~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/ipv4/raw.c

Version: ~ [ linux-5.12-rc6 ] ~ [ linux-5.11.13 ] ~ [ linux-5.10.29 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.111 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.186 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.230 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.266 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.266 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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 <linux/uaccess.h>
 45 #include <asm/ioctls.h>
 46 #include <linux/stddef.h>
 47 #include <linux/slab.h>
 48 #include <linux/errno.h>
 49 #include <linux/kernel.h>
 50 #include <linux/export.h>
 51 #include <linux/spinlock.h>
 52 #include <linux/sockios.h>
 53 #include <linux/socket.h>
 54 #include <linux/in.h>
 55 #include <linux/mroute.h>
 56 #include <linux/netdevice.h>
 57 #include <linux/in_route.h>
 58 #include <linux/route.h>
 59 #include <linux/skbuff.h>
 60 #include <linux/igmp.h>
 61 #include <net/net_namespace.h>
 62 #include <net/dst.h>
 63 #include <net/sock.h>
 64 #include <linux/ip.h>
 65 #include <linux/net.h>
 66 #include <net/ip.h>
 67 #include <net/icmp.h>
 68 #include <net/udp.h>
 69 #include <net/raw.h>
 70 #include <net/snmp.h>
 71 #include <net/tcp_states.h>
 72 #include <net/inet_common.h>
 73 #include <net/checksum.h>
 74 #include <net/xfrm.h>
 75 #include <linux/rtnetlink.h>
 76 #include <linux/proc_fs.h>
 77 #include <linux/seq_file.h>
 78 #include <linux/netfilter.h>
 79 #include <linux/netfilter_ipv4.h>
 80 #include <linux/compat.h>
 81 #include <linux/uio.h>
 82 
 83 struct raw_frag_vec {
 84         struct msghdr *msg;
 85         union {
 86                 struct icmphdr icmph;
 87                 char c[1];
 88         } hdr;
 89         int hlen;
 90 };
 91 
 92 struct raw_hashinfo raw_v4_hashinfo = {
 93         .lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
 94 };
 95 EXPORT_SYMBOL_GPL(raw_v4_hashinfo);
 96 
 97 int 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         return 0;
110 }
111 EXPORT_SYMBOL_GPL(raw_hash_sk);
112 
113 void raw_unhash_sk(struct sock *sk)
114 {
115         struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
116 
117         write_lock_bh(&h->lock);
118         if (sk_del_node_init(sk))
119                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
120         write_unlock_bh(&h->lock);
121 }
122 EXPORT_SYMBOL_GPL(raw_unhash_sk);
123 
124 struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
125                 unsigned short num, __be32 raddr, __be32 laddr, int dif)
126 {
127         sk_for_each_from(sk) {
128                 struct inet_sock *inet = inet_sk(sk);
129 
130                 if (net_eq(sock_net(sk), net) && inet->inet_num == num  &&
131                     !(inet->inet_daddr && inet->inet_daddr != raddr)    &&
132                     !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
133                     !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
134                         goto found; /* gotcha */
135         }
136         sk = NULL;
137 found:
138         return sk;
139 }
140 EXPORT_SYMBOL_GPL(__raw_v4_lookup);
141 
142 /*
143  *      0 - deliver
144  *      1 - block
145  */
146 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
147 {
148         struct icmphdr _hdr;
149         const struct icmphdr *hdr;
150 
151         hdr = skb_header_pointer(skb, skb_transport_offset(skb),
152                                  sizeof(_hdr), &_hdr);
153         if (!hdr)
154                 return 1;
155 
156         if (hdr->type < 32) {
157                 __u32 data = raw_sk(sk)->filter.data;
158 
159                 return ((1U << hdr->type) & data) != 0;
160         }
161 
162         /* Do not block unknown ICMP types */
163         return 0;
164 }
165 
166 /* IP input processing comes here for RAW socket delivery.
167  * Caller owns SKB, so we must make clones.
168  *
169  * RFC 1122: SHOULD pass TOS value up to the transport layer.
170  * -> It does. And not only TOS, but all IP header.
