?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[NET] NETNS: Omit namespace comparision without CONFIG_NET_NS.
[linux-2.6/libata-dev.git] / net / ipv4 / raw.c
blobd965f0a39c84e8562ee13958aea98023e8133e83
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * RAW - implementation of IP "raw" sockets.
7 *
8 * Version: $Id: raw.c,v 1.64 2002/02/01 22:01:04 davem Exp $
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 *
13 * Fixes:
14 * Alan Cox : verify_area() fixed up
15 * Alan Cox : ICMP error handling
16 * Alan Cox : EMSGSIZE if you send too big a packet
17 * Alan Cox : Now uses generic datagrams and shared
18 * skbuff library. No more peek crashes,
19 * no more backlogs
20 * Alan Cox : Checks sk->broadcast.
21 * Alan Cox : Uses skb_free_datagram/skb_copy_datagram
22 * Alan Cox : Raw passes ip options too
23 * Alan Cox : Setsocketopt added
24 * Alan Cox : Fixed error return for broadcasts
25 * Alan Cox : Removed wake_up calls
26 * Alan Cox : Use ttl/tos
27 * Alan Cox : Cleaned up old debugging
28 * Alan Cox : Use new kernel side addresses
29 * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets.
30 * Alan Cox : BSD style RAW socket demultiplexing.
31 * Alan Cox : Beginnings of mrouted support.
32 * Alan Cox : Added IP_HDRINCL option.
33 * Alan Cox : Skip broadcast check if BSDism set.
34 * David S. Miller : New socket lookup architecture.
35 *
36 * This program is free software; you can redistribute it and/or
37 * modify it under the terms of the GNU General Public License
38 * as published by the Free Software Foundation; either version
39 * 2 of the License, or (at your option) any later version.
40 */
41
42 #include <linux/types.h>
43 #include <asm/atomic.h>
44 #include <asm/byteorder.h>
45 #include <asm/current.h>
46 #include <asm/uaccess.h>
47 #include <asm/ioctls.h>
48 #include <linux/stddef.h>
49 #include <linux/slab.h>
50 #include <linux/errno.h>
51 #include <linux/aio.h>
52 #include <linux/kernel.h>
53 #include <linux/spinlock.h>
54 #include <linux/sockios.h>
55 #include <linux/socket.h>
56 #include <linux/in.h>
57 #include <linux/mroute.h>
58 #include <linux/netdevice.h>
59 #include <linux/in_route.h>
60 #include <linux/route.h>
61 #include <linux/skbuff.h>
62 #include <net/net_namespace.h>
63 #include <net/dst.h>
64 #include <net/sock.h>
65 #include <linux/gfp.h>
66 #include <linux/ip.h>
67 #include <linux/net.h>
68 #include <net/ip.h>
69 #include <net/icmp.h>
70 #include <net/udp.h>
71 #include <net/raw.h>
72 #include <net/snmp.h>
73 #include <net/tcp_states.h>
74 #include <net/inet_common.h>
75 #include <net/checksum.h>
76 #include <net/xfrm.h>
77 #include <linux/rtnetlink.h>
78 #include <linux/proc_fs.h>
79 #include <linux/seq_file.h>
80 #include <linux/netfilter.h>
81 #include <linux/netfilter_ipv4.h>
82
83 static struct raw_hashinfo raw_v4_hashinfo = {
84 .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
85 };
86
87 void raw_hash_sk(struct sock *sk)
88 {
89 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
90 struct hlist_head *head;
91
92 head = &h->ht[inet_sk(sk)->num & (RAW_HTABLE_SIZE - 1)];
93
94 write_lock_bh(&h->lock);
95 sk_add_node(sk, head);
96 sock_prot_inuse_add(sk->sk_prot, 1);
97 write_unlock_bh(&h->lock);
98 }
99 EXPORT_SYMBOL_GPL(raw_hash_sk);
100
101 void raw_unhash_sk(struct sock *sk)
102 {
103 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
104
105 write_lock_bh(&h->lock);
106 if (sk_del_node_init(sk))
107 sock_prot_inuse_add(sk->sk_prot, -1);
108 write_unlock_bh(&h->lock);
109 }
110 EXPORT_SYMBOL_GPL(raw_unhash_sk);
111
112 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
113 unsigned short num, __be32 raddr, __be32 laddr, int dif)
114 {
115 struct hlist_node *node;
116
117 sk_for_each_from(sk, node) {
118 struct inet_sock *inet = inet_sk(sk);
119
120 if (net_eq(sock_net(sk), net) && inet->num == num &&
121 !(inet->daddr && inet->daddr != raddr) &&
122 !(inet->rcv_saddr && inet->rcv_saddr != laddr) &&
123 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
124 goto found; /* gotcha */
125 }
126 sk = NULL;
127 found:
128 return sk;
129 }
130
131 /*
132 * 0 - deliver
133 * 1 - block
134 */
135 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
136 {
137 int type;
138
139 if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
140 return 1;
141
142 type = icmp_hdr(skb)->type;
143 if (type < 32) {
144 __u32 data = raw_sk(sk)->filter.data;
145
146 return ((1 << type) & data) != 0;
147 }
148
149 /* Do not block unknown ICMP types */
150 return 0;
151 }
152
153 /* IP input processing comes here for RAW socket delivery.
