Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * RAW - implementation of IP "raw" sockets.
8 *
9 * Authors: Ross Biro
10 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 *
12 * Fixes:
13 * Alan Cox : verify_area() fixed up
14 * Alan Cox : ICMP error handling
15 * Alan Cox : EMSGSIZE if you send too big a packet
16 * Alan Cox : Now uses generic datagrams and shared
17 * skbuff library. No more peek crashes,
18 * no more backlogs
19 * Alan Cox : Checks sk->broadcast.
20 * Alan Cox : Uses skb_free_datagram/skb_copy_datagram
21 * Alan Cox : Raw passes ip options too
22 * Alan Cox : Setsocketopt added
23 * Alan Cox : Fixed error return for broadcasts
24 * Alan Cox : Removed wake_up calls
25 * Alan Cox : Use ttl/tos
26 * Alan Cox : Cleaned up old debugging
27 * Alan Cox : Use new kernel side addresses
28 * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets.
29 * Alan Cox : BSD style RAW socket demultiplexing.
30 * Alan Cox : Beginnings of mrouted support.
31 * Alan Cox : Added IP_HDRINCL option.
32 * Alan Cox : Skip broadcast check if BSDism set.
33 * David S. Miller : New socket lookup architecture.
34 */
35
36#include <linux/types.h>
37#include <linux/atomic.h>
38#include <asm/byteorder.h>
39#include <asm/current.h>
40#include <linux/uaccess.h>
41#include <asm/ioctls.h>
42#include <linux/stddef.h>
43#include <linux/slab.h>
44#include <linux/errno.h>
45#include <linux/kernel.h>
46#include <linux/export.h>
47#include <linux/spinlock.h>
48#include <linux/sockios.h>
49#include <linux/socket.h>
50#include <linux/in.h>
51#include <linux/mroute.h>
52#include <linux/netdevice.h>
53#include <linux/in_route.h>
54#include <linux/route.h>
55#include <linux/skbuff.h>
56#include <linux/igmp.h>
57#include <net/net_namespace.h>
58#include <net/dst.h>
59#include <net/sock.h>
60#include <linux/ip.h>
61#include <linux/net.h>
62#include <net/ip.h>
63#include <net/icmp.h>
64#include <net/udp.h>
65#include <net/raw.h>
66#include <net/snmp.h>
67#include <net/tcp_states.h>
68#include <net/inet_common.h>
69#include <net/checksum.h>
70#include <net/xfrm.h>
71#include <linux/rtnetlink.h>
72#include <linux/proc_fs.h>
73#include <linux/seq_file.h>
74#include <linux/netfilter.h>
75#include <linux/netfilter_ipv4.h>
76#include <linux/compat.h>
77#include <linux/uio.h>
78
79struct raw_frag_vec {
80 struct msghdr *msg;
81 union {
82 struct icmphdr icmph;
83 char c[1];
84 } hdr;
85 int hlen;
86};
87
88struct raw_hashinfo raw_v4_hashinfo;
89EXPORT_SYMBOL_GPL(raw_v4_hashinfo);
90
91int raw_hash_sk(struct sock *sk)
92{
93 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
94 struct hlist_head *hlist;
95
96 hlist = &h->ht[raw_hashfunc(sock_net(sk), inet_sk(sk)->inet_num)];
97
98 spin_lock(&h->lock);
99 sk_add_node_rcu(sk, hlist);
100 sock_set_flag(sk, SOCK_RCU_FREE);
101 spin_unlock(&h->lock);
102 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
103
104 return 0;
105}
106EXPORT_SYMBOL_GPL(raw_hash_sk);
107
108void raw_unhash_sk(struct sock *sk)
109{
110 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
111
112 spin_lock(&h->lock);
113 if (sk_del_node_init_rcu(sk))
114 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
115 spin_unlock(&h->lock);
116}
117EXPORT_SYMBOL_GPL(raw_unhash_sk);
118
119bool raw_v4_match(struct net *net, const struct sock *sk, unsigned short num,
120 __be32 raddr, __be32 laddr, int dif, int sdif)
121{
122 const struct inet_sock *inet = inet_sk(sk);
123
124 if (net_eq(sock_net(sk), net) && inet->inet_num == num &&
125 !(inet->inet_daddr && inet->inet_daddr != raddr) &&
126 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
127 raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif))
128 return true;
129 return false;
130}
131EXPORT_SYMBOL_GPL(raw_v4_match);
132
133/*
134 * 0 - deliver
135 * 1 - block
136 */
137static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
138{
139 struct icmphdr _hdr;
140 const struct icmphdr *hdr;
141
142 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
143 sizeof(_hdr), &_hdr);
144 if (!hdr)
145 return 1;
146
147 if (hdr->type < 32) {
148 __u32 data = raw_sk(sk)->filter.data;
149
150 return ((1U << hdr->type) & data) != 0;
151 }
152
153 /* Do not block unknown ICMP types */
154 return 0;
155}
156
157/* IP input processing comes here for RAW socket delivery.
