Loading...
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 = {
89 .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
90};
91EXPORT_SYMBOL_GPL(raw_v4_hashinfo);
92
93int raw_hash_sk(struct sock *sk)
94{
95 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
96 struct hlist_head *head;
97
98 head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)];
99
100 write_lock_bh(&h->lock);
101 sk_add_node(sk, head);
102 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
103 write_unlock_bh(&h->lock);
104
105 return 0;
106}
107EXPORT_SYMBOL_GPL(raw_hash_sk);
108
109void raw_unhash_sk(struct sock *sk)
110{
111 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
112
113 write_lock_bh(&h->lock);
114 if (sk_del_node_init(sk))
115 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
116 write_unlock_bh(&h->lock);
117}
118EXPORT_SYMBOL_GPL(raw_unhash_sk);
119
120struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
121 unsigned short num, __be32 raddr, __be32 laddr,
122 int dif, int sdif)
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 raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if, dif, sdif))
131 goto found; /* gotcha */
132 }
133 sk = NULL;
134found:
135 return sk;
136}
137EXPORT_SYMBOL_GPL(__raw_v4_lookup);
138
139/*
140 * 0 - deliver
141 * 1 - block
142 */
143static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
144{
145 struct icmphdr _hdr;
146 const struct icmphdr *hdr;
147
148 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
149 sizeof(_hdr), &_hdr);
150 if (!hdr)
151 return 1;
152
153 if (hdr->type < 32) {
154 __u32 data = raw_sk(sk)->filter.data;
155
156 return ((1U << hdr->type) & data) != 0;
157 }
158
159 /* Do not block unknown ICMP types */
160 return 0;
161}
162
163/* IP input processing comes here for RAW socket delivery.
164 * Caller owns SKB, so we must make clones.
165 *
166 * RFC 1122: SHOULD pass TOS value up to the transport layer.
167 * -> It does. And not only TOS, but all IP header.
168 */
169static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
170{
171 int sdif = inet_sdif(skb);
172 int dif = inet_iif(skb);
173 struct sock *sk;
174 struct hlist_head *head;
175 int delivered = 0;
176 struct net *net;
177
178 read_lock(&raw_v4_hashinfo.lock);
179 head = &raw_v4_hashinfo.ht[hash];
180 if (hlist_empty(head))
181 goto out;
182
183 net = dev_net(skb->dev);
184 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
185 iph->saddr, iph->daddr, dif, sdif);
186
187 while (sk) {
188 delivered = 1;
189 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
190 ip_mc_sf_allow(sk, iph->daddr, iph->saddr,
191 skb->dev->ifindex, sdif)) {
192 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
193
194 /* Not releasing hash table! */
195 if (clone)
196 raw_rcv(sk, clone);
197 }
198 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
199 iph->saddr, iph->daddr,
200 dif, sdif);
201 }
202out:
203 read_unlock(&raw_v4_hashinfo.lock);
204 return delivered;
205}
206
207int raw_local_deliver(struct sk_buff *skb, int protocol)
208{
209 int hash;
210 struct sock *raw_sk;
211
212 hash = protocol & (RAW_HTABLE_SIZE - 1);
213 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
214
215 /* If there maybe a raw socket we must check - if not we
216 * don't care less
217 */
218 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
219 raw_sk = NULL;
220
221 return raw_sk != NULL;
222
223}
224
225static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
226{
227 struct inet_sock *inet = inet_sk(sk);
228 const int type = icmp_hdr(skb)->type;
229 const int code = icmp_hdr(skb)->code;
230 int err = 0;
231 int harderr = 0;
232
233 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
234 ipv4_sk_update_pmtu(skb, sk, info);
235 else if (type == ICMP_REDIRECT) {
236 ipv4_sk_redirect(skb, sk);
237 return;
238 }
239
240 /* Report error on raw socket, if:
241 1. User requested ip_recverr.
242 2. Socket is connected (otherwise the error indication
243 is useless without ip_recverr and error is hard.
244 */
245 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
246 return;
247
248 switch (type) {
249 default:
250 case ICMP_TIME_EXCEEDED:
251 err = EHOSTUNREACH;
252 break;
253 case ICMP_SOURCE_QUENCH:
254 return;
255 case ICMP_PARAMETERPROB:
256 err = EPROTO;
257 harderr = 1;
258 break;
259 case ICMP_DEST_UNREACH:
260 err = EHOSTUNREACH;
261 if (code > NR_ICMP_UNREACH)
262 break;
263 if (code == ICMP_FRAG_NEEDED) {
264 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
265 err = EMSGSIZE;
266 } else {
267 err = icmp_err_convert[code].errno;
268 harderr = icmp_err_convert[code].fatal;
269 }
270 }
271
272 if (inet->recverr) {
273 const struct iphdr *iph = (const struct iphdr *)skb->data;
274 u8 *payload = skb->data + (iph->ihl << 2);
275
276 if (inet->hdrincl)
277 payload = skb->data;
278 ip_icmp_error(sk, skb, err, 0, info, payload);
279 }
280
281 if (inet->recverr || harderr) {
282 sk->sk_err = err;
283 sk_error_report(sk);
284 }
285}
286
287void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
288{
289 int hash;
290 struct sock *raw_sk;
291 const struct iphdr *iph;
292 struct net *net;
293
294 hash = protocol & (RAW_HTABLE_SIZE - 1);
295
296 read_lock(&raw_v4_hashinfo.lock);
297 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
298 if (raw_sk) {
299 int dif = skb->dev->ifindex;
300 int sdif = inet_sdif(skb);
301
302 iph = (const struct iphdr *)skb->data;
303 net = dev_net(skb->dev);
304
305 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
306 iph->daddr, iph->saddr,
307 dif, sdif)) != NULL) {
308 raw_err(raw_sk, skb, info);
309 raw_sk = sk_next(raw_sk);
310 iph = (const struct iphdr *)skb->data;
311 }
312 }
313 read_unlock(&raw_v4_hashinfo.lock);
314}
315
316static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
317{
318 /* Charge it to the socket. */
319
320 ipv4_pktinfo_prepare(sk, skb);
321 if (sock_queue_rcv_skb(sk, skb) < 0) {
322 kfree_skb(skb);
323 return NET_RX_DROP;
324 }
325
326 return NET_RX_SUCCESS;
327}
328
329int raw_rcv(struct sock *sk, struct sk_buff *skb)
330{
331 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
332 atomic_inc(&sk->sk_drops);
333 kfree_skb(skb);
334 return NET_RX_DROP;
335 }
336 nf_reset_ct(skb);
337
