projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge tag 'nfs-for-4.4-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
[deliverable/linux.git] / net / ipv4 / raw.c
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * RAW - implementation of IP "raw" sockets.
7 *
8 * Authors: Ross Biro
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 *
11 * Fixes:
12 * Alan Cox : verify_area() fixed up
13 * Alan Cox : ICMP error handling
14 * Alan Cox : EMSGSIZE if you send too big a packet
15 * Alan Cox : Now uses generic datagrams and shared
16 * skbuff library. No more peek crashes,
17 * no more backlogs
18 * Alan Cox : Checks sk->broadcast.
19 * Alan Cox : Uses skb_free_datagram/skb_copy_datagram
20 * Alan Cox : Raw passes ip options too
21 * Alan Cox : Setsocketopt added
22 * Alan Cox : Fixed error return for broadcasts
23 * Alan Cox : Removed wake_up calls
24 * Alan Cox : Use ttl/tos
25 * Alan Cox : Cleaned up old debugging
26 * Alan Cox : Use new kernel side addresses
27 * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets.
28 * Alan Cox : BSD style RAW socket demultiplexing.
29 * Alan Cox : Beginnings of mrouted support.
30 * Alan Cox : Added IP_HDRINCL option.
31 * Alan Cox : Skip broadcast check if BSDism set.
32 * David S. Miller : New socket lookup architecture.
33 *
34 * This program is free software; you can redistribute it and/or
35 * modify it under the terms of the GNU General Public License
36 * as published by the Free Software Foundation; either version
37 * 2 of the License, or (at your option) any later version.
38 */
39
40 #include <linux/types.h>
41 #include <linux/atomic.h>
42 #include <asm/byteorder.h>
43 #include <asm/current.h>
44 #include <asm/uaccess.h>
45 #include <asm/ioctls.h>
46 #include <linux/stddef.h>
47 #include <linux/slab.h>
48 #include <linux/errno.h>
49 #include <linux/kernel.h>
50 #include <linux/export.h>
51 #include <linux/spinlock.h>
52 #include <linux/sockios.h>
53 #include <linux/socket.h>
54 #include <linux/in.h>
55 #include <linux/mroute.h>
56 #include <linux/netdevice.h>
57 #include <linux/in_route.h>
58 #include <linux/route.h>
59 #include <linux/skbuff.h>
60 #include <linux/igmp.h>
61 #include <net/net_namespace.h>
62 #include <net/dst.h>
63 #include <net/sock.h>
64 #include <linux/ip.h>
65 #include <linux/net.h>
66 #include <net/ip.h>
67 #include <net/icmp.h>
68 #include <net/udp.h>
69 #include <net/raw.h>
70 #include <net/snmp.h>
71 #include <net/tcp_states.h>
72 #include <net/inet_common.h>
73 #include <net/checksum.h>
74 #include <net/xfrm.h>
75 #include <linux/rtnetlink.h>
76 #include <linux/proc_fs.h>
77 #include <linux/seq_file.h>
78 #include <linux/netfilter.h>
79 #include <linux/netfilter_ipv4.h>
80 #include <linux/compat.h>
81 #include <linux/uio.h>
82
83 struct raw_frag_vec {
84 struct msghdr *msg;
85 union {
86 struct icmphdr icmph;
87 char c[1];
88 } hdr;
89 int hlen;
90 };
91
92 static struct raw_hashinfo raw_v4_hashinfo = {
93 .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
94 };
95
96 void raw_hash_sk(struct sock *sk)
97 {
98 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
99 struct hlist_head *head;
100
101 head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)];
102
103 write_lock_bh(&h->lock);
104 sk_add_node(sk, head);
105 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
106 write_unlock_bh(&h->lock);
107 }
108 EXPORT_SYMBOL_GPL(raw_hash_sk);
109
110 void raw_unhash_sk(struct sock *sk)
111 {
112 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
113
114 write_lock_bh(&h->lock);
115 if (sk_del_node_init(sk))
116 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
117 write_unlock_bh(&h->lock);
118 }
119 EXPORT_SYMBOL_GPL(raw_unhash_sk);
120
121 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
122 unsigned short num, __be32 raddr, __be32 laddr, int dif)
123 {
124 sk_for_each_from(sk) {
125 struct inet_sock *inet = inet_sk(sk);
126
127 if (net_eq(sock_net(sk), net) && inet->inet_num == num &&
128 !(inet->inet_daddr && inet->inet_daddr != raddr) &&
129 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
130 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
131 goto found; /* gotcha */
132 }
133 sk = NULL;
134 found:
135 return sk;
136 }
137
138 /*
139 * 0 - deliver
140 * 1 - block
141 */
142 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
143 {
144 struct icmphdr _hdr;
145 const struct icmphdr *hdr;
146
147 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
148 sizeof(_hdr), &_hdr);
149 if (!hdr)
150 return 1;
151
152 if (hdr->type < 32) {
153 __u32 data = raw_sk(sk)->filter.data;
154
155 return ((1U << hdr->type) & data) != 0;
156 }
157
158 /* Do not block unknown ICMP types */
159 return 0;
160 }
161
162 /* IP input processing comes here for RAW socket delivery.