154 * Caller owns SKB, so we must make clones.
155 *
156 * RFC 1122: SHOULD pass TOS value up to the transport layer.
157 * -> It does. And not only TOS, but all IP header.
158 */
159 static int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
160 {
161 struct sock *sk;
162 struct hlist_head *head;
163 int delivered = 0;
164 struct net *net;
165
166 read_lock(&raw_v4_hashinfo.lock);
167 head = &raw_v4_hashinfo.ht[hash];
168 if (hlist_empty(head))
169 goto out;
170
171 net = dev_net(skb->dev);
172 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
173 iph->saddr, iph->daddr,
174 skb->dev->ifindex);
175
176 while (sk) {
177 delivered = 1;
178 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
179 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
180
181 /* Not releasing hash table! */
182 if (clone)
183 raw_rcv(sk, clone);
184 }
185 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
186 iph->saddr, iph->daddr,
187 skb->dev->ifindex);
188 }
189 out:
190 read_unlock(&raw_v4_hashinfo.lock);
191 return delivered;
192 }
193
194 int raw_local_deliver(struct sk_buff *skb, int protocol)
195 {
196 int hash;
197 struct sock *raw_sk;
198
199 hash = protocol & (RAW_HTABLE_SIZE - 1);
200 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
201
202 /* If there maybe a raw socket we must check - if not we
203 * don't care less
204 */
205 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
206 raw_sk = NULL;
207
208 return raw_sk != NULL;
209
210 }
211
212 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
213 {
214 struct inet_sock *inet = inet_sk(sk);
215 const int type = icmp_hdr(skb)->type;
216 const int code = icmp_hdr(skb)->code;
217 int err = 0;
218 int harderr = 0;
219
220 /* Report error on raw socket, if:
221 1. User requested ip_recverr.
222 2. Socket is connected (otherwise the error indication
223 is useless without ip_recverr and error is hard.
224 */
225 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
226 return;
227
228 switch (type) {
229 default:
230 case ICMP_TIME_EXCEEDED:
231 err = EHOSTUNREACH;
232 break;
233 case ICMP_SOURCE_QUENCH:
234 return;
235 case ICMP_PARAMETERPROB:
236 err = EPROTO;
237 harderr = 1;
238 break;
239 case ICMP_DEST_UNREACH:
240 err = EHOSTUNREACH;
241 if (code > NR_ICMP_UNREACH)
242 break;
243 err = icmp_err_convert[code].errno;
244 harderr = icmp_err_convert[code].fatal;
245 if (code == ICMP_FRAG_NEEDED) {
246 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
247 err = EMSGSIZE;
248 }
249 }
250
251 if (inet->recverr) {
252 struct iphdr *iph = (struct iphdr*)skb->data;
253 u8 *payload = skb->data + (iph->ihl << 2);
254
255 if (inet->hdrincl)
256 payload = skb->data;
257 ip_icmp_error(sk, skb, err, 0, info, payload);
258 }
259
260 if (inet->recverr || harderr) {
261 sk->sk_err = err;
262 sk->sk_error_report(sk);
263 }
264 }
265
266 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
267 {
268 int hash;
269 struct sock *raw_sk;
270 struct iphdr *iph;
271 struct net *net;
272
273 hash = protocol & (RAW_HTABLE_SIZE - 1);
274
275 read_lock(&raw_v4_hashinfo.lock);
276 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
277 if (raw_sk != NULL) {
278 iph = (struct iphdr *)skb->data;
279 net = dev_net(skb->dev);
280
281 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
282 iph->daddr, iph->saddr,
283 skb->dev->ifindex)) != NULL) {
284 raw_err(raw_sk, skb, info);
285 raw_sk = sk_next(raw_sk);
286 iph = (struct iphdr *)skb->data;
287 }
288 }
289 read_unlock(&raw_v4_hashinfo.lock);
290 }
291
292 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
293 {
294 /* Charge it to the socket. */
295
296 if (sock_queue_rcv_skb(sk, skb) < 0) {
