Commit | Line | Data |
---|---|---|
1da177e4 LT |
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 | * | |
02c30a84 | 8 | * Authors: Ross Biro |
1da177e4 LT |
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 | ||
afd46503 JP |
68 | #define pr_fmt(fmt) "IPv4: " fmt |
69 | ||
f4c50d99 | 70 | #include <linux/err.h> |
1da177e4 LT |
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> | |
1da177e4 LT |
76 | #include <linux/module.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> | |
4fc268d2 | 82 | #include <linux/capability.h> |
1da177e4 LT |
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> | |
b3da2cf3 | 90 | #include <linux/random.h> |
5a0e3ad6 | 91 | #include <linux/slab.h> |
1da177e4 LT |
92 | |
93 | #include <asm/uaccess.h> | |
1da177e4 | 94 | |
1da177e4 LT |
95 | #include <linux/inet.h> |
96 | #include <linux/igmp.h> | |
14c85021 | 97 | #include <linux/inetdevice.h> |
1da177e4 | 98 | #include <linux/netdevice.h> |
73cc19f1 | 99 | #include <net/checksum.h> |
1da177e4 LT |
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> | |
295f7324 | 105 | #include <net/inet_connection_sock.h> |
1da177e4 LT |
106 | #include <net/tcp.h> |
107 | #include <net/udp.h> | |
ba4e58ec | 108 | #include <net/udplite.h> |
c319b4d7 | 109 | #include <net/ping.h> |
1da177e4 LT |
110 | #include <linux/skbuff.h> |
111 | #include <net/sock.h> | |
112 | #include <net/raw.h> | |
113 | #include <net/icmp.h> | |
1da177e4 LT |
114 | #include <net/inet_common.h> |
115 | #include <net/xfrm.h> | |
9b4661bd | 116 | #include <net/net_namespace.h> |
aebda156 | 117 | #include <net/secure_seq.h> |
1da177e4 LT |
118 | #ifdef CONFIG_IP_MROUTE |
119 | #include <linux/mroute.h> | |
120 | #endif | |
121 | ||
1da177e4 LT |
122 | |
123 | /* The inetsw table contains everything that inet_create needs to | |
124 | * build a new socket. | |
125 | */ | |
126 | static struct list_head inetsw[SOCK_MAX]; | |
127 | static DEFINE_SPINLOCK(inetsw_lock); | |
128 | ||
129 | /* New destruction routine */ | |
130 | ||
131 | void inet_sock_destruct(struct sock *sk) | |
132 | { | |
133 | struct inet_sock *inet = inet_sk(sk); | |
134 | ||
135 | __skb_queue_purge(&sk->sk_receive_queue); | |
136 | __skb_queue_purge(&sk->sk_error_queue); | |
137 | ||
95766fff HA |
138 | sk_mem_reclaim(sk); |
139 | ||
1da177e4 | 140 | if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) { |
3d1427f8 | 141 | pr_err("Attempt to release TCP socket in state %d %p\n", |
1da177e4 LT |
142 | sk->sk_state, sk); |
143 | return; | |
144 | } | |
145 | if (!sock_flag(sk, SOCK_DEAD)) { | |
3d1427f8 | 146 | pr_err("Attempt to release alive inet socket %p\n", sk); |
1da177e4 LT |
147 | return; |
148 | } | |
149 | ||
547b792c IJ |
150 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); |
151 | WARN_ON(atomic_read(&sk->sk_wmem_alloc)); | |
152 | WARN_ON(sk->sk_wmem_queued); | |
153 | WARN_ON(sk->sk_forward_alloc); | |
1da177e4 | 154 | |
f6d8bd05 | 155 | kfree(rcu_dereference_protected(inet->inet_opt, 1)); |
b6c6712a | 156 | dst_release(rcu_dereference_check(sk->sk_dst_cache, 1)); |
41063e9d | 157 | dst_release(sk->sk_rx_dst); |
e6848976 | 158 | sk_refcnt_debug_dec(sk); |
1da177e4 | 159 | } |
3d1427f8 | 160 | EXPORT_SYMBOL(inet_sock_destruct); |
1da177e4 LT |
161 | |
162 | /* | |
163 | * The routines beyond this point handle the behaviour of an AF_INET | |
164 | * socket object. Mostly it punts to the subprotocols of IP to do | |
165 | * the work. | |
166 | */ | |
167 | ||
168 | /* | |