158 * Caller owns SKB, so we must make clones.
159 *
160 * RFC 1122: SHOULD pass TOS value up to the transport layer.
161 * -> It does. And not only TOS, but all IP header.
162 */
163static int raw_v4_input(struct net *net, struct sk_buff *skb,
164 const struct iphdr *iph, int hash)
165{
166 int sdif = inet_sdif(skb);
167 struct hlist_head *hlist;
168 int dif = inet_iif(skb);
169 int delivered = 0;
170 struct sock *sk;
171
172 hlist = &raw_v4_hashinfo.ht[hash];
173 rcu_read_lock();
174 sk_for_each_rcu(sk, hlist) {
175 if (!raw_v4_match(net, sk, iph->protocol,
176 iph->saddr, iph->daddr, dif, sdif))
177 continue;
178
179 if (atomic_read(&sk->sk_rmem_alloc) >=
180 READ_ONCE(sk->sk_rcvbuf)) {
181 sk_drops_inc(sk);
182 continue;
183 }
184
185 delivered = 1;
186 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
187 ip_mc_sf_allow(sk, iph->daddr, iph->saddr,
188 skb->dev->ifindex, sdif)) {
189 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
190
191 /* Not releasing hash table! */
192 if (clone)
193 raw_rcv(sk, clone);
194 }
195 }
196 rcu_read_unlock();
197 return delivered;
198}
199
200int raw_local_deliver(struct sk_buff *skb, int protocol)
201{
202 struct net *net = dev_net(skb->dev);
203
204 return raw_v4_input(net, skb, ip_hdr(skb),
205 raw_hashfunc(net, protocol));
206}
207
208static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
209{
210 struct inet_sock *inet = inet_sk(sk);
211 const int type = icmp_hdr(skb)->type;
212 const int code = icmp_hdr(skb)->code;
213 int harderr = 0;
214 bool recverr;
215 int err = 0;
216
217 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
218 ipv4_sk_update_pmtu(skb, sk, info);
219 else if (type == ICMP_REDIRECT) {
220 ipv4_sk_redirect(skb, sk);
221 return;
222 }
223
224 /* Report error on raw socket, if:
225 1. User requested ip_recverr.
226 2. Socket is connected (otherwise the error indication
227 is useless without ip_recverr and error is hard.
228 */
229 recverr = inet_test_bit(RECVERR, sk);
230 if (!recverr && sk->sk_state != TCP_ESTABLISHED)
231 return;
232
233 switch (type) {
234 default:
235 case ICMP_TIME_EXCEEDED:
236 err = EHOSTUNREACH;
237 break;
238 case ICMP_SOURCE_QUENCH:
239 return;
240 case ICMP_PARAMETERPROB:
241 err = EPROTO;
242 harderr = 1;
243 break;
244 case ICMP_DEST_UNREACH:
245 err = EHOSTUNREACH;
246 if (code > NR_ICMP_UNREACH)
247 break;
248 if (code == ICMP_FRAG_NEEDED) {
249 harderr = READ_ONCE(inet->pmtudisc) != IP_PMTUDISC_DONT;
250 err = EMSGSIZE;
251 } else {
252 err = icmp_err_convert[code].errno;
253 harderr = icmp_err_convert[code].fatal;
254 }
255 }
256
257 if (recverr) {
258 const struct iphdr *iph = (const struct iphdr *)skb->data;
259 u8 *payload = skb->data + (iph->ihl << 2);
260
261 if (inet_test_bit(HDRINCL, sk))
262 payload = skb->data;
263 ip_icmp_error(sk, skb, err, 0, info, payload);
264 }
265
266 if (recverr || harderr) {
267 sk->sk_err = err;
268 sk_error_report(sk);
269 }
270}
271
272void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
273{
274 struct net *net = dev_net(skb->dev);
275 int dif = skb->dev->ifindex;
276 int sdif = inet_sdif(skb);
277 struct hlist_head *hlist;
278 const struct iphdr *iph;
279 struct sock *sk;
280 int hash;
281
282 hash = raw_hashfunc(net, protocol);
283 hlist = &raw_v4_hashinfo.ht[hash];
284
285 rcu_read_lock();
286 sk_for_each_rcu(sk, hlist) {
287 iph = (const struct iphdr *)skb->data;
288 if (!raw_v4_match(net, sk, iph->protocol,
289 iph->daddr, iph->saddr, dif, sdif))
