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

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

Version: ~ [ linux-5.6 ] ~ [ linux-5.5.13 ] ~ [ linux-5.4.28 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.113 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.174 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.217 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.217 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.82 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ 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  *              PF_INET protocol family socket handler.
  7  *
  8  * Authors:     Ross Biro
  9  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 10  *              Florian La Roche, <flla@stud.uni-sb.de>
 11  *              Alan Cox, <A.Cox@swansea.ac.uk>
 12  *
 13  * Changes (see also sock.c)
 14  *
 15  *              piggy,
 16  *              Karl Knutson    :       Socket protocol table
 17  *              A.N.Kuznetsov   :       Socket death error in accept().
 18  *              John Richardson :       Fix non blocking error in connect()
 19  *                                      so sockets that fail to connect
 20  *                                      don't return -EINPROGRESS.
 21  *              Alan Cox        :       Asynchronous I/O support
 22  *              Alan Cox        :       Keep correct socket pointer on sock
 23  *                                      structures
 24  *                                      when accept() ed
 25  *              Alan Cox        :       Semantics of SO_LINGER aren't state
 26  *                                      moved to close when you look carefully.
 27  *                                      With this fixed and the accept bug fixed
 28  *                                      some RPC stuff seems happier.
 29  *              Niibe Yutaka    :       4.4BSD style write async I/O
 30  *              Alan Cox,
 31  *              Tony Gale       :       Fixed reuse semantics.
 32  *              Alan Cox        :       bind() shouldn't abort existing but dead
 33  *                                      sockets. Stops FTP netin:.. I hope.
 34  *              Alan Cox        :       bind() works correctly for RAW sockets.
 35  *                                      Note that FreeBSD at least was broken
 36  *                                      in this respect so be careful with
 37  *                                      compatibility tests...
 38  *              Alan Cox        :       routing cache support
 39  *              Alan Cox        :       memzero the socket structure for
 40  *                                      compactness.
 41  *              Matt Day        :       nonblock connect error handler
 42  *              Alan Cox        :       Allow large numbers of pending sockets
 43  *                                      (eg for big web sites), but only if
 44  *                                      specifically application requested.
 45  *              Alan Cox        :       New buffering throughout IP. Used
 46  *                                      dumbly.
 47  *              Alan Cox        :       New buffering now used smartly.
 48  *              Alan Cox        :       BSD rather than common sense
 49  *                                      interpretation of listen.
 50  *              Germano Caronni :       Assorted small races.
 51  *              Alan Cox        :       sendmsg/recvmsg basic support.
 52  *              Alan Cox        :       Only sendmsg/recvmsg now supported.
 53  *              Alan Cox        :       Locked down bind (see security list).
 54  *              Alan Cox        :       Loosened bind a little.
 55  *              Mike McLagan    :       ADD/DEL DLCI Ioctls
 56  *      Willy Konynenberg       :       Transparent proxying support.
 57  *              David S. Miller :       New socket lookup architecture.
 58  *                                      Some other random speedups.
 59  *              Cyrus Durgin    :       Cleaned up file for kmod hacks.
 60  *              Andi Kleen      :       Fix inet_stream_connect TCP race.
 61  *
 62  *              This program is free software; you can redistribute it and/or
 63  *              modify it under the terms of the GNU General Public License
 64  *              as published by the Free Software Foundation; either version
 65  *              2 of the License, or (at your option) any later version.
 66  */
 67 
 68 #define pr_fmt(fmt) "IPv4: " fmt
 69 
 70 #include <linux/err.h>
 71 #include <linux/errno.h>
 72 #include <linux/types.h>
 73 #include <linux/socket.h>
 74 #include <linux/in.h>
 75 #include <linux/kernel.h>
 76 #include <linux/kmod.h>
 77 #include <linux/sched.h>
 78 #include <linux/timer.h>
 79 #include <linux/string.h>
 80 #include <linux/sockios.h>
 81 #include <linux/net.h>
 82 #include <linux/capability.h>
 83 #include <linux/fcntl.h>
 84 #include <linux/mm.h>
 85 #include <linux/interrupt.h>
 86 #include <linux/stat.h>
 87 #include <linux/init.h>
 88 #include <linux/poll.h>
 89 #include <linux/netfilter_ipv4.h>
 90 #include <linux/random.h>
 91 #include <linux/slab.h>
 92 
 93 #include <linux/uaccess.h>
 94 
 95 #include <linux/inet.h>
 96 #include <linux/igmp.h>
 97 #include <linux/inetdevice.h>
 98 #include <linux/netdevice.h>
 99 #include <net/checksum.h>
100 #include <net/ip.h>
101 #include <net/protocol.h>
102 #include <net/arp.h>
103 #include <net/route.h>
104 #include <net/ip_fib.h>
105 #include <net/inet_connection_sock.h>
106 #include <net/tcp.h>
107 #include <net/udp.h>
108 #include <net/udplite.h>
109 #include <net/ping.h>
110 #include <linux/skbuff.h>
111 #include <net/sock.h>
112 #include <net/raw.h>
113 #include <net/icmp.h>
114 #include <net/inet_common.h>
115 #include <net/ip_tunnels.h>
116 #include <net/xfrm.h>
117 #include <net/net_namespace.h>
118 #include <net/secure_seq.h>
119 #ifdef CONFIG_IP_MROUTE
120 #include <linux/mroute.h>
121 #endif
122 #include <net/l3mdev.h>
123 
124 
125 /* The inetsw table contains everything that inet_create needs to
126  * build a new socket.
127  */
128 static struct list_head inetsw[SOCK_MAX];
129 static DEFINE_SPINLOCK(inetsw_lock);
130 
131 /* New destruction routine */
132 
133 void inet_sock_destruct(struct sock *sk)
134 {
135         struct inet_sock *inet = inet_sk(sk);
136 
137         __skb_queue_purge(&sk->sk_receive_queue);
138         __skb_queue_purge(&sk->sk_error_queue);
139 
140         sk_mem_reclaim(sk);
141 
142         if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) {
143                 pr_err("Attempt to release TCP socket in state %d %p\n",
144                        sk->sk_state, sk);
145                 return;
146         }
147         if (!sock_flag(sk, SOCK_DEAD)) {
148                 pr_err("Attempt to release alive inet socket %p\n", sk);
149                 return;
150         }
151 
152         WARN_ON(atomic_read(&sk->sk_rmem_alloc));
153         WARN_ON(refcount_read(&sk->sk_wmem_alloc));
154         WARN_ON(sk->sk_wmem_queued);
155         WARN_ON(sk->sk_forward_alloc);
156 
157         kfree(rcu_dereference_protected(inet->inet_opt, 1));
158         dst_release(rcu_dereference_check(sk->sk_dst_cache, 1));
159         dst_release(sk->sk_rx_dst);
160         sk_refcnt_debug_dec(sk);
161 }
162 EXPORT_SYMBOL(inet_sock_destruct);
163 
164 /*
165  *      The routines beyond this point handle the behaviour of an AF_INET
166  *      socket object. Mostly it punts to the subprotocols of IP to do
167  *      the work.
168  */
169 
170 /*
171  *      Automatically bind an unbound socket.
172  */
173 
174 static int inet_autobind(struct sock *sk)
175 {
176         struct inet_sock *inet;
177         /* We may need to bind the socket. */
178         lock_sock(sk);
179         inet = inet_sk(sk);
180         if (!inet->inet_num) {
181                 if (sk->sk_prot->get_port(sk, 0)) {
182                         release_sock(sk);
183                         return -EAGAIN;
184                 }
185                 inet->inet_sport = htons(inet->inet_num);
186         }
187         release_sock(sk);
188         return 0;
189 }
190 
191 /*
192  *      Move a socket into listening state.
193  */
194 int inet_listen(struct socket *sock, int backlog)
195 {
196         struct sock *sk = sock->sk;
197         unsigned char old_state;
198         int err, tcp_fastopen;
199 
200         lock_sock(sk);
201 
202         err = -EINVAL;
203         if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM)
204                 goto out;
205 
206         old_state = sk->sk_state;
207         if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
208                 goto out;
209 
210         /* Really, if the socket is already in listen state
211          * we can only allow the backlog to be adjusted.
212          */
213         if (old_state != TCP_LISTEN) {
214                 /* Enable TFO w/o requiring TCP_FASTOPEN socket option.
215                  * Note that only TCP sockets (SOCK_STREAM) will reach here.
216                  * Also fastopen backlog may already been set via the option
217                  * because the socket was in TCP_LISTEN state previously but
218                  * was shutdown() rather than close().