171  */
172 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
173 {
174         struct sock *sk;
175         struct hlist_head *head;
176         int delivered = 0;
177         struct net *net;
178 
179         read_lock(&raw_v4_hashinfo.lock);
180         head = &raw_v4_hashinfo.ht[hash];
181         if (hlist_empty(head))
182                 goto out;
183 
184         net = dev_net(skb->dev);
185         sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
186                              iph->saddr, iph->daddr,
187                              skb->dev->ifindex);
188 
189         while (sk) {
190                 delivered = 1;
191                 if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
192                     ip_mc_sf_allow(sk, iph->daddr, iph->saddr,
193                                    skb->dev->ifindex)) {
194                         struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
195 
196                         /* Not releasing hash table! */
197                         if (clone)
198                                 raw_rcv(sk, clone);
199                 }
200                 sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
201                                      iph->saddr, iph->daddr,
202                                      skb->dev->ifindex);
203         }
204 out:
205         read_unlock(&raw_v4_hashinfo.lock);
206         return delivered;
207 }
208 
209 int raw_local_deliver(struct sk_buff *skb, int protocol)
210 {
211         int hash;
212         struct sock *raw_sk;
213 
214         hash = protocol & (RAW_HTABLE_SIZE - 1);
215         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
216 
217         /* If there maybe a raw socket we must check - if not we
218          * don't care less
219          */
220         if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
221                 raw_sk = NULL;
222 
223         return raw_sk != NULL;
224 
225 }
226 
227 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
228 {
229         struct inet_sock *inet = inet_sk(sk);
230         const int type = icmp_hdr(skb)->type;
231         const int code = icmp_hdr(skb)->code;
232         int err = 0;
233         int harderr = 0;
234 
235         if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
236                 ipv4_sk_update_pmtu(skb, sk, info);
237         else if (type == ICMP_REDIRECT) {
238                 ipv4_sk_redirect(skb, sk);
239                 return;
240         }
241 
242         /* Report error on raw socket, if:
243            1. User requested ip_recverr.
244            2. Socket is connected (otherwise the error indication
245               is useless without ip_recverr and error is hard.
246          */
247         if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
248                 return;
249 
250         switch (type) {
251         default:
252         case ICMP_TIME_EXCEEDED:
253                 err = EHOSTUNREACH;
254                 break;
255         case ICMP_SOURCE_QUENCH:
256                 return;
257         case ICMP_PARAMETERPROB:
258                 err = EPROTO;
259                 harderr = 1;
260                 break;
261         case ICMP_DEST_UNREACH:
262                 err = EHOSTUNREACH;
263                 if (code > NR_ICMP_UNREACH)
264                         break;
265                 err = icmp_err_convert[code].errno;
266                 harderr = icmp_err_convert[code].fatal;
267                 if (code == ICMP_FRAG_NEEDED) {
268                         harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
269                         err = EMSGSIZE;
270                 }
271         }
272 
273         if (inet->recverr) {
274                 const struct iphdr *iph = (const struct iphdr *)skb->data;
275                 u8 *payload = skb->data + (iph->ihl << 2);
276 
277                 if (inet->hdrincl)
278                         payload = skb->data;
279                 ip_icmp_error(sk, skb, err, 0, info, payload);
280         }
281 
282         if (inet->recverr || harderr) {
283                 sk->sk_err = err;
284                 sk->sk_error_report(sk);
285         }
286 }
287 
288 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
289 {
290         int hash;
291         struct sock *raw_sk;
292         const struct iphdr *iph;
293         struct net *net;
294 
295         hash = protocol & (RAW_HTABLE_SIZE - 1);
296 
297         read_lock(&raw_v4_hashinfo.lock);
298         raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
299         if (raw_sk) {
300                 iph = (const struct iphdr *)skb->data;
301                 net = dev_net(skb->dev);
302 
303                 while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
304                                                 iph->daddr, iph->saddr,
305                                                 skb->dev->ifindex)) != NULL) {
