]> bbs.cooldavid.org Git - net-next-2.6.git/blame - net/ipv4/raw.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[NIU]: Update driver version and release date.
[net-next-2.6.git] / net / ipv4 / raw.c
CommitLineData
1da177e4
LT		 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 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT		 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 */
bf0d5249 41
715b49ef 42#include <linux/types.h>
1da177e4
LT		 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>
1da177e4
LT		 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>
1da177e4 61#include <linux/skbuff.h>
457c4cbc 62#include <net/net_namespace.h>
1da177e4
LT		 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>
c752f073 73#include <net/tcp_states.h>
1da177e4
LT		 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
83struct hlist_head raw_v4_htable[RAWV4_HTABLE_SIZE];
84DEFINE_RWLOCK(raw_v4_lock);
85
86static void raw_v4_hash(struct sock *sk)
87{
88 struct hlist_head *head = &raw_v4_htable[inet_sk(sk)->num &
89 (RAWV4_HTABLE_SIZE - 1)];
90
91 write_lock_bh(&raw_v4_lock);
92 sk_add_node(sk, head);
93 sock_prot_inc_use(sk->sk_prot);
94 write_unlock_bh(&raw_v4_lock);
95}
96
97static void raw_v4_unhash(struct sock *sk)
98{
e905a9ed 99 write_lock_bh(&raw_v4_lock);
1da177e4
LT		 100 if (sk_del_node_init(sk))
101 sock_prot_dec_use(sk->sk_prot);
102 write_unlock_bh(&raw_v4_lock);
103}
104
105struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num,
6d741653 106 __be32 raddr, __be32 laddr,
1da177e4
LT		 107 int dif)
108{
109 struct hlist_node *node;
110
111 sk_for_each_from(sk, node) {
112 struct inet_sock *inet = inet_sk(sk);
113
114 if (inet->num == num &&
115 !(inet->daddr && inet->daddr != raddr) &&
116 !(inet->rcv_saddr && inet->rcv_saddr != laddr) &&
117 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
118 goto found; /* gotcha */
119 }
120 sk = NULL;
121found:
122 return sk;
123}
124
125/*
126 * 0 - deliver
127 * 1 - block
128 */
129static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
130{
131 int type;
132
133 if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
134 return 1;
135
88c7664f 136 type = icmp_hdr(skb)->type;
1da177e4
LT		 137 if (type < 32) {
138 __u32 data = raw_sk(sk)->filter.data;
139
140 return ((1 << type) & data) != 0;
141 }
142
143 /* Do not block unknown ICMP types */
144 return 0;
145}
146
147/* IP input processing comes here for RAW socket delivery.
148 * Caller owns SKB, so we must make clones.
149 *
150 * RFC 1122: SHOULD pass TOS value up to the transport layer.
151 * -> It does. And not only TOS, but all IP header.
152 */
d13964f4 153int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
1da177e4
LT		 154{
155 struct sock *sk;
156 struct hlist_head *head;
d13964f4 157 int delivered = 0;
1da177e4
LT		 158
159 read_lock(&raw_v4_lock);
160 head = &raw_v4_htable[hash];
161 if (hlist_empty(head))
162 goto out;
163 sk = __raw_v4_lookup(__sk_head(head), iph->protocol,
164 iph->saddr, iph->daddr,
165 skb->dev->ifindex);
166
167 while (sk) {
d13964f4 168 delivered = 1;
1da177e4
LT		 169 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
170 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
171
172 /* Not releasing hash table! */
173 if (clone)
174 raw_rcv(sk, clone);
175 }
176 sk = __raw_v4_lookup(sk_next(sk), iph->protocol,
177 iph->saddr, iph->daddr,
178 skb->dev->ifindex);
179 }
180out:
181 read_unlock(&raw_v4_lock);
d13964f4 182 return delivered;
1da177e4
LT		 183}
184
185void raw_err (struct sock *sk, struct sk_buff *skb, u32 info)
186{
187 struct inet_sock *inet = inet_sk(sk);
88c7664f
ACM		 188 const int type = icmp_hdr(skb)->type;
189 const int code = icmp_hdr(skb)->code;
1da177e4
LT		 190 int err = 0;
191 int harderr = 0;
192
193 /* Report error on raw socket, if:
194 1. User requested ip_recverr.
