]> 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
[NET] IEEE80211: Fix whitespace errors.
[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
LT		 61#include <linux/skbuff.h>
62#include <net/dst.h>
63#include <net/sock.h>
64#include <linux/gfp.h>
65#include <linux/ip.h>
66#include <linux/net.h>
67#include <net/ip.h>
68#include <net/icmp.h>
69#include <net/udp.h>
70#include <net/raw.h>
71#include <net/snmp.h>
c752f073 72#include <net/tcp_states.h>
1da177e4
LT		 73#include <net/inet_common.h>
74#include <net/checksum.h>
75#include <net/xfrm.h>
76#include <linux/rtnetlink.h>
77#include <linux/proc_fs.h>
78#include <linux/seq_file.h>
79#include <linux/netfilter.h>
80#include <linux/netfilter_ipv4.h>
81
82struct hlist_head raw_v4_htable[RAWV4_HTABLE_SIZE];
83DEFINE_RWLOCK(raw_v4_lock);
84
85static void raw_v4_hash(struct sock *sk)
86{
87 struct hlist_head *head = &raw_v4_htable[inet_sk(sk)->num &
88 (RAWV4_HTABLE_SIZE - 1)];
89
90 write_lock_bh(&raw_v4_lock);
91 sk_add_node(sk, head);
92 sock_prot_inc_use(sk->sk_prot);
93 write_unlock_bh(&raw_v4_lock);
94}
95
96static void raw_v4_unhash(struct sock *sk)
97{
98 write_lock_bh(&raw_v4_lock);
99 if (sk_del_node_init(sk))
100 sock_prot_dec_use(sk->sk_prot);
101 write_unlock_bh(&raw_v4_lock);
102}
103
104struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num,
6d741653 105 __be32 raddr, __be32 laddr,
1da177e4
LT		 106 int dif)
107{
108 struct hlist_node *node;
109
110 sk_for_each_from(sk, node) {
111 struct inet_sock *inet = inet_sk(sk);
112
113 if (inet->num == num &&
114 !(inet->daddr && inet->daddr != raddr) &&
115 !(inet->rcv_saddr && inet->rcv_saddr != laddr) &&
116 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
117 goto found; /* gotcha */
118 }
119 sk = NULL;
120found:
121 return sk;
122}
123
124/*
125 * 0 - deliver
126 * 1 - block
127 */
128static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
129{
130 int type;
131
132 if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
133 return 1;
134
135 type = skb->h.icmph->type;
136 if (type < 32) {
137 __u32 data = raw_sk(sk)->filter.data;
138
139 return ((1 << type) & data) != 0;
140 }
141
142 /* Do not block unknown ICMP types */
143 return 0;
144}
145
146/* IP input processing comes here for RAW socket delivery.
147 * Caller owns SKB, so we must make clones.
148 *
149 * RFC 1122: SHOULD pass TOS value up to the transport layer.
150 * -> It does. And not only TOS, but all IP header.
151 */
d13964f4 152int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
1da177e4
LT		 153{
154 struct sock *sk;
155 struct hlist_head *head;
d13964f4 156 int delivered = 0;
1da177e4
LT		 157
158 read_lock(&raw_v4_lock);
159 head = &raw_v4_htable[hash];
160 if (hlist_empty(head))
161 goto out;
162 sk = __raw_v4_lookup(__sk_head(head), iph->protocol,
163 iph->saddr, iph->daddr,
164 skb->dev->ifindex);
165
166 while (sk) {
d13964f4 167 delivered = 1;
1da177e4
LT		 168 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
169 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
170
171 /* Not releasing hash table! */
172 if (clone)
173 raw_rcv(sk, clone);
174 }
175 sk = __raw_v4_lookup(sk_next(sk), iph->protocol,
176 iph->saddr, iph->daddr,
177 skb->dev->ifindex);
178 }
179out:
180 read_unlock(&raw_v4_lock);
d13964f4 181 return delivered;
1da177e4
LT		 182}
183
184void raw_err (struct sock *sk, struct sk_buff *skb, u32 info)
185{
186 struct inet_sock *inet = inet_sk(sk);
187 int type = skb->h.icmph->type;
188 int code = skb->h.icmph->code;
189 int err = 0;
190 int harderr = 0;
191
192 /* Report error on raw socket, if:
193 1. User requested ip_recverr.
194 2. Socket is connected (otherwise the error indication
195 is useless without ip_recverr and error is hard.
