]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - net/ipv4/raw.c
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:


6d0fa8fb8af0afd652d99de84faa9ab1eb4e99f4
[mirror_ubuntu-jammy-kernel.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/aio.h>
50 #include <linux/kernel.h>
51 #include <linux/export.h>
52 #include <linux/spinlock.h>
53 #include <linux/sockios.h>
54 #include <linux/socket.h>
55 #include <linux/in.h>
56 #include <linux/mroute.h>
57 #include <linux/netdevice.h>
58 #include <linux/in_route.h>
59 #include <linux/route.h>
60 #include <linux/skbuff.h>
61 #include <linux/igmp.h>
62 #include <net/net_namespace.h>
63 #include <net/dst.h>
64 #include <net/sock.h>
65 #include <linux/ip.h>
66 #include <linux/net.h>
67 #include <net/ip.h>
68 #include <net/icmp.h>
69 #include <net/udp.h>
70 #include <net/raw.h>
71 #include <net/snmp.h>
72 #include <net/tcp_states.h>
73 #include <net/inet_common.h>
74 #include <net/checksum.h>
75 #include <net/xfrm.h>
76 #include <linux/rtnetlink.h>
77 #include <linux/proc_fs.h>
78 #include <linux/seq_file.h>
79 #include <linux/netfilter.h>
80 #include <linux/netfilter_ipv4.h>
81 #include <linux/compat.h>
82 #include <linux/uio.h>
83
84 struct raw_frag_vec {
85 struct msghdr *msg;
86 union {
87 struct icmphdr icmph;
88 char c[1];
89 } hdr;
90 int hlen;
91 };
92
93 static struct raw_hashinfo raw_v4_hashinfo = {
94 .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
95 };
96
97 void raw_hash_sk(struct sock *sk)
98 {
99 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
100 struct hlist_head *head;
101
102 head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)];
103
104 write_lock_bh(&h->lock);
105 sk_add_node(sk, head);
106 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
107 write_unlock_bh(&h->lock);
108 }
109 EXPORT_SYMBOL_GPL(raw_hash_sk);
110
111 void raw_unhash_sk(struct sock *sk)
112 {
113 struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
114
115 write_lock_bh(&h->lock);
116 if (sk_del_node_init(sk))
117 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
118 write_unlock_bh(&h->lock);
119 }
120 EXPORT_SYMBOL_GPL(raw_unhash_sk);
121
122 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
123 unsigned short num, __be32 raddr, __be32 laddr, int dif)
124 {
125 sk_for_each_from(sk) {
126 struct inet_sock *inet = inet_sk(sk);
127
128 if (net_eq(sock_net(sk), net) && inet->inet_num == num &&
129 !(inet->inet_daddr && inet->inet_daddr != raddr) &&
130 !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
131 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
132 goto found; /* gotcha */
133 }
134 sk = NULL;
135 found:
136 return sk;
137 }
138
139 /*
140 * 0 - deliver
141 * 1 - block
142 */
143 static 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 */
169 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
170 {
171 struct sock *sk;
172 struct hlist_head *head;
173 int delivered = 0;
174 struct net *net;
175
176 read_lock(&raw_v4_hashinfo.lock);
177 head = &raw_v4_hashinfo.ht[hash];
178 if (hlist_empty(head))
179 goto out;
180
181 net = dev_net(skb->dev);
182 sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
183 iph->saddr, iph->daddr,
184 skb->dev->ifindex);
185
186 while (sk) {
187 delivered = 1;
188 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
189 ip_mc_sf_allow(sk, iph->daddr, iph->saddr,
190 skb->dev->ifindex)) {
191 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
192
193 /* Not releasing hash table! */
194 if (clone)
195 raw_rcv(sk, clone);
196 }
197 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
198 iph->saddr, iph->daddr,
199 skb->dev->ifindex);
200 }
201 out:
202 read_unlock(&raw_v4_hashinfo.lock);
203 return delivered;
204 }
205
206 int raw_local_deliver(struct sk_buff *skb, int protocol)
207 {
208 int hash;
209 struct sock *raw_sk;
210
211 hash = protocol & (RAW_HTABLE_SIZE - 1);
212 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
213
214 /* If there maybe a raw socket we must check - if not we
215 * don't care less
216 */
217 if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
218 raw_sk = NULL;
219
220 return raw_sk != NULL;
221
222 }
223
224 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
225 {
226 struct inet_sock *inet = inet_sk(sk);
227 const int type = icmp_hdr(skb)->type;
228 const int code = icmp_hdr(skb)->code;
229 int err = 0;
230 int harderr = 0;
231
232 if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
233 ipv4_sk_update_pmtu(skb, sk, info);
234 else if (type == ICMP_REDIRECT) {
235 ipv4_sk_redirect(skb, sk);
236 return;
237 }
238
239 /* Report error on raw socket, if:
240 1. User requested ip_recverr.