195 2. Socket is connected (otherwise the error indication
196 is useless without ip_recverr and error is hard.
197 */
198 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
199 return;
200
201 switch (type) {
202 default:
203 case ICMP_TIME_EXCEEDED:
204 err = EHOSTUNREACH;
205 break;
206 case ICMP_SOURCE_QUENCH:
207 return;
208 case ICMP_PARAMETERPROB:
209 err = EPROTO;
210 harderr = 1;
211 break;
212 case ICMP_DEST_UNREACH:
213 err = EHOSTUNREACH;
214 if (code > NR_ICMP_UNREACH)
215 break;
216 err = icmp_err_convert[code].errno;
217 harderr = icmp_err_convert[code].fatal;
218 if (code == ICMP_FRAG_NEEDED) {
219 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
220 err = EMSGSIZE;
221 }
222 }
223
224 if (inet->recverr) {
225 struct iphdr *iph = (struct iphdr*)skb->data;
226 u8 *payload = skb->data + (iph->ihl << 2);
227
228 if (inet->hdrincl)
229 payload = skb->data;
230 ip_icmp_error(sk, skb, err, 0, info, payload);
231 }
232
233 if (inet->recverr || harderr) {
234 sk->sk_err = err;
235 sk->sk_error_report(sk);
236 }
237}
238
239static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
240{
241 /* Charge it to the socket. */
e905a9ed 242
1da177e4
LT		 243 if (sock_queue_rcv_skb(sk, skb) < 0) {
244 /* FIXME: increment a raw drops counter here */
245 kfree_skb(skb);
246 return NET_RX_DROP;
247 }
248
249 return NET_RX_SUCCESS;
250}
251
252int raw_rcv(struct sock *sk, struct sk_buff *skb)
253{
254 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
255 kfree_skb(skb);
256 return NET_RX_DROP;
257 }
b59c2701 258 nf_reset(skb);
1da177e4 259
d56f90a7 260 skb_push(skb, skb->data - skb_network_header(skb));
1da177e4
LT		 261
262 raw_rcv_skb(sk, skb);
263 return 0;
264}
265
f7d7fc03 266static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
e905a9ed 267 struct rtable *rt,
1da177e4
LT		 268 unsigned int flags)
269{
270 struct inet_sock *inet = inet_sk(sk);
271 int hh_len;
272 struct iphdr *iph;
273 struct sk_buff *skb;
274 int err;
275
276 if (length > rt->u.dst.dev->mtu) {
277 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
278 rt->u.dst.dev->mtu);
279 return -EMSGSIZE;
280 }
281 if (flags&MSG_PROBE)
282 goto out;
283
284 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
285
286 skb = sock_alloc_send_skb(sk, length+hh_len+15,
287 flags&MSG_DONTWAIT, &err);
288 if (skb == NULL)
e905a9ed 289 goto error;
1da177e4
LT		 290 skb_reserve(skb, hh_len);
291
292 skb->priority = sk->sk_priority;
293 skb->dst = dst_clone(&rt->u.dst);
294
7e28ecc2 295 skb_reset_network_header(skb);
eddc9ec5 296 iph = ip_hdr(skb);
7e28ecc2 297 skb_put(skb, length);
1da177e4
LT		 298
299 skb->ip_summed = CHECKSUM_NONE;
300
b0e380b1 301 skb->transport_header = skb->network_header;
1da177e4
LT		 302 err = memcpy_fromiovecend((void *)iph, from, 0, length);
303 if (err)
304 goto error_fault;
305
306 /* We don't modify invalid header */
f7d7fc03 307 if (length >= sizeof(*iph) && iph->ihl * 4U <= length) {
1da177e4
LT		 308 if (!iph->saddr)
309 iph->saddr = rt->rt_src;