338 skb_push(skb, skb->data - skb_network_header(skb));
339
340 raw_rcv_skb(sk, skb);
341 return 0;
342}
343
344static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
345 struct msghdr *msg, size_t length,
346 struct rtable **rtp, unsigned int flags,
347 const struct sockcm_cookie *sockc)
348{
349 struct inet_sock *inet = inet_sk(sk);
350 struct net *net = sock_net(sk);
351 struct iphdr *iph;
352 struct sk_buff *skb;
353 unsigned int iphlen;
354 int err;
355 struct rtable *rt = *rtp;
356 int hlen, tlen;
357
358 if (length > rt->dst.dev->mtu) {
359 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
360 rt->dst.dev->mtu);
361 return -EMSGSIZE;
362 }
363 if (length < sizeof(struct iphdr))
364 return -EINVAL;
365
366 if (flags&MSG_PROBE)
367 goto out;
368
369 hlen = LL_RESERVED_SPACE(rt->dst.dev);
370 tlen = rt->dst.dev->needed_tailroom;
371 skb = sock_alloc_send_skb(sk,
372 length + hlen + tlen + 15,
373 flags & MSG_DONTWAIT, &err);
374 if (!skb)
375 goto error;
376 skb_reserve(skb, hlen);
377
378 skb->priority = sk->sk_priority;
379 skb->mark = sockc->mark;
380 skb->tstamp = sockc->transmit_time;
381 skb_dst_set(skb, &rt->dst);
382 *rtp = NULL;
383
384 skb_reset_network_header(skb);
385 iph = ip_hdr(skb);
386 skb_put(skb, length);
387
388 skb->ip_summed = CHECKSUM_NONE;
389
390 skb_setup_tx_timestamp(skb, sockc->tsflags);
391
392 if (flags & MSG_CONFIRM)
393 skb_set_dst_pending_confirm(skb, 1);
394
395 skb->transport_header = skb->network_header;
396 err = -EFAULT;
397 if (memcpy_from_msg(iph, msg, length))
398 goto error_free;
399
400 iphlen = iph->ihl * 4;
401
402 /*
403 * We don't want to modify the ip header, but we do need to
404 * be sure that it won't cause problems later along the network
405 * stack. Specifically we want to make sure that iph->ihl is a
406 * sane value. If ihl points beyond the length of the buffer passed
407 * in, reject the frame as invalid
408 */
409 err = -EINVAL;
410 if (iphlen > length)
411 goto error_free;
412
413 if (iphlen >= sizeof(*iph)) {
414 if (!iph->saddr)
415 iph->saddr = fl4->saddr;
416 iph->check = 0;
417 iph->tot_len = htons(length);
418 if (!iph->id)
419 ip_select_ident(net, skb, NULL);
420
421 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
422 skb->transport_header += iphlen;
423 if (iph->protocol == IPPROTO_ICMP &&
424 length >= iphlen + sizeof(struct icmphdr))
425 icmp_out_count(net, ((struct icmphdr *)
426 skb_transport_header(skb))->type);
427 }
428
429 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
430 net, sk, skb, NULL, rt->dst.dev,
431 dst_output);
432 if (err > 0)
433 err = net_xmit_errno(err);
434 if (err)
435 goto error;
436out:
437 return 0;
438
439error_free:
440 kfree_skb(skb);
441error:
442 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
443 if (err == -ENOBUFS && !inet->recverr)
444 err = 0;
445 return err;
446}
447
448static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
449{
450 int err;
451
452 if (fl4->flowi4_proto != IPPROTO_ICMP)
453 return 0;
454
455 /* We only need the first two bytes. */
456 rfv->hlen = 2;
457
458 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
459 if (err)
460 return err;
461
462 fl4->fl4_icmp_type = rfv->hdr.icmph.type;
463 fl4->fl4_icmp_code = rfv->hdr.icmph.code;
464
465 return 0;
466}
467
468static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
469 struct sk_buff *skb)
470{
471 struct raw_frag_vec *rfv = from;
472
473 if (offset < rfv->hlen) {
474 int copy = min(rfv->hlen - offset, len);
475
476 if (skb->ip_summed == CHECKSUM_PARTIAL)
477 memcpy(to, rfv->hdr.c + offset, copy);
478 else
479 skb->csum = csum_block_add(
480 skb->csum,
481 csum_partial_copy_nocheck(rfv->hdr.c + offset,
482 to, copy),
483 odd);
484
485 odd = 0;
486 offset += copy;
487 to += copy;
488 len -= copy;
489
490 if (!len)
491 return 0;
492 }
493
494 offset -= rfv->hlen;
495
496 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
497}
498
499static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
500{
501 struct inet_sock *inet = inet_sk(sk);
502 struct net *net = sock_net(sk);
503 struct ipcm_cookie ipc;
504 struct rtable *rt = NULL;
505 struct flowi4 fl4;
506 int free = 0;
507 __be32 daddr;
508 __be32 saddr;
509 u8 tos;
510 int err;
511 struct ip_options_data opt_copy;
512 struct raw_frag_vec rfv;
513 int hdrincl;
514
515 err = -EMSGSIZE;
516 if (len > 0xFFFF)
517 goto out;
518
519 /* hdrincl should be READ_ONCE(inet->hdrincl)
520 * but READ_ONCE() doesn't work with bit fields.
521 * Doing this indirectly yields the same result.
522 */
523 hdrincl = inet->hdrincl;
524 hdrincl = READ_ONCE(hdrincl);
525 /*
526 * Check the flags.
527 */
528
529 err = -EOPNOTSUPP;
530 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
531 goto out; /* compatibility */
532
533 /*
534 * Get and verify the address.
535 */
536
537 if (msg->msg_namelen) {
538 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
539 err = -EINVAL;
540 if (msg->msg_namelen < sizeof(*usin))
541 goto out;
542 if (usin->sin_family != AF_INET) {
543 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
544 __func__, current->comm);
545 err = -EAFNOSUPPORT;
546 if (usin->sin_family)
547 goto out;
548 }
549 daddr = usin->sin_addr.s_addr;
550 /* ANK: I did not forget to get protocol from port field.
551 * I just do not know, who uses this weirdness.
552 * IP_HDRINCL is much more convenient.
553 */
554 } else {
555 err = -EDESTADDRREQ;
556 if (sk->sk_state != TCP_ESTABLISHED)
557 goto out;
558 daddr = inet->inet_daddr;
559 }
560
561 ipcm_init_sk(&ipc, inet);
562
563 if (msg->msg_controllen) {
564 err = ip_cmsg_send(sk, msg, &ipc, false);
565 if (unlikely(err)) {
566 kfree(ipc.opt);
567 goto out;
568 }
569 if (ipc.opt)
570 free = 1;
571 }
572
573 saddr = ipc.addr;
574 ipc.addr = daddr;
575
576 if (!ipc.opt) {
577 struct ip_options_rcu *inet_opt;
578
579 rcu_read_lock();
580 inet_opt = rcu_dereference(inet->inet_opt);
581 if (inet_opt) {
582 memcpy(&opt_copy, inet_opt,
583 sizeof(*inet_opt) + inet_opt->opt.optlen);
584 ipc.opt = &opt_copy.opt;
585 }
586 rcu_read_unlock();
587 }
588
589 if (ipc.opt) {
590 err = -EINVAL;
591 /* Linux does not mangle headers on raw sockets,
592 * so that IP options + IP_HDRINCL is non-sense.