297 atomic_inc(&sk->sk_drops);
298 kfree_skb(skb);
299 return NET_RX_DROP;
300 }
301
302 return NET_RX_SUCCESS;
303 }
304
305 int raw_rcv(struct sock *sk, struct sk_buff *skb)
306 {
307 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
308 atomic_inc(&sk->sk_drops);
309 kfree_skb(skb);
310 return NET_RX_DROP;
311 }
312 nf_reset(skb);
313
314 skb_push(skb, skb->data - skb_network_header(skb));
315
316 raw_rcv_skb(sk, skb);
317 return 0;
318 }
319
320 static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
321 struct rtable *rt,
322 unsigned int flags)
323 {
324 struct inet_sock *inet = inet_sk(sk);
325 int hh_len;
326 struct iphdr *iph;
327 struct sk_buff *skb;
328 unsigned int iphlen;
329 int err;
330
331 if (length > rt->u.dst.dev->mtu) {
332 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
333 rt->u.dst.dev->mtu);
334 return -EMSGSIZE;
335 }
336 if (flags&MSG_PROBE)
337 goto out;
338
339 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
340
341 skb = sock_alloc_send_skb(sk, length+hh_len+15,
342 flags&MSG_DONTWAIT, &err);
343 if (skb == NULL)
344 goto error;
345 skb_reserve(skb, hh_len);
346
347 skb->priority = sk->sk_priority;
348 skb->mark = sk->sk_mark;
349 skb->dst = dst_clone(&rt->u.dst);
350
351 skb_reset_network_header(skb);
352 iph = ip_hdr(skb);
353 skb_put(skb, length);
354
355 skb->ip_summed = CHECKSUM_NONE;
356
357 skb->transport_header = skb->network_header;
358 err = memcpy_fromiovecend((void *)iph, from, 0, length);
359 if (err)
360 goto error_fault;
361
362 /* We don't modify invalid header */
363 iphlen = iph->ihl * 4;
364 if (iphlen >= sizeof(*iph) && iphlen <= length) {
365 if (!iph->saddr)
366 iph->saddr = rt->rt_src;
367 iph->check = 0;
368 iph->tot_len = htons(length);
369 if (!iph->id)
370 ip_select_ident(iph, &rt->u.dst, NULL);
371
372 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
373 }
374 if (iph->protocol == IPPROTO_ICMP)
375 icmp_out_count(((struct icmphdr *)
376 skb_transport_header(skb))->type);
377
378 err = NF_HOOK(PF_INET, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
379 dst_output);
380 if (err > 0)
381 err = inet->recverr ? net_xmit_errno(err) : 0;
382 if (err)
383 goto error;
384 out:
385 return 0;
386
387 error_fault:
388 err = -EFAULT;
389 kfree_skb(skb);
390 error:
391 IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
392 return err;
393 }
394
395 static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
396 {
397 struct iovec *iov;
398 u8 __user *type = NULL;
399 u8 __user *code = NULL;
400 int probed = 0;
401 unsigned int i;
402
403 if (!msg->msg_iov)
404 return 0;
405
406 for (i = 0; i < msg->msg_iovlen; i++) {
407 iov = &msg->msg_iov[i];
408 if (!iov)
409 continue;
410
411 switch (fl->proto) {
412 case IPPROTO_ICMP:
413 /* check if one-byte field is readable or not. */
414 if (iov->iov_base && iov->iov_len < 1)
415 break;
416
417 if (!type) {
418 type = iov->iov_base;
419 /* check if code field is readable or not. */
420 if (iov->iov_len > 1)
421 code = type + 1;
422 } else if (!code)
423 code = iov->iov_base;
424
425 if (type && code) {
426 if (get_user(fl->fl_icmp_type, type) ||
427 get_user(fl->fl_icmp_code, code))
428 return -EFAULT;
429 probed = 1;
430 }
431 break;
432 default:
433 probed = 1;
434 break;
435 }
436 if (probed)
437 break;
438 }
439 return 0;
440 }
441
442 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
443 size_t len)
444 {
445 struct inet_sock *inet = inet_sk(sk);
446 struct ipcm_cookie ipc;
447 struct rtable *rt = NULL;
448 int free = 0;
449 __be32 daddr;
450 __be32 saddr;
451 u8 tos;
452 int err;
453
454 err = -EMSGSIZE;
455 if (len > 0xFFFF)
456 goto out;
457
458 /*
459 * Check the flags.