310 iph->check = 0;
311 iph->tot_len = htons(length);
312 if (!iph->id)
313 ip_select_ident(iph, &rt->u.dst, NULL);
314
315 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
316 }
96793b48
DS		 317 if (iph->protocol == IPPROTO_ICMP)
318 icmp_out_count(((struct icmphdr *)
319 skb_transport_header(skb))->type);
1da177e4
LT		 320
321 err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
322 dst_output);
323 if (err > 0)
324 err = inet->recverr ? net_xmit_errno(err) : 0;
325 if (err)
326 goto error;
327out:
328 return 0;
329
330error_fault:
331 err = -EFAULT;
332 kfree_skb(skb);
333error:
334 IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
e905a9ed 335 return err;
1da177e4
LT		 336}
337
a27b58fe 338static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
1da177e4
LT		 339{
340 struct iovec *iov;
341 u8 __user *type = NULL;
342 u8 __user *code = NULL;
343 int probed = 0;
93765d8a 344 unsigned int i;
1da177e4
LT		 345
346 if (!msg->msg_iov)
a27b58fe 347 return 0;
1da177e4
LT		 348
349 for (i = 0; i < msg->msg_iovlen; i++) {
350 iov = &msg->msg_iov[i];
351 if (!iov)
352 continue;
353
354 switch (fl->proto) {
355 case IPPROTO_ICMP:
356 /* check if one-byte field is readable or not. */
357 if (iov->iov_base && iov->iov_len < 1)
358 break;
359
360 if (!type) {
361 type = iov->iov_base;
362 /* check if code field is readable or not. */
363 if (iov->iov_len > 1)
364 code = type + 1;
365 } else if (!code)
366 code = iov->iov_base;
367
368 if (type && code) {
a27b58fe
HC		 369 if (get_user(fl->fl_icmp_type, type) ||
370 get_user(fl->fl_icmp_code, code))
371 return -EFAULT;
1da177e4
LT		 372 probed = 1;
373 }
374 break;
375 default:
376 probed = 1;
377 break;
378 }
379 if (probed)
380 break;
381 }
a27b58fe 382 return 0;
1da177e4
LT		 383}
384
385static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
386 size_t len)
387{
388 struct inet_sock *inet = inet_sk(sk);
389 struct ipcm_cookie ipc;
390 struct rtable *rt = NULL;
391 int free = 0;
3ca3c68e 392 __be32 daddr;
c1d18f9f 393 __be32 saddr;
1da177e4
LT		 394 u8 tos;
395 int err;
396
397 err = -EMSGSIZE;
926d4b81 398 if (len > 0xFFFF)
1da177e4
LT		 399 goto out;
400
401 /*
402 * Check the flags.
403 */
404
405 err = -EOPNOTSUPP;
406 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
407 goto out; /* compatibility */
e905a9ed 408
1da177e4 409 /*
e905a9ed 410 * Get and verify the address.
1da177e4
LT		 411 */
412
413 if (msg->msg_namelen) {
414 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
415 err = -EINVAL;
416 if (msg->msg_namelen < sizeof(*usin))
417 goto out;
418 if (usin->sin_family != AF_INET) {
419 static int complained;
420 if (!complained++)
421 printk(KERN_INFO "%s forgot to set AF_INET in "
422 "raw sendmsg. Fix it!\n",
423 current->comm);
424 err = -EAFNOSUPPORT;
425 if (usin->sin_family)
426 goto out;
427 }
428 daddr = usin->sin_addr.s_addr;
429 /* ANK: I did not forget to get protocol from port field.
430 * I just do not know, who uses this weirdness.