169 | * Automatically bind an unbound socket. | |
170 | */ | |
171 | ||
172 | static int inet_autobind(struct sock *sk) | |
173 | { | |
174 | struct inet_sock *inet; | |
175 | /* We may need to bind the socket. */ | |
176 | lock_sock(sk); | |
177 | inet = inet_sk(sk); | |
c720c7e8 | 178 | if (!inet->inet_num) { |
1da177e4 LT |
179 | if (sk->sk_prot->get_port(sk, 0)) { |
180 | release_sock(sk); | |
181 | return -EAGAIN; | |
182 | } | |
c720c7e8 | 183 | inet->inet_sport = htons(inet->inet_num); |
1da177e4 LT |
184 | } |
185 | release_sock(sk); | |
186 | return 0; | |
187 | } | |
188 | ||
189 | /* | |
190 | * Move a socket into listening state. | |
191 | */ | |
192 | int inet_listen(struct socket *sock, int backlog) | |
193 | { | |
194 | struct sock *sk = sock->sk; | |
195 | unsigned char old_state; | |
196 | int err; | |
197 | ||
198 | lock_sock(sk); | |
199 | ||
200 | err = -EINVAL; | |
201 | if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM) | |
202 | goto out; | |
203 | ||
204 | old_state = sk->sk_state; | |
205 | if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN))) | |
206 | goto out; | |
207 | ||
208 | /* Really, if the socket is already in listen state | |
209 | * we can only allow the backlog to be adjusted. | |
210 | */ | |
211 | if (old_state != TCP_LISTEN) { | |
8336886f JC |
212 | /* Check special setups for testing purpose to enable TFO w/o |
213 | * requiring TCP_FASTOPEN sockopt. | |
214 | * Note that only TCP sockets (SOCK_STREAM) will reach here. | |
215 | * Also fastopenq may already been allocated because this | |
216 | * socket was in TCP_LISTEN state previously but was | |
217 | * shutdown() (rather than close()). | |
218 | */ | |
219 | if ((sysctl_tcp_fastopen & TFO_SERVER_ENABLE) != 0 && | |
220 | inet_csk(sk)->icsk_accept_queue.fastopenq == NULL) { | |
221 | if ((sysctl_tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) != 0) | |
222 | err = fastopen_init_queue(sk, backlog); | |
223 | else if ((sysctl_tcp_fastopen & | |
224 | TFO_SERVER_WO_SOCKOPT2) != 0) | |
225 | err = fastopen_init_queue(sk, | |
226 | ((uint)sysctl_tcp_fastopen) >> 16); | |
227 | else | |
228 | err = 0; | |
229 | if (err) | |
230 | goto out; | |
231 | } | |
72a3effa | 232 | err = inet_csk_listen_start(sk, backlog); |
1da177e4 LT |
233 | if (err) |
234 | goto out; | |
235 | } | |
236 | sk->sk_max_ack_backlog = backlog; | |
237 | err = 0; | |
238 | ||
239 | out: | |
240 | release_sock(sk); | |
241 | return err; | |
242 | } | |
3d1427f8 | 243 | EXPORT_SYMBOL(inet_listen); |
1da177e4 LT |
244 | |
245 | /* | |
246 | * Create an inet socket. | |
247 | */ | |
248 | ||
3f378b68 EP |
249 | static int inet_create(struct net *net, struct socket *sock, int protocol, |
250 | int kern) | |
1da177e4 LT |
251 | { |
252 | struct sock *sk; | |
1da177e4 LT |
253 | struct inet_protosw *answer; |
254 | struct inet_sock *inet; | |
255 | struct proto *answer_prot; | |
256 | unsigned char answer_flags; | |
bb97d31f | 257 | int try_loading_module = 0; |
86c8f9d1 | 258 | int err; |
1da177e4 LT |
259 | |
260 | sock->state = SS_UNCONNECTED; | |
261 | ||
262 | /* Look for the requested type/protocol pair. */ | |
bb97d31f | 263 | lookup_protocol: |
86c8f9d1 | 264 | err = -ESOCKTNOSUPPORT; |
1da177e4 | 265 | rcu_read_lock(); |
696adfe8 | 266 | list_for_each_entry_rcu(answer, &inetsw[sock->type], list) { |
1da177e4 | 267 | |
696adfe8 | 268 | err = 0; |
1da177e4 LT |
269 | /* Check the non-wild match. */ |
270 | if (protocol == answer->protocol) { | |