241 2. Socket is connected (otherwise the error indication
242 is useless without ip_recverr and error is hard.
243 */
244 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
245 return;
246
247 switch (type) {
248 default:
249 case ICMP_TIME_EXCEEDED:
250 err = EHOSTUNREACH;
251 break;
252 case ICMP_SOURCE_QUENCH:
253 return;
254 case ICMP_PARAMETERPROB:
255 err = EPROTO;
256 harderr = 1;
257 break;
258 case ICMP_DEST_UNREACH:
259 err = EHOSTUNREACH;
260 if (code > NR_ICMP_UNREACH)
261 break;
262 err = icmp_err_convert[code].errno;
263 harderr = icmp_err_convert[code].fatal;
264 if (code == ICMP_FRAG_NEEDED) {
265 harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
266 err = EMSGSIZE;
267 }
268 }
269
270 if (inet->recverr) {
271 const struct iphdr *iph = (const struct iphdr *)skb->data;
272 u8 *payload = skb->data + (iph->ihl << 2);
273
274 if (inet->hdrincl)
275 payload = skb->data;
276 ip_icmp_error(sk, skb, err, 0, info, payload);
277 }
278
279 if (inet->recverr || harderr) {
280 sk->sk_err = err;
281 sk->sk_error_report(sk);
282 }
283 }
284
285 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
286 {
287 int hash;
288 struct sock *raw_sk;
289 const struct iphdr *iph;
290 struct net *net;
291
292 hash = protocol & (RAW_HTABLE_SIZE - 1);
293
294 read_lock(&raw_v4_hashinfo.lock);
295 raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
296 if (raw_sk) {
297 iph = (const struct iphdr *)skb->data;
298 net = dev_net(skb->dev);
299
300 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
301 iph->daddr, iph->saddr,
302 skb->dev->ifindex)) != NULL) {
303 raw_err(raw_sk, skb, info);
304 raw_sk = sk_next(raw_sk);
305 iph = (const struct iphdr *)skb->data;
306 }
307 }
308 read_unlock(&raw_v4_hashinfo.lock);
309 }
310
311 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
312 {
313 /* Charge it to the socket. */
314
315 ipv4_pktinfo_prepare(sk, skb);
316 if (sock_queue_rcv_skb(sk, skb) < 0) {
317 kfree_skb(skb);
318 return NET_RX_DROP;
319 }
320
321 return NET_RX_SUCCESS;
322 }
323
324 int raw_rcv(struct sock *sk, struct sk_buff *skb)
325 {
326 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
327 atomic_inc(&sk->sk_drops);
328 kfree_skb(skb);
329 return NET_RX_DROP;
330 }
331 nf_reset(skb);
332
333 skb_push(skb, skb->data - skb_network_header(skb));
334
335 raw_rcv_skb(sk, skb);
336 return 0;
337 }
338
339 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
340 struct msghdr *msg, size_t length,
341 struct rtable **rtp,
342 unsigned int flags)
343 {
344 struct inet_sock *inet = inet_sk(sk);
345 struct net *net = sock_net(sk);
346 struct iphdr *iph;
347 struct sk_buff *skb;
348 unsigned int iphlen;
349 int err;
350 struct rtable *rt = *rtp;
351 int hlen, tlen;
352
353 if (length > rt->dst.dev->mtu) {
354 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
355 rt->dst.dev->mtu);
356 return -EMSGSIZE;
357 }
358 if (flags&MSG_PROBE)
359 goto out;
360
361 hlen = LL_RESERVED_SPACE(rt->dst.dev);
362 tlen = rt->dst.dev->needed_tailroom;
363 skb = sock_alloc_send_skb(sk,
364 length + hlen + tlen + 15,
365 flags & MSG_DONTWAIT, &err);
366 if (!skb)
367 goto error;
368 skb_reserve(skb, hlen);
369
370 skb->priority = sk->sk_priority;
371 skb->mark = sk->sk_mark;