219                  */
220                 tcp_fastopen = sock_net(sk)->ipv4.sysctl_tcp_fastopen;
221                 if ((tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) &&
222                     (tcp_fastopen & TFO_SERVER_ENABLE) &&
223                     !inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) {
224                         fastopen_queue_tune(sk, backlog);
225                         tcp_fastopen_init_key_once(sock_net(sk));
226                 }
227 
228                 err = inet_csk_listen_start(sk, backlog);
229                 if (err)
230                         goto out;
231         }
232         sk->sk_max_ack_backlog = backlog;
233         err = 0;
234 
235 out:
236         release_sock(sk);
237         return err;
238 }
239 EXPORT_SYMBOL(inet_listen);
240 
241 /*
242  *      Create an inet socket.
243  */
244 
245 static int inet_create(struct net *net, struct socket *sock, int protocol,
246                        int kern)
247 {
248         struct sock *sk;
249         struct inet_protosw *answer;
250         struct inet_sock *inet;
251         struct proto *answer_prot;
252         unsigned char answer_flags;
253         int try_loading_module = 0;
254         int err;
255 
256         if (protocol < 0 || protocol >= IPPROTO_MAX)
257                 return -EINVAL;
258 
259         sock->state = SS_UNCONNECTED;
260 
261         /* Look for the requested type/protocol pair. */
262 lookup_protocol:
263         err = -ESOCKTNOSUPPORT;
264         rcu_read_lock();
265         list_for_each_entry_rcu(answer, &inetsw[sock->type], list) {
266 
267                 err = 0;
268                 /* Check the non-wild match. */
269                 if (protocol == answer->protocol) {
270                         if (protocol != IPPROTO_IP)
271                                 break;
272                 } else {
273                         /* Check for the two wild cases. */
274                         if (IPPROTO_IP == protocol) {
275                                 protocol = answer->protocol;
276                                 break;
277                         }
278                         if (IPPROTO_IP == answer->protocol)
279                                 break;
280                 }
281                 err = -EPROTONOSUPPORT;
282         }
283 
284         if (unlikely(err)) {
285                 if (try_loading_module < 2) {
286                         rcu_read_unlock();
287                         /*
288                          * Be more specific, e.g. net-pf-2-proto-132-type-1
289                          * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM)
290                          */
291                         if (++try_loading_module == 1)
292                                 request_module("net-pf-%d-proto-%d-type-%d",
293                                                PF_INET, protocol, sock->type);
294                         /*
295                          * Fall back to generic, e.g. net-pf-2-proto-132
296                          * (net-pf-PF_INET-proto-IPPROTO_SCTP)
297                          */
298                         else
299                                 request_module("net-pf-%d-proto-%d",
300                                                PF_INET, protocol);
301                         goto lookup_protocol;
302                 } else
303                         goto out_rcu_unlock;
304         }
305 
306         err = -EPERM;
307         if (sock->type == SOCK_RAW && !kern &&
308             !ns_capable(net->user_ns, CAP_NET_RAW))
309                 goto out_rcu_unlock;
310 
311         sock->ops = answer->ops;
312         answer_prot = answer->prot;
313         answer_flags = answer->flags;
314         rcu_read_unlock();
315 
316         WARN_ON(!answer_prot->slab);
317 
318         err = -ENOBUFS;
319         sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot, kern);
320         if (!sk)
321                 goto out;
322 
323         err = 0;
324         if (INET_PROTOSW_REUSE & answer_flags)
325                 sk->sk_reuse = SK_CAN_REUSE;
326 
327         inet = inet_sk(sk);
328         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
329 
330         inet->nodefrag = 0;
331 
332         if (SOCK_RAW == sock->type) {
333                 inet->inet_num = protocol;
334                 if (IPPROTO_RAW == protocol)
335                         inet->hdrincl = 1;
336         }
337 
338         if (net->ipv4.sysctl_ip_no_pmtu_disc)
339                 inet->pmtudisc = IP_PMTUDISC_DONT;
340         else
341                 inet->pmtudisc = IP_PMTUDISC_WANT;
342 
343         inet->inet_id = 0;
344 
345         sock_init_data(sock, sk);
346 
347         sk->sk_destruct    = inet_sock_destruct;
348         sk->sk_protocol    = protocol;
349         sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
350 
351         inet->uc_ttl    = -1;
352         inet->mc_loop   = 1;
353         inet->mc_ttl    = 1;
354         inet->mc_all    = 1;
355         inet->mc_index  = 0;
356         inet->mc_list   = NULL;
357         inet->rcv_tos   = 0;
358 
359         sk_refcnt_debug_inc(sk);
360 
361         if (inet->inet_num) {
362                 /* It assumes that any protocol which allows
363                  * the user to assign a number at socket
364                  * creation time automatically
365                  * shares.
366                  */
367                 inet->inet_sport = htons(inet->inet_num);
368                 /* Add to protocol hash chains. */
369                 err = sk->sk_prot->hash(sk);
370                 if (err) {
371                         sk_common_release(sk);
372                         goto out;
373                 }
374         }
375 
376         if (sk->sk_prot->init) {
377                 err = sk->sk_prot->init(sk);
378                 if (err) {
379                         sk_common_release(sk);
380                         goto out;
381                 }
382         }
383 
384         if (!kern) {
385                 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
386                 if (err) {
387                         sk_common_release(sk);
388                         goto out;
389                 }
390         }
391 out:
392         return err;
393 out_rcu_unlock:
394         rcu_read_unlock();
395         goto out;
396 }
397 
398 
399 /*
400  *      The peer socket should always be NULL (or else). When we call this
401  *      function we are destroying the object and from then on nobody
402  *      should refer to it.
403  */
404 int inet_release(struct socket *sock)
405 {
406         struct sock *sk = sock->sk;
407 
408         if (sk) {
409                 long timeout;
410 
411                 /* Applications forget to leave groups before exiting */
412                 ip_mc_drop_socket(sk);
413 
414                 /* If linger is set, we don't return until the close
415                  * is complete.  Otherwise we return immediately. The
416                  * actually closing is done the same either way.
417                  *
418                  * If the close is due to the process exiting, we never
419                  * linger..
420                  */
421                 timeout = 0;
422                 if (sock_flag(sk, SOCK_LINGER) &&
423                     !(current->flags & PF_EXITING))
424                         timeout = sk->sk_lingertime;
425                 sock->sk = NULL;
426                 sk->sk_prot->close(sk, timeout);
427         }
428         return 0;
429 }
430 EXPORT_SYMBOL(inet_release);
431 
432 int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
433 {
434         struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
435         struct sock *sk = sock->sk;
436         struct inet_sock *inet = inet_sk(sk);
437         struct net *net = sock_net(sk);
438         unsigned short snum;
439         int chk_addr_ret;
440         u32 tb_id = RT_TABLE_LOCAL;
441         int err;
442 
443         /* If the socket has its own bind function then use it. (RAW) */
444         if (sk->sk_prot->bind) {
445                 err = sk->sk_prot->bind(sk, uaddr, addr_len);
446                 goto out;
447         }
448         err = -EINVAL;
449         if (addr_len < sizeof(struct sockaddr_in))
450                 goto out;
451 
452         if (addr->sin_family != AF_INET) {
453                 /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET)
454                  * only if s_addr is INADDR_ANY.
455                  */
456                 err = -EAFNOSUPPORT;
457                 if (addr->sin_family != AF_UNSPEC ||
458                     addr->sin_addr.s_addr != htonl(INADDR_ANY))
459                         goto out;
460         }
461 
462         tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
463         chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
464 
465         /* Not specified by any standard per-se, however it breaks too
466          * many applications when removed.  It is unfortunate since
467          * allowing applications to make a non-local bind solves
468          * several problems with systems using dynamic addressing.
469          * (ie. your servers still start up even if your ISDN link
470          *  is temporarily down)
471          */
472         err = -EADDRNOTAVAIL;
473         if (!net->ipv4.sysctl_ip_nonlocal_bind &&
474             !(inet->freebind || inet->transparent) &&
475             addr->sin_addr.s_addr != htonl(INADDR_ANY) &&
476             chk_addr_ret != RTN_LOCAL &&
477             chk_addr_ret != RTN_MULTICAST &&
478             chk_addr_ret != RTN_BROADCAST)
479                 goto out;
480 
481         snum = ntohs(addr->sin_port);
482         err = -EACCES;
483         if (snum && snum < inet_prot_sock(net) &&
484             !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
485                 goto out;
486 
487         /*      We keep a pair of addresses. rcv_saddr is the one
488          *      used by hash lookups, and saddr is used for transmit.
489          *
490          *      In the BSD API these are the same except where it
491          *      would be illegal to use them (multicast/broadcast) in
492          *      which case the sending device address is used.