271 | if (protocol != IPPROTO_IP) | |
272 | break; | |
273 | } else { | |
274 | /* Check for the two wild cases. */ | |
275 | if (IPPROTO_IP == protocol) { | |
276 | protocol = answer->protocol; | |
277 | break; | |
278 | } | |
279 | if (IPPROTO_IP == answer->protocol) | |
280 | break; | |
281 | } | |
86c8f9d1 | 282 | err = -EPROTONOSUPPORT; |
1da177e4 LT |
283 | } |
284 | ||
696adfe8 | 285 | if (unlikely(err)) { |
bb97d31f ACM |
286 | if (try_loading_module < 2) { |
287 | rcu_read_unlock(); | |
288 | /* | |
289 | * Be more specific, e.g. net-pf-2-proto-132-type-1 | |
290 | * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM) | |
291 | */ | |
292 | if (++try_loading_module == 1) | |
293 | request_module("net-pf-%d-proto-%d-type-%d", | |
294 | PF_INET, protocol, sock->type); | |
295 | /* | |
296 | * Fall back to generic, e.g. net-pf-2-proto-132 | |
297 | * (net-pf-PF_INET-proto-IPPROTO_SCTP) | |
298 | */ | |
299 | else | |
300 | request_module("net-pf-%d-proto-%d", | |
301 | PF_INET, protocol); | |
302 | goto lookup_protocol; | |
303 | } else | |
304 | goto out_rcu_unlock; | |
305 | } | |
306 | ||
1da177e4 | 307 | err = -EPERM; |
52e804c6 EB |
308 | if (sock->type == SOCK_RAW && !kern && |
309 | !ns_capable(net->user_ns, CAP_NET_RAW)) | |
1da177e4 | 310 | goto out_rcu_unlock; |
1da177e4 LT |
311 | |
312 | sock->ops = answer->ops; | |
313 | answer_prot = answer->prot; | |
1da177e4 LT |
314 | answer_flags = answer->flags; |
315 | rcu_read_unlock(); | |
316 | ||
547b792c | 317 | WARN_ON(answer_prot->slab == NULL); |
1da177e4 LT |
318 | |
319 | err = -ENOBUFS; | |
6257ff21 | 320 | sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot); |
1da177e4 LT |
321 | if (sk == NULL) |
322 | goto out; | |
323 | ||
324 | err = 0; | |
1da177e4 | 325 | if (INET_PROTOSW_REUSE & answer_flags) |
4a17fd52 | 326 | sk->sk_reuse = SK_CAN_REUSE; |
1da177e4 LT |
327 | |
328 | inet = inet_sk(sk); | |
469de9b9 | 329 | inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; |
1da177e4 | 330 | |
7b2ff18e JO |
331 | inet->nodefrag = 0; |
332 | ||
1da177e4 | 333 | if (SOCK_RAW == sock->type) { |
c720c7e8 | 334 | inet->inet_num = protocol; |
1da177e4 LT |
335 | if (IPPROTO_RAW == protocol) |
336 | inet->hdrincl = 1; | |
337 | } | |
338 | ||
974eda11 | 339 | if (net->ipv4.sysctl_ip_no_pmtu_disc) |
1da177e4 LT |
340 | inet->pmtudisc = IP_PMTUDISC_DONT; |
341 | else | |
342 | inet->pmtudisc = IP_PMTUDISC_WANT; | |
343 | ||
c720c7e8 | 344 | inet->inet_id = 0; |
1da177e4 LT |
345 | |
346 | sock_init_data(sock, sk); | |
347 | ||
348 | sk->sk_destruct = inet_sock_destruct; | |
1da177e4 LT |
349 | sk->sk_protocol = protocol; |
350 | sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv; | |
351 | ||
352 | inet->uc_ttl = -1; | |
353 | inet->mc_loop = 1; | |
354 | inet->mc_ttl = 1; | |
f771bef9 | 355 | inet->mc_all = 1; |
1da177e4 LT |
356 | inet->mc_index = 0; |
357 | inet->mc_list = NULL; | |
4c507d28 | 358 | inet->rcv_tos = 0; |
1da177e4 | 359 | |
e6848976 | 360 | sk_refcnt_debug_inc(sk); |
1da177e4 | 361 | |
c720c7e8 | 362 | if (inet->inet_num) { |
1da177e4 LT |
363 | /* It assumes that any protocol which allows |
364 | * the user to assign a number at socket | |
365 | * creation time automatically | |
366 | * shares. | |
367 | */ | |
c720c7e8 | 368 | inet->inet_sport = htons(inet->inet_num); |
1da177e4 LT |
369 | /* Add to protocol hash chains. */ |
370 | sk->sk_prot->hash(sk); | |
371 | } | |
372 | ||
373 | if (sk->sk_prot->init) { | |