460 */
461
462 err = -EOPNOTSUPP;
463 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
464 goto out; /* compatibility */
465
466 /*
467 * Get and verify the address.
468 */
469
470 if (msg->msg_namelen) {
471 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
472 err = -EINVAL;
473 if (msg->msg_namelen < sizeof(*usin))
474 goto out;
475 if (usin->sin_family != AF_INET) {
476 static int complained;
477 if (!complained++)
478 printk(KERN_INFO "%s forgot to set AF_INET in "
479 "raw sendmsg. Fix it!\n",
480 current->comm);
481 err = -EAFNOSUPPORT;
482 if (usin->sin_family)
483 goto out;
484 }
485 daddr = usin->sin_addr.s_addr;
486 /* ANK: I did not forget to get protocol from port field.
487 * I just do not know, who uses this weirdness.
488 * IP_HDRINCL is much more convenient.
489 */
490 } else {
491 err = -EDESTADDRREQ;
492 if (sk->sk_state != TCP_ESTABLISHED)
493 goto out;
494 daddr = inet->daddr;
495 }
496
497 ipc.addr = inet->saddr;
498 ipc.opt = NULL;
499 ipc.oif = sk->sk_bound_dev_if;
500
501 if (msg->msg_controllen) {
502 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
503 if (err)
504 goto out;
505 if (ipc.opt)
506 free = 1;
507 }
508
509 saddr = ipc.addr;
510 ipc.addr = daddr;
511
512 if (!ipc.opt)
513 ipc.opt = inet->opt;
514
515 if (ipc.opt) {
516 err = -EINVAL;
517 /* Linux does not mangle headers on raw sockets,
518 * so that IP options + IP_HDRINCL is non-sense.
519 */
520 if (inet->hdrincl)
521 goto done;
522 if (ipc.opt->srr) {
523 if (!daddr)
524 goto done;
525 daddr = ipc.opt->faddr;
526 }
527 }
528 tos = RT_CONN_FLAGS(sk);
529 if (msg->msg_flags & MSG_DONTROUTE)
530 tos |= RTO_ONLINK;
531
532 if (ipv4_is_multicast(daddr)) {
533 if (!ipc.oif)
534 ipc.oif = inet->mc_index;
535 if (!saddr)
536 saddr = inet->mc_addr;
537 }
538
539 {
540 struct flowi fl = { .oif = ipc.oif,
541 .mark = sk->sk_mark,
542 .nl_u = { .ip4_u =
543 { .daddr = daddr,
544 .saddr = saddr,
545 .tos = tos } },
546 .proto = inet->hdrincl ? IPPROTO_RAW :
547 sk->sk_protocol,
548 };
549 if (!inet->hdrincl) {
550 err = raw_probe_proto_opt(&fl, msg);
551 if (err)
552 goto done;
553 }
554
555 security_sk_classify_flow(sk, &fl);
556 err = ip_route_output_flow(sock_net(sk), &rt, &fl, sk, 1);
557 }
558 if (err)
559 goto done;
560
561 err = -EACCES;
562 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
563 goto done;
564
565 if (msg->msg_flags & MSG_CONFIRM)
566 goto do_confirm;
567 back_from_confirm:
568
569 if (inet->hdrincl)
570 err = raw_send_hdrinc(sk, msg->msg_iov, len,
571 rt, msg->msg_flags);
572
573 else {
574 if (!ipc.addr)
575 ipc.addr = rt->rt_dst;
576 lock_sock(sk);
577 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
578 &ipc, rt, msg->msg_flags);
579 if (err)
580 ip_flush_pending_frames(sk);
581 else if (!(msg->msg_flags & MSG_MORE))
582 err = ip_push_pending_frames(sk);
583 release_sock(sk);
584 }
585 done:
586 if (free)
587 kfree(ipc.opt);
588 ip_rt_put(rt);
589
590 out:
591 if (err < 0)
592 return err;
593 return len;
594
595 do_confirm:
596 dst_confirm(&rt->u.dst);
597 if (!(msg->msg_flags & MSG_PROBE) || len)
598 goto back_from_confirm;
599 err = 0;
600 goto done;
601 }
602
603 static void raw_close(struct sock *sk, long timeout)
604 {
605 /*
606 * Raw sockets may have direct kernel refereneces. Kill them.