163 * Caller owns SKB, so we must make clones.
164 *
165 * RFC 1122: SHOULD pass TOS value up to the transport layer.
166 * -> It does. And not only TOS, but all IP header.
167 */
168 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
169 {
170 struct sock *sk;
171 struct hlist_head *head;
172 int delivered = 0;
173 struct net *net;
174
175 read_lock(&raw_v4_hashinfo.lock);
176 head = &raw_v4_hashinfo.ht[hash];
177 if (hlist_empty(head))
178 goto out;
179
180 net = dev_net(skb->dev);
181 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
182 iph->saddr, iph->daddr,
183 skb->dev->ifindex);
184
185 while (sk) {
186 delivered = 1;
187 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
188 ip_mc_sf_allow(sk, iph->daddr, iph->saddr,
189 skb->dev->ifindex)) {
190 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
191
192 /* Not releasing hash table! */
193 if (clone)
194 raw_rcv(sk, clone);
195 }
196 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
197 iph->saddr, iph->daddr,
198 skb->dev->ifindex);
199 }
200 out:
201 read_unlock(&raw_v4_hashinfo.lock);
202 return delivered;
203 }
204
205 int raw_local_deliver(struct sk_buff *skb, int protocol)
206 {
207 int hash;
208 struct sock *raw_sk;
209
210 hash = protocol & (RAW_HTABLE_SIZE - 1);
211 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
212
213 /* If there maybe a raw socket we must check - if not we
214 * don't care less
215 */
216 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
217 raw_sk = NULL;
218
219 return raw_sk != NULL;
220
221 }
222
223 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
224 {
225 struct inet_sock *inet = inet_sk(sk);
226 const int type = icmp_hdr(skb)->type;
227 const int code = icmp_hdr(skb)->code;
228 int err = 0;
229 int harderr = 0;
230
231 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
232 ipv4_sk_update_pmtu(skb, sk, info);
233 else if (type == ICMP_REDIRECT) {
234 ipv4_sk_redirect(skb, sk);
235 return;
236 }
237
238 /* Report error on raw socket, if:
239 1. User requested ip_recverr.
240 2. Socket is connected (otherwise the error indication
241 is useless without ip_recverr and error is hard.