431 * IP_HDRINCL is much more convenient.
432 */
433 } else {
434 err = -EDESTADDRREQ;
e905a9ed 435 if (sk->sk_state != TCP_ESTABLISHED)
1da177e4
LT		 436 goto out;
437 daddr = inet->daddr;
438 }
439
440 ipc.addr = inet->saddr;
441 ipc.opt = NULL;
442 ipc.oif = sk->sk_bound_dev_if;
443
444 if (msg->msg_controllen) {
445 err = ip_cmsg_send(msg, &ipc);
446 if (err)
447 goto out;
448 if (ipc.opt)
449 free = 1;
450 }
451
452 saddr = ipc.addr;
453 ipc.addr = daddr;
454
455 if (!ipc.opt)
456 ipc.opt = inet->opt;
457
458 if (ipc.opt) {
459 err = -EINVAL;
460 /* Linux does not mangle headers on raw sockets,
461 * so that IP options + IP_HDRINCL is non-sense.
462 */
463 if (inet->hdrincl)
464 goto done;
465 if (ipc.opt->srr) {
466 if (!daddr)
467 goto done;
468 daddr = ipc.opt->faddr;
469 }
470 }
471 tos = RT_CONN_FLAGS(sk);
472 if (msg->msg_flags & MSG_DONTROUTE)
473 tos |= RTO_ONLINK;
474
475 if (MULTICAST(daddr)) {
476 if (!ipc.oif)
477 ipc.oif = inet->mc_index;
478 if (!saddr)
479 saddr = inet->mc_addr;
480 }
481
482 {
483 struct flowi fl = { .oif = ipc.oif,
484 .nl_u = { .ip4_u =
485 { .daddr = daddr,
486 .saddr = saddr,
487 .tos = tos } },
488 .proto = inet->hdrincl ? IPPROTO_RAW :
e905a9ed 489 sk->sk_protocol,
1da177e4 490 };
a27b58fe
HC		 491 if (!inet->hdrincl) {
492 err = raw_probe_proto_opt(&fl, msg);
493 if (err)
494 goto done;
495 }
1da177e4 496
beb8d13b 497 security_sk_classify_flow(sk, &fl);
8eb9086f 498 err = ip_route_output_flow(&rt, &fl, sk, 1);
1da177e4
LT		 499 }
500 if (err)
501 goto done;
502
503 err = -EACCES;
504 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
505 goto done;
506
507 if (msg->msg_flags & MSG_CONFIRM)
508 goto do_confirm;
509back_from_confirm:
510
511 if (inet->hdrincl)
e905a9ed 512 err = raw_send_hdrinc(sk, msg->msg_iov, len,
1da177e4 513 rt, msg->msg_flags);
e905a9ed 514
1da177e4
LT		 515 else {
516 if (!ipc.addr)
517 ipc.addr = rt->rt_dst;
518 lock_sock(sk);
519 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
520 &ipc, rt, msg->msg_flags);
521 if (err)
522 ip_flush_pending_frames(sk);
523 else if (!(msg->msg_flags & MSG_MORE))
524 err = ip_push_pending_frames(sk);
525 release_sock(sk);
526 }
527done:
528 if (free)
529 kfree(ipc.opt);
530 ip_rt_put(rt);
531
5418c692
JJ		 532out:
533 if (err < 0)
534 return err;
535 return len;
1da177e4
LT		 536
537do_confirm:
538 dst_confirm(&rt->u.dst);
539 if (!(msg->msg_flags & MSG_PROBE) || len)
540 goto back_from_confirm;
541 err = 0;
542 goto done;
543}
544
545static void raw_close(struct sock *sk, long timeout)
546{
e905a9ed 547 /*
1da177e4
LT		 548 * Raw sockets may have direct kernel refereneces. Kill them.