372 skb_dst_set(skb, &rt->dst);
373 *rtp = NULL;
374
375 skb_reset_network_header(skb);
376 iph = ip_hdr(skb);
377 skb_put(skb, length);
378
379 skb->ip_summed = CHECKSUM_NONE;
380
381 sock_tx_timestamp(sk, &skb_shinfo(skb)->tx_flags);
382
383 skb->transport_header = skb->network_header;
384 err = -EFAULT;
385 if (memcpy_from_msg(iph, msg, length))
386 goto error_free;
387
388 iphlen = iph->ihl * 4;
389
390 /*
391 * We don't want to modify the ip header, but we do need to
392 * be sure that it won't cause problems later along the network
393 * stack. Specifically we want to make sure that iph->ihl is a
394 * sane value. If ihl points beyond the length of the buffer passed
395 * in, reject the frame as invalid
396 */
397 err = -EINVAL;
398 if (iphlen > length)
399 goto error_free;
400
401 if (iphlen >= sizeof(*iph)) {
402 if (!iph->saddr)
403 iph->saddr = fl4->saddr;
404 iph->check = 0;
405 iph->tot_len = htons(length);
406 if (!iph->id)
407 ip_select_ident(net, skb, NULL);
408
409 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
410 }
411 if (iph->protocol == IPPROTO_ICMP)
412 icmp_out_count(net, ((struct icmphdr *)
413 skb_transport_header(skb))->type);
414
415 err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT, skb, NULL,
416 rt->dst.dev, dst_output);
417 if (err > 0)
418 err = net_xmit_errno(err);
419 if (err)
420 goto error;
421 out:
422 return 0;
423
424 error_free:
425 kfree_skb(skb);
426 error:
427 IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
428 if (err == -ENOBUFS && !inet->recverr)
429 err = 0;
430 return err;
431 }
432
433 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
434 {
435 int err;
436
437 if (fl4->flowi4_proto != IPPROTO_ICMP)
438 return 0;
439
440 /* We only need the first two bytes. */
441 rfv->hlen = 2;
442
443 err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
444 if (err)
445 return err;
446
447 fl4->fl4_icmp_type = rfv->hdr.icmph.type;
448 fl4->fl4_icmp_code = rfv->hdr.icmph.code;
449
450 return 0;
451 }
452
453 static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
454 struct sk_buff *skb)
455 {
456 struct raw_frag_vec *rfv = from;
457
458 if (offset < rfv->hlen) {
459 int copy = min(rfv->hlen - offset, len);
460
461 if (skb->ip_summed == CHECKSUM_PARTIAL)
462 memcpy(to, rfv->hdr.c + offset, copy);
463 else
464 skb->csum = csum_block_add(
465 skb->csum,
466 csum_partial_copy_nocheck(rfv->hdr.c + offset,
467 to, copy, 0),
468 odd);
469
470 odd = 0;
471 offset += copy;
472 to += copy;
473 len -= copy;
474
475 if (!len)
476 return 0;
477 }
478
479 offset -= rfv->hlen;
480
481 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
482 }
483
484 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
485 {
486 struct inet_sock *inet = inet_sk(sk);
487 struct ipcm_cookie ipc;
488 struct rtable *rt = NULL;
489 struct flowi4 fl4;
490 int free = 0;
491 __be32 daddr;
492 __be32 saddr;
493 u8 tos;
494 int err;
495 struct ip_options_data opt_copy;
496 struct raw_frag_vec rfv;
497
498 err = -EMSGSIZE;
499 if (len > 0xFFFF)
500 goto out;
501
502 /*
503 * Check the flags.
504 */
505
506 err = -EOPNOTSUPP;
507 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */
508 goto out; /* compatibility */
509
510 /*
511 * Get and verify the address.
512 */
513
514 if (msg->msg_namelen) {
515 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
516 err = -EINVAL;
517 if (msg->msg_namelen < sizeof(*usin))
518 goto out;
519 if (usin->sin_family != AF_INET) {
520 pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
521 __func__, current->comm);
522 err = -EAFNOSUPPORT;
523 if (usin->sin_family)
524 goto out;
525 }
526 daddr = usin->sin_addr.s_addr;
527 /* ANK: I did not forget to get protocol from port field.
528 * I just do not know, who uses this weirdness.
529 * IP_HDRINCL is much more convenient.
530 */
531 } else {
532 err = -EDESTADDRREQ;
533 if (sk->sk_state != TCP_ESTABLISHED)
534 goto out;
535 daddr = inet->inet_daddr;
536 }
537
538 ipc.addr = inet->inet_saddr;
539 ipc.opt = NULL;
540 ipc.tx_flags = 0;
541 ipc.ttl = 0;
542 ipc.tos = -1;
543 ipc.oif = sk->sk_bound_dev_if;
544
545 if (msg->msg_controllen) {
546 err = ip_cmsg_send(sock_net(sk), msg, &ipc, false);
547 if (err)
548 goto out;
549 if (ipc.opt)
550 free = 1;
551 }
552
553 saddr = ipc.addr;
554 ipc.addr = daddr;
555
556 if (!ipc.opt) {
557 struct ip_options_rcu *inet_opt;
558
559 rcu_read_lock();
560 inet_opt = rcu_dereference(inet->inet_opt);
561 if (inet_opt) {
562 memcpy(&opt_copy, inet_opt,
563 sizeof(*inet_opt) + inet_opt->opt.optlen);
564 ipc.opt = &opt_copy.opt;
565 }
566 rcu_read_unlock();
567 }
568
569 if (ipc.opt) {
570 err = -EINVAL;
571 /* Linux does not mangle headers on raw sockets,
572 * so that IP options + IP_HDRINCL is non-sense.
573 */
574 if (inet->hdrincl)
575 goto done;
576 if (ipc.opt->opt.srr) {
577 if (!daddr)
578 goto done;
579 daddr = ipc.opt->opt.faddr;
580 }
581 }
582 tos = get_rtconn_flags(&ipc, sk);
583 if (msg->msg_flags & MSG_DONTROUTE)
584 tos |= RTO_ONLINK;
585
586 if (ipv4_is_multicast(daddr)) {
587 if (!ipc.oif)
588 ipc.oif = inet->mc_index;
589 if (!saddr)
590 saddr = inet->mc_addr;
591 } else if (!ipc.oif)
592 ipc.oif = inet->uc_index;
593
594 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
595 RT_SCOPE_UNIVERSE,
596 inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
597 inet_sk_flowi_flags(sk) |
598 (inet->hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
599 daddr, saddr, 0, 0);
600
601 if (!inet->hdrincl) {
602 rfv.msg = msg;
603 rfv.hlen = 0;
604
605 err = raw_probe_proto_opt(&rfv, &fl4);
606 if (err)
607 goto done;
608 }
609
610 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
611 rt = ip_route_output_flow(sock_net(sk), &fl4, sk);
612 if (IS_ERR(rt)) {
613 err = PTR_ERR(rt);
614 rt = NULL;
615 goto done;
616 }
617
618 err = -EACCES;
619 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
620 goto done;
621
622 if (msg->msg_flags & MSG_CONFIRM)
623 goto do_confirm;
624 back_from_confirm:
625
626 if (inet->hdrincl)
627 err = raw_send_hdrinc(sk, &fl4, msg, len,
628 &rt, msg->msg_flags);
629
630 else {
631 sock_tx_timestamp(sk, &ipc.tx_flags);
632
633 if (!ipc.addr)
634 ipc.addr = fl4.daddr;
635 lock_sock(sk);
636 err = ip_append_data(sk, &fl4, raw_getfrag,
637 &rfv, len, 0,
638 &ipc, &rt, msg->msg_flags);
639 if (err)
640 ip_flush_pending_frames(sk);
641 else if (!(msg->msg_flags & MSG_MORE)) {
642 err = ip_push_pending_frames(sk, &fl4);
643 if (err == -ENOBUFS && !inet->recverr)
644 err = 0;
645 }
646 release_sock(sk);
647 }
648 done:
649 if (free)
650 kfree(ipc.opt);
651 ip_rt_put(rt);
652
653 out:
654 if (err < 0)
655 return err;
656 return len;
657
658 do_confirm:
659 dst_confirm(&rt->dst);
660 if (!(msg->msg_flags & MSG_PROBE) || len)
661 goto back_from_confirm;
662 err = 0;
663 goto done;
664 }
665
666 static void raw_close(struct sock *sk, long timeout)
667 {
668 /*
669 * Raw sockets may have direct kernel references. Kill them.
670 */
671 ip_ra_control(sk, 0, NULL);
672
673 sk_common_release(sk);
674 }
675
676 static void raw_destroy(struct sock *sk)
677 {
678 lock_sock(sk);
679 ip_flush_pending_frames(sk);
680 release_sock(sk);
681 }
682
683 /* This gets rid of all the nasties in af_inet. -DaveM */
684 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
685 {
686 struct inet_sock *inet = inet_sk(sk);
687 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
688 int ret = -EINVAL;
689 int chk_addr_ret;
690
691 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
692 goto out;
693 chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
694 ret = -EADDRNOTAVAIL;
695 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
696 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
697 goto out;
698 inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
699 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
700 inet->inet_saddr = 0; /* Use device */
701 sk_dst_reset(sk);
702 ret = 0;
703 out: return ret;
704 }
705
706 /*
707 * This should be easy, if there is something there
708 * we return it, otherwise we block.
709 */
710
711 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
712 int noblock, int flags, int *addr_len)
713 {
714 struct inet_sock *inet = inet_sk(sk);
715 size_t copied = 0;
716 int err = -EOPNOTSUPP;
717 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
718 struct sk_buff *skb;
719
720 if (flags & MSG_OOB)
721 goto out;
722
723 if (flags & MSG_ERRQUEUE) {
724 err = ip_recv_error(sk, msg, len, addr_len);
725 goto out;
726 }
727
728 skb = skb_recv_datagram(sk, flags, noblock, &err);
729 if (!skb)
730 goto out;
731
732 copied = skb->len;
733 if (len < copied) {
734 msg->msg_flags |= MSG_TRUNC;
735 copied = len;
736 }
737
738 err = skb_copy_datagram_msg(skb, 0, msg, copied);
739 if (err)
740 goto done;
741
742 sock_recv_ts_and_drops(msg, sk, skb);
743
744 /* Copy the address. */
745 if (sin) {
746 sin->sin_family = AF_INET;
747 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
748 sin->sin_port = 0;
749 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
750 *addr_len = sizeof(*sin);
751 }
752 if (inet->cmsg_flags)
753 ip_cmsg_recv(msg, skb);
754 if (flags & MSG_TRUNC)
755 copied = skb->len;
756 done:
757 skb_free_datagram(sk, skb);
758 out:
759 if (err)
760 return err;
761 return copied;
762 }
763
764 static int raw_init(struct sock *sk)
765 {
766 struct raw_sock *rp = raw_sk(sk);
767
768 if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
769 memset(&rp->filter, 0, sizeof(rp->filter));
770 return 0;
771 }
772
773 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
774 {
775 if (optlen > sizeof(struct icmp_filter))
776 optlen = sizeof(struct icmp_filter);
777 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
778 return -EFAULT;
779 return 0;
780 }
781
782 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
783 {
784 int len, ret = -EFAULT;
785
786 if (get_user(len, optlen))
787 goto out;
788 ret = -EINVAL;
789 if (len < 0)
790 goto out;
791 if (len > sizeof(struct icmp_filter))
792 len = sizeof(struct icmp_filter);
793 ret = -EFAULT;
794 if (put_user(len, optlen) ||
795 copy_to_user(optval, &raw_sk(sk)->filter, len))
796 goto out;
797 ret = 0;
798 out: return ret;
799 }
800
801 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
802 char __user *optval, unsigned int optlen)
803 {
804 if (optname == ICMP_FILTER) {
805 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
806 return -EOPNOTSUPP;
807 else
808 return raw_seticmpfilter(sk, optval, optlen);
809 }
810 return -ENOPROTOOPT;
811 }
812
813 static int raw_setsockopt(struct sock *sk, int level, int optname,
814 char __user *optval, unsigned int optlen)
815 {
816 if (level != SOL_RAW)
817 return ip_setsockopt(sk, level, optname, optval, optlen);
818 return do_raw_setsockopt(sk, level, optname, optval, optlen);
819 }
820
821 #ifdef CONFIG_COMPAT
822 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
823 char __user *optval, unsigned int optlen)
824 {
825 if (level != SOL_RAW)
826 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
827 return do_raw_setsockopt(sk, level, optname, optval, optlen);
828 }
829 #endif
830
831 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
832 char __user *optval, int __user *optlen)
833 {
834 if (optname == ICMP_FILTER) {
835 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
836 return -EOPNOTSUPP;
837 else
838 return raw_geticmpfilter(sk, optval, optlen);
839 }
840 return -ENOPROTOOPT;
841 }
842
843 static int raw_getsockopt(struct sock *sk, int level, int optname,
844 char __user *optval, int __user *optlen)
845 {
846 if (level != SOL_RAW)
847 return ip_getsockopt(sk, level, optname, optval, optlen);
848 return do_raw_getsockopt(sk, level, optname, optval, optlen);
849 }
850
851 #ifdef CONFIG_COMPAT
852 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
853 char __user *optval, int __user *optlen)
854 {
855 if (level != SOL_RAW)
856 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
857 return do_raw_getsockopt(sk, level, optname, optval, optlen);
858 }
859 #endif
860
861 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
862 {
863 switch (cmd) {
864 case SIOCOUTQ: {
865 int amount = sk_wmem_alloc_get(sk);
866
867 return put_user(amount, (int __user *)arg);
868 }
869 case SIOCINQ: {
870 struct sk_buff *skb;
871 int amount = 0;
872
873 spin_lock_bh(&sk->sk_receive_queue.lock);
874 skb = skb_peek(&sk->sk_receive_queue);
875 if (skb)
876 amount = skb->len;
877 spin_unlock_bh(&sk->sk_receive_queue.lock);
878 return put_user(amount, (int __user *)arg);
879 }
880
881 default:
882 #ifdef CONFIG_IP_MROUTE
883 return ipmr_ioctl(sk, cmd, (void __user *)arg);
884 #else
885 return -ENOIOCTLCMD;
886 #endif
887 }
888 }
889
890 #ifdef CONFIG_COMPAT
891 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
892 {
893 switch (cmd) {
894 case SIOCOUTQ:
895 case SIOCINQ:
896 return -ENOIOCTLCMD;
897 default:
898 #ifdef CONFIG_IP_MROUTE
899 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
900 #else
901 return -ENOIOCTLCMD;
902 #endif
903 }
904 }
905 #endif
906
907 struct proto raw_prot = {
908 .name = "RAW",
909 .owner = THIS_MODULE,
910 .close = raw_close,
911 .destroy = raw_destroy,
912 .connect = ip4_datagram_connect,
913 .disconnect = udp_disconnect,
914 .ioctl = raw_ioctl,
915 .init = raw_init,
916 .setsockopt = raw_setsockopt,
917 .getsockopt = raw_getsockopt,
918 .sendmsg = raw_sendmsg,
919 .recvmsg = raw_recvmsg,
920 .bind = raw_bind,
921 .backlog_rcv = raw_rcv_skb,
922 .release_cb = ip4_datagram_release_cb,
923 .hash = raw_hash_sk,
924 .unhash = raw_unhash_sk,
925 .obj_size = sizeof(struct raw_sock),
926 .h.raw_hash = &raw_v4_hashinfo,
927 #ifdef CONFIG_COMPAT
928 .compat_setsockopt = compat_raw_setsockopt,
929 .compat_getsockopt = compat_raw_getsockopt,
930 .compat_ioctl = compat_raw_ioctl,
931 #endif
932 };
933
934 #ifdef CONFIG_PROC_FS
935 static struct sock *raw_get_first(struct seq_file *seq)
936 {
937 struct sock *sk;
938 struct raw_iter_state *state = raw_seq_private(seq);
939
940 for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
941 ++state->bucket) {
942 sk_for_each(sk, &state->h->ht[state->bucket])
943 if (sock_net(sk) == seq_file_net(seq))
944 goto found;
945 }
946 sk = NULL;
947 found:
948 return sk;
949 }
950
951 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
952 {
953 struct raw_iter_state *state = raw_seq_private(seq);
954
955 do {
956 sk = sk_next(sk);
957 try_again:
958 ;
959 } while (sk && sock_net(sk) != seq_file_net(seq));