242 */
243 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
244 return;
245
246 switch (type) {
247 default:
248 case ICMP_TIME_EXCEEDED:
249 err = EHOSTUNREACH;
250 break;
251 case ICMP_SOURCE_QUENCH:
252 return;
253 case ICMP_PARAMETERPROB:
254 err = EPROTO;
255 harderr = 1;
256 break;
257 case ICMP_DEST_UNREACH:
258 err = EHOSTUNREACH;
259 if (code > NR_ICMP_UNREACH)
260 break;
261 err = icmp_err_convert[code].errno;
262 harderr = icmp_err_convert[code].fatal;
263 if (code == ICMP_FRAG_NEEDED) {
264 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
265 err = EMSGSIZE;
266 }
267 }
268
269 if (inet->recverr) {
270 const struct iphdr *iph = (const struct iphdr *)skb->data;
271 u8 *payload = skb->data + (iph->ihl << 2);
272
273 if (inet->hdrincl)
274 payload = skb->data;
275 ip_icmp_error(sk, skb, err, 0, info, payload);
276 }
277
278 if (inet->recverr || harderr) {
279 sk->sk_err = err;
280 sk->sk_error_report(sk);
281 }
282 }
283
284 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
285 {
286 int hash;
287 struct sock *raw_sk;
288 const struct iphdr *iph;
289 struct net *net;
290
291 hash = protocol & (RAW_HTABLE_SIZE - 1);
292
293 read_lock(&raw_v4_hashinfo.lock);
294 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
295 if (raw_sk) {
296 iph = (const struct iphdr *)skb->data;
297 net = dev_net(skb->dev);
298
299 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
300 iph->daddr, iph->saddr,
301 skb->dev->ifindex)) != NULL) {
302 raw_err(raw_sk, skb, info);
303 raw_sk = sk_next(raw_sk);
304 iph = (const struct iphdr *)skb->data;
305 }
306 }
307 read_unlock(&raw_v4_hashinfo.lock);
308 }
309
310 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
311 {
312 /* Charge it to the socket. */
313
314 ipv4_pktinfo_prepare(sk, skb);
315 if (sock_queue_rcv_skb(sk, skb) < 0) {
316 kfree_skb(skb);
317 return NET_RX_DROP;
318 }
319
320 return NET_RX_SUCCESS;
321 }
322
323 int raw_rcv(struct sock *sk, struct sk_buff *skb)
324 {
325 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
326 atomic_inc(&sk->sk_drops);
327 kfree_skb(skb);
328 return NET_RX_DROP;
329 }
330 nf_reset(skb);
331
332 skb_push(skb, skb->data - skb_network_header(skb));
333
334 raw_rcv_skb(sk, skb);
335 return 0;
336 }
337
338 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
339 struct msghdr *msg, size_t length,
340 struct rtable **rtp,
341 unsigned int flags)
342 {
343 struct inet_sock *inet = inet_sk(sk);
344 struct net *net = sock_net(sk);
345 struct iphdr *iph;
346 struct sk_buff *skb;
347 unsigned int iphlen;
348 int err;
349 struct rtable *rt = *rtp;
350 int hlen, tlen;
351
352 if (length > rt->dst.dev->mtu) {
353 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
354 rt->dst.dev->mtu);
355 return -EMSGSIZE;
356 }
357 if (flags&MSG_PROBE)
358 goto out;
359
360 hlen = LL_RESERVED_SPACE(rt->dst.dev);
361 tlen = rt->dst.dev->needed_tailroom;
362 skb = sock_alloc_send_skb(sk,
363 length + hlen + tlen + 15,
364 flags & MSG_DONTWAIT, &err);
365 if (!skb)
366 goto error;
367 skb_reserve(skb, hlen);
368
369 skb->priority = sk->sk_priority;
370 skb->mark = sk->sk_mark;
371 skb_dst_set(skb, &rt->dst);
372 *rtp = NULL;
373
374 skb_reset_network_header(skb);
375 iph = ip_hdr(skb);
376 skb_put(skb, length);
377
378 skb->ip_summed = CHECKSUM_NONE;
379
380 sock_tx_timestamp(sk, &skb_shinfo(skb)->tx_flags);
381
382 skb->transport_header = skb->network_header;
383 err = -EFAULT;
384 if (memcpy_from_msg(iph, msg, length))
385 goto error_free;
386
387 iphlen = iph->ihl * 4;
388