306                         raw_err(raw_sk, skb, info);
307                         raw_sk = sk_next(raw_sk);
308                         iph = (const struct iphdr *)skb->data;
309                 }
310         }
311         read_unlock(&raw_v4_hashinfo.lock);
312 }
313 
314 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
315 {
316         /* Charge it to the socket. */
317 
318         ipv4_pktinfo_prepare(sk, skb);
319         if (sock_queue_rcv_skb(sk, skb) < 0) {
320                 kfree_skb(skb);
321                 return NET_RX_DROP;
322         }
323 
324         return NET_RX_SUCCESS;
325 }
326 
327 int raw_rcv(struct sock *sk, struct sk_buff *skb)
328 {
329         if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
330                 atomic_inc(&sk->sk_drops);
331                 kfree_skb(skb);
332                 return NET_RX_DROP;
333         }
334         nf_reset(skb);
335 
336         skb_push(skb, skb->data - skb_network_header(skb));
337 
338         raw_rcv_skb(sk, skb);
339         return 0;
340 }
341 
342 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
343                            struct msghdr *msg, size_t length,
344                            struct rtable **rtp, unsigned int flags,
345                            const struct sockcm_cookie *sockc)
346 {
347         struct inet_sock *inet = inet_sk(sk);
348         struct net *net = sock_net(sk);
349         struct iphdr *iph;
350         struct sk_buff *skb;
351         unsigned int iphlen;
352         int err;
353         struct rtable *rt = *rtp;
354         int hlen, tlen;
355 
356         if (length > rt->dst.dev->mtu) {
357                 ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
358                                rt->dst.dev->mtu);
359                 return -EMSGSIZE;
360         }
361         if (length < sizeof(struct iphdr))
362                 return -EINVAL;
363 
364         if (flags&MSG_PROBE)
365                 goto out;
366 
367         hlen = LL_RESERVED_SPACE(rt->dst.dev);
368         tlen = rt->dst.dev->needed_tailroom;
369         skb = sock_alloc_send_skb(sk,
370                                   length + hlen + tlen + 15,
371                                   flags & MSG_DONTWAIT, &err);
372         if (!skb)
373                 goto error;
374         skb_reserve(skb, hlen);
375 
376         skb->priority = sk->sk_priority;
377         skb->mark = sk->sk_mark;
378         skb_dst_set(skb, &rt->dst);
379         *rtp = NULL;
380 
381         skb_reset_network_header(skb);
382         iph = ip_hdr(skb);
383         skb_put(skb, length);
384 
385         skb->ip_summed = CHECKSUM_NONE;
386 
387         sock_tx_timestamp(sk, sockc->tsflags, &skb_shinfo(skb)->tx_flags);
388 
389         if (flags & MSG_CONFIRM)
390                 skb_set_dst_pending_confirm(skb, 1);
391 
392         skb->transport_header = skb->network_header;
393         err = -EFAULT;
394         if (memcpy_from_msg(iph, msg, length))
395                 goto error_free;
396 
397         iphlen = iph->ihl * 4;
398 
399         /*
400          * We don't want to modify the ip header, but we do need to
401          * be sure that it won't cause problems later along the network
402          * stack.  Specifically we want to make sure that iph->ihl is a
403          * sane value.  If ihl points beyond the length of the buffer passed
404          * in, reject the frame as invalid
405          */
406         err = -EINVAL;
407         if (iphlen > length)
408                 goto error_free;
409 
410         if (iphlen >= sizeof(*iph)) {
411                 if (!iph->saddr)
412                         iph->saddr = fl4->saddr;
413                 iph->check   = 0;
414                 iph->tot_len = htons(length);
415                 if (!iph->id)
416                         ip_select_ident(net, skb, NULL);
417 
418                 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
419                 skb->transport_header += iphlen;
420                 if (iph->protocol == IPPROTO_ICMP &&
421                     length >= iphlen + sizeof(struct icmphdr))
422                         icmp_out_count(net, ((struct icmphdr *)
423                                 skb_transport_header(skb))->type);
424         }
425 
426         err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
427                       net, sk, skb, NULL, rt->dst.dev,
428                       dst_output);
429         if (err > 0)
430                 err = net_xmit_errno(err);
431         if (err)
432                 goto error;
433 out:
434         return 0;
435 
436 error_free:
437         kfree_skb(skb);
438 error:
439         IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
440         if (err == -ENOBUFS && !inet->recverr)
441                 err = 0;
442         return err;
443 }
444 
445 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