593 */
594 if (hdrincl)
595 goto done;
596 if (ipc.opt->opt.srr) {
597 if (!daddr)
598 goto done;
599 daddr = ipc.opt->opt.faddr;
600 }
601 }
602 tos = get_rtconn_flags(&ipc, sk);
603 if (msg->msg_flags & MSG_DONTROUTE)
604 tos |= RTO_ONLINK;
605
606 if (ipv4_is_multicast(daddr)) {
607 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
608 ipc.oif = inet->mc_index;
609 if (!saddr)
610 saddr = inet->mc_addr;
611 } else if (!ipc.oif) {
612 ipc.oif = inet->uc_index;
613 } else if (ipv4_is_lbcast(daddr) && inet->uc_index) {
614 /* oif is set, packet is to local broadcast
615 * and uc_index is set. oif is most likely set
616 * by sk_bound_dev_if. If uc_index != oif check if the
617 * oif is an L3 master and uc_index is an L3 slave.
618 * If so, we want to allow the send using the uc_index.
619 */
620 if (ipc.oif != inet->uc_index &&
621 ipc.oif == l3mdev_master_ifindex_by_index(sock_net(sk),
622 inet->uc_index)) {
623 ipc.oif = inet->uc_index;
624 }
625 }
626
627 flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
628 RT_SCOPE_UNIVERSE,
629 hdrincl ? IPPROTO_RAW : sk->sk_protocol,
630 inet_sk_flowi_flags(sk) |
631 (hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
632 daddr, saddr, 0, 0, sk->sk_uid);
633
634 if (!hdrincl) {
635 rfv.msg = msg;
636 rfv.hlen = 0;
637
638 err = raw_probe_proto_opt(&rfv, &fl4);
639 if (err)
640 goto done;
641 }
642
643 security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
644 rt = ip_route_output_flow(net, &fl4, sk);
645 if (IS_ERR(rt)) {
646 err = PTR_ERR(rt);
647 rt = NULL;
648 goto done;
649 }
650
651 err = -EACCES;
652 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
653 goto done;
654
655 if (msg->msg_flags & MSG_CONFIRM)
656 goto do_confirm;
657back_from_confirm:
658
659 if (hdrincl)
660 err = raw_send_hdrinc(sk, &fl4, msg, len,
661 &rt, msg->msg_flags, &ipc.sockc);
662
663 else {
664 if (!ipc.addr)
665 ipc.addr = fl4.daddr;
666 lock_sock(sk);
667 err = ip_append_data(sk, &fl4, raw_getfrag,
668 &rfv, len, 0,
669 &ipc, &rt, msg->msg_flags);
670 if (err)
671 ip_flush_pending_frames(sk);
672 else if (!(msg->msg_flags & MSG_MORE)) {
673 err = ip_push_pending_frames(sk, &fl4);
674 if (err == -ENOBUFS && !inet->recverr)
675 err = 0;
676 }
677 release_sock(sk);
678 }
679done:
680 if (free)
681 kfree(ipc.opt);
682 ip_rt_put(rt);
683
684out:
685 if (err < 0)
686 return err;
687 return len;
688
689do_confirm:
690 if (msg->msg_flags & MSG_PROBE)
691 dst_confirm_neigh(&rt->dst, &fl4.daddr);
692 if (!(msg->msg_flags & MSG_PROBE) || len)
693 goto back_from_confirm;
694 err = 0;
695 goto done;
696}
697
698static void raw_close(struct sock *sk, long timeout)
699{
700 /*
701 * Raw sockets may have direct kernel references. Kill them.
702 */
703 ip_ra_control(sk, 0, NULL);
704
705 sk_common_release(sk);
706}
707
708static void raw_destroy(struct sock *sk)
709{
710 lock_sock(sk);
711 ip_flush_pending_frames(sk);
712 release_sock(sk);
713}
714
715/* This gets rid of all the nasties in af_inet. -DaveM */
716static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
717{
718 struct inet_sock *inet = inet_sk(sk);
719 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
720 u32 tb_id = RT_TABLE_LOCAL;
721 int ret = -EINVAL;
722 int chk_addr_ret;
723
724 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
725 goto out;
726
727 if (sk->sk_bound_dev_if)
728 tb_id = l3mdev_fib_table_by_index(sock_net(sk),
729 sk->sk_bound_dev_if) ? : tb_id;
730
731 chk_addr_ret = inet_addr_type_table(sock_net(sk), addr->sin_addr.s_addr,
732 tb_id);
733
734 ret = -EADDRNOTAVAIL;
735 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
736 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
737 goto out;
738 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
739 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
740 inet->inet_saddr = 0; /* Use device */
741 sk_dst_reset(sk);
742 ret = 0;
743out: return ret;
744}
745
746/*
747 * This should be easy, if there is something there
748 * we return it, otherwise we block.