389 /*
390 * We don't want to modify the ip header, but we do need to
391 * be sure that it won't cause problems later along the network
392 * stack. Specifically we want to make sure that iph->ihl is a
393 * sane value. If ihl points beyond the length of the buffer passed
394 * in, reject the frame as invalid
395 */
396 err = -EINVAL;
397 if (iphlen > length)
398 goto error_free;
399
400 if (iphlen >= sizeof(*iph)) {
401 if (!iph->saddr)
402 iph->saddr = fl4->saddr;
403 iph->check = 0;
404 iph->tot_len = htons(length);
405 if (!iph->id)
406 ip_select_ident(net, skb, NULL);
407
408 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
409 skb->transport_header += iphlen;
410 if (iph->protocol == IPPROTO_ICMP &&
411 length >= iphlen + sizeof(struct icmphdr))
412 icmp_out_count(net, ((struct icmphdr *)
413 skb_transport_header(skb))->type);
414 }
415
416 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
417 net, sk, skb, NULL, rt->dst.dev,
418 dst_output);
419 if (err > 0)
420 err = net_xmit_errno(err);
421 if (err)
422 goto error;
423 out:
424 return 0;
425
426 error_free:
427 kfree_skb(skb);
428 error:
429 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
430 if (err == -ENOBUFS && !inet->recverr)
431 err = 0;
432 return err;
433 }
434
435 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
436 {
437 int err;
438
439 if (fl4->flowi4_proto != IPPROTO_ICMP)
440 return 0;
441
442 /* We only need the first two bytes. */
443 rfv->hlen = 2;
444
445 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
446 if (err)
447 return err;
448
449 fl4->fl4_icmp_type = rfv->hdr.icmph.type;
450 fl4->fl4_icmp_code = rfv->hdr.icmph.code;
451
452 return 0;
453 }
454
455 static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
456 struct sk_buff *skb)
457 {
458 struct raw_frag_vec *rfv = from;
459
460 if (offset < rfv->hlen) {
461 int copy = min(rfv->hlen - offset, len);
462
463 if (skb->ip_summed == CHECKSUM_PARTIAL)
464 memcpy(to, rfv->hdr.c + offset, copy);
465 else
466 skb->csum = csum_block_add(
467 skb->csum,
468 csum_partial_copy_nocheck(rfv->hdr.c + offset,
469 to, copy, 0),
470 odd);
471
472 odd = 0;
473 offset += copy;
474 to += copy;
475 len -= copy;
476
477 if (!len)
478 return 0;
479 }
480
481 offset -= rfv->hlen;
482
483 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
484 }
485
486 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
487 {
488 struct inet_sock *inet = inet_sk(sk);
489 struct net *net = sock_net(sk);
490 struct ipcm_cookie ipc;
491 struct rtable *rt = NULL;
492 struct flowi4 fl4;
493 int free = 0;
494 __be32 daddr;
495 __be32 saddr;
496 u8 tos;
497 int err;
498 struct ip_options_data opt_copy;
499 struct raw_frag_vec rfv;
500
501 err = -EMSGSIZE;
502 if (len > 0xFFFF)
503 goto out;
504
505 /*
506 * Check the flags.
507 */
508
509 err = -EOPNOTSUPP;
510 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
511 goto out; /* compatibility */
512
513 /*
514 * Get and verify the address.
515 */
516
517 if (msg->msg_namelen) {
518 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
519 err = -EINVAL;
520 if (msg->msg_namelen < sizeof(*usin))
521 goto out;
522 if (usin->sin_family != AF_INET) {
523 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
524 __func__, current->comm);
525 err = -EAFNOSUPPORT;
526 if (usin->sin_family)
527 goto out;
528 }
529 daddr = usin->sin_addr.s_addr;
530 /* ANK: I did not forget to get protocol from port field.
531 * I just do not know, who uses this weirdness.
532 * IP_HDRINCL is much more convenient.