493          */
494         lock_sock(sk);
495 
496         /* Check these errors (active socket, double bind). */
497         err = -EINVAL;
498         if (sk->sk_state != TCP_CLOSE || inet->inet_num)
499                 goto out_release_sock;
500 
501         inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
502         if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
503                 inet->inet_saddr = 0;  /* Use device */
504 
505         /* Make sure we are allowed to bind here. */
506         if ((snum || !inet->bind_address_no_port) &&
507             sk->sk_prot->get_port(sk, snum)) {
508                 inet->inet_saddr = inet->inet_rcv_saddr = 0;
509                 err = -EADDRINUSE;
510                 goto out_release_sock;
511         }
512 
513         if (inet->inet_rcv_saddr)
514                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
515         if (snum)
516                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
517         inet->inet_sport = htons(inet->inet_num);
518         inet->inet_daddr = 0;
519         inet->inet_dport = 0;
520         sk_dst_reset(sk);
521         err = 0;
522 out_release_sock:
523         release_sock(sk);
524 out:
525         return err;
526 }
527 EXPORT_SYMBOL(inet_bind);
528 
529 int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr,
530                        int addr_len, int flags)
531 {
532         struct sock *sk = sock->sk;
533 
534         if (addr_len < sizeof(uaddr->sa_family))
535                 return -EINVAL;
536         if (uaddr->sa_family == AF_UNSPEC)
537                 return sk->sk_prot->disconnect(sk, flags);
538 
539         if (!inet_sk(sk)->inet_num && inet_autobind(sk))
540                 return -EAGAIN;
541         return sk->sk_prot->connect(sk, uaddr, addr_len);
542 }
543 EXPORT_SYMBOL(inet_dgram_connect);
544 
545 static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias)
546 {
547         DEFINE_WAIT_FUNC(wait, woken_wake_function);
548 
549         add_wait_queue(sk_sleep(sk), &wait);
550         sk->sk_write_pending += writebias;
551 
552         /* Basic assumption: if someone sets sk->sk_err, he _must_
553          * change state of the socket from TCP_SYN_*.
554          * Connect() does not allow to get error notifications
555          * without closing the socket.
556          */
557         while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
558                 release_sock(sk);
559                 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
560                 lock_sock(sk);
561                 if (signal_pending(current) || !timeo)
562                         break;
563         }
564         remove_wait_queue(sk_sleep(sk), &wait);
565         sk->sk_write_pending -= writebias;
566         return timeo;
567 }
568 
569 /*
570  *      Connect to a remote host. There is regrettably still a little
571  *      TCP 'magic' in here.
572  */
573 int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
574                           int addr_len, int flags, int is_sendmsg)
575 {
576         struct sock *sk = sock->sk;
577         int err;
578         long timeo;
579 
580         /*
581          * uaddr can be NULL and addr_len can be 0 if:
582          * sk is a TCP fastopen active socket and
583          * TCP_FASTOPEN_CONNECT sockopt is set and
584          * we already have a valid cookie for this socket.
585          * In this case, user can call write() after connect().
586          * write() will invoke tcp_sendmsg_fastopen() which calls
587          * __inet_stream_connect().
588          */
589         if (uaddr) {
590                 if (addr_len < sizeof(uaddr->sa_family))
591                         return -EINVAL;
592 
593                 if (uaddr->sa_family == AF_UNSPEC) {
594                         err = sk->sk_prot->disconnect(sk, flags);
595                         sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
596                         goto out;
597                 }
598         }
599 
600         switch (sock->state) {
601         default:
602                 err = -EINVAL;
603                 goto out;
604         case SS_CONNECTED:
605                 err = -EISCONN;
606                 goto out;
607         case SS_CONNECTING:
608                 if (inet_sk(sk)->defer_connect)
609                         err = is_sendmsg ? -EINPROGRESS : -EISCONN;
610                 else
611                         err = -EALREADY;
612                 /* Fall out of switch with err, set for this state */
613                 break;
614         case SS_UNCONNECTED:
615                 err = -EISCONN;
616                 if (sk->sk_state != TCP_CLOSE)
617                         goto out;
618 
619                 err = sk->sk_prot->connect(sk, uaddr, addr_len);
620                 if (err < 0)
621                         goto out;
622 
623                 sock->state = SS_CONNECTING;
624 
625                 if (!err && inet_sk(sk)->defer_connect)
626                         goto out;
627 
628                 /* Just entered SS_CONNECTING state; the only
629                  * difference is that return value in non-blocking
630                  * case is EINPROGRESS, rather than EALREADY.
631                  */
632                 err = -EINPROGRESS;
633                 break;
634         }
635 
636         timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
637 
638         if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
639                 int writebias = (sk->sk_protocol == IPPROTO_TCP) &&
640                                 tcp_sk(sk)->fastopen_req &&
641                                 tcp_sk(sk)->fastopen_req->data ? 1 : 0;
642 
643                 /* Error code is set above */
644                 if (!timeo || !inet_wait_for_connect(sk, timeo, writebias))
645                         goto out;
646 
647                 err = sock_intr_errno(timeo);
648                 if (signal_pending(current))
649                         goto out;
650         }
651 
652         /* Connection was closed by RST, timeout, ICMP error
653          * or another process disconnected us.
654          */
655         if (sk->sk_state == TCP_CLOSE)
656                 goto sock_error;
657 
658         /* sk->sk_err may be not zero now, if RECVERR was ordered by user
659          * and error was received after socket entered established state.
660          * Hence, it is handled normally after connect() return successfully.
661          */
662 
663         sock->state = SS_CONNECTED;
664         err = 0;
665 out:
666         return err;
667 
668 sock_error:
669         err = sock_error(sk) ? : -ECONNABORTED;
670         sock->state = SS_UNCONNECTED;
671         if (sk->sk_prot->disconnect(sk, flags))
672                 sock->state = SS_DISCONNECTING;
673         goto out;
674 }
675 EXPORT_SYMBOL(__inet_stream_connect);
676 
677 int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr,
678                         int addr_len, int flags)
679 {
680         int err;
681 
682         lock_sock(sock->sk);
683         err = __inet_stream_connect(sock, uaddr, addr_len, flags, 0);
684         release_sock(sock->sk);
685         return err;
686 }
687 EXPORT_SYMBOL(inet_stream_connect);
688 
689 /*
690  *      Accept a pending connection. The TCP layer now gives BSD semantics.
691  */
692 
693 int inet_accept(struct socket *sock, struct socket *newsock, int flags,
694                 bool kern)
695 {
696         struct sock *sk1 = sock->sk;
697         int err = -EINVAL;
698         struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err, kern);
699 
700         if (!sk2)
701                 goto do_err;
702 
703         lock_sock(sk2);
704 
705         sock_rps_record_flow(sk2);
706         WARN_ON(!((1 << sk2->sk_state) &
707                   (TCPF_ESTABLISHED | TCPF_SYN_RECV |
708                   TCPF_CLOSE_WAIT | TCPF_CLOSE)));
709 
710         sock_graft(sk2, newsock);
711 
712         newsock->state = SS_CONNECTED;
713         err = 0;
714         release_sock(sk2);
715 do_err:
716         return err;
717 }
718 EXPORT_SYMBOL(inet_accept);
719 
720 
721 /*
722  *      This does both peername and sockname.
723  */
724 int inet_getname(struct socket *sock, struct sockaddr *uaddr,
725                         int *uaddr_len, int peer)
726 {
727         struct sock *sk         = sock->sk;
728         struct inet_sock *inet  = inet_sk(sk);
729         DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr);
730 
731         sin->sin_family = AF_INET;
732         if (peer) {
733                 if (!inet->inet_dport ||
734                     (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
735                      peer == 1))
736                         return -ENOTCONN;
737                 sin->sin_port = inet->inet_dport;
738                 sin->sin_addr.s_addr = inet->inet_daddr;
739         } else {
740                 __be32 addr = inet->inet_rcv_saddr;
741                 if (!addr)
742                         addr = inet->inet_saddr;
743                 sin->sin_port = inet->inet_sport;
744                 sin->sin_addr.s_addr = addr;
745         }
746         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
747         *uaddr_len = sizeof(*sin);
748         return 0;
749 }
750 EXPORT_SYMBOL(inet_getname);
751 
752 int inet_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
753 {
754         struct sock *sk = sock->sk;
755 
756         sock_rps_record_flow(sk);
757 
758         /* We may need to bind the socket. */
759         if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
760             inet_autobind(sk))
761                 return -EAGAIN;
762 
763         return sk->sk_prot->sendmsg(sk, msg, size);
764 }
765 EXPORT_SYMBOL(inet_sendmsg);
766 
767 ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset,
768                       size_t size, int flags)
769 {
770         struct sock *sk = sock->sk;
771 
772         sock_rps_record_flow(sk);
773 
774         /* We may need to bind the socket. */
775         if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind &&
776             inet_autobind(sk))
777                 return -EAGAIN;
778 
779         if (sk->sk_prot->sendpage)
780                 return sk->sk_prot->sendpage(sk, page, offset, size, flags);
781         return sock_no_sendpage(sock, page, offset, size, flags);
782 }
783 EXPORT_SYMBOL(inet_sendpage);
784 
785 int inet_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
786                  int flags)
787 {
788         struct sock *sk = sock->sk;
789         int addr_len = 0;
790         int err;
791 
792         sock_rps_record_flow(sk);
793 
794         err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
795                                    flags & ~MSG_DONTWAIT, &addr_len);
796         if (err >= 0)
797                 msg->msg_namelen = addr_len;
798         return err;
799 }
800 EXPORT_SYMBOL(inet_recvmsg);
801 
802 int inet_shutdown(struct socket *sock, int how)
803 {
804         struct sock *sk = sock->sk;
805         int err = 0;
806 
807         /* This should really check to make sure
808          * the socket is a TCP socket. (WHY AC...)