607 */
608 ip_ra_control(sk, 0, NULL);
609
610 sk_common_release(sk);
611 }
612
613 /* This gets rid of all the nasties in af_inet. -DaveM */
614 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
615 {
616 struct inet_sock *inet = inet_sk(sk);
617 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
618 int ret = -EINVAL;
619 int chk_addr_ret;
620
621 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
622 goto out;
623 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
624 ret = -EADDRNOTAVAIL;
625 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
626 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
627 goto out;
628 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
629 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
630 inet->saddr = 0; /* Use device */
631 sk_dst_reset(sk);
632 ret = 0;
633 out: return ret;
634 }
635
636 /*
637 * This should be easy, if there is something there
638 * we return it, otherwise we block.
639 */
640
641 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
642 size_t len, int noblock, int flags, int *addr_len)
643 {
644 struct inet_sock *inet = inet_sk(sk);
645 size_t copied = 0;
646 int err = -EOPNOTSUPP;
647 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
648 struct sk_buff *skb;
649
650 if (flags & MSG_OOB)
651 goto out;
652
653 if (addr_len)
654 *addr_len = sizeof(*sin);
655
656 if (flags & MSG_ERRQUEUE) {
657 err = ip_recv_error(sk, msg, len);
658 goto out;
659 }
660
661 skb = skb_recv_datagram(sk, flags, noblock, &err);
662 if (!skb)
663 goto out;
664
665 copied = skb->len;
666 if (len < copied) {
667 msg->msg_flags |= MSG_TRUNC;
668 copied = len;
669 }
670
671 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
672 if (err)
673 goto done;
674
675 sock_recv_timestamp(msg, sk, skb);
676
677 /* Copy the address. */
678 if (sin) {
679 sin->sin_family = AF_INET;
680 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
681 sin->sin_port = 0;
682 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
683 }
684 if (inet->cmsg_flags)
685 ip_cmsg_recv(msg, skb);
686 if (flags & MSG_TRUNC)
687 copied = skb->len;
688 done:
689 skb_free_datagram(sk, skb);
690 out:
691 if (err)
692 return err;
693 return copied;
694 }
695
696 static int raw_init(struct sock *sk)
697 {
698 struct raw_sock *rp = raw_sk(sk);
699
700 if (inet_sk(sk)->num == IPPROTO_ICMP)
701 memset(&rp->filter, 0, sizeof(rp->filter));
702 return 0;
703 }
704
705 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
706 {
707 if (optlen > sizeof(struct icmp_filter))
708 optlen = sizeof(struct icmp_filter);
709 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
710 return -EFAULT;
711 return 0;
712 }
713
714 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
715 {
716 int len, ret = -EFAULT;
717
718 if (get_user(len, optlen))
719 goto out;
720 ret = -EINVAL;
721 if (len < 0)
722 goto out;
723 if (len > sizeof(struct icmp_filter))
724 len = sizeof(struct icmp_filter);
725 ret = -EFAULT;
726 if (put_user(len, optlen) ||
727 copy_to_user(optval, &raw_sk(sk)->filter, len))
728 goto out;
729 ret = 0;
730 out: return ret;
731 }
732
733 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
734 char __user *optval, int optlen)
735 {
736 if (optname == ICMP_FILTER) {
737 if (inet_sk(sk)->num != IPPROTO_ICMP)
738 return -EOPNOTSUPP;
739 else
740 return raw_seticmpfilter(sk, optval, optlen);
741 }
742 return -ENOPROTOOPT;
743 }
744
745 static int raw_setsockopt(struct sock *sk, int level, int optname,
746 char __user *optval, int optlen)
747 {