533 */
534 } else {
535 err = -EDESTADDRREQ;
536 if (sk->sk_state != TCP_ESTABLISHED)
537 goto out;
538 daddr = inet->inet_daddr;
539 }
540
541 ipc.addr = inet->inet_saddr;
542 ipc.opt = NULL;
543 ipc.tx_flags = 0;
544 ipc.ttl = 0;
545 ipc.tos = -1;
546 ipc.oif = sk->sk_bound_dev_if;
547
548 if (msg->msg_controllen) {
549 err = ip_cmsg_send(net, msg, &ipc, false);
550 if (err)
551 goto out;
552 if (ipc.opt)
553 free = 1;
554 }
555
556 saddr = ipc.addr;
557 ipc.addr = daddr;
558
559 if (!ipc.opt) {
560 struct ip_options_rcu *inet_opt;
561
562 rcu_read_lock();
563 inet_opt = rcu_dereference(inet->inet_opt);
564 if (inet_opt) {
565 memcpy(&opt_copy, inet_opt,
566 sizeof(*inet_opt) + inet_opt->opt.optlen);
567 ipc.opt = &opt_copy.opt;
568 }
569 rcu_read_unlock();
570 }
571
572 if (ipc.opt) {
573 err = -EINVAL;
574 /* Linux does not mangle headers on raw sockets,
575 * so that IP options + IP_HDRINCL is non-sense.
576 */
577 if (inet->hdrincl)
578 goto done;
579 if (ipc.opt->opt.srr) {
580 if (!daddr)
581 goto done;
582 daddr = ipc.opt->opt.faddr;
583 }
584 }
585 tos = get_rtconn_flags(&ipc, sk);
586 if (msg->msg_flags & MSG_DONTROUTE)
587 tos |= RTO_ONLINK;
588
589 if (ipv4_is_multicast(daddr)) {
590 if (!ipc.oif)
591 ipc.oif = inet->mc_index;
592 if (!saddr)
593 saddr = inet->mc_addr;
594 } else if (!ipc.oif)
595 ipc.oif = inet->uc_index;
596
597 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
598 RT_SCOPE_UNIVERSE,
599 inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
600 inet_sk_flowi_flags(sk) |
601 (inet->hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
602 daddr, saddr, 0, 0);
603
604 if (!saddr && ipc.oif)
605 l3mdev_get_saddr(net, ipc.oif, &fl4);
606
607 if (!inet->hdrincl) {
608 rfv.msg = msg;
609 rfv.hlen = 0;
610
611 err = raw_probe_proto_opt(&rfv, &fl4);
612 if (err)
613 goto done;
614 }
615
616 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
617 rt = ip_route_output_flow(net, &fl4, sk);
618 if (IS_ERR(rt)) {
619 err = PTR_ERR(rt);
620 rt = NULL;
621 goto done;
622 }
623
624 err = -EACCES;
625 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
626 goto done;
627
628 if (msg->msg_flags & MSG_CONFIRM)
629 goto do_confirm;
630 back_from_confirm:
631
632 if (inet->hdrincl)
633 err = raw_send_hdrinc(sk, &fl4, msg, len,
634 &rt, msg->msg_flags);
635
636 else {
637 sock_tx_timestamp(sk, &ipc.tx_flags);
638
639 if (!ipc.addr)
640 ipc.addr = fl4.daddr;
641 lock_sock(sk);
642 err = ip_append_data(sk, &fl4, raw_getfrag,
643 &rfv, len, 0,
644 &ipc, &rt, msg->msg_flags);
645 if (err)
646 ip_flush_pending_frames(sk);
647 else if (!(msg->msg_flags & MSG_MORE)) {
648 err = ip_push_pending_frames(sk, &fl4);
649 if (err == -ENOBUFS && !inet->recverr)
650 err = 0;
651 }
652 release_sock(sk);
653 }
654 done:
655 if (free)
656 kfree(ipc.opt);
657 ip_rt_put(rt);
658
659 out:
660 if (err < 0)
661 return err;
662 return len;
663
664 do_confirm:
665 dst_confirm(&rt->dst);
666 if (!(msg->msg_flags & MSG_PROBE) || len)
667 goto back_from_confirm;
668 err = 0;
669 goto done;
670 }
671
672 static void raw_close(struct sock *sk, long timeout)
673 {
674 /*
675 * Raw sockets may have direct kernel references. Kill them.