809          */
810         how++; /* maps 0->1 has the advantage of making bit 1 rcvs and
811                        1->2 bit 2 snds.
812                        2->3 */
813         if ((how & ~SHUTDOWN_MASK) || !how)     /* MAXINT->0 */
814                 return -EINVAL;
815 
816         lock_sock(sk);
817         if (sock->state == SS_CONNECTING) {
818                 if ((1 << sk->sk_state) &
819                     (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE))
820                         sock->state = SS_DISCONNECTING;
821                 else
822                         sock->state = SS_CONNECTED;
823         }
824 
825         switch (sk->sk_state) {
826         case TCP_CLOSE:
827                 err = -ENOTCONN;
828                 /* Hack to wake up other listeners, who can poll for
829                    POLLHUP, even on eg. unconnected UDP sockets -- RR */
830                 /* fall through */
831         default:
832                 sk->sk_shutdown |= how;
833                 if (sk->sk_prot->shutdown)
834                         sk->sk_prot->shutdown(sk, how);
835                 break;
836 
837         /* Remaining two branches are temporary solution for missing
838          * close() in multithreaded environment. It is _not_ a good idea,
839          * but we have no choice until close() is repaired at VFS level.
840          */
841         case TCP_LISTEN:
842                 if (!(how & RCV_SHUTDOWN))
843                         break;
844                 /* fall through */
845         case TCP_SYN_SENT:
846                 err = sk->sk_prot->disconnect(sk, O_NONBLOCK);
847                 sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED;
848                 break;
849         }
850 
851         /* Wake up anyone sleeping in poll. */
852         sk->sk_state_change(sk);
853         release_sock(sk);
854         return err;
855 }
856 EXPORT_SYMBOL(inet_shutdown);
857 
858 /*
859  *      ioctl() calls you can issue on an INET socket. Most of these are
860  *      device configuration and stuff and very rarely used. Some ioctls
861  *      pass on to the socket itself.
862  *
863  *      NOTE: I like the idea of a module for the config stuff. ie ifconfig
864  *      loads the devconfigure module does its configuring and unloads it.
865  *      There's a good 20K of config code hanging around the kernel.
866  */
867 
868 int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
869 {
870         struct sock *sk = sock->sk;
871         int err = 0;
872         struct net *net = sock_net(sk);
873 
874         switch (cmd) {
875         case SIOCGSTAMP:
876                 err = sock_get_timestamp(sk, (struct timeval __user *)arg);
877                 break;
878         case SIOCGSTAMPNS:
879                 err = sock_get_timestampns(sk, (struct timespec __user *)arg);
880                 break;
881         case SIOCADDRT:
882         case SIOCDELRT:
883         case SIOCRTMSG:
884                 err = ip_rt_ioctl(net, cmd, (void __user *)arg);
885                 break;
886         case SIOCDARP:
887         case SIOCGARP:
888         case SIOCSARP:
889                 err = arp_ioctl(net, cmd, (void __user *)arg);
890                 break;
891         case SIOCGIFADDR:
892         case SIOCSIFADDR:
893         case SIOCGIFBRDADDR:
894         case SIOCSIFBRDADDR:
895         case SIOCGIFNETMASK:
896         case SIOCSIFNETMASK:
897         case SIOCGIFDSTADDR:
898         case SIOCSIFDSTADDR:
899         case SIOCSIFPFLAGS:
900         case SIOCGIFPFLAGS:
901         case SIOCSIFFLAGS:
902                 err = devinet_ioctl(net, cmd, (void __user *)arg);
903                 break;
904         default:
905                 if (sk->sk_prot->ioctl)
906                         err = sk->sk_prot->ioctl(sk, cmd, arg);
907                 else
908                         err = -ENOIOCTLCMD;
909                 break;
910         }
911         return err;
912 }
913 EXPORT_SYMBOL(inet_ioctl);
914 
915 #ifdef CONFIG_COMPAT
916 static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
917 {
918         struct sock *sk = sock->sk;
919         int err = -ENOIOCTLCMD;
920 
921         if (sk->sk_prot->compat_ioctl)
922                 err = sk->sk_prot->compat_ioctl(sk, cmd, arg);
923 
924         return err;
925 }
926 #endif
927 
928 const struct proto_ops inet_stream_ops = {
929         .family            = PF_INET,
930         .owner             = THIS_MODULE,
931         .release           = inet_release,
932         .bind              = inet_bind,
933         .connect           = inet_stream_connect,
934         .socketpair        = sock_no_socketpair,
935         .accept            = inet_accept,
936         .getname           = inet_getname,
937         .poll              = tcp_poll,
938         .ioctl             = inet_ioctl,
939         .listen            = inet_listen,
940         .shutdown          = inet_shutdown,
941         .setsockopt        = sock_common_setsockopt,
942         .getsockopt        = sock_common_getsockopt,
943         .sendmsg           = inet_sendmsg,
944         .recvmsg           = inet_recvmsg,
945         .mmap              = sock_no_mmap,
946         .sendpage          = inet_sendpage,
947         .splice_read       = tcp_splice_read,
948         .read_sock         = tcp_read_sock,
949         .sendmsg_locked    = tcp_sendmsg_locked,
950         .sendpage_locked   = tcp_sendpage_locked,
951         .peek_len          = tcp_peek_len,
952 #ifdef CONFIG_COMPAT
953         .compat_setsockopt = compat_sock_common_setsockopt,
954         .compat_getsockopt = compat_sock_common_getsockopt,
955         .compat_ioctl      = inet_compat_ioctl,
956 #endif
957 };
958 EXPORT_SYMBOL(inet_stream_ops);
959 
960 const struct proto_ops inet_dgram_ops = {
961         .family            = PF_INET,
962         .owner             = THIS_MODULE,
963         .release           = inet_release,
964         .bind              = inet_bind,
965         .connect           = inet_dgram_connect,
966         .socketpair        = sock_no_socketpair,
967         .accept            = sock_no_accept,
968         .getname           = inet_getname,
969         .poll              = udp_poll,
970         .ioctl             = inet_ioctl,
971         .listen            = sock_no_listen,
972         .shutdown          = inet_shutdown,
973         .setsockopt        = sock_common_setsockopt,
974         .getsockopt        = sock_common_getsockopt,
975         .sendmsg           = inet_sendmsg,
976         .recvmsg           = inet_recvmsg,
977         .mmap              = sock_no_mmap,
978         .sendpage          = inet_sendpage,
979         .set_peek_off      = sk_set_peek_off,
980 #ifdef CONFIG_COMPAT
981         .compat_setsockopt = compat_sock_common_setsockopt,
982         .compat_getsockopt = compat_sock_common_getsockopt,
983         .compat_ioctl      = inet_compat_ioctl,
984 #endif
985 };
986 EXPORT_SYMBOL(inet_dgram_ops);
987 
988 /*
989  * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without
990  * udp_poll
991  */
992 static const struct proto_ops inet_sockraw_ops = {
993         .family            = PF_INET,
994         .owner             = THIS_MODULE,
995         .release           = inet_release,
996         .bind              = inet_bind,
997         .connect           = inet_dgram_connect,
998         .socketpair        = sock_no_socketpair,
999         .accept            = sock_no_accept,
1000         .getname           = inet_getname,
1001         .poll              = datagram_poll,
1002         .ioctl             = inet_ioctl,
1003         .listen            = sock_no_listen,
1004         .shutdown          = inet_shutdown,
1005         .setsockopt        = sock_common_setsockopt,
1006         .getsockopt        = sock_common_getsockopt,
1007         .sendmsg           = inet_sendmsg,
1008         .recvmsg           = inet_recvmsg,
1009         .mmap              = sock_no_mmap,
1010         .sendpage          = inet_sendpage,
1011 #ifdef CONFIG_COMPAT
1012         .compat_setsockopt = compat_sock_common_setsockopt,
1013         .compat_getsockopt = compat_sock_common_getsockopt,
1014         .compat_ioctl      = inet_compat_ioctl,
1015 #endif
1016 };
1017 
1018 static const struct net_proto_family inet_family_ops = {
1019         .family = PF_INET,
1020         .create = inet_create,
1021         .owner  = THIS_MODULE,
1022 };
1023 
1024 /* Upon startup we insert all the elements in inetsw_array[] into
1025  * the linked list inetsw.