748 if (level != SOL_RAW)
749 return ip_setsockopt(sk, level, optname, optval, optlen);
750 return do_raw_setsockopt(sk, level, optname, optval, optlen);
751 }
752
753 #ifdef CONFIG_COMPAT
754 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
755 char __user *optval, int optlen)
756 {
757 if (level != SOL_RAW)
758 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
759 return do_raw_setsockopt(sk, level, optname, optval, optlen);
760 }
761 #endif
762
763 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
764 char __user *optval, int __user *optlen)
765 {
766 if (optname == ICMP_FILTER) {
767 if (inet_sk(sk)->num != IPPROTO_ICMP)
768 return -EOPNOTSUPP;
769 else
770 return raw_geticmpfilter(sk, optval, optlen);
771 }
772 return -ENOPROTOOPT;
773 }
774
775 static int raw_getsockopt(struct sock *sk, int level, int optname,
776 char __user *optval, int __user *optlen)
777 {
778 if (level != SOL_RAW)
779 return ip_getsockopt(sk, level, optname, optval, optlen);
780 return do_raw_getsockopt(sk, level, optname, optval, optlen);
781 }
782
783 #ifdef CONFIG_COMPAT
784 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
785 char __user *optval, int __user *optlen)
786 {
787 if (level != SOL_RAW)
788 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
789 return do_raw_getsockopt(sk, level, optname, optval, optlen);
790 }
791 #endif
792
793 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
794 {
795 switch (cmd) {
796 case SIOCOUTQ: {
797 int amount = atomic_read(&sk->sk_wmem_alloc);
798 return put_user(amount, (int __user *)arg);
799 }
800 case SIOCINQ: {
801 struct sk_buff *skb;
802 int amount = 0;
803
804 spin_lock_bh(&sk->sk_receive_queue.lock);
805 skb = skb_peek(&sk->sk_receive_queue);
806 if (skb != NULL)
807 amount = skb->len;
808 spin_unlock_bh(&sk->sk_receive_queue.lock);
809 return put_user(amount, (int __user *)arg);
810 }
811
812 default:
813 #ifdef CONFIG_IP_MROUTE
814 return ipmr_ioctl(sk, cmd, (void __user *)arg);
815 #else
816 return -ENOIOCTLCMD;
817 #endif
818 }
819 }
820
821 DEFINE_PROTO_INUSE(raw)
822
823 struct proto raw_prot = {
824 .name = "RAW",
825 .owner = THIS_MODULE,
826 .close = raw_close,
827 .connect = ip4_datagram_connect,
828 .disconnect = udp_disconnect,
829 .ioctl = raw_ioctl,
830 .init = raw_init,
831 .setsockopt = raw_setsockopt,
832 .getsockopt = raw_getsockopt,
833 .sendmsg = raw_sendmsg,
834 .recvmsg = raw_recvmsg,
835 .bind = raw_bind,
836 .backlog_rcv = raw_rcv_skb,
837 .hash = raw_hash_sk,
838 .unhash = raw_unhash_sk,
839 .obj_size = sizeof(struct raw_sock),
840 .h.raw_hash = &raw_v4_hashinfo,
841 #ifdef CONFIG_COMPAT
842 .compat_setsockopt = compat_raw_setsockopt,
843 .compat_getsockopt = compat_raw_getsockopt,
844 #endif
845 REF_PROTO_INUSE(raw)
846 };
847
848 #ifdef CONFIG_PROC_FS
849 static struct sock *raw_get_first(struct seq_file *seq)
850 {
851 struct sock *sk;
852 struct raw_iter_state* state = raw_seq_private(seq);
853
854 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
855 ++state->bucket) {
856 struct hlist_node *node;
857
858 sk_for_each(sk, node, &state->h->ht[state->bucket])
859 if (sock_net(sk) == seq_file_net(seq))
860 goto found;
861 }
862 sk = NULL;
863 found:
864 return sk;
865 }
866
867 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
868 {
869 struct raw_iter_state* state = raw_seq_private(seq);
870
871 do {
872 sk = sk_next(sk);
873 try_again:
874 ;
875 } while (sk && sock_net(sk) != seq_file_net(seq));
876
877 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
878 sk = sk_head(&state->h->ht[state->bucket]);