676 */
677 ip_ra_control(sk, 0, NULL);
678
679 sk_common_release(sk);
680 }
681
682 static void raw_destroy(struct sock *sk)
683 {
684 lock_sock(sk);
685 ip_flush_pending_frames(sk);
686 release_sock(sk);
687 }
688
689 /* This gets rid of all the nasties in af_inet. -DaveM */
690 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
691 {
692 struct inet_sock *inet = inet_sk(sk);
693 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
694 int ret = -EINVAL;
695 int chk_addr_ret;
696
697 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
698 goto out;
699 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
700 ret = -EADDRNOTAVAIL;
701 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
702 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
703 goto out;
704 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
705 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
706 inet->inet_saddr = 0; /* Use device */
707 sk_dst_reset(sk);
708 ret = 0;
709 out: return ret;
710 }
711
712 /*
713 * This should be easy, if there is something there
714 * we return it, otherwise we block.
715 */
716
717 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
718 int noblock, int flags, int *addr_len)
719 {
720 struct inet_sock *inet = inet_sk(sk);
721 size_t copied = 0;
722 int err = -EOPNOTSUPP;
723 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
724 struct sk_buff *skb;
725
726 if (flags & MSG_OOB)
727 goto out;
728
729 if (flags & MSG_ERRQUEUE) {
730 err = ip_recv_error(sk, msg, len, addr_len);
731 goto out;
732 }
733
734 skb = skb_recv_datagram(sk, flags, noblock, &err);
735 if (!skb)
736 goto out;
737
738 copied = skb->len;
739 if (len < copied) {
740 msg->msg_flags |= MSG_TRUNC;
741 copied = len;
742 }
743
744 err = skb_copy_datagram_msg(skb, 0, msg, copied);
745 if (err)
746 goto done;
747
748 sock_recv_ts_and_drops(msg, sk, skb);
749
750 /* Copy the address. */
751 if (sin) {
752 sin->sin_family = AF_INET;
753 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
754 sin->sin_port = 0;
755 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
756 *addr_len = sizeof(*sin);
757 }
758 if (inet->cmsg_flags)
759 ip_cmsg_recv(msg, skb);
760 if (flags & MSG_TRUNC)
761 copied = skb->len;
762 done:
763 skb_free_datagram(sk, skb);
764 out:
765 if (err)
766 return err;
767 return copied;
768 }
769
770 static int raw_init(struct sock *sk)
771 {
772 struct raw_sock *rp = raw_sk(sk);
773
774 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
775 memset(&rp->filter, 0, sizeof(rp->filter));
776 return 0;
777 }
778
779 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
780 {
781 if (optlen > sizeof(struct icmp_filter))
782 optlen = sizeof(struct icmp_filter);
783 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
784 return -EFAULT;
785 return 0;
786 }
787
788 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
789 {
790 int len, ret = -EFAULT;
791
792 if (get_user(len, optlen))
793 goto out;
794 ret = -EINVAL;
795 if (len < 0)
796 goto out;
797 if (len > sizeof(struct icmp_filter))
798 len = sizeof(struct icmp_filter);
799 ret = -EFAULT;
800 if (put_user(len, optlen) ||
801 copy_to_user(optval, &raw_sk(sk)->filter, len))
802 goto out;
803 ret = 0;
804 out: return ret;
805 }
806
807 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
808 char __user *optval, unsigned int optlen)
809 {
810 if (optname == ICMP_FILTER) {
811 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
812 return -EOPNOTSUPP;
813 else
814 return raw_seticmpfilter(sk, optval, optlen);
815 }
816 return -ENOPROTOOPT;
817 }
818
819 static int raw_setsockopt(struct sock *sk, int level, int optname,
820 char __user *optval, unsigned int optlen)
821 {
822 if (level != SOL_RAW)
823 return ip_setsockopt(sk, level, optname, optval, optlen);