1026  */
1027 static struct inet_protosw inetsw_array[] =
1028 {
1029         {
1030                 .type =       SOCK_STREAM,
1031                 .protocol =   IPPROTO_TCP,
1032                 .prot =       &tcp_prot,
1033                 .ops =        &inet_stream_ops,
1034                 .flags =      INET_PROTOSW_PERMANENT |
1035                               INET_PROTOSW_ICSK,
1036         },
1037 
1038         {
1039                 .type =       SOCK_DGRAM,
1040                 .protocol =   IPPROTO_UDP,
1041                 .prot =       &udp_prot,
1042                 .ops =        &inet_dgram_ops,
1043                 .flags =      INET_PROTOSW_PERMANENT,
1044        },
1045 
1046        {
1047                 .type =       SOCK_DGRAM,
1048                 .protocol =   IPPROTO_ICMP,
1049                 .prot =       &ping_prot,
1050                 .ops =        &inet_sockraw_ops,
1051                 .flags =      INET_PROTOSW_REUSE,
1052        },
1053 
1054        {
1055                .type =       SOCK_RAW,
1056                .protocol =   IPPROTO_IP,        /* wild card */
1057                .prot =       &raw_prot,
1058                .ops =        &inet_sockraw_ops,
1059                .flags =      INET_PROTOSW_REUSE,
1060        }
1061 };
1062 
1063 #define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array)
1064 
1065 void inet_register_protosw(struct inet_protosw *p)
1066 {
1067         struct list_head *lh;
1068         struct inet_protosw *answer;
1069         int protocol = p->protocol;
1070         struct list_head *last_perm;
1071 
1072         spin_lock_bh(&inetsw_lock);
1073 
1074         if (p->type >= SOCK_MAX)
1075                 goto out_illegal;
1076 
1077         /* If we are trying to override a permanent protocol, bail. */
1078         last_perm = &inetsw[p->type];
1079         list_for_each(lh, &inetsw[p->type]) {
1080                 answer = list_entry(lh, struct inet_protosw, list);
1081                 /* Check only the non-wild match. */
1082                 if ((INET_PROTOSW_PERMANENT & answer->flags) == 0)
1083                         break;
1084                 if (protocol == answer->protocol)
1085                         goto out_permanent;
1086                 last_perm = lh;
1087         }
1088 
1089         /* Add the new entry after the last permanent entry if any, so that
1090          * the new entry does not override a permanent entry when matched with
1091          * a wild-card protocol. But it is allowed to override any existing
1092          * non-permanent entry.  This means that when we remove this entry, the
1093          * system automatically returns to the old behavior.
1094          */
1095         list_add_rcu(&p->list, last_perm);
1096 out:
1097         spin_unlock_bh(&inetsw_lock);
1098 
1099         return;
1100 
1101 out_permanent:
1102         pr_err("Attempt to override permanent protocol %d\n", protocol);
1103         goto out;
1104 
1105 out_illegal:
1106         pr_err("Ignoring attempt to register invalid socket type %d\n",
1107                p->type);
1108         goto out;
1109 }
1110 EXPORT_SYMBOL(inet_register_protosw);
1111 
1112 void inet_unregister_protosw(struct inet_protosw *p)
1113 {
1114         if (INET_PROTOSW_PERMANENT & p->flags) {
1115                 pr_err("Attempt to unregister permanent protocol %d\n",
1116                        p->protocol);
1117         } else {
1118                 spin_lock_bh(&inetsw_lock);
1119                 list_del_rcu(&p->list);
1120                 spin_unlock_bh(&inetsw_lock);
1121 
1122                 synchronize_net();
1123         }
1124 }
1125 EXPORT_SYMBOL(inet_unregister_protosw);
1126 
1127 static int inet_sk_reselect_saddr(struct sock *sk)
1128 {
1129         struct inet_sock *inet = inet_sk(sk);
1130         __be32 old_saddr = inet->inet_saddr;
1131         __be32 daddr = inet->inet_daddr;
1132         struct flowi4 *fl4;
1133         struct rtable *rt;
1134         __be32 new_saddr;
1135         struct ip_options_rcu *inet_opt;
1136 
1137         inet_opt = rcu_dereference_protected(inet->inet_opt,
1138                                              lockdep_sock_is_held(sk));
1139         if (inet_opt && inet_opt->opt.srr)
1140                 daddr = inet_opt->opt.faddr;
1141 
1142         /* Query new route. */
1143         fl4 = &inet->cork.fl.u.ip4;
1144         rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk),
1145                               sk->sk_bound_dev_if, sk->sk_protocol,
1146                               inet->inet_sport, inet->inet_dport, sk);
1147         if (IS_ERR(rt))
1148                 return PTR_ERR(rt);
1149 
1150         sk_setup_caps(sk, &rt->dst);
1151 
1152         new_saddr = fl4->saddr;
1153 
1154         if (new_saddr == old_saddr)
1155                 return 0;
1156 
1157         if (sock_net(sk)->ipv4.sysctl_ip_dynaddr > 1) {
1158                 pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n",
1159                         __func__, &old_saddr, &new_saddr);
1160         }
1161 
1162         inet->inet_saddr = inet->inet_rcv_saddr = new_saddr;
1163 
1164         /*
1165          * XXX The only one ugly spot where we need to
1166          * XXX really change the sockets identity after
1167          * XXX it has entered the hashes. -DaveM
1168          *
1169          * Besides that, it does not check for connection
1170          * uniqueness. Wait for troubles.
1171          */
1172         return __sk_prot_rehash(sk);
1173 }
1174 
1175 int inet_sk_rebuild_header(struct sock *sk)
1176 {
1177         struct inet_sock *inet = inet_sk(sk);
1178         struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0);
1179         __be32 daddr;
1180         struct ip_options_rcu *inet_opt;
1181         struct flowi4 *fl4;
1182         int err;
1183 
1184         /* Route is OK, nothing to do. */
1185         if (rt)
1186                 return 0;
1187 
1188         /* Reroute. */
1189         rcu_read_lock();
1190         inet_opt = rcu_dereference(inet->inet_opt);
1191         daddr = inet->inet_daddr;
1192         if (inet_opt && inet_opt->opt.srr)
1193                 daddr = inet_opt->opt.faddr;
1194         rcu_read_unlock();
1195         fl4 = &inet->cork.fl.u.ip4;
1196         rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr,
1197                                    inet->inet_dport, inet->inet_sport,
1198                                    sk->sk_protocol, RT_CONN_FLAGS(sk),
1199                                    sk->sk_bound_dev_if);
1200         if (!IS_ERR(rt)) {
1201                 err = 0;
1202                 sk_setup_caps(sk, &rt->dst);
1203         } else {
1204                 err = PTR_ERR(rt);
1205 
1206                 /* Routing failed... */
1207                 sk->sk_route_caps = 0;
1208                 /*
1209                  * Other protocols have to map its equivalent state to TCP_SYN_SENT.