290 continue;
291 raw_err(sk, skb, info);
292 }
293 rcu_read_unlock();
294}
295
296static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
297{
298 enum skb_drop_reason reason;
299
300 /* Charge it to the socket. */
301
302 ipv4_pktinfo_prepare(sk, skb, true);
303 reason = sock_queue_rcv_skb_reason(sk, skb);
304 if (reason) {
305 sk_skb_reason_drop(sk, skb, reason);
306 return NET_RX_DROP;
307 }
308
309 return NET_RX_SUCCESS;
310}
311
312int raw_rcv(struct sock *sk, struct sk_buff *skb)
313{
314 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
315 sk_drops_inc(sk);
316 sk_skb_reason_drop(sk, skb, SKB_DROP_REASON_XFRM_POLICY);
317 return NET_RX_DROP;
318 }
319 nf_reset_ct(skb);
320
321 skb_push(skb, -skb_network_offset(skb));
322
323 raw_rcv_skb(sk, skb);
324 return 0;
325}
326
327static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
328 struct msghdr *msg, size_t length,
329 struct rtable **rtp, unsigned int flags,
330 const struct sockcm_cookie *sockc)
331{
332 struct inet_sock *inet = inet_sk(sk);
333 struct net *net = sock_net(sk);
334 struct iphdr *iph;
335 struct sk_buff *skb;
336 unsigned int iphlen;
337 int err;
338 struct rtable *rt = *rtp;
339 int hlen, tlen;
340
341 if (length > rt->dst.dev->mtu) {
342 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
343 rt->dst.dev->mtu);
344 return -EMSGSIZE;
345 }
346 if (length < sizeof(struct iphdr))
347 return -EINVAL;
348
349 if (flags&MSG_PROBE)
350 goto out;
351
352 hlen = LL_RESERVED_SPACE(rt->dst.dev);
353 tlen = rt->dst.dev->needed_tailroom;
354 skb = sock_alloc_send_skb(sk,
355 length + hlen + tlen + 15,
356 flags & MSG_DONTWAIT, &err);
357 if (!skb)
358 goto error;
359 skb_reserve(skb, hlen);
360
361 skb->protocol = htons(ETH_P_IP);
362 skb->priority = sockc->priority;
363 skb->mark = sockc->mark;
364 skb_set_delivery_type_by_clockid(skb, sockc->transmit_time, sk->sk_clockid);
365 skb_dst_set(skb, &rt->dst);
366 *rtp = NULL;
367
368 skb_reset_network_header(skb);
369 iph = ip_hdr(skb);
370 skb_put(skb, length);
371
372 skb->ip_summed = CHECKSUM_NONE;
373
374 skb_setup_tx_timestamp(skb, sockc);
375
376 if (flags & MSG_CONFIRM)
377 skb_set_dst_pending_confirm(skb, 1);
378
379 skb->transport_header = skb->network_header;
380 err = -EFAULT;
381 if (memcpy_from_msg(iph, msg, length))
382 goto error_free;
383
384 iphlen = iph->ihl * 4;
385
386 /*
387 * We don't want to modify the ip header, but we do need to
388 * be sure that it won't cause problems later along the network
389 * stack. Specifically we want to make sure that iph->ihl is a
390 * sane value. If ihl points beyond the length of the buffer passed
391 * in, reject the frame as invalid
392 */
393 err = -EINVAL;
394 if (iphlen > length)
395 goto error_free;
396
397 if (iphlen >= sizeof(*iph)) {
398 if (!iph->saddr)
399 iph->saddr = fl4->saddr;
400 iph->check = 0;
401 iph->tot_len = htons(length);
402 if (!iph->id)
403 ip_select_ident(net, skb, NULL);
404
405 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
406 skb->transport_header += iphlen;
407 if (iph->protocol == IPPROTO_ICMP &&
408 length >= iphlen + sizeof(struct icmphdr))
409 icmp_out_count(net, ((struct icmphdr *)
410 skb_transport_header(skb))->type);
411 }
412
413 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
414 net, sk, skb, NULL, rt->dst.dev,
415 dst_output);
416 if (err > 0)
417 err = net_xmit_errno(err);
418 if (err)
419 goto error;
420out:
421 return 0;
422
423error_free:
424 kfree_skb(skb);
425error:
