015
1016 void raw_seq_stop(struct seq_file *seq, void *v)
1017 {
1018 struct raw_iter_state *state = raw_seq_private(seq);
1019
1020 read_unlock(&state->h->lock);
1021 }
1022 EXPORT_SYMBOL_GPL(raw_seq_stop);
1023
1024 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1025 {
1026 struct inet_sock *inet = inet_sk(sp);
1027 __be32 dest = inet->inet_daddr,
1028 src = inet->inet_rcv_saddr;
1029 __u16 destp = 0,
1030 srcp = inet->inet_num;
1031
1032 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1033 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1034 i, src, srcp, dest, destp, sp->sk_state,
1035 sk_wmem_alloc_get(sp),
1036 sk_rmem_alloc_get(sp),
1037 0, 0L, 0,
1038 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1039 0, sock_i_ino(sp),
1040 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1041 }
1042
1043 static int raw_seq_show(struct seq_file *seq, void *v)
1044 {
1045 if (v == SEQ_START_TOKEN)
1046 seq_printf(seq, " sl local_address rem_address st tx_queue "
1047 "rx_queue tr tm->when retrnsmt uid timeout "
1048 "inode ref pointer drops\n");
1049 else
1050 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1051 return 0;
1052 }
1053
1054 static const struct seq_operations raw_seq_ops = {
1055 .start = raw_seq_start,
1056 .next = raw_seq_next,
1057 .stop = raw_seq_stop,
1058 .show = raw_seq_show,
1059 };
1060
1061 int raw_seq_open(struct inode *ino, struct file *file,
1062 struct raw_hashinfo *h, const struct seq_operations *ops)
1063 {
1064 int err;
1065 struct raw_iter_state *i;
1066
1067 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1068 if (err < 0)
1069 return err;
1070
1071 i = raw_seq_private((struct seq_file *)file->private_data);
1072 i->h = h;
1073 return 0;
1074 }
1075 EXPORT_SYMBOL_GPL(raw_seq_open);
1076
1077 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1078 {
1079 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1080 }
1081
1082 static const struct file_operations raw_seq_fops = {
1083 .owner = THIS_MODULE,
1084 .open = raw_v4_seq_open,
1085 .read = seq_read,
1086 .llseek = seq_lseek,
1087 .release = seq_release_net,
1088 };
1089
1090 static __net_init int raw_init_net(struct net *net)
1091 {
1092 if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops))
1093 return -ENOMEM;
1094
1095 return 0;
1096 }
1097
1098 static __net_exit void raw_exit_net(struct net *net)
1099 {
1100 remove_proc_entry("raw", net->proc_net);
1101 }
1102
1103 static __net_initdata struct pernet_operations raw_net_ops = {
1104 .init = raw_init_net,
1105 .exit = raw_exit_net,
1106 };
1107
1108 int __init raw_proc_init(void)
1109 {
1110 return register_pernet_subsys(&raw_net_ops);
1111 }
1112
1113 void __init raw_proc_exit(void)
1114 {
1115 unregister_pernet_subsys(&raw_net_ops);
1116 }
1117 #endif /* CONFIG_PROC_FS */
Linux with added FPC-III support