446 {
447         int err;
448 
449         if (fl4->flowi4_proto != IPPROTO_ICMP)
450                 return 0;
451 
452         /* We only need the first two bytes. */
453         rfv->hlen = 2;
454 
455         err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
456         if (err)
457                 return err;
458 
459         fl4->fl4_icmp_type = rfv->hdr.icmph.type;
460         fl4->fl4_icmp_code = rfv->hdr.icmph.code;
461 
462         return 0;
463 }
464 
465 static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
466                        struct sk_buff *skb)
467 {
468         struct raw_frag_vec *rfv = from;
469 
470         if (offset < rfv->hlen) {
471                 int copy = min(rfv->hlen - offset, len);
472 
473                 if (skb->ip_summed == CHECKSUM_PARTIAL)
474                         memcpy(to, rfv->hdr.c + offset, copy);
475                 else
476                         skb->csum = csum_block_add(
477                                 skb->csum,
478                                 csum_partial_copy_nocheck(rfv->hdr.c + offset,
479                                                           to, copy, 0),
480                                 odd);
481 
482                 odd = 0;
483                 offset += copy;
484                 to += copy;
485                 len -= copy;
486 
487                 if (!len)
488                         return 0;
489         }
490 
491         offset -= rfv->hlen;
492 
493         return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
494 }
495 
496 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
497 {
498         struct inet_sock *inet = inet_sk(sk);
499         struct net *net = sock_net(sk);
500         struct ipcm_cookie ipc;
501         struct rtable *rt = NULL;
502         struct flowi4 fl4;
503         int free = 0;
504         __be32 daddr;
505         __be32 saddr;
506         u8  tos;
507         int err;
508         struct ip_options_data opt_copy;
509         struct raw_frag_vec rfv;
510 
511         err = -EMSGSIZE;
512         if (len > 0xFFFF)
513                 goto out;
514 
515         /*
516          *      Check the flags.
517          */
518 
519         err = -EOPNOTSUPP;
520         if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
521                 goto out;               /* compatibility */
522 
523         /*
524          *      Get and verify the address.
525          */
526 
527         if (msg->msg_namelen) {
528                 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
529                 err = -EINVAL;
530                 if (msg->msg_namelen < sizeof(*usin))
531                         goto out;
532                 if (usin->sin_family != AF_INET) {
533                         pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
534                                      __func__, current->comm);
535                         err = -EAFNOSUPPORT;
536                         if (usin->sin_family)
537                                 goto out;
538                 }
539                 daddr = usin->sin_addr.s_addr;
540                 /* ANK: I did not forget to get protocol from port field.
541                  * I just do not know, who uses this weirdness.
542                  * IP_HDRINCL is much more convenient.
543                  */
544         } else {
545                 err = -EDESTADDRREQ;
546                 if (sk->sk_state != TCP_ESTABLISHED)
547                         goto out;
548                 daddr = inet->inet_daddr;
549         }
550 
551         ipc.sockc.tsflags = sk->sk_tsflags;
552         ipc.addr = inet->inet_saddr;
553         ipc.opt = NULL;
554         ipc.tx_flags = 0;
555         ipc.ttl = 0;
556         ipc.tos = -1;
557         ipc.oif = sk->sk_bound_dev_if;
558 
559         if (msg->msg_controllen) {
560                 err = ip_cmsg_send(sk, msg, &ipc, false);
561                 if (unlikely(err)) {
562                         kfree(ipc.opt);
563                         goto out;
564                 }
565                 if (ipc.opt)
566                         free = 1;
567         }
568 
569         saddr = ipc.addr;
570         ipc.addr = daddr;
571 
572         if (!ipc.opt) {
573                 struct ip_options_rcu *inet_opt;
574 
575                 rcu_read_lock();
576                 inet_opt = rcu_dereference(inet->inet_opt);
577                 if (inet_opt) {
578                         memcpy(&opt_copy, inet_opt,
579                                sizeof(*inet_opt) + inet_opt->opt.optlen);
580                         ipc.opt = &opt_copy.opt;
581                 }
582                 rcu_read_unlock();
583         }
584 
585         if (ipc.opt) {
586                 err = -EINVAL;
587                 /* Linux does not mangle headers on raw sockets,
588                  * so that IP options + IP_HDRINCL is non-sense.