426 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
427 if (err == -ENOBUFS && !inet_test_bit(RECVERR, sk))
428 err = 0;
429 return err;
430}
431
432static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
433{
434 int err;
435
436 if (fl4->flowi4_proto != IPPROTO_ICMP)
437 return 0;
438
439 /* We only need the first two bytes. */
440 rfv->hlen = 2;
441
442 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
443 if (err)
444 return err;
445
446 fl4->fl4_icmp_type = rfv->hdr.icmph.type;
447 fl4->fl4_icmp_code = rfv->hdr.icmph.code;
448
449 return 0;
450}
451
452static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
453 struct sk_buff *skb)
454{
455 struct raw_frag_vec *rfv = from;
456
457 if (offset < rfv->hlen) {
458 int copy = min(rfv->hlen - offset, len);
459
460 if (skb->ip_summed == CHECKSUM_PARTIAL)
461 memcpy(to, rfv->hdr.c + offset, copy);
462 else
463 skb->csum = csum_block_add(
464 skb->csum,
465 csum_partial_copy_nocheck(rfv->hdr.c + offset,
466 to, copy),
467 odd);
468
469 odd = 0;
470 offset += copy;
471 to += copy;
472 len -= copy;
473
474 if (!len)
475 return 0;
476 }
477
478 offset -= rfv->hlen;
479
480 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
481}
482
483static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
484{
485 DEFINE_RAW_FLEX(struct ip_options_rcu, opt_copy, opt.__data,
486 IP_OPTIONS_DATA_FIXED_SIZE);
487 struct inet_sock *inet = inet_sk(sk);
488 struct net *net = sock_net(sk);
489 struct ipcm_cookie ipc;
490 struct rtable *rt = NULL;
491 struct flowi4 fl4;
492 u8 scope;
493 int free = 0;
494 __be32 daddr;
495 __be32 saddr;
496 int uc_index, err;
497 struct raw_frag_vec rfv;
498 int hdrincl;
499
500 err = -EMSGSIZE;
501 if (len > 0xFFFF)
502 goto out;
503
504 hdrincl = inet_test_bit(HDRINCL, sk);
505
506 /*
507 * Check the flags.
508 */
509
510 err = -EOPNOTSUPP;
511 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
512 goto out; /* compatibility */
513
514 /*
515 * Get and verify the address.
516 */
517
518 if (msg->msg_namelen) {
519 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
520 err = -EINVAL;
521 if (msg->msg_namelen < sizeof(*usin))
522 goto out;
523 if (usin->sin_family != AF_INET) {
524 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
525 __func__, current->comm);
526 err = -EAFNOSUPPORT;
527 if (usin->sin_family)
528 goto out;
529 }
530 daddr = usin->sin_addr.s_addr;
531 /* ANK: I did not forget to get protocol from port field.
532 * I just do not know, who uses this weirdness.
533 * IP_HDRINCL is much more convenient.
534 */
535 } else {
536 err = -EDESTADDRREQ;
537 if (sk->sk_state != TCP_ESTABLISHED)
538 goto out;
539 daddr = inet->inet_daddr;
540 }
541
542 ipcm_init_sk(&ipc, inet);
543 /* Keep backward compat */
544 if (hdrincl)
545 ipc.protocol = IPPROTO_RAW;
546
547 if (msg->msg_controllen) {
548 err = ip_cmsg_send(sk, msg, &ipc, false);
549 if (unlikely(err)) {
550 kfree(ipc.opt);
551 goto out;
552 }
553 if (ipc.opt)
554 free = 1;
555 }
556
557 saddr = ipc.addr;
558 ipc.addr = daddr;
559
560 if (!ipc.opt) {
561 struct ip_options_rcu *inet_opt;
562
563 rcu_read_lock();
564 inet_opt = rcu_dereference(inet->inet_opt);
565 if (inet_opt) {
566 memcpy(opt_copy, inet_opt,
567 sizeof(*inet_opt) + inet_opt->opt.optlen);
568 ipc.opt = opt_copy;
569 }
570 rcu_read_unlock();
571 }
572
573 if (ipc.opt) {
574 err = -EINVAL;
575 /* Linux does not mangle headers on raw sockets,
576 * so that IP options + IP_HDRINCL is non-sense.