549 */
550 ip_ra_control(sk, 0, NULL);
551
552 sk_common_release(sk);
553}
554
555/* This gets rid of all the nasties in af_inet. -DaveM */
556static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
557{
558 struct inet_sock *inet = inet_sk(sk);
559 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
560 int ret = -EINVAL;
561 int chk_addr_ret;
562
563 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
564 goto out;
565 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
566 ret = -EADDRNOTAVAIL;
567 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
568 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
569 goto out;
570 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
571 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
572 inet->saddr = 0; /* Use device */
573 sk_dst_reset(sk);
574 ret = 0;
575out: return ret;
576}
577
578/*
579 * This should be easy, if there is something there
580 * we return it, otherwise we block.
581 */
582
583static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
584 size_t len, int noblock, int flags, int *addr_len)
585{
586 struct inet_sock *inet = inet_sk(sk);
587 size_t copied = 0;
588 int err = -EOPNOTSUPP;
589 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
590 struct sk_buff *skb;
591
592 if (flags & MSG_OOB)
593 goto out;
594
595 if (addr_len)
596 *addr_len = sizeof(*sin);
597
598 if (flags & MSG_ERRQUEUE) {
599 err = ip_recv_error(sk, msg, len);
600 goto out;
601 }
602
603 skb = skb_recv_datagram(sk, flags, noblock, &err);
604 if (!skb)
605 goto out;
606
607 copied = skb->len;
608 if (len < copied) {
609 msg->msg_flags |= MSG_TRUNC;
610 copied = len;
611 }
612
613 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
614 if (err)
615 goto done;
616
617 sock_recv_timestamp(msg, sk, skb);
618
619 /* Copy the address. */
620 if (sin) {
621 sin->sin_family = AF_INET;
eddc9ec5 622 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
f59fc7f3 623 sin->sin_port = 0;
1da177e4
LT		 624 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
625 }
626 if (inet->cmsg_flags)
627 ip_cmsg_recv(msg, skb);
628 if (flags & MSG_TRUNC)
629 copied = skb->len;
630done:
631 skb_free_datagram(sk, skb);
5418c692
JJ		 632out:
633 if (err)
634 return err;
635 return copied;
1da177e4
LT		 636}
637
638static int raw_init(struct sock *sk)
639{
640 struct raw_sock *rp = raw_sk(sk);
641
642 if (inet_sk(sk)->num == IPPROTO_ICMP)
643 memset(&rp->filter, 0, sizeof(rp->filter));
644 return 0;
645}
646
647static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
648{
649 if (optlen > sizeof(struct icmp_filter))
650 optlen = sizeof(struct icmp_filter);
651 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
652 return -EFAULT;
653 return 0;
654}
655
656static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
657{
658 int len, ret = -EFAULT;
659
660 if (get_user(len, optlen))
661 goto out;
662 ret = -EINVAL;
663 if (len < 0)
664 goto out;
665 if (len > sizeof(struct icmp_filter))
666 len = sizeof(struct icmp_filter);
667 ret = -EFAULT;
668 if (put_user(len, optlen) ||
669 copy_to_user(optval, &raw_sk(sk)->filter, len))
670 goto out;
671 ret = 0;
672out: return ret;
673}
674
3fdadf7d 675static int do_raw_setsockopt(struct sock *sk, int level, int optname,
1da177e4
LT		 676 char __user *optval, int optlen)
677{
1da177e4
LT		 678 if (optname == ICMP_FILTER) {
679 if (inet_sk(sk)->num != IPPROTO_ICMP)
680 return -EOPNOTSUPP;
681 else