196 */
197 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
198 return;
199
200 switch (type) {
201 default:
202 case ICMP_TIME_EXCEEDED:
203 err = EHOSTUNREACH;
204 break;
205 case ICMP_SOURCE_QUENCH:
206 return;
207 case ICMP_PARAMETERPROB:
208 err = EPROTO;
209 harderr = 1;
210 break;
211 case ICMP_DEST_UNREACH:
212 err = EHOSTUNREACH;
213 if (code > NR_ICMP_UNREACH)
214 break;
215 err = icmp_err_convert[code].errno;
216 harderr = icmp_err_convert[code].fatal;
217 if (code == ICMP_FRAG_NEEDED) {
218 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
219 err = EMSGSIZE;
220 }
221 }
222
223 if (inet->recverr) {
224 struct iphdr *iph = (struct iphdr*)skb->data;
225 u8 *payload = skb->data + (iph->ihl << 2);
226
227 if (inet->hdrincl)
228 payload = skb->data;
229 ip_icmp_error(sk, skb, err, 0, info, payload);
230 }
231
232 if (inet->recverr || harderr) {
233 sk->sk_err = err;
234 sk->sk_error_report(sk);
235 }
236}
237
238static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
239{
240 /* Charge it to the socket. */
241
242 if (sock_queue_rcv_skb(sk, skb) < 0) {
243 /* FIXME: increment a raw drops counter here */
244 kfree_skb(skb);
245 return NET_RX_DROP;
246 }
247
248 return NET_RX_SUCCESS;
249}
250
251int raw_rcv(struct sock *sk, struct sk_buff *skb)
252{
253 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
254 kfree_skb(skb);
255 return NET_RX_DROP;
256 }
b59c2701 257 nf_reset(skb);
1da177e4
LT		 258
259 skb_push(skb, skb->data - skb->nh.raw);
260
261 raw_rcv_skb(sk, skb);
262 return 0;
263}
264
f7d7fc03 265static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
1da177e4
LT		 266 struct rtable *rt,
267 unsigned int flags)
268{
269 struct inet_sock *inet = inet_sk(sk);
270 int hh_len;
271 struct iphdr *iph;
272 struct sk_buff *skb;
273 int err;
274
275 if (length > rt->u.dst.dev->mtu) {
276 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
277 rt->u.dst.dev->mtu);
278 return -EMSGSIZE;
279 }
280 if (flags&MSG_PROBE)
281 goto out;
282
283 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
284
285 skb = sock_alloc_send_skb(sk, length+hh_len+15,
286 flags&MSG_DONTWAIT, &err);
287 if (skb == NULL)
288 goto error;
289 skb_reserve(skb, hh_len);
290
291 skb->priority = sk->sk_priority;
292 skb->dst = dst_clone(&rt->u.dst);
293
294 skb->nh.iph = iph = (struct iphdr *)skb_put(skb, length);
295
296 skb->ip_summed = CHECKSUM_NONE;
297
298 skb->h.raw = skb->nh.raw;
299 err = memcpy_fromiovecend((void *)iph, from, 0, length);
300 if (err)
301 goto error_fault;
302
303 /* We don't modify invalid header */
f7d7fc03 304 if (length >= sizeof(*iph) && iph->ihl * 4U <= length) {
1da177e4
LT		 305 if (!iph->saddr)
306 iph->saddr = rt->rt_src;
307 iph->check = 0;
308 iph->tot_len = htons(length);
309 if (!iph->id)
310 ip_select_ident(iph, &rt->u.dst, NULL);
311
312 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
313 }
314
315 err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
316 dst_output);
317 if (err > 0)
318 err = inet->recverr ? net_xmit_errno(err) : 0;
319 if (err)