589                  */
590                 if (inet->hdrincl)
591                         goto done;
592                 if (ipc.opt->opt.srr) {
593                         if (!daddr)
594                                 goto done;
595                         daddr = ipc.opt->opt.faddr;
596                 }
597         }
598         tos = get_rtconn_flags(&ipc, sk);
599         if (msg->msg_flags & MSG_DONTROUTE)
600                 tos |= RTO_ONLINK;
601 
602         if (ipv4_is_multicast(daddr)) {
603                 if (!ipc.oif)
604                         ipc.oif = inet->mc_index;
605                 if (!saddr)
606                         saddr = inet->mc_addr;
607         } else if (!ipc.oif)
608                 ipc.oif = inet->uc_index;
609 
610         flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
611                            RT_SCOPE_UNIVERSE,
612                            inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
613                            inet_sk_flowi_flags(sk) |
614                             (inet->hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
615                            daddr, saddr, 0, 0, sk->sk_uid);
616 
617         if (!inet->hdrincl) {
618                 rfv.msg = msg;
619                 rfv.hlen = 0;
620 
621                 err = raw_probe_proto_opt(&rfv, &fl4);
622                 if (err)
623                         goto done;
624         }
625 
626         security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
627         rt = ip_route_output_flow(net, &fl4, sk);
628         if (IS_ERR(rt)) {
629                 err = PTR_ERR(rt);
630                 rt = NULL;
631                 goto done;
632         }
633 
634         err = -EACCES;
635         if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
636                 goto done;
637 
638         if (msg->msg_flags & MSG_CONFIRM)
639                 goto do_confirm;
640 back_from_confirm:
641 
642         if (inet->hdrincl)
643                 err = raw_send_hdrinc(sk, &fl4, msg, len,
644                                       &rt, msg->msg_flags, &ipc.sockc);
645 
646          else {
647                 sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
648 
649                 if (!ipc.addr)
650                         ipc.addr = fl4.daddr;
651                 lock_sock(sk);
652                 err = ip_append_data(sk, &fl4, raw_getfrag,
653                                      &rfv, len, 0,
654                                      &ipc, &rt, msg->msg_flags);
655                 if (err)
656                         ip_flush_pending_frames(sk);
657                 else if (!(msg->msg_flags & MSG_MORE)) {
658                         err = ip_push_pending_frames(sk, &fl4);
659                         if (err == -ENOBUFS && !inet->recverr)
660                                 err = 0;
661                 }
662                 release_sock(sk);
663         }
664 done:
665         if (free)
666                 kfree(ipc.opt);
667         ip_rt_put(rt);
668 
669 out:
670         if (err < 0)
671                 return err;
672         return len;
673 
674 do_confirm:
675         if (msg->msg_flags & MSG_PROBE)
676                 dst_confirm_neigh(&rt->dst, &fl4.daddr);
677         if (!(msg->msg_flags & MSG_PROBE) || len)
678                 goto back_from_confirm;
679         err = 0;
680         goto done;
681 }
682 
683 static void raw_close(struct sock *sk, long timeout)
684 {
685         /*
686          * Raw sockets may have direct kernel references. Kill them.
687          */
688         rtnl_lock();
689         ip_ra_control(sk, 0, NULL);
690         rtnl_unlock();
691 
692         sk_common_release(sk);
693 }
694 
695 static void raw_destroy(struct sock *sk)
696 {
697         lock_sock(sk);
698         ip_flush_pending_frames(sk);
699         release_sock(sk);
700 }
701 
702 /* This gets rid of all the nasties in af_inet. -DaveM */
703 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
704 {
705         struct inet_sock *inet = inet_sk(sk);
706         struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
707         u32 tb_id = RT_TABLE_LOCAL;
708         int ret = -EINVAL;
709         int chk_addr_ret;
710 
711         if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
712                 goto out;
713 
714         if (sk->sk_bound_dev_if)
715                 tb_id = l3mdev_fib_table_by_index(sock_net(sk),
716                                                  sk->sk_bound_dev_if) ? : tb_id;
717 
718         chk_addr_ret = inet_addr_type_table(sock_net(sk), addr->sin_addr.s_addr,
719                                             tb_id);
720 
721         ret = -EADDRNOTAVAIL;
722         if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
723             chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
724                 goto out;
725         inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
726         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
727                 inet->inet_saddr = 0;  /* Use device */
728         sk_dst_reset(sk);
729         ret = 0;
730 out:    return ret;
731 }
732 
733 /*
734  *      This should be easy, if there is something there
735  *      we return it, otherwise we block.