1210                  * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme
1211                  */
1212                 if (!sock_net(sk)->ipv4.sysctl_ip_dynaddr ||
1213                     sk->sk_state != TCP_SYN_SENT ||
1214                     (sk->sk_userlocks & SOCK_BINDADDR_LOCK) ||
1215                     (err = inet_sk_reselect_saddr(sk)) != 0)
1216                         sk->sk_err_soft = -err;
1217         }
1218 
1219         return err;
1220 }
1221 EXPORT_SYMBOL(inet_sk_rebuild_header);
1222 
1223 struct sk_buff *inet_gso_segment(struct sk_buff *skb,
1224                                  netdev_features_t features)
1225 {
1226         bool udpfrag = false, fixedid = false, gso_partial, encap;
1227         struct sk_buff *segs = ERR_PTR(-EINVAL);
1228         const struct net_offload *ops;
1229         unsigned int offset = 0;
1230         struct iphdr *iph;
1231         int proto, tot_len;
1232         int nhoff;
1233         int ihl;
1234         int id;
1235 
1236         skb_reset_network_header(skb);
1237         nhoff = skb_network_header(skb) - skb_mac_header(skb);
1238         if (unlikely(!pskb_may_pull(skb, sizeof(*iph))))
1239                 goto out;
1240 
1241         iph = ip_hdr(skb);
1242         ihl = iph->ihl * 4;
1243         if (ihl < sizeof(*iph))
1244                 goto out;
1245 
1246         id = ntohs(iph->id);
1247         proto = iph->protocol;
1248 
1249         /* Warning: after this point, iph might be no longer valid */
1250         if (unlikely(!pskb_may_pull(skb, ihl)))
1251                 goto out;
1252         __skb_pull(skb, ihl);
1253 
1254         encap = SKB_GSO_CB(skb)->encap_level > 0;
1255         if (encap)
1256                 features &= skb->dev->hw_enc_features;
1257         SKB_GSO_CB(skb)->encap_level += ihl;
1258 
1259         skb_reset_transport_header(skb);
1260 
1261         segs = ERR_PTR(-EPROTONOSUPPORT);
1262 
1263         if (!skb->encapsulation || encap) {
1264                 udpfrag = !!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP);
1265                 fixedid = !!(skb_shinfo(skb)->gso_type & SKB_GSO_TCP_FIXEDID);
1266 
1267                 /* fixed ID is invalid if DF bit is not set */
1268                 if (fixedid && !(ip_hdr(skb)->frag_off & htons(IP_DF)))
1269                         goto out;
1270         }
1271 
1272         ops = rcu_dereference(inet_offloads[proto]);
1273         if (likely(ops && ops->callbacks.gso_segment))
1274                 segs = ops->callbacks.gso_segment(skb, features);
1275 
1276         if (IS_ERR_OR_NULL(segs))
1277                 goto out;
1278 
1279         gso_partial = !!(skb_shinfo(segs)->gso_type & SKB_GSO_PARTIAL);
1280 
1281         skb = segs;
1282         do {
1283                 iph = (struct iphdr *)(skb_mac_header(skb) + nhoff);
1284                 if (udpfrag) {
1285                         iph->frag_off = htons(offset >> 3);
1286                         if (skb->next)
1287                                 iph->frag_off |= htons(IP_MF);
1288                         offset += skb->len - nhoff - ihl;
1289                         tot_len = skb->len - nhoff;
1290                 } else if (skb_is_gso(skb)) {
1291                         if (!fixedid) {
1292                                 iph->id = htons(id);
1293                                 id += skb_shinfo(skb)->gso_segs;
1294                         }
1295 
1296                         if (gso_partial)
1297                                 tot_len = skb_shinfo(skb)->gso_size +
1298                                           SKB_GSO_CB(skb)->data_offset +
1299                                           skb->head - (unsigned char *)iph;
1300                         else
1301                                 tot_len = skb->len - nhoff;
1302                 } else {
1303                         if (!fixedid)
1304                                 iph->id = htons(id++);
1305                         tot_len = skb->len - nhoff;
1306                 }
1307                 iph->tot_len = htons(tot_len);
1308                 ip_send_check(iph);
1309                 if (encap)
1310                         skb_reset_inner_headers(skb);
1311                 skb->network_header = (u8 *)iph - skb->head;
1312         } while ((skb = skb->next));
1313 
1314 out:
1315         return segs;
1316 }
1317 EXPORT_SYMBOL(inet_gso_segment);
1318 
1319 struct sk_buff **inet_gro_receive(struct sk_buff **head, struct sk_buff *skb)
1320 {
1321         const struct net_offload *ops;
1322         struct sk_buff **pp = NULL;
1323         struct sk_buff *p;
1324         const struct iphdr *iph;
1325         unsigned int hlen;
1326         unsigned int off;
1327         unsigned int id;
1328         int flush = 1;
1329         int proto;
1330 
1331         off = skb_gro_offset(skb);
1332         hlen = off + sizeof(*iph);
1333         iph = skb_gro_header_fast(skb, off);
1334         if (skb_gro_header_hard(skb, hlen)) {
1335                 iph = skb_gro_header_slow(skb, hlen, off);
1336                 if (unlikely(!iph))
1337                         goto out;
1338         }
1339 
1340         proto = iph->protocol;
1341 
1342         rcu_read_lock();
1343         ops = rcu_dereference(inet_offloads[proto]);
1344         if (!ops || !ops->callbacks.gro_receive)
1345                 goto out_unlock;
1346 
1347         if (*(u8 *)iph != 0x45)
1348                 goto out_unlock;
1349 
1350         if (ip_is_fragment(iph))
1351                 goto out_unlock;
1352 
1353         if (unlikely(ip_fast_csum((u8 *)iph, 5)))
1354                 goto out_unlock;
1355 
1356         id = ntohl(*(__be32 *)&iph->id);
1357         flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF));
1358         id >>= 16;
1359 
1360         for (p = *head; p; p = p->next) {
1361                 struct iphdr *iph2;
1362                 u16 flush_id;
1363 
1364                 if (!NAPI_GRO_CB(p)->same_flow)
1365                         continue;
1366 
1367                 iph2 = (struct iphdr *)(p->data + off);
1368                 /* The above works because, with the exception of the top
1369                  * (inner most) layer, we only aggregate pkts with the same
1370                  * hdr length so all the hdrs we'll need to verify will start
1371                  * at the same offset.
1372                  */
1373                 if ((iph->protocol ^ iph2->protocol) |
1374                     ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) |
1375                     ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) {
1376                         NAPI_GRO_CB(p)->same_flow = 0;
1377                         continue;
1378                 }
1379 
1380                 /* All fields must match except length and checksum. */
1381                 NAPI_GRO_CB(p)->flush |=
1382                         (iph->ttl ^ iph2->ttl) |
1383                         (iph->tos ^ iph2->tos) |
1384                         ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF));
1385 
1386                 NAPI_GRO_CB(p)->flush |= flush;
1387 
1388                 /* We need to store of the IP ID check to be included later
1389                  * when we can verify that this packet does in fact belong
1390                  * to a given flow.
1391                  */
1392                 flush_id = (u16)(id - ntohs(iph2->id));
1393 
1394                 /* This bit of code makes it much easier for us to identify
1395                  * the cases where we are doing atomic vs non-atomic IP ID
1396                  * checks.  Specifically an atomic check can return IP ID
1397                  * values 0 - 0xFFFF, while a non-atomic check can only
1398                  * return 0 or 0xFFFF.
1399                  */
1400                 if (!NAPI_GRO_CB(p)->is_atomic ||
1401                     !(iph->frag_off & htons(IP_DF))) {
1402                         flush_id ^= NAPI_GRO_CB(p)->count;
1403                         flush_id = flush_id ? 0xFFFF : 0;
1404                 }
1405 
1406                 /* If the previous IP ID value was based on an atomic
1407                  * datagram we can overwrite the value and ignore it.
1408                  */
1409                 if (NAPI_GRO_CB(skb)->is_atomic)
1410                         NAPI_GRO_CB(p)->flush_id = flush_id;
1411                 else
1412                         NAPI_GRO_CB(p)->flush_id |= flush_id;
1413         }
1414 
1415         NAPI_GRO_CB(skb)->is_atomic = !!(iph->frag_off & htons(IP_DF));
1416         NAPI_GRO_CB(skb)->flush |= flush;
1417         skb_set_network_header(skb, off);
1418         /* The above will be needed by the transport layer if there is one
1419          * immediately following this IP hdr.
1420          */
1421 
1422         /* Note : No need to call skb_gro_postpull_rcsum() here,
1423          * as we already checked checksum over ipv4 header was 0
1424          */
1425         skb_gro_pull(skb, sizeof(*iph));
1426         skb_set_transport_header(skb, skb_gro_offset(skb));
1427 
1428         pp = call_gro_receive(ops->callbacks.gro_receive, head, skb);
1429 
1430 out_unlock:
1431         rcu_read_unlock();
1432 
1433 out:
1434         skb_gro_flush_final(skb, pp, flush);
1435 
1436         return pp;
1437 }
1438 EXPORT_SYMBOL(inet_gro_receive);
1439 
1440 static struct sk_buff **ipip_gro_receive(struct sk_buff **head,
1441                                          struct sk_buff *skb)
1442 {
1443         if (NAPI_GRO_CB(skb)->encap_mark) {
1444                 NAPI_GRO_CB(skb)->flush = 1;
1445                 return NULL;
1446         }
1447 
1448         NAPI_GRO_CB(skb)->encap_mark = 1;
1449 
1450         return inet_gro_receive(head, skb);
1451 }
1452 
1453 #define SECONDS_PER_DAY 86400
1454 
1455 /* inet_current_timestamp - Return IP network timestamp
1456  *
1457  * Return milliseconds since midnight in network byte order.