682 return raw_seticmpfilter(sk, optval, optlen);
683 }
684 return -ENOPROTOOPT;
685}
686
3fdadf7d
DM		 687static int raw_setsockopt(struct sock *sk, int level, int optname,
688 char __user *optval, int optlen)
1da177e4
LT		 689{
690 if (level != SOL_RAW)
3fdadf7d
DM		 691 return ip_setsockopt(sk, level, optname, optval, optlen);
692 return do_raw_setsockopt(sk, level, optname, optval, optlen);
693}
1da177e4 694
3fdadf7d
DM		 695#ifdef CONFIG_COMPAT
696static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
543d9cfe 697 char __user *optval, int optlen)
3fdadf7d
DM		 698{
699 if (level != SOL_RAW)
543d9cfe 700 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM		 701 return do_raw_setsockopt(sk, level, optname, optval, optlen);
702}
703#endif
704
705static int do_raw_getsockopt(struct sock *sk, int level, int optname,
706 char __user *optval, int __user *optlen)
707{
1da177e4
LT		 708 if (optname == ICMP_FILTER) {
709 if (inet_sk(sk)->num != IPPROTO_ICMP)
710 return -EOPNOTSUPP;
711 else
712 return raw_geticmpfilter(sk, optval, optlen);
713 }
714 return -ENOPROTOOPT;
715}
716
3fdadf7d
DM		 717static int raw_getsockopt(struct sock *sk, int level, int optname,
718 char __user *optval, int __user *optlen)
719{
720 if (level != SOL_RAW)
721 return ip_getsockopt(sk, level, optname, optval, optlen);
722 return do_raw_getsockopt(sk, level, optname, optval, optlen);
723}
724
725#ifdef CONFIG_COMPAT
726static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
543d9cfe 727 char __user *optval, int __user *optlen)
3fdadf7d
DM		 728{
729 if (level != SOL_RAW)
543d9cfe 730 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM		 731 return do_raw_getsockopt(sk, level, optname, optval, optlen);
732}
733#endif
734
1da177e4
LT		 735static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
736{
737 switch (cmd) {
738 case SIOCOUTQ: {
739 int amount = atomic_read(&sk->sk_wmem_alloc);
740 return put_user(amount, (int __user *)arg);
741 }
742 case SIOCINQ: {
743 struct sk_buff *skb;
744 int amount = 0;
745
e0f9f858 746 spin_lock_bh(&sk->sk_receive_queue.lock);
1da177e4
LT		 747 skb = skb_peek(&sk->sk_receive_queue);
748 if (skb != NULL)
749 amount = skb->len;
e0f9f858 750 spin_unlock_bh(&sk->sk_receive_queue.lock);
1da177e4
LT		 751 return put_user(amount, (int __user *)arg);
752 }
753
754 default:
755#ifdef CONFIG_IP_MROUTE
756 return ipmr_ioctl(sk, cmd, (void __user *)arg);
757#else
758 return -ENOIOCTLCMD;
759#endif
760 }
761}
762
47a31a6f
ED		 763DEFINE_PROTO_INUSE(raw)
764
1da177e4 765struct proto raw_prot = {
543d9cfe
ACM		 766 .name = "RAW",
767 .owner = THIS_MODULE,
768 .close = raw_close,
769 .connect = ip4_datagram_connect,
770 .disconnect = udp_disconnect,
771 .ioctl = raw_ioctl,
772 .init = raw_init,
773 .setsockopt = raw_setsockopt,
774 .getsockopt = raw_getsockopt,
775 .sendmsg = raw_sendmsg,
776 .recvmsg = raw_recvmsg,
777 .bind = raw_bind,
778 .backlog_rcv = raw_rcv_skb,
779 .hash = raw_v4_hash,
780 .unhash = raw_v4_unhash,
781 .obj_size = sizeof(struct raw_sock),
3fdadf7d 782#ifdef CONFIG_COMPAT
543d9cfe
ACM		 783 .compat_setsockopt = compat_raw_setsockopt,
784 .compat_getsockopt = compat_raw_getsockopt,
3fdadf7d 785#endif
47a31a6f 786 REF_PROTO_INUSE(raw)
1da177e4
LT		 787};
788
789#ifdef CONFIG_PROC_FS
790struct raw_iter_state {
791 int bucket;
792};
793
794#define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
795
796static struct sock *raw_get_first(struct seq_file *seq)
797{
798 struct sock *sk;