736  */
737 
738 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
739                        int noblock, int flags, int *addr_len)
740 {
741         struct inet_sock *inet = inet_sk(sk);
742         size_t copied = 0;
743         int err = -EOPNOTSUPP;
744         DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
745         struct sk_buff *skb;
746 
747         if (flags & MSG_OOB)
748                 goto out;
749 
750         if (flags & MSG_ERRQUEUE) {
751                 err = ip_recv_error(sk, msg, len, addr_len);
752                 goto out;
753         }
754 
755         skb = skb_recv_datagram(sk, flags, noblock, &err);
756         if (!skb)
757                 goto out;
758         if (ccs_socket_post_recvmsg_permission(sk, skb, flags)) {
759                 err = -EAGAIN; /* Hope less harmful than -EPERM. */
760                 goto out;
761         }
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_ts_and_drops(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                 *addr_len = sizeof(*sin);
782         }
783         if (inet->cmsg_flags)
784                 ip_cmsg_recv(msg, skb);
785         if (flags & MSG_TRUNC)
786                 copied = skb->len;
787 done:
788         skb_free_datagram(sk, skb);
789 out:
790         if (err)
791                 return err;
792         return copied;
793 }
794 
795 static int raw_init(struct sock *sk)
796 {
797         struct raw_sock *rp = raw_sk(sk);
798 
799         if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
800                 memset(&rp->filter, 0, sizeof(rp->filter));
801         return 0;
802 }
803 
804 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
805 {
806         if (optlen > sizeof(struct icmp_filter))
807                 optlen = sizeof(struct icmp_filter);
808         if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
809                 return -EFAULT;
810         return 0;
811 }
812 
813 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
814 {
815         int len, ret = -EFAULT;
816 
817         if (get_user(len, optlen))
818                 goto out;
819         ret = -EINVAL;
820         if (len < 0)
821                 goto out;
822         if (len > sizeof(struct icmp_filter))
823                 len = sizeof(struct icmp_filter);
824         ret = -EFAULT;
825         if (put_user(len, optlen) ||
826             copy_to_user(optval, &raw_sk(sk)->filter, len))
827                 goto out;
828         ret = 0;
829 out:    return ret;
830 }
831 
832 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
833                           char __user *optval, unsigned int optlen)
834 {
835         if (optname == ICMP_FILTER) {
836                 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
837                         return -EOPNOTSUPP;
838                 else
839                         return raw_seticmpfilter(sk, optval, optlen);
840         }
841         return -ENOPROTOOPT;
842 }
843 
844 static int raw_setsockopt(struct sock *sk, int level, int optname,
845                           char __user *optval, unsigned int optlen)
846 {
847         if (level != SOL_RAW)
848                 return ip_setsockopt(sk, level, optname, optval, optlen);
849         return do_raw_setsockopt(sk, level, optname, optval, optlen);
850 }
851 
852 #ifdef CONFIG_COMPAT
853 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
854                                  char __user *optval, unsigned int optlen)
855 {
856         if (level != SOL_RAW)
857                 return compat_ip_setsockopt(sk, level, optname, optval, optlen);
858         return do_raw_setsockopt(sk, level, optname, optval, optlen);
859 }
860 #endif
861 
862 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
863                           char __user *optval, int __user *optlen)
864 {
865         if (optname == ICMP_FILTER) {
866                 if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
867                         return -EOPNOTSUPP;
868                 else
869                         return raw_geticmpfilter(sk, optval, optlen);
870         }
871         return -ENOPROTOOPT;
872 }
873 
874 static int raw_getsockopt(struct sock *sk, int level, int optname,
875                           char __user *optval, int __user *optlen)
876 {
877         if (level != SOL_RAW)
878                 return ip_getsockopt(sk, level, optname, optval, optlen);