1458  */
1459 __be32 inet_current_timestamp(void)
1460 {
1461         u32 secs;
1462         u32 msecs;
1463         struct timespec64 ts;
1464 
1465         ktime_get_real_ts64(&ts);
1466 
1467         /* Get secs since midnight. */
1468         (void)div_u64_rem(ts.tv_sec, SECONDS_PER_DAY, &secs);
1469         /* Convert to msecs. */
1470         msecs = secs * MSEC_PER_SEC;
1471         /* Convert nsec to msec. */
1472         msecs += (u32)ts.tv_nsec / NSEC_PER_MSEC;
1473 
1474         /* Convert to network byte order. */
1475         return htonl(msecs);
1476 }
1477 EXPORT_SYMBOL(inet_current_timestamp);
1478 
1479 int inet_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
1480 {
1481         if (sk->sk_family == AF_INET)
1482                 return ip_recv_error(sk, msg, len, addr_len);
1483 #if IS_ENABLED(CONFIG_IPV6)
1484         if (sk->sk_family == AF_INET6)
1485                 return pingv6_ops.ipv6_recv_error(sk, msg, len, addr_len);
1486 #endif
1487         return -EINVAL;
1488 }
1489 
1490 int inet_gro_complete(struct sk_buff *skb, int nhoff)
1491 {
1492         __be16 newlen = htons(skb->len - nhoff);
1493         struct iphdr *iph = (struct iphdr *)(skb->data + nhoff);
1494         const struct net_offload *ops;
1495         int proto = iph->protocol;
1496         int err = -ENOSYS;
1497 
1498         if (skb->encapsulation) {
1499                 skb_set_inner_protocol(skb, cpu_to_be16(ETH_P_IP));
1500                 skb_set_inner_network_header(skb, nhoff);
1501         }
1502 
1503         csum_replace2(&iph->check, iph->tot_len, newlen);
1504         iph->tot_len = newlen;
1505 
1506         rcu_read_lock();
1507         ops = rcu_dereference(inet_offloads[proto]);
1508         if (WARN_ON(!ops || !ops->callbacks.gro_complete))
1509                 goto out_unlock;
1510 
1511         /* Only need to add sizeof(*iph) to get to the next hdr below
1512          * because any hdr with option will have been flushed in
1513          * inet_gro_receive().
1514          */
1515         err = ops->callbacks.gro_complete(skb, nhoff + sizeof(*iph));
1516 
1517 out_unlock:
1518         rcu_read_unlock();
1519 
1520         return err;
1521 }
1522 EXPORT_SYMBOL(inet_gro_complete);
1523 
1524 static int ipip_gro_complete(struct sk_buff *skb, int nhoff)
1525 {
1526         skb->encapsulation = 1;
1527         skb_shinfo(skb)->gso_type |= SKB_GSO_IPXIP4;
1528         return inet_gro_complete(skb, nhoff);
1529 }
1530 
1531 int inet_ctl_sock_create(struct sock **sk, unsigned short family,
1532                          unsigned short type, unsigned char protocol,
1533                          struct net *net)
1534 {
1535         struct socket *sock;
1536         int rc = sock_create_kern(net, family, type, protocol, &sock);
1537 
1538         if (rc == 0) {
1539                 *sk = sock->sk;
1540                 (*sk)->sk_allocation = GFP_ATOMIC;
1541                 /*
1542                  * Unhash it so that IP input processing does not even see it,
1543                  * we do not wish this socket to see incoming packets.
1544                  */
1545                 (*sk)->sk_prot->unhash(*sk);
1546         }
1547         return rc;
1548 }
1549 EXPORT_SYMBOL_GPL(inet_ctl_sock_create);
1550 
1551 u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offt)
1552 {
1553         return  *(((unsigned long *)per_cpu_ptr(mib, cpu)) + offt);
1554 }
1555 EXPORT_SYMBOL_GPL(snmp_get_cpu_field);
1556 
1557 unsigned long snmp_fold_field(void __percpu *mib, int offt)
1558 {
1559         unsigned long res = 0;
1560         int i;
1561 
1562         for_each_possible_cpu(i)
1563                 res += snmp_get_cpu_field(mib, i, offt);
1564         return res;
1565 }
1566 EXPORT_SYMBOL_GPL(snmp_fold_field);
1567 
1568 #if BITS_PER_LONG==32
1569 
1570 u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offt,
1571                          size_t syncp_offset)
1572 {
1573         void *bhptr;
1574         struct u64_stats_sync *syncp;
1575         u64 v;
1576         unsigned int start;
1577 
1578         bhptr = per_cpu_ptr(mib, cpu);
1579         syncp = (struct u64_stats_sync *)(bhptr + syncp_offset);
1580         do {
1581                 start = u64_stats_fetch_begin_irq(syncp);
1582                 v = *(((u64 *)bhptr) + offt);
1583         } while (u64_stats_fetch_retry_irq(syncp, start));
1584 
1585         return v;
1586 }
1587 EXPORT_SYMBOL_GPL(snmp_get_cpu_field64);
1588 
1589 u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset)
1590 {
1591         u64 res = 0;
1592         int cpu;
1593 
1594         for_each_possible_cpu(cpu) {
1595                 res += snmp_get_cpu_field64(mib, cpu, offt, syncp_offset);
1596         }
1597         return res;
1598 }
1599 EXPORT_SYMBOL_GPL(snmp_fold_field64);
1600 #endif
1601 
1602 #ifdef CONFIG_IP_MULTICAST
1603 static const struct net_protocol igmp_protocol = {
1604         .handler =      igmp_rcv,
1605         .netns_ok =     1,
1606 };
1607 #endif
1608 
1609 /* thinking of making this const? Don't.
1610  * early_demux can change based on sysctl.
1611  */
1612 static struct net_protocol tcp_protocol = {
1613         .early_demux    =       tcp_v4_early_demux,
1614         .early_demux_handler =  tcp_v4_early_demux,
1615         .handler        =       tcp_v4_rcv,
1616         .err_handler    =       tcp_v4_err,
1617         .no_policy      =       1,
1618         .netns_ok       =       1,
1619         .icmp_strict_tag_validation = 1,
1620 };
1621 
1622 /* thinking of making this const? Don't.
1623  * early_demux can change based on sysctl.
1624  */
1625 static struct net_protocol udp_protocol = {
1626         .early_demux =  udp_v4_early_demux,
1627         .early_demux_handler =  udp_v4_early_demux,
1628         .handler =      udp_rcv,
1629         .err_handler =  udp_err,
1630         .no_policy =    1,
1631         .netns_ok =     1,
1632 };
1633 
1634 static const struct net_protocol icmp_protocol = {
1635         .handler =      icmp_rcv,
1636         .err_handler =  icmp_err,
1637         .no_policy =    1,
1638         .netns_ok =     1,
1639 };
1640 
1641 static __net_init int ipv4_mib_init_net(struct net *net)
1642 {
1643         int i;
1644 
1645         net->mib.tcp_statistics = alloc_percpu(struct tcp_mib);
1646         if (!net->mib.tcp_statistics)
1647                 goto err_tcp_mib;
1648         net->mib.ip_statistics = alloc_percpu(struct ipstats_mib);
1649         if (!net->mib.ip_statistics)
1650                 goto err_ip_mib;
1651 
1652         for_each_possible_cpu(i) {
1653                 struct ipstats_mib *af_inet_stats;
1654                 af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i);
1655                 u64_stats_init(&af_inet_stats->syncp);
1656         }
1657 
1658         net->mib.net_statistics = alloc_percpu(struct linux_mib);
1659         if (!net->mib.net_statistics)
1660                 goto err_net_mib;
1661         net->mib.udp_statistics = alloc_percpu(struct udp_mib);
1662         if (!net->mib.udp_statistics)
1663                 goto err_udp_mib;
1664         net->mib.udplite_statistics = alloc_percpu(struct udp_mib);
1665         if (!net->mib.udplite_statistics)
1666                 goto err_udplite_mib;
1667         net->mib.icmp_statistics = alloc_percpu(struct icmp_mib);
1668         if (!net->mib.icmp_statistics)
1669                 goto err_icmp_mib;
1670         net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib),
1671                                               GFP_KERNEL);
1672         if (!net->mib.icmpmsg_statistics)
1673                 goto err_icmpmsg_mib;
1674 
1675         tcp_mib_init(net);
1676         return 0;
1677 
1678 err_icmpmsg_mib:
1679         free_percpu(net->mib.icmp_statistics);
1680 err_icmp_mib:
1681         free_percpu(net->mib.udplite_statistics);
1682 err_udplite_mib:
1683         free_percpu(net->mib.udp_statistics);
1684 err_udp_mib:
1685         free_percpu(net->mib.net_statistics);
1686 err_net_mib:
1687         free_percpu(net->mib.ip_statistics);
1688 err_ip_mib:
1689         free_percpu(net->mib.tcp_statistics);
1690 err_tcp_mib:
1691         return -ENOMEM;
1692 }
1693 
1694 static __net_exit void ipv4_mib_exit_net(struct net *net)
1695 {
1696         kfree(net->mib.icmpmsg_statistics);
1697         free_percpu(net->mib.icmp_statistics);
1698         free_percpu(net->mib.udplite_statistics);
1699         free_percpu(net->mib.udp_statistics);
1700         free_percpu(net->mib.net_statistics);
1701         free_percpu(net->mib.ip_statistics);
1702         free_percpu(net->mib.tcp_statistics);
1703 }
1704 
1705 static __net_initdata struct pernet_operations ipv4_mib_ops = {
1706         .init = ipv4_mib_init_net,
1707         .exit = ipv4_mib_exit_net,
1708 };
1709 
1710 static int __init init_ipv4_mibs(void)
1711 {
1712         return register_pernet_subsys(&ipv4_mib_ops);
1713 }
1714 
1715 static __net_init int inet_init_net(struct net *net)
1716 {
1717         /*
1718          * Set defaults for local port range
1719          */
1720         seqlock_init(&net->ipv4.ip_local_ports.lock);
1721         net->ipv4.ip_local_ports.range[0] =  32768;
1722         net->ipv4.ip_local_ports.range[1] =  60999;
1723 
1724         seqlock_init(&net->ipv4.ping_group_range.lock);
1725         /*
1726          * Sane defaults - nobody may create ping sockets.