879 goto try_again;
880 }
881 return sk;
882 }
883
884 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
885 {
886 struct sock *sk = raw_get_first(seq);
887
888 if (sk)
889 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
890 --pos;
891 return pos ? NULL : sk;
892 }
893
894 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
895 {
896 struct raw_iter_state *state = raw_seq_private(seq);
897
898 read_lock(&state->h->lock);
899 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
900 }
901 EXPORT_SYMBOL_GPL(raw_seq_start);
902
903 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
904 {
905 struct sock *sk;
906
907 if (v == SEQ_START_TOKEN)
908 sk = raw_get_first(seq);
909 else
910 sk = raw_get_next(seq, v);
911 ++*pos;
912 return sk;
913 }
914 EXPORT_SYMBOL_GPL(raw_seq_next);
915
916 void raw_seq_stop(struct seq_file *seq, void *v)
917 {
918 struct raw_iter_state *state = raw_seq_private(seq);
919
920 read_unlock(&state->h->lock);
921 }
922 EXPORT_SYMBOL_GPL(raw_seq_stop);
923
924 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
925 {
926 struct inet_sock *inet = inet_sk(sp);
927 __be32 dest = inet->daddr,
928 src = inet->rcv_saddr;
929 __u16 destp = 0,
930 srcp = inet->num;
931
932 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
933 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d",
934 i, src, srcp, dest, destp, sp->sk_state,
935 atomic_read(&sp->sk_wmem_alloc),
936 atomic_read(&sp->sk_rmem_alloc),
937 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
938 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
939 }
940
941 static int raw_seq_show(struct seq_file *seq, void *v)
942 {
943 if (v == SEQ_START_TOKEN)
944 seq_printf(seq, " sl local_address rem_address st tx_queue "
945 "rx_queue tr tm->when retrnsmt uid timeout "
946 "inode drops\n");
947 else
948 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
949 return 0;
950 }
951
952 static const struct seq_operations raw_seq_ops = {
953 .start = raw_seq_start,
954 .next = raw_seq_next,
955 .stop = raw_seq_stop,
956 .show = raw_seq_show,
957 };
958
959 int raw_seq_open(struct inode *ino, struct file *file,
960 struct raw_hashinfo *h, const struct seq_operations *ops)
961 {
962 int err;
963 struct raw_iter_state *i;
964
965 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
966 if (err < 0)
967 return err;
968
969 i = raw_seq_private((struct seq_file *)file->private_data);
970 i->h = h;
971 return 0;
972 }
973 EXPORT_SYMBOL_GPL(raw_seq_open);
974
975 static int raw_v4_seq_open(struct inode *inode, struct file *file)
976 {
977 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
978 }
979
980 static const struct file_operations raw_seq_fops = {
981 .owner = THIS_MODULE,
982 .open = raw_v4_seq_open,
983 .read = seq_read,
984 .llseek = seq_lseek,
985 .release = seq_release_net,
986 };
987
988 static __net_init int raw_init_net(struct net *net)
989 {
990 if (!proc_net_fops_create(net, "raw", S_IRUGO, &raw_seq_fops))
991 return -ENOMEM;
992
993 return 0;
994 }
995
996 static __net_exit void raw_exit_net(struct net *net)
997 {
998 proc_net_remove(net, "raw");
999 }
1000
1001 static __net_initdata struct pernet_operations raw_net_ops = {
1002 .init = raw_init_net,
1003 .exit = raw_exit_net,
1004 };
1005
1006 int __init raw_proc_init(void)
1007 {
1008 return register_pernet_subsys(&raw_net_ops);
1009 }
1010
1011 void __init raw_proc_exit(void)
1012 {
1013 unregister_pernet_subsys(&raw_net_ops);
1014 }
1015 #endif /* CONFIG_PROC_FS */
Linux 2.6 libata-dev (mirror)