577 */
578 if (hdrincl)
579 goto done;
580 if (ipc.opt->opt.srr) {
581 if (!daddr)
582 goto done;
583 daddr = ipc.opt->opt.faddr;
584 }
585 }
586 scope = ip_sendmsg_scope(inet, &ipc, msg);
587
588 uc_index = READ_ONCE(inet->uc_index);
589 if (ipv4_is_multicast(daddr)) {
590 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
591 ipc.oif = READ_ONCE(inet->mc_index);
592 if (!saddr)
593 saddr = READ_ONCE(inet->mc_addr);
594 } else if (!ipc.oif) {
595 ipc.oif = uc_index;
596 } else if (ipv4_is_lbcast(daddr) && uc_index) {
597 /* oif is set, packet is to local broadcast
598 * and uc_index is set. oif is most likely set
599 * by sk_bound_dev_if. If uc_index != oif check if the
600 * oif is an L3 master and uc_index is an L3 slave.
601 * If so, we want to allow the send using the uc_index.
602 */
603 if (ipc.oif != uc_index &&
604 ipc.oif == l3mdev_master_ifindex_by_index(sock_net(sk),
605 uc_index)) {
606 ipc.oif = uc_index;
607 }
608 }
609
610 flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark,
611 ipc.tos & INET_DSCP_MASK, scope,
612 hdrincl ? ipc.protocol : sk->sk_protocol,
613 inet_sk_flowi_flags(sk) |
614 (hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
615 daddr, saddr, 0, 0, sk_uid(sk));
616
617 fl4.fl4_icmp_type = 0;
618 fl4.fl4_icmp_code = 0;
619
620 if (!hdrincl) {
621 rfv.msg = msg;
622 rfv.hlen = 0;
623
624 err = raw_probe_proto_opt(&rfv, &fl4);
625 if (err)
626 goto done;
627 }
628
629 security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
630 rt = ip_route_output_flow(net, &fl4, sk);
631 if (IS_ERR(rt)) {
632 err = PTR_ERR(rt);
633 rt = NULL;
634 goto done;
635 }
636
637 err = -EACCES;
638 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
639 goto done;
640
641 if (msg->msg_flags & MSG_CONFIRM)
642 goto do_confirm;
643back_from_confirm:
644
645 if (hdrincl)
646 err = raw_send_hdrinc(sk, &fl4, msg, len,
647 &rt, msg->msg_flags, &ipc.sockc);
648
649 else {
650 if (!ipc.addr)
651 ipc.addr = fl4.daddr;
652 lock_sock(sk);
653 err = ip_append_data(sk, &fl4, raw_getfrag,
654 &rfv, len, 0,
655 &ipc, &rt, msg->msg_flags);
656 if (err)
657 ip_flush_pending_frames(sk);
658 else if (!(msg->msg_flags & MSG_MORE)) {
659 err = ip_push_pending_frames(sk, &fl4);
660 if (err == -ENOBUFS && !inet_test_bit(RECVERR, sk))
661 err = 0;
662 }
663 release_sock(sk);
664 }
665done:
666 if (free)
667 kfree(ipc.opt);
668 ip_rt_put(rt);
669
670out:
671 if (err < 0)
672 return err;
673 return len;
674
675do_confirm:
676 if (msg->msg_flags & MSG_PROBE)
677 dst_confirm_neigh(&rt->dst, &fl4.daddr);
678 if (!(msg->msg_flags & MSG_PROBE) || len)
679 goto back_from_confirm;
680 err = 0;
681 goto done;
682}
683
684static void raw_close(struct sock *sk, long timeout)
685{
686 /*
687 * Raw sockets may have direct kernel references. Kill them.