799 struct raw_iter_state* state = raw_seq_private(seq);
800
801 for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
802 struct hlist_node *node;
803
804 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
805 if (sk->sk_family == PF_INET)
806 goto found;
807 }
808 sk = NULL;
809found:
810 return sk;
811}
812
813static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
814{
815 struct raw_iter_state* state = raw_seq_private(seq);
816
817 do {
818 sk = sk_next(sk);
819try_again:
820 ;
821 } while (sk && sk->sk_family != PF_INET);
822
823 if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
824 sk = sk_head(&raw_v4_htable[state->bucket]);
825 goto try_again;
826 }
827 return sk;
828}
829
830static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
831{
832 struct sock *sk = raw_get_first(seq);
833
834 if (sk)
835 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
836 --pos;
837 return pos ? NULL : sk;
838}
839
840static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
841{
842 read_lock(&raw_v4_lock);
843 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
844}
845
846static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
847{
848 struct sock *sk;
849
850 if (v == SEQ_START_TOKEN)
851 sk = raw_get_first(seq);
852 else
853 sk = raw_get_next(seq, v);
854 ++*pos;
855 return sk;
856}
857
858static void raw_seq_stop(struct seq_file *seq, void *v)
859{
860 read_unlock(&raw_v4_lock);
861}
862
863static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
864{
865 struct inet_sock *inet = inet_sk(sp);
714e85be
AV		 866 __be32 dest = inet->daddr,
867 src = inet->rcv_saddr;
1da177e4
LT		 868 __u16 destp = 0,
869 srcp = inet->num;
870
871 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
872 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
e905a9ed 873 i, src, srcp, dest, destp, sp->sk_state,
1da177e4
LT		 874 atomic_read(&sp->sk_wmem_alloc),
875 atomic_read(&sp->sk_rmem_alloc),
876 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
877 atomic_read(&sp->sk_refcnt), sp);
878 return tmpbuf;
879}
880
881static int raw_seq_show(struct seq_file *seq, void *v)
882{
883 char tmpbuf[129];
884
885 if (v == SEQ_START_TOKEN)
886 seq_printf(seq, "%-127s\n",
887 " sl local_address rem_address st tx_queue "
888 "rx_queue tr tm->when retrnsmt uid timeout "
889 "inode");
890 else {
891 struct raw_iter_state *state = raw_seq_private(seq);
892
893 seq_printf(seq, "%-127s\n",
894 get_raw_sock(v, tmpbuf, state->bucket));
895 }
896 return 0;
897}
898
f690808e 899static const struct seq_operations raw_seq_ops = {
1da177e4
LT		 900 .start = raw_seq_start,
901 .next = raw_seq_next,
902 .stop = raw_seq_stop,
903 .show = raw_seq_show,
904};
905
906static int raw_seq_open(struct inode *inode, struct file *file)
907{
cf7732e4
PE		 908 return seq_open_private(file, &raw_seq_ops,
909 sizeof(struct raw_iter_state));
1da177e4
LT		 910}
911
9a32144e 912static const struct file_operations raw_seq_fops = {
1da177e4
LT		 913 .owner = THIS_MODULE,
914 .open = raw_seq_open,
915 .read = seq_read,
916 .llseek = seq_lseek,
917 .release = seq_release_private,
918};
919
920int __init raw_proc_init(void)
921{
457c4cbc 922 if (!proc_net_fops_create(&init_net, "raw", S_IRUGO, &raw_seq_fops))
1da177e4
LT		 923 return -ENOMEM;
924 return 0;
925}
926
927void __init raw_proc_exit(void)
928{
457c4cbc 929 proc_net_remove(&init_net, "raw");
1da177e4
LT		 930}
931#endif /* CONFIG_PROC_FS */
Clone of davem's net-next-2.6 tree