879         return do_raw_getsockopt(sk, level, optname, optval, optlen);
880 }
881 
882 #ifdef CONFIG_COMPAT
883 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
884                                  char __user *optval, int __user *optlen)
885 {
886         if (level != SOL_RAW)
887                 return compat_ip_getsockopt(sk, level, optname, optval, optlen);
888         return do_raw_getsockopt(sk, level, optname, optval, optlen);
889 }
890 #endif
891 
892 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
893 {
894         switch (cmd) {
895         case SIOCOUTQ: {
896                 int amount = sk_wmem_alloc_get(sk);
897 
898                 return put_user(amount, (int __user *)arg);
899         }
900         case SIOCINQ: {
901                 struct sk_buff *skb;
902                 int amount = 0;
903 
904                 spin_lock_bh(&sk->sk_receive_queue.lock);
905                 skb = skb_peek(&sk->sk_receive_queue);
906                 if (skb)
907                         amount = skb->len;
908                 spin_unlock_bh(&sk->sk_receive_queue.lock);
909                 return put_user(amount, (int __user *)arg);
910         }
911 
912         default:
913 #ifdef CONFIG_IP_MROUTE
914                 return ipmr_ioctl(sk, cmd, (void __user *)arg);
915 #else
916                 return -ENOIOCTLCMD;
917 #endif
918         }
919 }
920 
921 #ifdef CONFIG_COMPAT
922 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
923 {
924         switch (cmd) {
925         case SIOCOUTQ:
926         case SIOCINQ:
927                 return -ENOIOCTLCMD;
928         default:
929 #ifdef CONFIG_IP_MROUTE
930                 return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
931 #else
932                 return -ENOIOCTLCMD;
933 #endif
934         }
935 }
936 #endif
937 
938 int raw_abort(struct sock *sk, int err)
939 {
940         lock_sock(sk);
941 
942         sk->sk_err = err;
943         sk->sk_error_report(sk);
944         __udp_disconnect(sk, 0);
945 
946         release_sock(sk);
947 
948         return 0;
949 }
950 EXPORT_SYMBOL_GPL(raw_abort);
951 
952 struct proto raw_prot = {
953         .name              = "RAW",
954         .owner             = THIS_MODULE,
955         .close             = raw_close,
956         .destroy           = raw_destroy,
957         .connect           = ip4_datagram_connect,
958         .disconnect        = __udp_disconnect,
959         .ioctl             = raw_ioctl,
960         .init              = raw_init,
961         .setsockopt        = raw_setsockopt,
962         .getsockopt        = raw_getsockopt,
963         .sendmsg           = raw_sendmsg,
964         .recvmsg           = raw_recvmsg,
965         .bind              = raw_bind,
966         .backlog_rcv       = raw_rcv_skb,
967         .release_cb        = ip4_datagram_release_cb,
968         .hash              = raw_hash_sk,
969         .unhash            = raw_unhash_sk,
970         .obj_size          = sizeof(struct raw_sock),
971         .h.raw_hash        = &raw_v4_hashinfo,
972 #ifdef CONFIG_COMPAT
973         .compat_setsockopt = compat_raw_setsockopt,
974         .compat_getsockopt = compat_raw_getsockopt,
975         .compat_ioctl      = compat_raw_ioctl,
976 #endif
977         .diag_destroy      = raw_abort,
978 };
979 
980 #ifdef CONFIG_PROC_FS
981 static struct sock *raw_get_first(struct seq_file *seq)
982 {
983         struct sock *sk;
984         struct raw_iter_state *state = raw_seq_private(seq);
985 
986         for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
987                         ++state->bucket) {
988                 sk_for_each(sk, &state->h->ht[state->bucket])
989                         if (sock_net(sk) == seq_file_net(seq))
990                                 goto found;
991         }
992         sk = NULL;
993 found:
994         return sk;
995 }
996 
997 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
998 {
999         struct raw_iter_state *state = raw_seq_private(seq);
1000 
1001         do {
1002                 sk = sk_next(sk);
1003 try_again:
1004                 ;
1005         } while (sk && sock_net(sk) != seq_file_net(seq));
1006 
1007         if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
1008                 sk = sk_head(&state->h->ht[state->bucket]);
1009                 goto try_again;
1010         }
1011         return sk;
1012 }
1013 