688 */
689 ip_ra_control(sk, 0, NULL);
690
691 sk_common_release(sk);
692}
693
694static void raw_destroy(struct sock *sk)
695{
696 lock_sock(sk);
697 ip_flush_pending_frames(sk);
698 release_sock(sk);
699}
700
701/* This gets rid of all the nasties in af_inet. -DaveM */
702static int raw_bind(struct sock *sk, struct sockaddr_unsized *uaddr,
703 int addr_len)
704{
705 struct inet_sock *inet = inet_sk(sk);
706 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
707 struct net *net = sock_net(sk);
708 u32 tb_id = RT_TABLE_LOCAL;
709 int ret = -EINVAL;
710 int chk_addr_ret;
711
712 lock_sock(sk);
713 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
714 goto out;
715
716 if (sk->sk_bound_dev_if)
717 tb_id = l3mdev_fib_table_by_index(net,
718 sk->sk_bound_dev_if) ? : tb_id;
719
720 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
721
722 ret = -EADDRNOTAVAIL;
723 if (!inet_addr_valid_or_nonlocal(net, inet, addr->sin_addr.s_addr,
724 chk_addr_ret))
725 goto out;
726
727 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
728 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
729 inet->inet_saddr = 0; /* Use device */
730 sk_dst_reset(sk);
731 ret = 0;
732out:
733 release_sock(sk);
734 return ret;
735}
736
737/*
738 * This should be easy, if there is something there
739 * we return it, otherwise we block.
740 */
741
742static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
743 int flags)
744{
745 struct inet_sock *inet = inet_sk(sk);
746 size_t copied = 0;
747 int err = -EOPNOTSUPP;
748 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
749 struct sk_buff *skb;
750
751 if (flags & MSG_OOB)
752 goto out;
753
754 if (flags & MSG_ERRQUEUE) {
755 err = ip_recv_error(sk, msg, len);
756 goto out;
757 }
758
759 skb = skb_recv_datagram(sk, flags, &err);
760 if (!skb)
761 goto out;
762
763 copied = skb->len;
764 if (len < copied) {
765 msg->msg_flags |= MSG_TRUNC;
766 copied = len;
767 }
768
769 err = skb_copy_datagram_msg(skb, 0, msg, copied);
770 if (err)
771 goto done;
772
773 sock_recv_cmsgs(msg, sk, skb);
774
775 /* Copy the address. */
776 if (sin) {
777 sin->sin_family = AF_INET;
778 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
779 sin->sin_port = 0;
780 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
781 msg->msg_namelen = sizeof(*sin);
782 }
783 if (inet_cmsg_flags(inet))
784 ip_cmsg_recv(msg, skb);
785 if (flags & MSG_TRUNC)
786 copied = skb->len;
787done:
788 skb_free_datagram(sk, skb);
789out:
790 if (err)
791 return err;
792 return copied;
793}
794
795static int raw_sk_init(struct sock *sk)
796{
797 struct raw_sock *rp = raw_sk(sk);
798
799 sk->sk_drop_counters = &rp->drop_counters;
800 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
801 memset(&rp->filter, 0, sizeof(rp->filter));
802 return 0;
803}
804
805static int raw_seticmpfilter(struct sock *sk, sockptr_t optval, int optlen)
806{
807 if (optlen > sizeof(struct icmp_filter))
808 optlen = sizeof(struct icmp_filter);
809 if (copy_from_sockptr(&raw_sk(sk)->filter, optval, optlen))
810 return -EFAULT;
811 return 0;
812}
813
814static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
815{
816 int len, ret = -EFAULT;
817
818 if (get_user(len, optlen))
819 goto out;
820 ret = -EINVAL;
821 if (len < 0)
822 goto out;
823 if (len > sizeof(struct icmp_filter))
824 len = sizeof(struct icmp_filter);
825 ret = -EFAULT;
826 if (put_user(len, optlen) ||
827 copy_to_user(optval, &raw_sk(sk)->filter, len))
828 goto out;
829 ret = 0;
830out: return ret;
831}
832
833static int do_raw_setsockopt(struct sock *sk, int optname,
834 sockptr_t optval, unsigned int optlen)
835{
836 if (optname == ICMP_FILTER) {