320 goto error;
321out:
322 return 0;
323
324error_fault:
325 err = -EFAULT;
326 kfree_skb(skb);
327error:
328 IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
329 return err;
330}
331
a27b58fe 332static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
1da177e4
LT		 333{
334 struct iovec *iov;
335 u8 __user *type = NULL;
336 u8 __user *code = NULL;
337 int probed = 0;
93765d8a 338 unsigned int i;
1da177e4
LT		 339
340 if (!msg->msg_iov)
a27b58fe 341 return 0;
1da177e4
LT		 342
343 for (i = 0; i < msg->msg_iovlen; i++) {
344 iov = &msg->msg_iov[i];
345 if (!iov)
346 continue;
347
348 switch (fl->proto) {
349 case IPPROTO_ICMP:
350 /* check if one-byte field is readable or not. */
351 if (iov->iov_base && iov->iov_len < 1)
352 break;
353
354 if (!type) {
355 type = iov->iov_base;
356 /* check if code field is readable or not. */
357 if (iov->iov_len > 1)
358 code = type + 1;
359 } else if (!code)
360 code = iov->iov_base;
361
362 if (type && code) {
a27b58fe
HC		 363 if (get_user(fl->fl_icmp_type, type) ||
364 get_user(fl->fl_icmp_code, code))
365 return -EFAULT;
1da177e4
LT		 366 probed = 1;
367 }
368 break;
369 default:
370 probed = 1;
371 break;
372 }
373 if (probed)
374 break;
375 }
a27b58fe 376 return 0;
1da177e4
LT		 377}
378
379static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
380 size_t len)
381{
382 struct inet_sock *inet = inet_sk(sk);
383 struct ipcm_cookie ipc;
384 struct rtable *rt = NULL;
385 int free = 0;
3ca3c68e 386 __be32 daddr;
c1d18f9f 387 __be32 saddr;
1da177e4
LT		 388 u8 tos;
389 int err;
390
391 err = -EMSGSIZE;
926d4b81 392 if (len > 0xFFFF)
1da177e4
LT		 393 goto out;
394
395 /*
396 * Check the flags.
397 */
398
399 err = -EOPNOTSUPP;
400 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
401 goto out; /* compatibility */
402
403 /*
404 * Get and verify the address.
405 */
406
407 if (msg->msg_namelen) {
408 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
409 err = -EINVAL;
410 if (msg->msg_namelen < sizeof(*usin))
411 goto out;
412 if (usin->sin_family != AF_INET) {
413 static int complained;
414 if (!complained++)
415 printk(KERN_INFO "%s forgot to set AF_INET in "
416 "raw sendmsg. Fix it!\n",
417 current->comm);
418 err = -EAFNOSUPPORT;
419 if (usin->sin_family)
420 goto out;
421 }
422 daddr = usin->sin_addr.s_addr;
423 /* ANK: I did not forget to get protocol from port field.
424 * I just do not know, who uses this weirdness.
425 * IP_HDRINCL is much more convenient.
426 */
427 } else {
428 err = -EDESTADDRREQ;
429 if (sk->sk_state != TCP_ESTABLISHED)
430 goto out;
431 daddr = inet->daddr;
432 }
433
434 ipc.addr = inet->saddr;
435 ipc.opt = NULL;
436 ipc.oif = sk->sk_bound_dev_if;
437
438 if (msg->msg_controllen) {
439 err = ip_cmsg_send(msg, &ipc);
440 if (err)
441 goto out;
442 if (ipc.opt)
443 free = 1;
444 }
445
446 saddr = ipc.addr;
447 ipc.addr = daddr;
448
449 if (!ipc.opt)
450 ipc.opt = inet->opt;
451
452 if (ipc.opt) {
453 err = -EINVAL;
454 /* Linux does not mangle headers on raw sockets,
455 * so that IP options + IP_HDRINCL is non-sense.