1727          * Boot scripts should set this to distro-specific group.
1728          */
1729         net->ipv4.ping_group_range.range[0] = make_kgid(&init_user_ns, 1);
1730         net->ipv4.ping_group_range.range[1] = make_kgid(&init_user_ns, 0);
1731 
1732         /* Default values for sysctl-controlled parameters.
1733          * We set them here, in case sysctl is not compiled.
1734          */
1735         net->ipv4.sysctl_ip_default_ttl = IPDEFTTL;
1736         net->ipv4.sysctl_ip_dynaddr = 0;
1737         net->ipv4.sysctl_ip_early_demux = 1;
1738         net->ipv4.sysctl_udp_early_demux = 1;
1739         net->ipv4.sysctl_tcp_early_demux = 1;
1740 #ifdef CONFIG_SYSCTL
1741         net->ipv4.sysctl_ip_prot_sock = PROT_SOCK;
1742 #endif
1743 
1744         /* Some igmp sysctl, whose values are always used */
1745         net->ipv4.sysctl_igmp_max_memberships = 20;
1746         net->ipv4.sysctl_igmp_max_msf = 10;
1747         /* IGMP reports for link-local multicast groups are enabled by default */
1748         net->ipv4.sysctl_igmp_llm_reports = 1;
1749         net->ipv4.sysctl_igmp_qrv = 2;
1750 
1751         return 0;
1752 }
1753 
1754 static __net_exit void inet_exit_net(struct net *net)
1755 {
1756 }
1757 
1758 static __net_initdata struct pernet_operations af_inet_ops = {
1759         .init = inet_init_net,
1760         .exit = inet_exit_net,
1761 };
1762 
1763 static int __init init_inet_pernet_ops(void)
1764 {
1765         return register_pernet_subsys(&af_inet_ops);
1766 }
1767 
1768 static int ipv4_proc_init(void);
1769 
1770 /*
1771  *      IP protocol layer initialiser
1772  */
1773 
1774 static struct packet_offload ip_packet_offload __read_mostly = {
1775         .type = cpu_to_be16(ETH_P_IP),
1776         .callbacks = {
1777                 .gso_segment = inet_gso_segment,
1778                 .gro_receive = inet_gro_receive,
1779                 .gro_complete = inet_gro_complete,
1780         },
1781 };
1782 
1783 static const struct net_offload ipip_offload = {
1784         .callbacks = {
1785                 .gso_segment    = inet_gso_segment,
1786                 .gro_receive    = ipip_gro_receive,
1787                 .gro_complete   = ipip_gro_complete,
1788         },
1789 };
1790 
1791 static int __init ipip_offload_init(void)
1792 {
1793         return inet_add_offload(&ipip_offload, IPPROTO_IPIP);
1794 }
1795 
1796 static int __init ipv4_offload_init(void)
1797 {
1798         /*
1799          * Add offloads
1800          */
1801         if (udpv4_offload_init() < 0)
1802                 pr_crit("%s: Cannot add UDP protocol offload\n", __func__);
1803         if (tcpv4_offload_init() < 0)
1804                 pr_crit("%s: Cannot add TCP protocol offload\n", __func__);
1805         if (ipip_offload_init() < 0)
1806                 pr_crit("%s: Cannot add IPIP protocol offload\n", __func__);
1807 
1808         dev_add_offload(&ip_packet_offload);
1809         return 0;
1810 }
1811 
1812 fs_initcall(ipv4_offload_init);
1813 
1814 static struct packet_type ip_packet_type __read_mostly = {
1815         .type = cpu_to_be16(ETH_P_IP),
1816         .func = ip_rcv,
1817 };
1818 
1819 static int __init inet_init(void)
1820 {
1821         struct inet_protosw *q;
1822         struct list_head *r;
1823         int rc = -EINVAL;
1824 
1825         sock_skb_cb_check_size(sizeof(struct inet_skb_parm));
1826 
1827         rc = proto_register(&tcp_prot, 1);
1828         if (rc)
1829                 goto out;
1830 
1831         rc = proto_register(&udp_prot, 1);
1832         if (rc)
1833                 goto out_unregister_tcp_proto;
1834 
1835         rc = proto_register(&raw_prot, 1);
1836         if (rc)
1837                 goto out_unregister_udp_proto;
1838 
1839         rc = proto_register(&ping_prot, 1);
1840         if (rc)
1841                 goto out_unregister_raw_proto;
1842 
1843         /*
1844          *      Tell SOCKET that we are alive...
1845          */
1846 
1847         (void)sock_register(&inet_family_ops);
1848 
1849 #ifdef CONFIG_SYSCTL
1850         ip_static_sysctl_init();
1851 #endif
1852 
1853         /*
1854          *      Add all the base protocols.
1855          */
1856 
1857         if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0)
1858                 pr_crit("%s: Cannot add ICMP protocol\n", __func__);
1859         if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0)
1860                 pr_crit("%s: Cannot add UDP protocol\n", __func__);
1861         if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0)
1862                 pr_crit("%s: Cannot add TCP protocol\n", __func__);
1863 #ifdef CONFIG_IP_MULTICAST
1864         if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0)
1865                 pr_crit("%s: Cannot add IGMP protocol\n", __func__);
1866 #endif
1867 
1868         /* Register the socket-side information for inet_create. */
1869         for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r)
1870                 INIT_LIST_HEAD(r);
1871 
1872         for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q)
1873                 inet_register_protosw(q);
1874 
1875         /*
1876          *      Set the ARP module up
1877          */
1878 
1879         arp_init();
1880 
1881         /*
1882          *      Set the IP module up
1883          */
1884 
1885         ip_init();
1886 
1887         /* Setup TCP slab cache for open requests. */
1888         tcp_init();
1889 
1890         /* Setup UDP memory threshold */
1891         udp_init();
1892 
1893         /* Add UDP-Lite (RFC 3828) */
1894         udplite4_register();
1895 
1896         ping_init();
1897 
1898         /*
1899          *      Set the ICMP layer up
1900          */
1901 
1902         if (icmp_init() < 0)
1903                 panic("Failed to create the ICMP control socket.\n");
1904 
1905         /*
1906          *      Initialise the multicast router
1907          */
1908 #if defined(CONFIG_IP_MROUTE)
1909         if (ip_mr_init())
1910                 pr_crit("%s: Cannot init ipv4 mroute\n", __func__);
1911 #endif
1912 
1913         if (init_inet_pernet_ops())
1914                 pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__);
1915         /*
1916          *      Initialise per-cpu ipv4 mibs
1917          */
1918 
1919         if (init_ipv4_mibs())
1920                 pr_crit("%s: Cannot init ipv4 mibs\n", __func__);
1921 
1922         ipv4_proc_init();
1923 
1924         ipfrag_init();
1925 
1926         dev_add_pack(&ip_packet_type);
1927 
1928         ip_tunnel_core_init();
1929 
1930         rc = 0;
1931 out:
1932         return rc;
1933 out_unregister_raw_proto:
1934         proto_unregister(&raw_prot);
1935 out_unregister_udp_proto:
1936         proto_unregister(&udp_prot);
1937 out_unregister_tcp_proto:
1938         proto_unregister(&tcp_prot);
1939         goto out;
1940 }
1941 
1942 fs_initcall(inet_init);
1943 
1944 /* ------------------------------------------------------------------------ */
1945 
1946 #ifdef CONFIG_PROC_FS
1947 static int __init ipv4_proc_init(void)
1948 {
1949         int rc = 0;
1950 
1951         if (raw_proc_init())
1952                 goto out_raw;
1953         if (tcp4_proc_init())
1954                 goto out_tcp;
1955         if (udp4_proc_init())
1956                 goto out_udp;
1957         if (ping_proc_init())
1958                 goto out_ping;
1959         if (ip_misc_proc_init())
1960                 goto out_misc;
1961 out:
1962         return rc;
1963 out_misc:
1964         ping_proc_exit();
1965 out_ping:
1966         udp4_proc_exit();
1967 out_udp:
1968         tcp4_proc_exit();
1969 out_tcp:
1970         raw_proc_exit();
1971 out_raw:
1972         rc = -ENOMEM;
1973         goto out;
1974 }
1975 
1976 #else /* CONFIG_PROC_FS */
1977 static int __init ipv4_proc_init(void)
1978 {
1979         return 0;
1980 }
1981 #endif /* CONFIG_PROC_FS */
1982 

~ [ 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