749 */
750
751static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
752 int noblock, int flags, int *addr_len)
753{
754 struct inet_sock *inet = inet_sk(sk);
755 size_t copied = 0;
756 int err = -EOPNOTSUPP;
757 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
758 struct sk_buff *skb;
759
760 if (flags & MSG_OOB)
761 goto out;
762
763 if (flags & MSG_ERRQUEUE) {
764 err = ip_recv_error(sk, msg, len, addr_len);
765 goto out;
766 }
767
768 skb = skb_recv_datagram(sk, flags, noblock, &err);
769 if (!skb)
770 goto out;
771
772 copied = skb->len;
773 if (len < copied) {
774 msg->msg_flags |= MSG_TRUNC;
775 copied = len;
776 }
777
778 err = skb_copy_datagram_msg(skb, 0, msg, copied);
779 if (err)
780 goto done;
781
782 sock_recv_ts_and_drops(msg, sk, skb);
783
784 /* Copy the address. */
785 if (sin) {
786 sin->sin_family = AF_INET;
787 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
788 sin->sin_port = 0;
789 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
790 *addr_len = sizeof(*sin);
791 }
792 if (inet->cmsg_flags)
793 ip_cmsg_recv(msg, skb);
794 if (flags & MSG_TRUNC)
795 copied = skb->len;
796done:
797 skb_free_datagram(sk, skb);
798out:
799 if (err)
800 return err;
801 return copied;
802}
803
804static int raw_sk_init(struct sock *sk)
805{
806 struct raw_sock *rp = raw_sk(sk);
807
808 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
809 memset(&rp->filter, 0, sizeof(rp->filter));
810 return 0;
811}
812
813static int raw_seticmpfilter(struct sock *sk, sockptr_t optval, int optlen)
814{
815 if (optlen > sizeof(struct icmp_filter))
816 optlen = sizeof(struct icmp_filter);
817 if (copy_from_sockptr(&raw_sk(sk)->filter, optval, optlen))
818 return -EFAULT;
819 return 0;
820}
821
822static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
823{
824 int len, ret = -EFAULT;
825
826 if (get_user(len, optlen))
827 goto out;
828 ret = -EINVAL;
829 if (len < 0)
830 goto out;
831 if (len > sizeof(struct icmp_filter))
832 len = sizeof(struct icmp_filter);
833 ret = -EFAULT;
834 if (put_user(len, optlen) ||
835 copy_to_user(optval, &raw_sk(sk)->filter, len))
836 goto out;
837 ret = 0;
838out: return ret;
839}
840
841static int do_raw_setsockopt(struct sock *sk, int level, int optname,
842 sockptr_t optval, unsigned int optlen)
843{
844 if (optname == ICMP_FILTER) {
845 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
846 return -EOPNOTSUPP;
847 else
848 return raw_seticmpfilter(sk, optval, optlen);
849 }
850 return -ENOPROTOOPT;
851}
852
853static int raw_setsockopt(struct sock *sk, int level, int optname,
854 sockptr_t optval, unsigned int optlen)
855{
856 if (level != SOL_RAW)
857 return ip_setsockopt(sk, level, optname, optval, optlen);
858 return do_raw_setsockopt(sk, level, optname, optval, optlen);
859}
860
861static int do_raw_getsockopt(struct sock *sk, int level, int optname,
862 char __user *optval, int __user *optlen)
863{
864 if (optname == ICMP_FILTER) {
865 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
866 return -EOPNOTSUPP;
867 else
868 return raw_geticmpfilter(sk, optval, optlen);
869 }
870 return -ENOPROTOOPT;
871}
872
873static int raw_getsockopt(struct sock *sk, int level, int optname,
874 char __user *optval, int __user *optlen)
875{
876 if (level != SOL_RAW)
877 return ip_getsockopt(sk, level, optname, optval, optlen);
878 return do_raw_getsockopt(sk, level, optname, optval, optlen);
879}
880
881static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
882{
883 switch (cmd) {
884 case SIOCOUTQ: {
885 int amount = sk_wmem_alloc_get(sk);
886
887 return put_user(amount, (int __user *)arg);
888 }
889 case SIOCINQ: {
890 struct sk_buff *skb;
891 int amount = 0;
892
893 spin_lock_bh(&sk->sk_receive_queue.lock);
894 skb = skb_peek(&sk->sk_receive_queue);
895 if (skb)
896 amount = skb->len;
897 spin_unlock_bh(&sk->sk_receive_queue.lock);
898 return put_user(amount, (int __user *)arg);
899 }
900
901 default:
902#ifdef CONFIG_IP_MROUTE
903 return ipmr_ioctl(sk, cmd, (void __user *)arg);
904#else
905 return -ENOIOCTLCMD;
906#endif
907 }
908}
909
910#ifdef CONFIG_COMPAT
911static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
912{
913 switch (cmd) {
914 case SIOCOUTQ:
915 case SIOCINQ:
916 return -ENOIOCTLCMD;
917 default:
918#ifdef CONFIG_IP_MROUTE
919 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
920#else
921 return -ENOIOCTLCMD;
922#endif
923 }
924}
925#endif
926
927int raw_abort(struct sock *sk, int err)
928{
929 lock_sock(sk);
930
931 sk->sk_err = err;
932 sk_error_report(sk);
933 __udp_disconnect(sk, 0);
934
935 release_sock(sk);
936
937 return 0;
938}
939EXPORT_SYMBOL_GPL(raw_abort);
940
941struct proto raw_prot = {
942 .name = "RAW",
943 .owner = THIS_MODULE,
944 .close = raw_close,
945 .destroy = raw_destroy,
946 .connect = ip4_datagram_connect,
947 .disconnect = __udp_disconnect,
948 .ioctl = raw_ioctl,
949 .init = raw_sk_init,
950 .setsockopt = raw_setsockopt,
951 .getsockopt = raw_getsockopt,
952 .sendmsg = raw_sendmsg,
953 .recvmsg = raw_recvmsg,
954 .bind = raw_bind,
955 .backlog_rcv = raw_rcv_skb,
956 .release_cb = ip4_datagram_release_cb,
957 .hash = raw_hash_sk,
958 .unhash = raw_unhash_sk,
959 .obj_size = sizeof(struct raw_sock),
960 .useroffset = offsetof(struct raw_sock, filter),
961 .usersize = sizeof_field(struct raw_sock, filter),
962 .h.raw_hash = &raw_v4_hashinfo,
963#ifdef CONFIG_COMPAT
964 .compat_ioctl = compat_raw_ioctl,
965#endif
966 .diag_destroy = raw_abort,
967};
968
969#ifdef CONFIG_PROC_FS
970static struct sock *raw_get_first(struct seq_file *seq)
971{
972 struct sock *sk;
973 struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file));
974 struct raw_iter_state *state = raw_seq_private(seq);
975
976 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
977 ++state->bucket) {
978 sk_for_each(sk, &h->ht[state->bucket])
979 if (sock_net(sk) == seq_file_net(seq))
980 goto found;
981 }
982 sk = NULL;
983found:
984 return sk;
985}
986
987static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
988{
989 struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file));
990 struct raw_iter_state *state = raw_seq_private(seq);
991
992 do {
993 sk = sk_next(sk);
994try_again:
995 ;
996 } while (sk && sock_net(sk) != seq_file_net(seq));
997
998 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
999 sk = sk_head(&h->ht[state->bucket]);
1000 goto try_again;
1001 }
1002 return sk;
1003}
1004
1005static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
1006{
1007 struct sock *sk = raw_get_first(seq);
1008
1009 if (sk)
1010 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
1011 --pos;
1012 return pos ? NULL : sk;
1013}
1014
1015void *raw_seq_start(struct seq_file *seq, loff_t *pos)
1016 __acquires(&h->lock)
1017{
1018 struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file));
1019
1020 read_lock(&h->lock);
1021 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1022}
1023EXPORT_SYMBOL_GPL(raw_seq_start);
1024
1025void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1026{
1027 struct sock *sk;
1028
1029 if (v == SEQ_START_TOKEN)
1030 sk = raw_get_first(seq);
1031 else
1032 sk = raw_get_next(seq, v);
1033 ++*pos;
1034 return sk;
1035}
1036EXPORT_SYMBOL_GPL(raw_seq_next);
1037
1038void raw_seq_stop(struct seq_file *seq, void *v)
1039 __releases(&h->lock)
1040{
1041 struct raw_hashinfo *h = PDE_DATA(file_inode(seq->file));
1042
1043 read_unlock(&h->lock);
1044}
1045EXPORT_SYMBOL_GPL(raw_seq_stop);
1046
1047static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1048{