837 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
838 return -EOPNOTSUPP;
839 else
840 return raw_seticmpfilter(sk, optval, optlen);
841 }
842 return -ENOPROTOOPT;
843}
844
845static int raw_setsockopt(struct sock *sk, int level, int optname,
846 sockptr_t optval, unsigned int optlen)
847{
848 if (level != SOL_RAW)
849 return ip_setsockopt(sk, level, optname, optval, optlen);
850 return do_raw_setsockopt(sk, optname, optval, optlen);
851}
852
853static int do_raw_getsockopt(struct sock *sk, int optname,
854 char __user *optval, int __user *optlen)
855{
856 if (optname == ICMP_FILTER) {
857 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
858 return -EOPNOTSUPP;
859 else
860 return raw_geticmpfilter(sk, optval, optlen);
861 }
862 return -ENOPROTOOPT;
863}
864
865static int raw_getsockopt(struct sock *sk, int level, int optname,
866 char __user *optval, int __user *optlen)
867{
868 if (level != SOL_RAW)
869 return ip_getsockopt(sk, level, optname, optval, optlen);
870 return do_raw_getsockopt(sk, optname, optval, optlen);
871}
872
873static int raw_ioctl(struct sock *sk, int cmd, int *karg)
874{
875 switch (cmd) {
876 case SIOCOUTQ: {
877 *karg = sk_wmem_alloc_get(sk);
878 return 0;
879 }
880 case SIOCINQ: {
881 struct sk_buff *skb;
882
883 spin_lock_bh(&sk->sk_receive_queue.lock);
884 skb = skb_peek(&sk->sk_receive_queue);
885 if (skb)
886 *karg = skb->len;
887 else
888 *karg = 0;
889 spin_unlock_bh(&sk->sk_receive_queue.lock);
890 return 0;
891 }
892
893 default:
894#ifdef CONFIG_IP_MROUTE
895 return ipmr_ioctl(sk, cmd, karg);
896#else
897 return -ENOIOCTLCMD;
898#endif
899 }
900}
901
902#ifdef CONFIG_COMPAT
903static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
904{
905 switch (cmd) {
906 case SIOCOUTQ:
907 case SIOCINQ:
908 return -ENOIOCTLCMD;
909 default:
910#ifdef CONFIG_IP_MROUTE
911 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
912#else
913 return -ENOIOCTLCMD;
914#endif
915 }
916}
917#endif
918
919int raw_abort(struct sock *sk, int err)
920{
921 lock_sock(sk);
922
923 sk->sk_err = err;
924 sk_error_report(sk);
925 __udp_disconnect(sk, 0);
926
927 release_sock(sk);
928
929 return 0;
930}
931EXPORT_SYMBOL_GPL(raw_abort);
932
933struct proto raw_prot = {
934 .name = "RAW",
935 .owner = THIS_MODULE,
936 .close = raw_close,
937 .destroy = raw_destroy,
938 .connect = ip4_datagram_connect,
939 .disconnect = __udp_disconnect,
940 .ioctl = raw_ioctl,
941 .init = raw_sk_init,
942 .setsockopt = raw_setsockopt,
943 .getsockopt = raw_getsockopt,
944 .sendmsg = raw_sendmsg,
945 .recvmsg = raw_recvmsg,
946 .bind = raw_bind,
947 .backlog_rcv = raw_rcv_skb,
948 .release_cb = ip4_datagram_release_cb,
949 .hash = raw_hash_sk,
950 .unhash = raw_unhash_sk,
951 .obj_size = sizeof(struct raw_sock),
952 .useroffset = offsetof(struct raw_sock, filter),
953 .usersize = sizeof_field(struct raw_sock, filter),
954 .h.raw_hash = &raw_v4_hashinfo,
955#ifdef CONFIG_COMPAT
956 .compat_ioctl = compat_raw_ioctl,
957#endif
958 .diag_destroy = raw_abort,
959};
960
961#ifdef CONFIG_PROC_FS
962static struct sock *raw_get_first(struct seq_file *seq, int bucket)
963{
964 struct raw_hashinfo *h = pde_data(file_inode(seq->file));
965 struct raw_iter_state *state = raw_seq_private(seq);
966 struct hlist_head *hlist;
967 struct sock *sk;
968
969 for (state->bucket = bucket; state->bucket < RAW_HTABLE_SIZE;
970 ++state->bucket) {
971 hlist = &h->ht[state->bucket];
972 sk_for_each(sk, hlist) {
973 if (sock_net(sk) == seq_file_net(seq))
974 return sk;
975 }
976 }
977 return NULL;
978}
979
980static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
981{
982 struct raw_iter_state *state = raw_seq_private(seq);
983
984 do {
985 sk = sk_next(sk);
986 } while (sk && sock_net(sk) != seq_file_net(seq));
987
988 if (!sk)
989 return raw_get_first(seq, state->bucket + 1);
990 return sk;
991}
992
993static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
994{
995 struct sock *sk = raw_get_first(seq, 0);
996
997 if (sk)
998 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
999 --pos;
1000 return pos ? NULL : sk;
1001}
1002
1003void *raw_seq_start(struct seq_file *seq, loff_t *pos)
1004 __acquires(&h->lock)
1005{
1006 struct raw_hashinfo *h = pde_data(file_inode(seq->file));
1007
1008 spin_lock(&h->lock);
1009
1010 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1011}
1012EXPORT_SYMBOL_GPL(raw_seq_start);
1013
1014void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1015{
1016 struct sock *sk;
1017
1018 if (v == SEQ_START_TOKEN)
1019 sk = raw_get_first(seq, 0);
1020 else
1021 sk = raw_get_next(seq, v);
1022 ++*pos;
1023 return sk;
1024}
1025EXPORT_SYMBOL_GPL(raw_seq_next);
1026
1027void raw_seq_stop(struct seq_file *seq, void *v)
1028 __releases(&h->lock)
1029{
1030 struct raw_hashinfo *h = pde_data(file_inode(seq->file));
1031
1032 spin_unlock(&h->lock);
1033}
1034EXPORT_SYMBOL_GPL(raw_seq_stop);
1035
1036static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1037{
1038 struct inet_sock *inet = inet_sk(sp);
1039 __be32 dest = inet->inet_daddr,
1040 src = inet->inet_rcv_saddr;
1041 __u16 destp = 0,
1042 srcp = inet->inet_num;
1043
1044 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1045 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %llu %d %pK %u\n",
1046 i, src, srcp, dest, destp, sp->sk_state,
1047 sk_wmem_alloc_get(sp),
1048 sk_rmem_alloc_get(sp),
1049 0, 0L, 0,
1050 from_kuid_munged(seq_user_ns(seq), sk_uid(sp)),
1051 0, sock_i_ino(sp),
1052 refcount_read(&sp->sk_refcnt), sp, sk_drops_read(sp));
1053}
1054
1055static int raw_seq_show(struct seq_file *seq, void *v)
1056{
1057 if (v == SEQ_START_TOKEN)
1058 seq_printf(seq, " sl local_address rem_address st tx_queue "
1059 "rx_queue tr tm->when retrnsmt uid timeout "
1060 "inode ref pointer drops\n");
1061 else
1062 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1063 return 0;
1064}
1065
1066static const struct seq_operations raw_seq_ops = {
1067 .start = raw_seq_start,
1068 .next = raw_seq_next,
1069 .stop = raw_seq_stop,
1070 .show = raw_seq_show,
1071};
1072
1073static __net_init int raw_init_net(struct net *net)
1074{
1075 if (!proc_create_net_data("raw", 0444, net->proc_net, &raw_seq_ops,
1076 sizeof(struct raw_iter_state), &raw_v4_hashinfo))
1077 return -ENOMEM;
1078
1079 return 0;
1080}
1081
1082static __net_exit void raw_exit_net(struct net *net)
1083{
1084 remove_proc_entry("raw", net->proc_net);
1085}
1086
1087static __net_initdata struct pernet_operations raw_net_ops = {
1088 .init = raw_init_net,
1089 .exit = raw_exit_net,
1090};
1091
1092int __init raw_proc_init(void)
1093{
1094
1095 return register_pernet_subsys(&raw_net_ops);
1096}
1097
1098void __init raw_proc_exit(void)
1099{
1100 unregister_pernet_subsys(&raw_net_ops);
1101}
1102#endif /* CONFIG_PROC_FS */
1103
1104static void raw_sysctl_init_net(struct net *net)
1105{
1106#ifdef CONFIG_NET_L3_MASTER_DEV
1107 net->ipv4.sysctl_raw_l3mdev_accept = 1;
1108#endif
1109}
1110
1111static int __net_init raw_sysctl_init(struct net *net)
1112{
1113 raw_sysctl_init_net(net);
1114 return 0;
1115}
1116
1117static struct pernet_operations __net_initdata raw_sysctl_ops = {
1118 .init = raw_sysctl_init,
1119};
1120
1121void __init raw_init(void)
1122{
1123 raw_sysctl_init_net(&init_net);
1124 if (register_pernet_subsys(&raw_sysctl_ops))
1125 panic("RAW: failed to init sysctl parameters.\n");
1126}