960
961 if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
962 sk = sk_head(&state->h->ht[state->bucket]);
963 goto try_again;
964 }
965 return sk;
966 }
967
968 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
969 {
970 struct sock *sk = raw_get_first(seq);
971
972 if (sk)
973 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
974 --pos;
975 return pos ? NULL : sk;
976 }
977
978 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
979 {
980 struct raw_iter_state *state = raw_seq_private(seq);
981
982 read_lock(&state->h->lock);
983 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
984 }
985 EXPORT_SYMBOL_GPL(raw_seq_start);
986
987 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
988 {
989 struct sock *sk;
990
991 if (v == SEQ_START_TOKEN)
992 sk = raw_get_first(seq);
993 else
994 sk = raw_get_next(seq, v);
995 ++*pos;
996 return sk;
997 }
998 EXPORT_SYMBOL_GPL(raw_seq_next);
999
1000 void raw_seq_stop(struct seq_file *seq, void *v)
1001 {
1002 struct raw_iter_state *state = raw_seq_private(seq);
1003
1004 read_unlock(&state->h->lock);
1005 }
1006 EXPORT_SYMBOL_GPL(raw_seq_stop);
1007
1008 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1009 {
1010 struct inet_sock *inet = inet_sk(sp);
1011 __be32 dest = inet->inet_daddr,
1012 src = inet->inet_rcv_saddr;
1013 __u16 destp = 0,
1014 srcp = inet->inet_num;
1015
1016 seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1017 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1018 i, src, srcp, dest, destp, sp->sk_state,
1019 sk_wmem_alloc_get(sp),
1020 sk_rmem_alloc_get(sp),
1021 0, 0L, 0,
1022 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1023 0, sock_i_ino(sp),
1024 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1025 }
1026
1027 static int raw_seq_show(struct seq_file *seq, void *v)
1028 {
1029 if (v == SEQ_START_TOKEN)
1030 seq_printf(seq, " sl local_address rem_address st tx_queue "
1031 "rx_queue tr tm->when retrnsmt uid timeout "
1032 "inode ref pointer drops\n");
1033 else
1034 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1035 return 0;
1036 }
1037
1038 static const struct seq_operations raw_seq_ops = {
1039 .start = raw_seq_start,
1040 .next = raw_seq_next,
1041 .stop = raw_seq_stop,
1042 .show = raw_seq_show,
1043 };
1044
1045 int raw_seq_open(struct inode *ino, struct file *file,
1046 struct raw_hashinfo *h, const struct seq_operations *ops)
1047 {
1048 int err;
1049 struct raw_iter_state *i;
1050
1051 err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1052 if (err < 0)
1053 return err;
1054
1055 i = raw_seq_private((struct seq_file *)file->private_data);
1056 i->h = h;
1057 return 0;
1058 }
1059 EXPORT_SYMBOL_GPL(raw_seq_open);
1060
1061 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1062 {
1063 return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1064 }
1065
1066 static const struct file_operations raw_seq_fops = {
1067 .owner = THIS_MODULE,
1068 .open = raw_v4_seq_open,
1069 .read = seq_read,
1070 .llseek = seq_lseek,
1071 .release = seq_release_net,
1072 };
1073
1074 static __net_init int raw_init_net(struct net *net)
1075 {
1076 if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops))
1077 return -ENOMEM;
1078
1079 return 0;
1080 }
1081
1082 static __net_exit void raw_exit_net(struct net *net)
1083 {
1084 remove_proc_entry("raw", net->proc_net);
1085 }
1086
1087 static __net_initdata struct pernet_operations raw_net_ops = {
1088 .init = raw_init_net,
1089 .exit = raw_exit_net,
1090 };
1091
1092 int __init raw_proc_init(void)
1093 {
1094 return register_pernet_subsys(&raw_net_ops);
1095 }
1096
1097 void __init raw_proc_exit(void)
1098 {
1099 unregister_pernet_subsys(&raw_net_ops);
1100 }
1101 #endif /* CONFIG_PROC_FS */
Ubuntu Jammy Linux kernel mirror