1049 struct inet_sock *inet = inet_sk(sp);
1050 __be32 dest = inet->inet_daddr,
1051 src = inet->inet_rcv_saddr;
1052 __u16 destp = 0,
1053 srcp = inet->inet_num;
1054
1055 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1056 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u\n",
1057 i, src, srcp, dest, destp, sp->sk_state,
1058 sk_wmem_alloc_get(sp),
1059 sk_rmem_alloc_get(sp),
1060 0, 0L, 0,
1061 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1062 0, sock_i_ino(sp),
1063 refcount_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1064}
1065
1066static int raw_seq_show(struct seq_file *seq, void *v)
1067{
1068 if (v == SEQ_START_TOKEN)
1069 seq_printf(seq, " sl local_address rem_address st tx_queue "
1070 "rx_queue tr tm->when retrnsmt uid timeout "
1071 "inode ref pointer drops\n");
1072 else
1073 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1074 return 0;
1075}
1076
1077static const struct seq_operations raw_seq_ops = {
1078 .start = raw_seq_start,
1079 .next = raw_seq_next,
1080 .stop = raw_seq_stop,
1081 .show = raw_seq_show,
1082};
1083
1084static __net_init int raw_init_net(struct net *net)
1085{
1086 if (!proc_create_net_data("raw", 0444, net->proc_net, &raw_seq_ops,
1087 sizeof(struct raw_iter_state), &raw_v4_hashinfo))
1088 return -ENOMEM;
1089
1090 return 0;
1091}
1092
1093static __net_exit void raw_exit_net(struct net *net)
1094{
1095 remove_proc_entry("raw", net->proc_net);
1096}
1097
1098static __net_initdata struct pernet_operations raw_net_ops = {
1099 .init = raw_init_net,
1100 .exit = raw_exit_net,
1101};
1102
1103int __init raw_proc_init(void)
1104{
1105 return register_pernet_subsys(&raw_net_ops);
1106}
1107
1108void __init raw_proc_exit(void)
1109{
1110 unregister_pernet_subsys(&raw_net_ops);
1111}
1112#endif /* CONFIG_PROC_FS */
1113
1114static void raw_sysctl_init_net(struct net *net)
1115{
1116#ifdef CONFIG_NET_L3_MASTER_DEV
1117 net->ipv4.sysctl_raw_l3mdev_accept = 1;
1118#endif
1119}
1120
1121static int __net_init raw_sysctl_init(struct net *net)
1122{
1123 raw_sysctl_init_net(net);
1124 return 0;
1125}
1126
1127static struct pernet_operations __net_initdata raw_sysctl_ops = {
1128 .init = raw_sysctl_init,
1129};
1130
1131void __init raw_init(void)
1132{
1133 raw_sysctl_init_net(&init_net);
1134 if (register_pernet_subsys(&raw_sysctl_ops))
1135 panic("RAW: failed to init sysctl parameters.\n");
1136}
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 atomic_inc(&sk->sk_drops);
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 if (sock_queue_rcv_skb_reason(sk, skb, &reason) < 0) {
304 sk_skb_reason_drop(sk, skb, reason);
305 return NET_RX_DROP;
306 }
307
308 return NET_RX_SUCCESS;
309}
310
311int raw_rcv(struct sock *sk, struct sk_buff *skb)
312{
313 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
314 atomic_inc(&sk->sk_drops);
315 sk_skb_reason_drop(sk, skb, SKB_DROP_REASON_XFRM_POLICY);
316 return NET_RX_DROP;
317 }
318 nf_reset_ct(skb);
319
320 skb_push(skb, -skb_network_offset(skb));
321
322 raw_rcv_skb(sk, skb);
323 return 0;
324}
325
326static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
327 struct msghdr *msg, size_t length,
328 struct rtable **rtp, unsigned int flags,
329 const struct sockcm_cookie *sockc)
330{
331 struct inet_sock *inet = inet_sk(sk);
332 struct net *net = sock_net(sk);
333 struct iphdr *iph;
334 struct sk_buff *skb;
335 unsigned int iphlen;
336 int err;
337 struct rtable *rt = *rtp;
338 int hlen, tlen;
339
340 if (length > rt->dst.dev->mtu) {
341 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
342 rt->dst.dev->mtu);
343 return -EMSGSIZE;
344 }
345 if (length < sizeof(struct iphdr))
346 return -EINVAL;
347
348 if (flags&MSG_PROBE)
349 goto out;
350
351 hlen = LL_RESERVED_SPACE(rt->dst.dev);
352 tlen = rt->dst.dev->needed_tailroom;
353 skb = sock_alloc_send_skb(sk,
354 length + hlen + tlen + 15,
355 flags & MSG_DONTWAIT, &err);
356 if (!skb)
357 goto error;
358 skb_reserve(skb, hlen);
359
360 skb->protocol = htons(ETH_P_IP);
361 skb->priority = READ_ONCE(sk->sk_priority);
362 skb->mark = sockc->mark;
363 skb_set_delivery_type_by_clockid(skb, sockc->transmit_time, sk->sk_clockid);
364 skb_dst_set(skb, &rt->dst);
365 *rtp = NULL;
366
367 skb_reset_network_header(skb);
368 iph = ip_hdr(skb);
369 skb_put(skb, length);
370
371 skb->ip_summed = CHECKSUM_NONE;
372
373 skb_setup_tx_timestamp(skb, sockc);
374
375 if (flags & MSG_CONFIRM)
376 skb_set_dst_pending_confirm(skb, 1);
377
378 skb->transport_header = skb->network_header;
379 err = -EFAULT;
380 if (memcpy_from_msg(iph, msg, length))
381 goto error_free;
382
383 iphlen = iph->ihl * 4;
384
385 /*
386 * We don't want to modify the ip header, but we do need to
387 * be sure that it won't cause problems later along the network
388 * stack. Specifically we want to make sure that iph->ihl is a
389 * sane value. If ihl points beyond the length of the buffer passed
390 * in, reject the frame as invalid
391 */
392 err = -EINVAL;
393 if (iphlen > length)
394 goto error_free;
395
396 if (iphlen >= sizeof(*iph)) {
397 if (!iph->saddr)
398 iph->saddr = fl4->saddr;
399 iph->check = 0;
400 iph->tot_len = htons(length);
401 if (!iph->id)
402 ip_select_ident(net, skb, NULL);
403
404 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
405 skb->transport_header += iphlen;
406 if (iph->protocol == IPPROTO_ICMP &&
407 length >= iphlen + sizeof(struct icmphdr))
408 icmp_out_count(net, ((struct icmphdr *)
409 skb_transport_header(skb))->type);
410 }
411
412 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
413 net, sk, skb, NULL, rt->dst.dev,
414 dst_output);
415 if (err > 0)
416 err = net_xmit_errno(err);
417 if (err)
418 goto error;
419out:
420 return 0;
421
422error_free:
423 kfree_skb(skb);
424error:
425 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
426 if (err == -ENOBUFS && !inet_test_bit(RECVERR, sk))
427 err = 0;
428 return err;
429}
430
431static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
432{
433 int err;
434
435 if (fl4->flowi4_proto != IPPROTO_ICMP)
436 return 0;
437
438 /* We only need the first two bytes. */
439 rfv->hlen = 2;
440
441 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
442 if (err)
443 return err;
444
445 fl4->fl4_icmp_type = rfv->hdr.icmph.type;
446 fl4->fl4_icmp_code = rfv->hdr.icmph.code;
447
448 return 0;
449}
450
451static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
452 struct sk_buff *skb)
453{
454 struct raw_frag_vec *rfv = from;
455
456 if (offset < rfv->hlen) {
457 int copy = min(rfv->hlen - offset, len);
458
459 if (skb->ip_summed == CHECKSUM_PARTIAL)
460 memcpy(to, rfv->hdr.c + offset, copy);
461 else
462 skb->csum = csum_block_add(
463 skb->csum,
464 csum_partial_copy_nocheck(rfv->hdr.c + offset,
465 to, copy),
466 odd);
467
468 odd = 0;
469 offset += copy;
470 to += copy;
471 len -= copy;
472
473 if (!len)
474 return 0;
475 }
476
477 offset -= rfv->hlen;
478
479 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
480}
481
482static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
483{
484 struct inet_sock *inet = inet_sk(sk);
485 struct net *net = sock_net(sk);
486 struct ipcm_cookie ipc;
487 struct rtable *rt = NULL;
488 struct flowi4 fl4;
489 u8 tos, scope;
490 int free = 0;
491 __be32 daddr;
492 __be32 saddr;
493 int uc_index, err;
494 struct ip_options_data opt_copy;
495 struct raw_frag_vec rfv;
496 int hdrincl;
497
498 err = -EMSGSIZE;
499 if (len > 0xFFFF)
500 goto out;
501
502 hdrincl = inet_test_bit(HDRINCL, sk);
503
504 /*
505 * Check the flags.