456 */
457 if (inet->hdrincl)
458 goto done;
459 if (ipc.opt->srr) {
460 if (!daddr)
461 goto done;
462 daddr = ipc.opt->faddr;
463 }
464 }
465 tos = RT_CONN_FLAGS(sk);
466 if (msg->msg_flags & MSG_DONTROUTE)
467 tos |= RTO_ONLINK;
468
469 if (MULTICAST(daddr)) {
470 if (!ipc.oif)
471 ipc.oif = inet->mc_index;
472 if (!saddr)
473 saddr = inet->mc_addr;
474 }
475
476 {
477 struct flowi fl = { .oif = ipc.oif,
478 .nl_u = { .ip4_u =
479 { .daddr = daddr,
480 .saddr = saddr,
481 .tos = tos } },
482 .proto = inet->hdrincl ? IPPROTO_RAW :
483 sk->sk_protocol,
484 };
a27b58fe
HC		 485 if (!inet->hdrincl) {
486 err = raw_probe_proto_opt(&fl, msg);
487 if (err)
488 goto done;
489 }
1da177e4 490
beb8d13b 491 security_sk_classify_flow(sk, &fl);
8eb9086f 492 err = ip_route_output_flow(&rt, &fl, sk, 1);
1da177e4
LT		 493 }
494 if (err)
495 goto done;
496
497 err = -EACCES;
498 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
499 goto done;
500
501 if (msg->msg_flags & MSG_CONFIRM)
502 goto do_confirm;
503back_from_confirm:
504
505 if (inet->hdrincl)
506 err = raw_send_hdrinc(sk, msg->msg_iov, len,
507 rt, msg->msg_flags);
508
509 else {
510 if (!ipc.addr)
511 ipc.addr = rt->rt_dst;
512 lock_sock(sk);
513 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
514 &ipc, rt, msg->msg_flags);
515 if (err)
516 ip_flush_pending_frames(sk);
517 else if (!(msg->msg_flags & MSG_MORE))
518 err = ip_push_pending_frames(sk);
519 release_sock(sk);
520 }
521done:
522 if (free)
523 kfree(ipc.opt);
524 ip_rt_put(rt);
525
5418c692
JJ		 526out:
527 if (err < 0)
528 return err;
529 return len;
1da177e4
LT		 530
531do_confirm:
532 dst_confirm(&rt->u.dst);
533 if (!(msg->msg_flags & MSG_PROBE) || len)
534 goto back_from_confirm;
535 err = 0;
536 goto done;
537}
538
539static void raw_close(struct sock *sk, long timeout)
540{
541 /*
542 * Raw sockets may have direct kernel refereneces. Kill them.
543 */
544 ip_ra_control(sk, 0, NULL);
545
546 sk_common_release(sk);
547}
548
549/* This gets rid of all the nasties in af_inet. -DaveM */
550static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
551{
552 struct inet_sock *inet = inet_sk(sk);
553 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
554 int ret = -EINVAL;
555 int chk_addr_ret;
556
557 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
558 goto out;
559 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
560 ret = -EADDRNOTAVAIL;
561 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
562 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
563 goto out;
564 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
565 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
566 inet->saddr = 0; /* Use device */
567 sk_dst_reset(sk);
568 ret = 0;
569out: return ret;
570}
571
572/*
573 * This should be easy, if there is something there
574 * we return it, otherwise we block.
575 */
576
577static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
578 size_t len, int noblock, int flags, int *addr_len)
579{
580 struct inet_sock *inet = inet_sk(sk);
581 size_t copied = 0;
582 int err = -EOPNOTSUPP;
583 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
584 struct sk_buff *skb;
585
586 if (flags & MSG_OOB)
587 goto out;
588
589 if (addr_len)
590 *addr_len = sizeof(*sin);
591
592 if (flags & MSG_ERRQUEUE) {
593 err = ip_recv_error(sk, msg, len);
594 goto out;
595 }
596
597 skb = skb_recv_datagram(sk, flags, noblock, &err);
598 if (!skb)
599 goto out;
600
601 copied = skb->len;
602 if (len < copied) {
603 msg->msg_flags |= MSG_TRUNC;
604 copied = len;
605 }
606
607 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
608 if (err)
609 goto done;
610
611 sock_recv_timestamp(msg, sk, skb);
612
613 /* Copy the address. */
614 if (sin) {
615 sin->sin_family = AF_INET;
616 sin->sin_addr.s_addr = skb->nh.iph->saddr;
f59fc7f3 617 sin->sin_port = 0;
1da177e4
LT		 618 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
619 }
620 if (inet->cmsg_flags)
621 ip_cmsg_recv(msg, skb);
622 if (flags & MSG_TRUNC)
623 copied = skb->len;
624done:
625 skb_free_datagram(sk, skb);
5418c692
JJ		 626out:
627 if (err)
628 return err;
629 return copied;
1da177e4
LT		 630}
631
632static int raw_init(struct sock *sk)
633{
634 struct raw_sock *rp = raw_sk(sk);
635
636 if (inet_sk(sk)->num == IPPROTO_ICMP)
637 memset(&rp->filter, 0, sizeof(rp->filter));
638 return 0;
639}
640
641static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
642{
643 if (optlen > sizeof(struct icmp_filter))
644 optlen = sizeof(struct icmp_filter);
645 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
646 return -EFAULT;
647 return 0;
648}
649
650static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
651{
652 int len, ret = -EFAULT;
653
654 if (get_user(len, optlen))
655 goto out;
656 ret = -EINVAL;
657 if (len < 0)
658 goto out;
659 if (len > sizeof(struct icmp_filter))
660 len = sizeof(struct icmp_filter);
661 ret = -EFAULT;
662 if (put_user(len, optlen) ||
663 copy_to_user(optval, &raw_sk(sk)->filter, len))
664 goto out;
665 ret = 0;
666out: return ret;
667}
668
3fdadf7d 669static int do_raw_setsockopt(struct sock *sk, int level, int optname,
1da177e4
LT		 670 char __user *optval, int optlen)
671{
1da177e4
LT		 672 if (optname == ICMP_FILTER) {
673 if (inet_sk(sk)->num != IPPROTO_ICMP)
674 return -EOPNOTSUPP;
675 else
676 return raw_seticmpfilter(sk, optval, optlen);
677 }
678 return -ENOPROTOOPT;
679}
680
3fdadf7d
DM		 681static int raw_setsockopt(struct sock *sk, int level, int optname,
682 char __user *optval, int optlen)
1da177e4
LT		 683{
684 if (level != SOL_RAW)
3fdadf7d
DM		 685 return ip_setsockopt(sk, level, optname, optval, optlen);
686 return do_raw_setsockopt(sk, level, optname, optval, optlen);
687}
1da177e4 688
3fdadf7d
DM		 689#ifdef CONFIG_COMPAT
690static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
543d9cfe 691 char __user *optval, int optlen)
3fdadf7d
DM		 692{
693 if (level != SOL_RAW)
543d9cfe 694 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM		 695 return do_raw_setsockopt(sk, level, optname, optval, optlen);
696}
697#endif
698
699static int do_raw_getsockopt(struct sock *sk, int level, int optname,
700 char __user *optval, int __user *optlen)
701{
1da177e4
LT		 702 if (optname == ICMP_FILTER) {
703 if (inet_sk(sk)->num != IPPROTO_ICMP)
704 return -EOPNOTSUPP;
705 else
706 return raw_geticmpfilter(sk, optval, optlen);
707 }
708 return -ENOPROTOOPT;
709}
710
3fdadf7d
DM		 711static int raw_getsockopt(struct sock *sk, int level, int optname,
712 char __user *optval, int __user *optlen)
713{
714 if (level != SOL_RAW)
715 return ip_getsockopt(sk, level, optname, optval, optlen);
716 return do_raw_getsockopt(sk, level, optname, optval, optlen);
717}
718
719#ifdef CONFIG_COMPAT
720static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
543d9cfe 721 char __user *optval, int __user *optlen)
3fdadf7d
DM		 722{
723 if (level != SOL_RAW)
543d9cfe 724 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
3fdadf7d
DM		 725 return do_raw_getsockopt(sk, level, optname, optval, optlen);
726}
727#endif
728
1da177e4
LT		 729static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
730{
731 switch (cmd) {
732 case SIOCOUTQ: {
733 int amount = atomic_read(&sk->sk_wmem_alloc);
734 return put_user(amount, (int __user *)arg);
735 }
736 case SIOCINQ: {
737 struct sk_buff *skb;
738 int amount = 0;
739
e0f9f858 740 spin_lock_bh(&sk->sk_receive_queue.lock);
1da177e4
LT		 741 skb = skb_peek(&sk->sk_receive_queue);
742 if (skb != NULL)
743 amount = skb->len;
e0f9f858 744 spin_unlock_bh(&sk->sk_receive_queue.lock);
1da177e4
LT		 745 return put_user(amount, (int __user *)arg);
746 }
747
748 default:
749#ifdef CONFIG_IP_MROUTE
750 return ipmr_ioctl(sk, cmd, (void __user *)arg);
751#else
752 return -ENOIOCTLCMD;
753#endif
754 }
755}
756
757struct proto raw_prot = {
543d9cfe
ACM		 758 .name = "RAW",
759 .owner = THIS_MODULE,
760 .close = raw_close,
761 .connect = ip4_datagram_connect,
762 .disconnect = udp_disconnect,
763 .ioctl = raw_ioctl,
764 .init = raw_init,
765 .setsockopt = raw_setsockopt,
766 .getsockopt = raw_getsockopt,
767 .sendmsg = raw_sendmsg,
768 .recvmsg = raw_recvmsg,
769 .bind = raw_bind,
770 .backlog_rcv = raw_rcv_skb,
771 .hash = raw_v4_hash,
772 .unhash = raw_v4_unhash,
773 .obj_size = sizeof(struct raw_sock),
3fdadf7d 774#ifdef CONFIG_COMPAT
543d9cfe
ACM		 775 .compat_setsockopt = compat_raw_setsockopt,
776 .compat_getsockopt = compat_raw_getsockopt,
3fdadf7d 777#endif
1da177e4
LT		 778};
779
780#ifdef CONFIG_PROC_FS
781struct raw_iter_state {
782 int bucket;
783};
784
785#define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
786
787static struct sock *raw_get_first(struct seq_file *seq)
788{
789 struct sock *sk;
790 struct raw_iter_state* state = raw_seq_private(seq);
791
792 for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
793 struct hlist_node *node;
794
795 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
796 if (sk->sk_family == PF_INET)
797 goto found;
798 }
799 sk = NULL;
800found:
801 return sk;
802}
803
804static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
805{
806 struct raw_iter_state* state = raw_seq_private(seq);
807
808 do {
809 sk = sk_next(sk);
810try_again:
811 ;
812 } while (sk && sk->sk_family != PF_INET);
813
814 if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
815 sk = sk_head(&raw_v4_htable[state->bucket]);
816 goto try_again;
817 }
818 return sk;
819}
820
821static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
822{
823 struct sock *sk = raw_get_first(seq);
824
825 if (sk)
826 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
827 --pos;
828 return pos ? NULL : sk;
829}
830
831static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
832{
833 read_lock(&raw_v4_lock);
834 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
835}
836
837static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
838{
839 struct sock *sk;
840
841 if (v == SEQ_START_TOKEN)
842 sk = raw_get_first(seq);
843 else
844 sk = raw_get_next(seq, v);
845 ++*pos;
846 return sk;
847}
848
849static void raw_seq_stop(struct seq_file *seq, void *v)
850{
851 read_unlock(&raw_v4_lock);
852}
853
854static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
855{
856 struct inet_sock *inet = inet_sk(sp);
714e85be
AV		 857 __be32 dest = inet->daddr,
858 src = inet->rcv_saddr;
1da177e4
LT		 859 __u16 destp = 0,
860 srcp = inet->num;
861
862 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
863 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
864 i, src, srcp, dest, destp, sp->sk_state,
865 atomic_read(&sp->sk_wmem_alloc),
866 atomic_read(&sp->sk_rmem_alloc),
867 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
868 atomic_read(&sp->sk_refcnt), sp);
869 return tmpbuf;
870}
871
872static int raw_seq_show(struct seq_file *seq, void *v)
873{
874 char tmpbuf[129];
875
876 if (v == SEQ_START_TOKEN)
877 seq_printf(seq, "%-127s\n",
878 " sl local_address rem_address st tx_queue "
879 "rx_queue tr tm->when retrnsmt uid timeout "
880 "inode");
881 else {
882 struct raw_iter_state *state = raw_seq_private(seq);
883
884 seq_printf(seq, "%-127s\n",
885 get_raw_sock(v, tmpbuf, state->bucket));
886 }
887 return 0;
888}
889
890static struct seq_operations raw_seq_ops = {
891 .start = raw_seq_start,
892 .next = raw_seq_next,
893 .stop = raw_seq_stop,
894 .show = raw_seq_show,
895};
896
897static int raw_seq_open(struct inode *inode, struct file *file)
898{
899 struct seq_file *seq;
900 int rc = -ENOMEM;
901 struct raw_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
902
903 if (!s)
904 goto out;
905 rc = seq_open(file, &raw_seq_ops);
906 if (rc)
907 goto out_kfree;
908
909 seq = file->private_data;
910 seq->private = s;
911 memset(s, 0, sizeof(*s));
912out:
913 return rc;
914out_kfree:
915 kfree(s);
916 goto out;
917}
918
919static struct file_operations raw_seq_fops = {
920 .owner = THIS_MODULE,
921 .open = raw_seq_open,
922 .read = seq_read,
923 .llseek = seq_lseek,
924 .release = seq_release_private,
925};
926
927int __init raw_proc_init(void)
928{
929 if (!proc_net_fops_create("raw", S_IRUGO, &raw_seq_fops))
930 return -ENOMEM;
931 return 0;
932}
933
934void __init raw_proc_exit(void)
935{
936 proc_net_remove("raw");
937}
938#endif /* CONFIG_PROC_FS */
Clone of davem's net-next-2.6 tree