824 return do_raw_setsockopt(sk, level, optname, optval, optlen);
825 }
826
827 #ifdef CONFIG_COMPAT
828 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
829 char __user *optval, unsigned int optlen)
830 {
831 if (level != SOL_RAW)
832 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
833 return do_raw_setsockopt(sk, level, optname, optval, optlen);
834 }
835 #endif
836
837 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
838 char __user *optval, int __user *optlen)
839 {
840 if (optname == ICMP_FILTER) {
841 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
842 return -EOPNOTSUPP;
843 else
844 return raw_geticmpfilter(sk, optval, optlen);
845 }
846 return -ENOPROTOOPT;
847 }
848
849 static int raw_getsockopt(struct sock *sk, int level, int optname,
850 char __user *optval, int __user *optlen)
851 {
852 if (level != SOL_RAW)
853 return ip_getsockopt(sk, level, optname, optval, optlen);
854 return do_raw_getsockopt(sk, level, optname, optval, optlen);
855 }
856
857 #ifdef CONFIG_COMPAT
858 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
859 char __user *optval, int __user *optlen)
860 {
861 if (level != SOL_RAW)
862 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
863 return do_raw_getsockopt(sk, level, optname, optval, optlen);
864 }
865 #endif
866
867 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
868 {
869 switch (cmd) {
870 case SIOCOUTQ: {
871 int amount = sk_wmem_alloc_get(sk);
872
873 return put_user(amount, (int __user *)arg);
874 }
875 case SIOCINQ: {
876 struct sk_buff *skb;
877 int amount = 0;
878
879 spin_lock_bh(&sk->sk_receive_queue.lock);
880 skb = skb_peek(&sk->sk_receive_queue);
881 if (skb)
882 amount = skb->len;
883 spin_unlock_bh(&sk->sk_receive_queue.lock);
884 return put_user(amount, (int __user *)arg);
885 }
886
887 default:
888 #ifdef CONFIG_IP_MROUTE
889 return ipmr_ioctl(sk, cmd, (void __user *)arg);
890 #else
891 return -ENOIOCTLCMD;
892 #endif
893 }
894 }
895
896 #ifdef CONFIG_COMPAT
897 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
898 {
899 switch (cmd) {
900 case SIOCOUTQ:
901 case SIOCINQ:
902 return -ENOIOCTLCMD;
903 default:
904 #ifdef CONFIG_IP_MROUTE
905 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
906 #else
907 return -ENOIOCTLCMD;
908 #endif
909 }
910 }
911 #endif
912
913 struct proto raw_prot = {
914 .name = "RAW",
915 .owner = THIS_MODULE,
916 .close = raw_close,
917 .destroy = raw_destroy,
918 .connect = ip4_datagram_connect,
919 .disconnect = udp_disconnect,
920 .ioctl = raw_ioctl,
921 .init = raw_init,
922 .setsockopt = raw_setsockopt,
923 .getsockopt = raw_getsockopt,
924 .sendmsg = raw_sendmsg,
925 .recvmsg = raw_recvmsg,
926 .bind = raw_bind,
927 .backlog_rcv = raw_rcv_skb,
928 .release_cb = ip4_datagram_release_cb,
929 .hash = raw_hash_sk,
930 .unhash = raw_unhash_sk,
931 .obj_size = sizeof(struct raw_sock),
932 .h.raw_hash = &raw_v4_hashinfo,
933 #ifdef CONFIG_COMPAT
934 .compat_setsockopt = compat_raw_setsockopt,
935 .compat_getsockopt = compat_raw_getsockopt,
936 .compat_ioctl = compat_raw_ioctl,
937 #endif
938 };
939
940 #ifdef CONFIG_PROC_FS
941 static struct sock *raw_get_first(struct seq_file *seq)
942 {
943 struct sock *sk;
944 struct raw_iter_state *state = raw_seq_private(seq);
945
946 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
947 ++state->bucket) {
948 sk_for_each(sk, &state->h->ht[state->bucket])
949 if (sock_net(sk) == seq_file_net(seq))
950 goto found;
951 }
952 sk = NULL;
953 found:
954 return sk;
955 }
956
957 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
958 {
959 struct raw_iter_state *state = raw_seq_private(seq);
960
961 do {
962 sk = sk_next(sk);
963 try_again:
964 ;
965 } while (sk && sock_net(sk) != seq_file_net(seq));
966
967 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