1014 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
1015 {
1016         struct sock *sk = raw_get_first(seq);
1017 
1018         if (sk)
1019                 while (pos && (sk = raw_get_next(seq, sk)) != NULL)
1020                         --pos;
1021         return pos ? NULL : sk;
1022 }
1023 
1024 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
1025 {
1026         struct raw_iter_state *state = raw_seq_private(seq);
1027 
1028         read_lock(&state->h->lock);
1029         return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1030 }
1031 EXPORT_SYMBOL_GPL(raw_seq_start);
1032 
1033 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1034 {
1035         struct sock *sk;
1036 
1037         if (v == SEQ_START_TOKEN)
1038                 sk = raw_get_first(seq);
1039         else
1040                 sk = raw_get_next(seq, v);
1041         ++*pos;
1042         return sk;
1043 }
1044 EXPORT_SYMBOL_GPL(raw_seq_next);
1045 
1046 void raw_seq_stop(struct seq_file *seq, void *v)
1047 {
1048         struct raw_iter_state *state = raw_seq_private(seq);
1049 
1050         read_unlock(&state->h->lock);
1051 }
1052 EXPORT_SYMBOL_GPL(raw_seq_stop);
1053 
1054 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1055 {
1056         struct inet_sock *inet = inet_sk(sp);
1057         __be32 dest = inet->inet_daddr,
1058                src = inet->inet_rcv_saddr;
1059         __u16 destp = 0,
1060               srcp  = inet->inet_num;
1061 
1062         seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1063                 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1064                 i, src, srcp, dest, destp, sp->sk_state,
1065                 sk_wmem_alloc_get(sp),
1066                 sk_rmem_alloc_get(sp),
1067                 0, 0L, 0,
1068                 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1069                 0, sock_i_ino(sp),
1070                 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1071 }
1072 
1073 static int raw_seq_show(struct seq_file *seq, void *v)
1074 {
1075         if (v == SEQ_START_TOKEN)
1076                 seq_printf(seq, "  sl  local_address rem_address   st tx_queue "
1077                                 "rx_queue tr tm->when retrnsmt   uid  timeout "
1078                                 "inode ref pointer drops\n");
1079         else
1080                 raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1081         return 0;
1082 }
1083 
1084 static const struct seq_operations raw_seq_ops = {
1085         .start = raw_seq_start,
1086         .next  = raw_seq_next,
1087         .stop  = raw_seq_stop,
1088         .show  = raw_seq_show,
1089 };
1090 
1091 int raw_seq_open(struct inode *ino, struct file *file,
1092                  struct raw_hashinfo *h, const struct seq_operations *ops)
1093 {
1094         int err;
1095         struct raw_iter_state *i;
1096 
1097         err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1098         if (err < 0)
1099                 return err;
1100 
1101         i = raw_seq_private((struct seq_file *)file->private_data);
1102         i->h = h;
1103         return 0;
1104 }
1105 EXPORT_SYMBOL_GPL(raw_seq_open);
1106 
1107 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1108 {
1109         return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1110 }
1111 
1112 static const struct file_operations raw_seq_fops = {
1113         .owner   = THIS_MODULE,
1114         .open    = raw_v4_seq_open,
1115         .read    = seq_read,
1116         .llseek  = seq_lseek,
1117         .release = seq_release_net,
1118 };
1119 
1120 static __net_init int raw_init_net(struct net *net)
1121 {
1122         if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops))
1123                 return -ENOMEM;
1124 
1125         return 0;
1126 }
1127 
1128 static __net_exit void raw_exit_net(struct net *net)
1129 {
1130         remove_proc_entry("raw", net->proc_net);
1131 }
1132 
1133 static __net_initdata struct pernet_operations raw_net_ops = {
1134         .init = raw_init_net,
1135         .exit = raw_exit_net,
1136 };
1137 
1138 int __init raw_proc_init(void)
1139 {
1140         return register_pernet_subsys(&raw_net_ops);
1141 }
1142 
1143 void __init raw_proc_exit(void)
1144 {
1145         unregister_pernet_subsys(&raw_net_ops);
1146 }
1147 #endif /* CONFIG_PROC_FS */
1148 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp