blob: b430cf2a4f6609db27f6ea137e3f18b0281bd442 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * 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 */
Dave Jonesbf0d5242006-09-22 14:00:29 -070041
Alan Cox715b49e2006-01-18 17:44:07 -080042#include <linux/types.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#include <asm/atomic.h>
44#include <asm/byteorder.h>
45#include <asm/current.h>
46#include <asm/uaccess.h>
47#include <asm/ioctls.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#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>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -070072#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070073#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,
Alexey Dobriyan6d741652006-06-05 21:06:41 -0700105 __be32 raddr, __be32 laddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106 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 */
Patrick McHardyd13964f2005-08-09 19:45:02 -0700152int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153{
154 struct sock *sk;
155 struct hlist_head *head;
Patrick McHardyd13964f2005-08-09 19:45:02 -0700156 int delivered = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157
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) {
Patrick McHardyd13964f2005-08-09 19:45:02 -0700167 delivered = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 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);
Patrick McHardyd13964f2005-08-09 19:45:02 -0700181 return delivered;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182}
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 }
Patrick McHardyb59c2702006-01-06 23:06:10 -0800257 nf_reset(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258
259 skb_push(skb, skb->data - skb->nh.raw);
260
261 raw_rcv_skb(sk, skb);
262 return 0;
263}
264
Jesper Juhlf7d7fc02005-06-18 23:00:34 -0700265static int raw_send_hdrinc(struct sock *sk, void *from, size_t length,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 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 */
Jesper Juhlf7d7fc02005-06-18 23:00:34 -0700304 if (length >= sizeof(*iph) && iph->ihl * 4U <= length) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 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
332static void raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
333{
334 struct iovec *iov;
335 u8 __user *type = NULL;
336 u8 __user *code = NULL;
337 int probed = 0;
Jesper Juhl93765d82005-06-18 23:00:15 -0700338 unsigned int i;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339
340 if (!msg->msg_iov)
341 return;
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) {
363 get_user(fl->fl_icmp_type, type);
Mark J Cox6d1cfe32005-09-19 17:55:30 -0700364 get_user(fl->fl_icmp_code, code);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 probed = 1;
366 }
367 break;
368 default:
369 probed = 1;
370 break;
371 }
372 if (probed)
373 break;
374 }
375}
376
377static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
378 size_t len)
379{
380 struct inet_sock *inet = inet_sk(sk);
381 struct ipcm_cookie ipc;
382 struct rtable *rt = NULL;
383 int free = 0;
Al Viro3ca3c682006-09-27 18:28:07 -0700384 __be32 daddr;
Al Viroc1d18f92006-09-27 18:28:28 -0700385 __be32 saddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 u8 tos;
387 int err;
388
389 err = -EMSGSIZE;
Jesper Juhl926d4b82005-06-18 23:00:00 -0700390 if (len > 0xFFFF)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 goto out;
392
393 /*
394 * Check the flags.
395 */
396
397 err = -EOPNOTSUPP;
398 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
399 goto out; /* compatibility */
400
401 /*
402 * Get and verify the address.
403 */
404
405 if (msg->msg_namelen) {
406 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
407 err = -EINVAL;
408 if (msg->msg_namelen < sizeof(*usin))
409 goto out;
410 if (usin->sin_family != AF_INET) {
411 static int complained;
412 if (!complained++)
413 printk(KERN_INFO "%s forgot to set AF_INET in "
414 "raw sendmsg. Fix it!\n",
415 current->comm);
416 err = -EAFNOSUPPORT;
417 if (usin->sin_family)
418 goto out;
419 }
420 daddr = usin->sin_addr.s_addr;
421 /* ANK: I did not forget to get protocol from port field.
422 * I just do not know, who uses this weirdness.
423 * IP_HDRINCL is much more convenient.
424 */
425 } else {
426 err = -EDESTADDRREQ;
427 if (sk->sk_state != TCP_ESTABLISHED)
428 goto out;
429 daddr = inet->daddr;
430 }
431
432 ipc.addr = inet->saddr;
433 ipc.opt = NULL;
434 ipc.oif = sk->sk_bound_dev_if;
435
436 if (msg->msg_controllen) {
437 err = ip_cmsg_send(msg, &ipc);
438 if (err)
439 goto out;
440 if (ipc.opt)
441 free = 1;
442 }
443
444 saddr = ipc.addr;
445 ipc.addr = daddr;
446
447 if (!ipc.opt)
448 ipc.opt = inet->opt;
449
450 if (ipc.opt) {
451 err = -EINVAL;
452 /* Linux does not mangle headers on raw sockets,
453 * so that IP options + IP_HDRINCL is non-sense.
454 */
455 if (inet->hdrincl)
456 goto done;
457 if (ipc.opt->srr) {
458 if (!daddr)
459 goto done;
460 daddr = ipc.opt->faddr;
461 }
462 }
463 tos = RT_CONN_FLAGS(sk);
464 if (msg->msg_flags & MSG_DONTROUTE)
465 tos |= RTO_ONLINK;
466
467 if (MULTICAST(daddr)) {
468 if (!ipc.oif)
469 ipc.oif = inet->mc_index;
470 if (!saddr)
471 saddr = inet->mc_addr;
472 }
473
474 {
475 struct flowi fl = { .oif = ipc.oif,
476 .nl_u = { .ip4_u =
477 { .daddr = daddr,
478 .saddr = saddr,
479 .tos = tos } },
480 .proto = inet->hdrincl ? IPPROTO_RAW :
481 sk->sk_protocol,
482 };
483 if (!inet->hdrincl)
484 raw_probe_proto_opt(&fl, msg);
485
Venkat Yekkiralabeb8d132006-08-04 23:12:42 -0700486 security_sk_classify_flow(sk, &fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
488 }
489 if (err)
490 goto done;
491
492 err = -EACCES;
493 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
494 goto done;
495
496 if (msg->msg_flags & MSG_CONFIRM)
497 goto do_confirm;
498back_from_confirm:
499
500 if (inet->hdrincl)
501 err = raw_send_hdrinc(sk, msg->msg_iov, len,
502 rt, msg->msg_flags);
503
504 else {
505 if (!ipc.addr)
506 ipc.addr = rt->rt_dst;
507 lock_sock(sk);
508 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
509 &ipc, rt, msg->msg_flags);
510 if (err)
511 ip_flush_pending_frames(sk);
512 else if (!(msg->msg_flags & MSG_MORE))
513 err = ip_push_pending_frames(sk);
514 release_sock(sk);
515 }
516done:
517 if (free)
518 kfree(ipc.opt);
519 ip_rt_put(rt);
520
Jesper Juhl5418c6922005-06-18 22:59:45 -0700521out:
522 if (err < 0)
523 return err;
524 return len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525
526do_confirm:
527 dst_confirm(&rt->u.dst);
528 if (!(msg->msg_flags & MSG_PROBE) || len)
529 goto back_from_confirm;
530 err = 0;
531 goto done;
532}
533
534static void raw_close(struct sock *sk, long timeout)
535{
536 /*
537 * Raw sockets may have direct kernel refereneces. Kill them.
538 */
539 ip_ra_control(sk, 0, NULL);
540
541 sk_common_release(sk);
542}
543
544/* This gets rid of all the nasties in af_inet. -DaveM */
545static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
546{
547 struct inet_sock *inet = inet_sk(sk);
548 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
549 int ret = -EINVAL;
550 int chk_addr_ret;
551
552 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
553 goto out;
554 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
555 ret = -EADDRNOTAVAIL;
556 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
557 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
558 goto out;
559 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
560 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
561 inet->saddr = 0; /* Use device */
562 sk_dst_reset(sk);
563 ret = 0;
564out: return ret;
565}
566
567/*
568 * This should be easy, if there is something there
569 * we return it, otherwise we block.
570 */
571
572static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
573 size_t len, int noblock, int flags, int *addr_len)
574{
575 struct inet_sock *inet = inet_sk(sk);
576 size_t copied = 0;
577 int err = -EOPNOTSUPP;
578 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
579 struct sk_buff *skb;
580
581 if (flags & MSG_OOB)
582 goto out;
583
584 if (addr_len)
585 *addr_len = sizeof(*sin);
586
587 if (flags & MSG_ERRQUEUE) {
588 err = ip_recv_error(sk, msg, len);
589 goto out;
590 }
591
592 skb = skb_recv_datagram(sk, flags, noblock, &err);
593 if (!skb)
594 goto out;
595
596 copied = skb->len;
597 if (len < copied) {
598 msg->msg_flags |= MSG_TRUNC;
599 copied = len;
600 }
601
602 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
603 if (err)
604 goto done;
605
606 sock_recv_timestamp(msg, sk, skb);
607
608 /* Copy the address. */
609 if (sin) {
610 sin->sin_family = AF_INET;
611 sin->sin_addr.s_addr = skb->nh.iph->saddr;
Tetsuo Handaf59fc7f2006-07-25 17:05:35 -0700612 sin->sin_port = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
614 }
615 if (inet->cmsg_flags)
616 ip_cmsg_recv(msg, skb);
617 if (flags & MSG_TRUNC)
618 copied = skb->len;
619done:
620 skb_free_datagram(sk, skb);
Jesper Juhl5418c6922005-06-18 22:59:45 -0700621out:
622 if (err)
623 return err;
624 return copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625}
626
627static int raw_init(struct sock *sk)
628{
629 struct raw_sock *rp = raw_sk(sk);
630
631 if (inet_sk(sk)->num == IPPROTO_ICMP)
632 memset(&rp->filter, 0, sizeof(rp->filter));
633 return 0;
634}
635
636static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
637{
638 if (optlen > sizeof(struct icmp_filter))
639 optlen = sizeof(struct icmp_filter);
640 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
641 return -EFAULT;
642 return 0;
643}
644
645static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
646{
647 int len, ret = -EFAULT;
648
649 if (get_user(len, optlen))
650 goto out;
651 ret = -EINVAL;
652 if (len < 0)
653 goto out;
654 if (len > sizeof(struct icmp_filter))
655 len = sizeof(struct icmp_filter);
656 ret = -EFAULT;
657 if (put_user(len, optlen) ||
658 copy_to_user(optval, &raw_sk(sk)->filter, len))
659 goto out;
660 ret = 0;
661out: return ret;
662}
663
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800664static int do_raw_setsockopt(struct sock *sk, int level, int optname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 char __user *optval, int optlen)
666{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667 if (optname == ICMP_FILTER) {
668 if (inet_sk(sk)->num != IPPROTO_ICMP)
669 return -EOPNOTSUPP;
670 else
671 return raw_seticmpfilter(sk, optval, optlen);
672 }
673 return -ENOPROTOOPT;
674}
675
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800676static int raw_setsockopt(struct sock *sk, int level, int optname,
677 char __user *optval, int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678{
679 if (level != SOL_RAW)
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800680 return ip_setsockopt(sk, level, optname, optval, optlen);
681 return do_raw_setsockopt(sk, level, optname, optval, optlen);
682}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800684#ifdef CONFIG_COMPAT
685static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800686 char __user *optval, int optlen)
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800687{
688 if (level != SOL_RAW)
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800689 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800690 return do_raw_setsockopt(sk, level, optname, optval, optlen);
691}
692#endif
693
694static int do_raw_getsockopt(struct sock *sk, int level, int optname,
695 char __user *optval, int __user *optlen)
696{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 if (optname == ICMP_FILTER) {
698 if (inet_sk(sk)->num != IPPROTO_ICMP)
699 return -EOPNOTSUPP;
700 else
701 return raw_geticmpfilter(sk, optval, optlen);
702 }
703 return -ENOPROTOOPT;
704}
705
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800706static int raw_getsockopt(struct sock *sk, int level, int optname,
707 char __user *optval, int __user *optlen)
708{
709 if (level != SOL_RAW)
710 return ip_getsockopt(sk, level, optname, optval, optlen);
711 return do_raw_getsockopt(sk, level, optname, optval, optlen);
712}
713
714#ifdef CONFIG_COMPAT
715static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800716 char __user *optval, int __user *optlen)
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800717{
718 if (level != SOL_RAW)
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800719 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800720 return do_raw_getsockopt(sk, level, optname, optval, optlen);
721}
722#endif
723
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
725{
726 switch (cmd) {
727 case SIOCOUTQ: {
728 int amount = atomic_read(&sk->sk_wmem_alloc);
729 return put_user(amount, (int __user *)arg);
730 }
731 case SIOCINQ: {
732 struct sk_buff *skb;
733 int amount = 0;
734
Herbert Xue0f9f852005-06-18 22:56:18 -0700735 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 skb = skb_peek(&sk->sk_receive_queue);
737 if (skb != NULL)
738 amount = skb->len;
Herbert Xue0f9f852005-06-18 22:56:18 -0700739 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 return put_user(amount, (int __user *)arg);
741 }
742
743 default:
744#ifdef CONFIG_IP_MROUTE
745 return ipmr_ioctl(sk, cmd, (void __user *)arg);
746#else
747 return -ENOIOCTLCMD;
748#endif
749 }
750}
751
752struct proto raw_prot = {
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800753 .name = "RAW",
754 .owner = THIS_MODULE,
755 .close = raw_close,
756 .connect = ip4_datagram_connect,
757 .disconnect = udp_disconnect,
758 .ioctl = raw_ioctl,
759 .init = raw_init,
760 .setsockopt = raw_setsockopt,
761 .getsockopt = raw_getsockopt,
762 .sendmsg = raw_sendmsg,
763 .recvmsg = raw_recvmsg,
764 .bind = raw_bind,
765 .backlog_rcv = raw_rcv_skb,
766 .hash = raw_v4_hash,
767 .unhash = raw_v4_unhash,
768 .obj_size = sizeof(struct raw_sock),
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800769#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -0800770 .compat_setsockopt = compat_raw_setsockopt,
771 .compat_getsockopt = compat_raw_getsockopt,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800772#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773};
774
775#ifdef CONFIG_PROC_FS
776struct raw_iter_state {
777 int bucket;
778};
779
780#define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
781
782static struct sock *raw_get_first(struct seq_file *seq)
783{
784 struct sock *sk;
785 struct raw_iter_state* state = raw_seq_private(seq);
786
787 for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
788 struct hlist_node *node;
789
790 sk_for_each(sk, node, &raw_v4_htable[state->bucket])
791 if (sk->sk_family == PF_INET)
792 goto found;
793 }
794 sk = NULL;
795found:
796 return sk;
797}
798
799static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
800{
801 struct raw_iter_state* state = raw_seq_private(seq);
802
803 do {
804 sk = sk_next(sk);
805try_again:
806 ;
807 } while (sk && sk->sk_family != PF_INET);
808
809 if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
810 sk = sk_head(&raw_v4_htable[state->bucket]);
811 goto try_again;
812 }
813 return sk;
814}
815
816static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
817{
818 struct sock *sk = raw_get_first(seq);
819
820 if (sk)
821 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
822 --pos;
823 return pos ? NULL : sk;
824}
825
826static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
827{
828 read_lock(&raw_v4_lock);
829 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
830}
831
832static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
833{
834 struct sock *sk;
835
836 if (v == SEQ_START_TOKEN)
837 sk = raw_get_first(seq);
838 else
839 sk = raw_get_next(seq, v);
840 ++*pos;
841 return sk;
842}
843
844static void raw_seq_stop(struct seq_file *seq, void *v)
845{
846 read_unlock(&raw_v4_lock);
847}
848
849static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
850{
851 struct inet_sock *inet = inet_sk(sp);
852 unsigned int dest = inet->daddr,
853 src = inet->rcv_saddr;
854 __u16 destp = 0,
855 srcp = inet->num;
856
857 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
858 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
859 i, src, srcp, dest, destp, sp->sk_state,
860 atomic_read(&sp->sk_wmem_alloc),
861 atomic_read(&sp->sk_rmem_alloc),
862 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
863 atomic_read(&sp->sk_refcnt), sp);
864 return tmpbuf;
865}
866
867static int raw_seq_show(struct seq_file *seq, void *v)
868{
869 char tmpbuf[129];
870
871 if (v == SEQ_START_TOKEN)
872 seq_printf(seq, "%-127s\n",
873 " sl local_address rem_address st tx_queue "
874 "rx_queue tr tm->when retrnsmt uid timeout "
875 "inode");
876 else {
877 struct raw_iter_state *state = raw_seq_private(seq);
878
879 seq_printf(seq, "%-127s\n",
880 get_raw_sock(v, tmpbuf, state->bucket));
881 }
882 return 0;
883}
884
885static struct seq_operations raw_seq_ops = {
886 .start = raw_seq_start,
887 .next = raw_seq_next,
888 .stop = raw_seq_stop,
889 .show = raw_seq_show,
890};
891
892static int raw_seq_open(struct inode *inode, struct file *file)
893{
894 struct seq_file *seq;
895 int rc = -ENOMEM;
896 struct raw_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
897
898 if (!s)
899 goto out;
900 rc = seq_open(file, &raw_seq_ops);
901 if (rc)
902 goto out_kfree;
903
904 seq = file->private_data;
905 seq->private = s;
906 memset(s, 0, sizeof(*s));
907out:
908 return rc;
909out_kfree:
910 kfree(s);
911 goto out;
912}
913
914static struct file_operations raw_seq_fops = {
915 .owner = THIS_MODULE,
916 .open = raw_seq_open,
917 .read = seq_read,
918 .llseek = seq_lseek,
919 .release = seq_release_private,
920};
921
922int __init raw_proc_init(void)
923{
924 if (!proc_net_fops_create("raw", S_IRUGO, &raw_seq_fops))
925 return -ENOMEM;
926 return 0;
927}
928
929void __init raw_proc_exit(void)
930{
931 proc_net_remove("raw");
932}
933#endif /* CONFIG_PROC_FS */