968 sk = sk_head(&state->h->ht[state->bucket]);
969 goto try_again;
970 }
971 return sk;
972 }
973
974 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
975 {
976 struct sock *sk = raw_get_first(seq);
977
978 if (sk)
979 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
980 --pos;
981 return pos ? NULL : sk;
982 }
983
984 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
985 {
986 struct raw_iter_state *state = raw_seq_private(seq);
987
988 read_lock(&state->h->lock);
989 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
990 }
991 EXPORT_SYMBOL_GPL(raw_seq_start);
992
993 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
994 {
995 struct sock *sk;
996
997 if (v == SEQ_START_TOKEN)
998 sk = raw_get_first(seq);
999 else
1000 sk = raw_get_next(seq, v);
1001 ++*pos;
1002 return sk;
1003 }
1004 EXPORT_SYMBOL_GPL(raw_seq_next);
1005
1006 void raw_seq_stop(struct seq_file *seq, void *v)
1007 {
1008 struct raw_iter_state *state = raw_seq_private(seq);
1009
1010 read_unlock(&state->h->lock);
1011 }
1012 EXPORT_SYMBOL_GPL(raw_seq_stop);
1013
1014 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1015 {
1016 struct inet_sock *inet = inet_sk(sp);
1017 __be32 dest = inet->inet_daddr,
1018 src = inet->inet_rcv_saddr;
1019 __u16 destp = 0,
1020 srcp = inet->inet_num;
1021
1022 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1023 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1024 i, src, srcp, dest, destp, sp->sk_state,
1025 sk_wmem_alloc_get(sp),
1026 sk_rmem_alloc_get(sp),
1027 0, 0L, 0,
1028 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1029 0, sock_i_ino(sp),
1030 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1031 }
1032
1033 static int raw_seq_show(struct seq_file *seq, void *v)
1034 {
1035 if (v == SEQ_START_TOKEN)
1036 seq_printf(seq, " sl local_address rem_address st tx_queue "
1037 "rx_queue tr tm->when retrnsmt uid timeout "
1038 "inode ref pointer drops\n");
1039 else
1040 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1041 return 0;
1042 }
1043
1044 static const struct seq_operations raw_seq_ops = {
1045 .start = raw_seq_start,
1046 .next = raw_seq_next,
1047 .stop = raw_seq_stop,
1048 .show = raw_seq_show,
1049 };
1050
1051 int raw_seq_open(struct inode *ino, struct file *file,
1052 struct raw_hashinfo *h, const struct seq_operations *ops)
1053 {
1054 int err;
1055 struct raw_iter_state *i;
1056
1057 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1058 if (err < 0)
1059 return err;
1060
1061 i = raw_seq_private((struct seq_file *)file->private_data);
1062 i->h = h;
1063 return 0;
1064 }
1065 EXPORT_SYMBOL_GPL(raw_seq_open);
1066
1067 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1068 {
1069 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1070 }
1071
1072 static const struct file_operations raw_seq_fops = {
1073 .owner = THIS_MODULE,
1074 .open = raw_v4_seq_open,
1075 .read = seq_read,
1076 .llseek = seq_lseek,
1077 .release = seq_release_net,
1078 };
1079
1080 static __net_init int raw_init_net(struct net *net)
1081 {
1082 if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops))
1083 return -ENOMEM;
1084
1085 return 0;
1086 }
1087
1088 static __net_exit void raw_exit_net(struct net *net)
1089 {
1090 remove_proc_entry("raw", net->proc_net);
1091 }
1092
1093 static __net_initdata struct pernet_operations raw_net_ops = {
1094 .init = raw_init_net,
1095 .exit = raw_exit_net,
1096 };
1097
1098 int __init raw_proc_init(void)
1099 {
1100 return register_pernet_subsys(&raw_net_ops);
1101 }
1102
1103 void __init raw_proc_exit(void)
1104 {
1105 unregister_pernet_subsys(&raw_net_ops);
1106 }
1107 #endif /* CONFIG_PROC_FS */
Linux kernel - Deliverables
This page took 0.05254 seconds and 5 git commands to generate.