506 */
507
508 err = -EOPNOTSUPP;
509 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
510 goto out; /* compatibility */
511
512 /*
513 * Get and verify the address.
514 */
515
516 if (msg->msg_namelen) {
517 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
518 err = -EINVAL;
519 if (msg->msg_namelen < sizeof(*usin))
520 goto out;
521 if (usin->sin_family != AF_INET) {
522 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
523 __func__, current->comm);
524 err = -EAFNOSUPPORT;
525 if (usin->sin_family)
526 goto out;
527 }
528 daddr = usin->sin_addr.s_addr;
529 /* ANK: I did not forget to get protocol from port field.
530 * I just do not know, who uses this weirdness.
531 * IP_HDRINCL is much more convenient.
532 */
533 } else {
534 err = -EDESTADDRREQ;
535 if (sk->sk_state != TCP_ESTABLISHED)
536 goto out;
537 daddr = inet->inet_daddr;
538 }
539
540 ipcm_init_sk(&ipc, inet);
541 /* Keep backward compat */
542 if (hdrincl)
543 ipc.protocol = IPPROTO_RAW;
544
545 if (msg->msg_controllen) {
546 err = ip_cmsg_send(sk, msg, &ipc, false);
547 if (unlikely(err)) {
548 kfree(ipc.opt);
549 goto out;
550 }
551 if (ipc.opt)
552 free = 1;
553 }
554
555 saddr = ipc.addr;
556 ipc.addr = daddr;
557
558 if (!ipc.opt) {
559 struct ip_options_rcu *inet_opt;
560
561 rcu_read_lock();
562 inet_opt = rcu_dereference(inet->inet_opt);
563 if (inet_opt) {
564 memcpy(&opt_copy, inet_opt,
565 sizeof(*inet_opt) + inet_opt->opt.optlen);
566 ipc.opt = &opt_copy.opt;
567 }
568 rcu_read_unlock();
569 }
570
571 if (ipc.opt) {
572 err = -EINVAL;
573 /* Linux does not mangle headers on raw sockets,
574 * so that IP options + IP_HDRINCL is non-sense.
575 */
576 if (hdrincl)
577 goto done;
578 if (ipc.opt->opt.srr) {
579 if (!daddr)
580 goto done;
581 daddr = ipc.opt->opt.faddr;
582 }
583 }
584 tos = get_rttos(&ipc, inet);
585 scope = ip_sendmsg_scope(inet, &ipc, msg);
586
587 uc_index = READ_ONCE(inet->uc_index);
588 if (ipv4_is_multicast(daddr)) {
589 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
590 ipc.oif = READ_ONCE(inet->mc_index);
591 if (!saddr)
592 saddr = READ_ONCE(inet->mc_addr);
593 } else if (!ipc.oif) {
594 ipc.oif = uc_index;
595 } else if (ipv4_is_lbcast(daddr) && uc_index) {
596 /* oif is set, packet is to local broadcast
597 * and uc_index is set. oif is most likely set
598 * by sk_bound_dev_if. If uc_index != oif check if the
599 * oif is an L3 master and uc_index is an L3 slave.
600 * If so, we want to allow the send using the uc_index.
601 */
602 if (ipc.oif != uc_index &&
603 ipc.oif == l3mdev_master_ifindex_by_index(sock_net(sk),
604 uc_index)) {
605 ipc.oif = uc_index;
606 }
607 }
608
609 flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos, scope,
610 hdrincl ? ipc.protocol : sk->sk_protocol,
611 inet_sk_flowi_flags(sk) |
612 (hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
613 daddr, saddr, 0, 0, sk->sk_uid);
614
615 fl4.fl4_icmp_type = 0;
616 fl4.fl4_icmp_code = 0;
617
618 if (!hdrincl) {
619 rfv.msg = msg;
620 rfv.hlen = 0;
621
622 err = raw_probe_proto_opt(&rfv, &fl4);
623 if (err)
624 goto done;
625 }
626
627 security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
628 rt = ip_route_output_flow(net, &fl4, sk);
629 if (IS_ERR(rt)) {
630 err = PTR_ERR(rt);
631 rt = NULL;
632 goto done;
633 }
634
635 err = -EACCES;
636 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
637 goto done;
638
639 if (msg->msg_flags & MSG_CONFIRM)
640 goto do_confirm;
641back_from_confirm:
642
643 if (hdrincl)
644 err = raw_send_hdrinc(sk, &fl4, msg, len,
645 &rt, msg->msg_flags, &ipc.sockc);
646
647 else {
648 if (!ipc.addr)
649 ipc.addr = fl4.daddr;
650 lock_sock(sk);
651 err = ip_append_data(sk, &fl4, raw_getfrag,
652 &rfv, len, 0,
653 &ipc, &rt, msg->msg_flags);
654 if (err)
655 ip_flush_pending_frames(sk);
656 else if (!(msg->msg_flags & MSG_MORE)) {
657 err = ip_push_pending_frames(sk, &fl4);
658 if (err == -ENOBUFS && !inet_test_bit(RECVERR, sk))
659 err = 0;
660 }
661 release_sock(sk);
662 }
663done:
664 if (free)
665 kfree(ipc.opt);
666 ip_rt_put(rt);
667
668out:
669 if (err < 0)
670 return err;
671 return len;
672
673do_confirm:
674 if (msg->msg_flags & MSG_PROBE)
675 dst_confirm_neigh(&rt->dst, &fl4.daddr);
676 if (!(msg->msg_flags & MSG_PROBE) || len)
677 goto back_from_confirm;
678 err = 0;
679 goto done;
680}
681
682static void raw_close(struct sock *sk, long timeout)
683{
684 /*
685 * Raw sockets may have direct kernel references. Kill them.
686 */
687 ip_ra_control(sk, 0, NULL);
688
689 sk_common_release(sk);
690}
691
692static void raw_destroy(struct sock *sk)
693{
694 lock_sock(sk);
695 ip_flush_pending_frames(sk);
696 release_sock(sk);
697}
698
699/* This gets rid of all the nasties in af_inet. -DaveM */
700static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
701{
702 struct inet_sock *inet = inet_sk(sk);
703 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
704 struct net *net = sock_net(sk);
705 u32 tb_id = RT_TABLE_LOCAL;
706 int ret = -EINVAL;
707 int chk_addr_ret;
708
709 lock_sock(sk);
710 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
711 goto out;
712
713 if (sk->sk_bound_dev_if)
714 tb_id = l3mdev_fib_table_by_index(net,
715 sk->sk_bound_dev_if) ? : tb_id;
716
717 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
718
719 ret = -EADDRNOTAVAIL;
720 if (!inet_addr_valid_or_nonlocal(net, inet, addr->sin_addr.s_addr,
721 chk_addr_ret))
722 goto out;
723
724 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
725 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
726 inet->inet_saddr = 0; /* Use device */
727 sk_dst_reset(sk);
728 ret = 0;
729out:
730 release_sock(sk);
731 return ret;
732}
733
734/*
735 * This should be easy, if there is something there
736 * we return it, otherwise we block.
737 */
738
739static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
740 int flags, int *addr_len)
741{
742 struct inet_sock *inet = inet_sk(sk);
743 size_t copied = 0;
744 int err = -EOPNOTSUPP;
745 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
746 struct sk_buff *skb;
747
748 if (flags & MSG_OOB)
749 goto out;
750
751 if (flags & MSG_ERRQUEUE) {
752 err = ip_recv_error(sk, msg, len, addr_len);
753 goto out;
754 }
755
756 skb = skb_recv_datagram(sk, flags, &err);
757 if (!skb)
758 goto out;
759
760 copied = skb->len;
761 if (len < copied) {
762 msg->msg_flags |= MSG_TRUNC;
763 copied = len;
764 }
765
766 err = skb_copy_datagram_msg(skb, 0, msg, copied);
767 if (err)
768 goto done;
769
770 sock_recv_cmsgs(msg, sk, skb);
771
772 /* Copy the address. */
773 if (sin) {
774 sin->sin_family = AF_INET;
775 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
776 sin->sin_port = 0;
777 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
778 *addr_len = sizeof(*sin);
779 }
780 if (inet_cmsg_flags(inet))
781 ip_cmsg_recv(msg, skb);
782 if (flags & MSG_TRUNC)
783 copied = skb->len;
784done:
785 skb_free_datagram(sk, skb);
786out:
787 if (err)
788 return err;
789 return copied;
790}
791
792static int raw_sk_init(struct sock *sk)
793{
794 struct raw_sock *rp = raw_sk(sk);
795
796 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
797 memset(&rp->filter, 0, sizeof(rp->filter));
798 return 0;
799}
800
801static int raw_seticmpfilter(struct sock *sk, sockptr_t optval, int optlen)
802{
803 if (optlen > sizeof(struct icmp_filter))
804 optlen = sizeof(struct icmp_filter);
805 if (copy_from_sockptr(&raw_sk(sk)->filter, optval, optlen))
806 return -EFAULT;
807 return 0;
808}
809
810static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
811{
812 int len, ret = -EFAULT;
813
814 if (get_user(len, optlen))
815 goto out;
816 ret = -EINVAL;
817 if (len < 0)
818 goto out;
819 if (len > sizeof(struct icmp_filter))
820 len = sizeof(struct icmp_filter);
821 ret = -EFAULT;
822 if (put_user(len, optlen) ||
823 copy_to_user(optval, &raw_sk(sk)->filter, len))
824 goto out;
825 ret = 0;
826out: return ret;
827}
828
829static int do_raw_setsockopt(struct sock *sk, int optname,
830 sockptr_t optval, unsigned int optlen)
831{
832 if (optname == ICMP_FILTER) {
833 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
834 return -EOPNOTSUPP;
835 else
836 return raw_seticmpfilter(sk, optval, optlen);
837 }
838 return -ENOPROTOOPT;
839}
840
841static int raw_setsockopt(struct sock *sk, int level, int optname,
842 sockptr_t optval, unsigned int optlen)
843{
844 if (level != SOL_RAW)
845 return ip_setsockopt(sk, level, optname, optval, optlen);
846 return do_raw_setsockopt(sk, optname, optval, optlen);
847}
848
849static int do_raw_getsockopt(struct sock *sk, int optname,
850 char __user *optval, int __user *optlen)
851{
852 if (optname == ICMP_FILTER) {
853 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
854 return -EOPNOTSUPP;
855 else
856 return raw_geticmpfilter(sk, optval, optlen);
857 }
858 return -ENOPROTOOPT;
859}
860
861static int raw_getsockopt(struct sock *sk, int level, int optname,
862 char __user *optval, int __user *optlen)
863{
864 if (level != SOL_RAW)
865 return ip_getsockopt(sk, level, optname, optval, optlen);
866 return do_raw_getsockopt(sk, optname, optval, optlen);
867}
868
869static int raw_ioctl(struct sock *sk, int cmd, int *karg)
870{
871 switch (cmd) {
872 case SIOCOUTQ: {
873 *karg = sk_wmem_alloc_get(sk);
874 return 0;
875 }
876 case SIOCINQ: {
877 struct sk_buff *skb;
878
879 spin_lock_bh(&sk->sk_receive_queue.lock);
880 skb = skb_peek(&sk->sk_receive_queue);
881 if (skb)
882 *karg = skb->len;
883 else
884 *karg = 0;
885 spin_unlock_bh(&sk->sk_receive_queue.lock);
886 return 0;
887 }
888
889 default:
890#ifdef CONFIG_IP_MROUTE
891 return ipmr_ioctl(sk, cmd, karg);
892#else
893 return -ENOIOCTLCMD;
894#endif
895 }
896}
897
898#ifdef CONFIG_COMPAT
899static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
900{
901 switch (cmd) {
902 case SIOCOUTQ:
903 case SIOCINQ:
904 return -ENOIOCTLCMD;
905 default:
906#ifdef CONFIG_IP_MROUTE
907 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
908#else
909 return -ENOIOCTLCMD;
910#endif
911 }
912}
913#endif
914
915int raw_abort(struct sock *sk, int err)
916{
917 lock_sock(sk);
918
919 sk->sk_err = err;
920 sk_error_report(sk);
921 __udp_disconnect(sk, 0);
922
923 release_sock(sk);
924
925 return 0;
926}
927EXPORT_SYMBOL_GPL(raw_abort);
928
929struct proto raw_prot = {
930 .name = "RAW",
931 .owner = THIS_MODULE,
932 .close = raw_close,
933 .destroy = raw_destroy,
934 .connect = ip4_datagram_connect,
935 .disconnect = __udp_disconnect,
936 .ioctl = raw_ioctl,
937 .init = raw_sk_init,
938 .setsockopt = raw_setsockopt,
939 .getsockopt = raw_getsockopt,
940 .sendmsg = raw_sendmsg,
941 .recvmsg = raw_recvmsg,
942 .bind = raw_bind,
943 .backlog_rcv = raw_rcv_skb,
944 .release_cb = ip4_datagram_release_cb,
945 .hash = raw_hash_sk,
946 .unhash = raw_unhash_sk,
947 .obj_size = sizeof(struct raw_sock),
948 .useroffset = offsetof(struct raw_sock, filter),
949 .usersize = sizeof_field(struct raw_sock, filter),
950 .h.raw_hash = &raw_v4_hashinfo,
951#ifdef CONFIG_COMPAT
952 .compat_ioctl = compat_raw_ioctl,
953#endif
954 .diag_destroy = raw_abort,
955};
956
957#ifdef CONFIG_PROC_FS
958static struct sock *raw_get_first(struct seq_file *seq, int bucket)
959{
960 struct raw_hashinfo *h = pde_data(file_inode(seq->file));
961 struct raw_iter_state *state = raw_seq_private(seq);
962 struct hlist_head *hlist;
963 struct sock *sk;
964
965 for (state->bucket = bucket; state->bucket < RAW_HTABLE_SIZE;
966 ++state->bucket) {
967 hlist = &h->ht[state->bucket];
968 sk_for_each(sk, hlist) {
969 if (sock_net(sk) == seq_file_net(seq))
970 return sk;
971 }
972 }
973 return NULL;
974}
975
976static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
977{
978 struct raw_iter_state *state = raw_seq_private(seq);
979
980 do {
981 sk = sk_next(sk);
982 } while (sk && sock_net(sk) != seq_file_net(seq));
983
984 if (!sk)
985 return raw_get_first(seq, state->bucket + 1);
986 return sk;
987}
988
989static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
990{
991 struct sock *sk = raw_get_first(seq, 0);
992
993 if (sk)
994 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
995 --pos;
996 return pos ? NULL : sk;
997}
998
999void *raw_seq_start(struct seq_file *seq, loff_t *pos)
1000 __acquires(&h->lock)
1001{
1002 struct raw_hashinfo *h = pde_data(file_inode(seq->file));
1003
1004 spin_lock(&h->lock);
1005
1006 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1007}
1008EXPORT_SYMBOL_GPL(raw_seq_start);
1009
1010void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1011{
1012 struct sock *sk;
1013
1014 if (v == SEQ_START_TOKEN)
1015 sk = raw_get_first(seq, 0);
1016 else
1017 sk = raw_get_next(seq, v);
1018 ++*pos;
1019 return sk;
1020}
1021EXPORT_SYMBOL_GPL(raw_seq_next);
1022
1023void raw_seq_stop(struct seq_file *seq, void *v)
1024 __releases(&h->lock)
1025{
1026 struct raw_hashinfo *h = pde_data(file_inode(seq->file));
1027
1028 spin_unlock(&h->lock);
1029}
1030EXPORT_SYMBOL_GPL(raw_seq_stop);
1031
1032static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1033{
1034 struct inet_sock *inet = inet_sk(sp);
1035 __be32 dest = inet->inet_daddr,
1036 src = inet->inet_rcv_saddr;
1037 __u16 destp = 0,
1038 srcp = inet->inet_num;
1039
1040 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1041 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u\n",
1042 i, src, srcp, dest, destp, sp->sk_state,
1043 sk_wmem_alloc_get(sp),
1044 sk_rmem_alloc_get(sp),
1045 0, 0L, 0,
1046 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1047 0, sock_i_ino(sp),
1048 refcount_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1049}
1050
1051static int raw_seq_show(struct seq_file *seq, void *v)
1052{
1053 if (v == SEQ_START_TOKEN)
1054 seq_printf(seq, " sl local_address rem_address st tx_queue "
1055 "rx_queue tr tm->when retrnsmt uid timeout "
1056 "inode ref pointer drops\n");
1057 else
1058 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1059 return 0;
1060}
1061
1062static const struct seq_operations raw_seq_ops = {
1063 .start = raw_seq_start,
1064 .next = raw_seq_next,
1065 .stop = raw_seq_stop,
1066 .show = raw_seq_show,
1067};
1068
1069static __net_init int raw_init_net(struct net *net)
1070{
1071 if (!proc_create_net_data("raw", 0444, net->proc_net, &raw_seq_ops,
1072 sizeof(struct raw_iter_state), &raw_v4_hashinfo))
1073 return -ENOMEM;
1074
1075 return 0;
1076}
1077
1078static __net_exit void raw_exit_net(struct net *net)
1079{
1080 remove_proc_entry("raw", net->proc_net);
1081}
1082
1083static __net_initdata struct pernet_operations raw_net_ops = {
1084 .init = raw_init_net,
1085 .exit = raw_exit_net,
1086};
1087
1088int __init raw_proc_init(void)
1089{
1090
1091 return register_pernet_subsys(&raw_net_ops);
1092}
1093
1094void __init raw_proc_exit(void)
1095{
1096 unregister_pernet_subsys(&raw_net_ops);
1097}
1098#endif /* CONFIG_PROC_FS */
1099
1100static void raw_sysctl_init_net(struct net *net)
1101{
1102#ifdef CONFIG_NET_L3_MASTER_DEV
1103 net->ipv4.sysctl_raw_l3mdev_accept = 1;
1104#endif
1105}
1106
1107static int __net_init raw_sysctl_init(struct net *net)
1108{
1109 raw_sysctl_init_net(net);
1110 return 0;
1111}
1112
1113static struct pernet_operations __net_initdata raw_sysctl_ops = {
1114 .init = raw_sysctl_init,
1115};
1116
1117void __init raw_init(void)
1118{
1119 raw_sysctl_init_net(&init_net);
1120 if (register_pernet_subsys(&raw_sysctl_ops))
1121 panic("RAW: failed to init sysctl parameters.\n");
1122}