374 | err = sk->sk_prot->init(sk); | |
375 | if (err) | |
376 | sk_common_release(sk); | |
377 | } | |
378 | out: | |
379 | return err; | |
380 | out_rcu_unlock: | |
381 | rcu_read_unlock(); | |
382 | goto out; | |
383 | } | |
384 | ||
385 | ||
386 | /* | |
387 | * The peer socket should always be NULL (or else). When we call this | |
388 | * function we are destroying the object and from then on nobody | |
389 | * should refer to it. | |
390 | */ | |
391 | int inet_release(struct socket *sock) | |
392 | { | |
393 | struct sock *sk = sock->sk; | |
394 | ||
395 | if (sk) { | |
396 | long timeout; | |
397 | ||
c58dc01b | 398 | sock_rps_reset_flow(sk); |
fec5e652 | 399 | |
1da177e4 LT |
400 | /* Applications forget to leave groups before exiting */ |
401 | ip_mc_drop_socket(sk); | |
402 | ||
403 | /* If linger is set, we don't return until the close | |
404 | * is complete. Otherwise we return immediately. The | |
405 | * actually closing is done the same either way. | |
406 | * | |
407 | * If the close is due to the process exiting, we never | |
408 | * linger.. | |
409 | */ | |
410 | timeout = 0; | |
411 | if (sock_flag(sk, SOCK_LINGER) && | |
412 | !(current->flags & PF_EXITING)) | |
413 | timeout = sk->sk_lingertime; | |
414 | sock->sk = NULL; | |
415 | sk->sk_prot->close(sk, timeout); | |
416 | } | |
417 | return 0; | |
418 | } | |
3d1427f8 | 419 | EXPORT_SYMBOL(inet_release); |
1da177e4 | 420 | |
1da177e4 LT |
421 | int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) |
422 | { | |
423 | struct sockaddr_in *addr = (struct sockaddr_in *)uaddr; | |
424 | struct sock *sk = sock->sk; | |
425 | struct inet_sock *inet = inet_sk(sk); | |
3594698a | 426 | struct net *net = sock_net(sk); |
1da177e4 LT |
427 | unsigned short snum; |
428 | int chk_addr_ret; | |
429 | int err; | |
430 | ||
431 | /* If the socket has its own bind function then use it. (RAW) */ | |
432 | if (sk->sk_prot->bind) { | |
433 | err = sk->sk_prot->bind(sk, uaddr, addr_len); | |
434 | goto out; | |
435 | } | |
436 | err = -EINVAL; | |
437 | if (addr_len < sizeof(struct sockaddr_in)) | |
438 | goto out; | |
439 | ||
c349a528 | 440 | if (addr->sin_family != AF_INET) { |
29c486df ED |
441 | /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET) |
442 | * only if s_addr is INADDR_ANY. | |
443 | */ | |
c349a528 | 444 | err = -EAFNOSUPPORT; |
29c486df ED |
445 | if (addr->sin_family != AF_UNSPEC || |
446 | addr->sin_addr.s_addr != htonl(INADDR_ANY)) | |
447 | goto out; | |
c349a528 | 448 | } |
d0733d2e | 449 | |
3594698a | 450 | chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr); |
1da177e4 LT |
451 | |
452 | /* Not specified by any standard per-se, however it breaks too | |
453 | * many applications when removed. It is unfortunate since | |
454 | * allowing applications to make a non-local bind solves | |
455 | * several problems with systems using dynamic addressing. | |
456 | * (ie. your servers still start up even if your ISDN link | |
457 | * is temporarily down) | |
458 | */ | |
459 | err = -EADDRNOTAVAIL; | |
49a60158 | 460 | if (!net->ipv4.sysctl_ip_nonlocal_bind && |
b9fb1506 | 461 | !(inet->freebind || inet->transparent) && |
e6f1cebf | 462 | addr->sin_addr.s_addr != htonl(INADDR_ANY) && |
1da177e4 LT |
463 | chk_addr_ret != RTN_LOCAL && |
464 | chk_addr_ret != RTN_MULTICAST && | |
465 | chk_addr_ret != RTN_BROADCAST) | |
466 | goto out; | |
467 | ||
468 | snum = ntohs(addr->sin_port); | |
469 | err = -EACCES; | |
3594698a EB |
470 | if (snum && snum < PROT_SOCK && |
471 | !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) | |
1da177e4 LT |
472 | goto out; |
473 | ||
474 | /* We keep a pair of addresses. rcv_saddr is the one | |
475 | * used by hash lookups, and saddr is used for transmit. | |
476 | * | |
477 | * In the BSD API these are the same except where it | |
478 | * would be illegal to use them (multicast/broadcast) in | |
479 | * which case the sending device address is used. | |
480 | */ | |
481 | lock_sock(sk); | |
482 | ||
483 | /* Check these errors (active socket, double bind). */ | |
484 | err = -EINVAL; | |
c720c7e8 | 485 | if (sk->sk_state != TCP_CLOSE || inet->inet_num) |
1da177e4 LT |
486 | goto out_release_sock; |
487 | ||
c720c7e8 | 488 | inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; |
1da177e4 | 489 | if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) |
c720c7e8 | 490 | inet->inet_saddr = 0; /* Use device */ |
1da177e4 LT |
491 | |
492 | /* Make sure we are allowed to bind here. */ | |
493 | if (sk->sk_prot->get_port(sk, snum)) { | |
c720c7e8 | 494 | inet->inet_saddr = inet->inet_rcv_saddr = 0; |
1da177e4 LT |
495 | err = -EADDRINUSE; |
496 | goto out_release_sock; | |
497 | } | |
498 | ||
c720c7e8 | 499 | if (inet->inet_rcv_saddr) |
1da177e4 LT |
500 | sk->sk_userlocks |= SOCK_BINDADDR_LOCK; |
501 | if (snum) | |
502 | sk->sk_userlocks |= SOCK_BINDPORT_LOCK; | |
c720c7e8 ED |
503 | inet->inet_sport = htons(inet->inet_num); |
504 | inet->inet_daddr = 0; | |
505 | inet->inet_dport = 0; | |
1da177e4 LT |
506 | sk_dst_reset(sk); |
507 | err = 0; | |
508 | out_release_sock: | |
509 | release_sock(sk); | |
510 | out: | |
511 | return err; | |
512 | } | |
3d1427f8 | 513 | EXPORT_SYMBOL(inet_bind); |
1da177e4 | 514 | |
5e73ea1a | 515 | int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr, |
1da177e4 LT |
516 | int addr_len, int flags) |
517 | { | |
518 | struct sock *sk = sock->sk; | |
519 | ||
6503d961 CG |
520 | if (addr_len < sizeof(uaddr->sa_family)) |
521 | return -EINVAL; | |
1da177e4 LT |
522 | if (uaddr->sa_family == AF_UNSPEC) |
523 | return sk->sk_prot->disconnect(sk, flags); | |
524 | ||
c720c7e8 | 525 | if (!inet_sk(sk)->inet_num && inet_autobind(sk)) |
1da177e4 | 526 | return -EAGAIN; |
e3192690 | 527 | return sk->sk_prot->connect(sk, uaddr, addr_len); |
1da177e4 | 528 | } |
3d1427f8 | 529 | EXPORT_SYMBOL(inet_dgram_connect); |
1da177e4 | 530 | |
783237e8 | 531 | static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias) |
1da177e4 LT |
532 | { |
533 | DEFINE_WAIT(wait); | |
534 | ||
aa395145 | 535 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
783237e8 | 536 | sk->sk_write_pending += writebias; |
1da177e4 LT |
537 | |
538 | /* Basic assumption: if someone sets sk->sk_err, he _must_ | |
539 | * change state of the socket from TCP_SYN_*. | |
540 | * Connect() does not allow to get error notifications | |
541 | * without closing the socket. | |
542 | */ | |
543 | while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
544 | release_sock(sk); | |
545 | timeo = schedule_timeout(timeo); | |
546 | lock_sock(sk); | |
547 | if (signal_pending(current) || !timeo) | |
548 | break; | |
aa395145 | 549 | prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE); |
1da177e4 | 550 | } |
aa395145 | 551 | finish_wait(sk_sleep(sk), &wait); |
783237e8 | 552 | sk->sk_write_pending -= writebias; |
1da177e4 LT |
553 | return timeo; |
554 | } | |
555 | ||
556 | /* | |
557 | * Connect to a remote host. There is regrettably still a little | |
558 | * TCP 'magic' in here. | |
559 | */ | |
cf60af03 YC |
560 | int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, |
561 | int addr_len, int flags) | |
1da177e4 LT |
562 | { |
563 | struct sock *sk = sock->sk; | |
564 | int err; | |
565 | long timeo; | |
566 | ||
6503d961 CG |
567 | if (addr_len < sizeof(uaddr->sa_family)) |
568 | return -EINVAL; | |
569 | ||
1da177e4 LT |
570 | if (uaddr->sa_family == AF_UNSPEC) { |
571 | err = sk->sk_prot->disconnect(sk, flags); | |
572 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
573 | goto out; | |
574 | } | |
575 | ||
576 | switch (sock->state) { | |
577 | default: | |
578 | err = -EINVAL; | |
579 | goto out; | |
580 | case SS_CONNECTED: | |
581 | err = -EISCONN; | |
582 | goto out; | |
583 | case SS_CONNECTING: | |
584 | err = -EALREADY; | |
585 | /* Fall out of switch with err, set for this state */ | |
586 | break; | |
587 | case SS_UNCONNECTED: | |
588 | err = -EISCONN; | |
589 | if (sk->sk_state != TCP_CLOSE) | |
590 | goto out; | |
591 | ||
592 | err = sk->sk_prot->connect(sk, uaddr, addr_len); | |
593 | if (err < 0) | |
594 | goto out; | |
595 | ||
e905a9ed | 596 | sock->state = SS_CONNECTING; |
1da177e4 LT |
597 | |
598 | /* Just entered SS_CONNECTING state; the only | |
599 | * difference is that return value in non-blocking | |
600 | * case is EINPROGRESS, rather than EALREADY. | |
601 | */ | |
602 | err = -EINPROGRESS; | |
603 | break; | |
604 | } | |
605 | ||
606 | timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); | |
607 | ||
608 | if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
783237e8 YC |
609 | int writebias = (sk->sk_protocol == IPPROTO_TCP) && |
610 | tcp_sk(sk)->fastopen_req && | |
611 | tcp_sk(sk)->fastopen_req->data ? 1 : 0; | |
612 | ||
1da177e4 | 613 | /* Error code is set above */ |
783237e8 | 614 | if (!timeo || !inet_wait_for_connect(sk, timeo, writebias)) |
1da177e4 LT |
615 | goto out; |
616 | ||
617 | err = sock_intr_errno(timeo); | |
618 | if (signal_pending(current)) | |
619 | goto out; | |
620 | } | |
621 | ||
622 | /* Connection was closed by RST, timeout, ICMP error | |
623 | * or another process disconnected us. | |
624 | */ | |
625 | if (sk->sk_state == TCP_CLOSE) | |
626 | goto sock_error; | |
627 | ||
628 | /* sk->sk_err may be not zero now, if RECVERR was ordered by user | |
629 | * and error was received after socket entered established state. | |
630 | * Hence, it is handled normally after connect() return successfully. | |
631 | */ | |
632 | ||
633 | sock->state = SS_CONNECTED; | |
634 | err = 0; | |
635 | out: | |
1da177e4 LT |
636 | return err; |
637 | ||
638 | sock_error: | |
639 | err = sock_error(sk) ? : -ECONNABORTED; | |
640 | sock->state = SS_UNCONNECTED; | |
641 | if (sk->sk_prot->disconnect(sk, flags)) | |
642 | sock->state = SS_DISCONNECTING; | |
643 | goto out; | |
644 | } | |
cf60af03 YC |
645 | EXPORT_SYMBOL(__inet_stream_connect); |
646 | ||
647 | int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, | |
648 | int addr_len, int flags) | |
649 | { | |
650 | int err; | |
651 | ||
652 | lock_sock(sock->sk); | |
653 | err = __inet_stream_connect(sock, uaddr, addr_len, flags); | |
654 | release_sock(sock->sk); | |
655 | return err; | |
656 | } | |
3d1427f8 | 657 | EXPORT_SYMBOL(inet_stream_connect); |
1da177e4 LT |
658 | |
659 | /* | |
660 | * Accept a pending connection. The TCP layer now gives BSD semantics. | |
661 | */ | |
662 | ||
663 | int inet_accept(struct socket *sock, struct socket *newsock, int flags) | |
664 | { | |
665 | struct sock *sk1 = sock->sk; | |
666 | int err = -EINVAL; | |
667 | struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err); | |
668 | ||
669 | if (!sk2) | |
670 | goto do_err; | |
671 | ||
672 | lock_sock(sk2); | |
673 | ||
1eddcead | 674 | sock_rps_record_flow(sk2); |
547b792c | 675 | WARN_ON(!((1 << sk2->sk_state) & |
8336886f JC |
676 | (TCPF_ESTABLISHED | TCPF_SYN_RECV | |
677 | TCPF_CLOSE_WAIT | TCPF_CLOSE))); | |
1da177e4 LT |
678 | |
679 | sock_graft(sk2, newsock); | |
680 | ||
681 | newsock->state = SS_CONNECTED; | |
682 | err = 0; | |
683 | release_sock(sk2); | |
684 | do_err: | |
685 | return err; | |
686 | } | |
3d1427f8 | 687 | EXPORT_SYMBOL(inet_accept); |
1da177e4 LT |
688 | |
689 | ||
690 | /* | |
691 | * This does both peername and sockname. | |
692 | */ | |
693 | int inet_getname(struct socket *sock, struct sockaddr *uaddr, | |
694 | int *uaddr_len, int peer) | |
695 | { | |
696 | struct sock *sk = sock->sk; | |
697 | struct inet_sock *inet = inet_sk(sk); | |
38bfd8f5 | 698 | DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr); |
1da177e4 LT |
699 | |
700 | sin->sin_family = AF_INET; | |
701 | if (peer) { | |
c720c7e8 | 702 | if (!inet->inet_dport || |
1da177e4 LT |
703 | (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && |
704 | peer == 1)) | |
705 | return -ENOTCONN; | |
c720c7e8 ED |
706 | sin->sin_port = inet->inet_dport; |
707 | sin->sin_addr.s_addr = inet->inet_daddr; | |
1da177e4 | 708 | } else { |
c720c7e8 | 709 | __be32 addr = inet->inet_rcv_saddr; |
1da177e4 | 710 | if (!addr) |
c720c7e8 ED |
711 | addr = inet->inet_saddr; |
712 | sin->sin_port = inet->inet_sport; | |
1da177e4 LT |
713 | sin->sin_addr.s_addr = addr; |
714 | } | |
715 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
716 | *uaddr_len = sizeof(*sin); | |
717 | return 0; | |
718 | } | |
3d1427f8 | 719 | EXPORT_SYMBOL(inet_getname); |
1da177e4 LT |
720 | |
721 | int inet_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg, | |
722 | size_t size) | |
723 | { | |
724 | struct sock *sk = sock->sk; | |
725 | ||
c58dc01b | 726 | sock_rps_record_flow(sk); |
fec5e652 | 727 | |
1da177e4 | 728 | /* We may need to bind the socket. */ |
7ba42910 CG |
729 | if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind && |
730 | inet_autobind(sk)) | |
1da177e4 LT |
731 | return -EAGAIN; |
732 | ||
733 | return sk->sk_prot->sendmsg(iocb, sk, msg, size); | |
734 | } | |
3d1427f8 | 735 | EXPORT_SYMBOL(inet_sendmsg); |
1da177e4 | 736 | |
7ba42910 CG |
737 | ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, |
738 | size_t size, int flags) | |
1da177e4 LT |
739 | { |
740 | struct sock *sk = sock->sk; | |
741 | ||
c58dc01b | 742 | sock_rps_record_flow(sk); |
fec5e652 | 743 | |
1da177e4 | 744 | /* We may need to bind the socket. */ |
7ba42910 CG |
745 | if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind && |
746 | inet_autobind(sk)) | |
1da177e4 LT |
747 | return -EAGAIN; |
748 | ||
749 | if (sk->sk_prot->sendpage) | |
750 | return sk->sk_prot->sendpage(sk, page, offset, size, flags); | |
751 | return sock_no_sendpage(sock, page, offset, size, flags); | |
752 | } | |
7ba42910 | 753 | EXPORT_SYMBOL(inet_sendpage); |
1da177e4 | 754 | |
fec5e652 TH |
755 | int inet_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *msg, |
756 | size_t size, int flags) | |
757 | { | |
758 | struct sock *sk = sock->sk; | |
759 | int addr_len = 0; | |
760 | int err; | |
761 | ||
c58dc01b | 762 | sock_rps_record_flow(sk); |
fec5e652 TH |
763 | |
764 | err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT, | |
765 | flags & ~MSG_DONTWAIT, &addr_len); | |
766 | if (err >= 0) | |
767 | msg->msg_namelen = addr_len; | |
768 | return err; | |
769 | } | |
770 | EXPORT_SYMBOL(inet_recvmsg); | |
1da177e4 LT |
771 | |
772 | int inet_shutdown(struct socket *sock, int how) | |
773 | { | |
774 | struct sock *sk = sock->sk; | |
775 | int err = 0; | |
776 | ||
777 | /* This should really check to make sure | |
778 | * the socket is a TCP socket. (WHY AC...) | |
779 | */ | |
780 | how++; /* maps 0->1 has the advantage of making bit 1 rcvs and | |
781 | 1->2 bit 2 snds. | |
782 | 2->3 */ | |
783 | if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */ | |
784 | return -EINVAL; | |
785 | ||
786 | lock_sock(sk); | |
787 | if (sock->state == SS_CONNECTING) { | |
788 | if ((1 << sk->sk_state) & | |
789 | (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE)) | |
790 | sock->state = SS_DISCONNECTING; | |
791 | else | |
792 | sock->state = SS_CONNECTED; | |
793 | } | |
794 | ||
795 | switch (sk->sk_state) { | |
796 | case TCP_CLOSE: | |
797 | err = -ENOTCONN; | |
798 | /* Hack to wake up other listeners, who can poll for | |
799 | POLLHUP, even on eg. unconnected UDP sockets -- RR */ | |
800 | default: | |
801 | sk->sk_shutdown |= how; | |
802 | if (sk->sk_prot->shutdown) | |
803 | sk->sk_prot->shutdown(sk, how); | |
804 | break; | |
805 | ||
806 | /* Remaining two branches are temporary solution for missing | |
807 | * close() in multithreaded environment. It is _not_ a good idea, | |
808 | * but we have no choice until close() is repaired at VFS level. | |
809 | */ | |
810 | case TCP_LISTEN: | |
811 | if (!(how & RCV_SHUTDOWN)) | |
812 | break; | |
813 | /* Fall through */ | |
814 | case TCP_SYN_SENT: | |
815 | err = sk->sk_prot->disconnect(sk, O_NONBLOCK); | |
816 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
817 | break; | |
818 | } | |
819 | ||
820 | /* Wake up anyone sleeping in poll. */ | |
821 | sk->sk_state_change(sk); | |
822 | release_sock(sk); | |
823 | return err; | |
824 | } | |
3d1427f8 | 825 | EXPORT_SYMBOL(inet_shutdown); |
1da177e4 LT |
826 | |
827 | /* | |
828 | * ioctl() calls you can issue on an INET socket. Most of these are | |
829 | * device configuration and stuff and very rarely used. Some ioctls | |
830 | * pass on to the socket itself. | |
831 | * | |
832 | * NOTE: I like the idea of a module for the config stuff. ie ifconfig | |
833 | * loads the devconfigure module does its configuring and unloads it. | |
834 | * There's a good 20K of config code hanging around the kernel. | |
835 | */ | |
836 | ||
837 | int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) | |
838 | { | |
839 | struct sock *sk = sock->sk; | |
840 | int err = 0; | |
3b1e0a65 | 841 | struct net *net = sock_net(sk); |
1da177e4 LT |
842 | |
843 | switch (cmd) { | |
3d1427f8 ED |
844 | case SIOCGSTAMP: |
845 | err = sock_get_timestamp(sk, (struct timeval __user *)arg); | |
846 | break; | |
847 | case SIOCGSTAMPNS: | |
848 | err = sock_get_timestampns(sk, (struct timespec __user *)arg); | |
849 | break; | |
850 | case SIOCADDRT: | |
851 | case SIOCDELRT: | |
852 | case SIOCRTMSG: | |
853 | err = ip_rt_ioctl(net, cmd, (void __user *)arg); | |
854 | break; | |
855 | case SIOCDARP: | |
856 | case SIOCGARP: | |
857 | case SIOCSARP: | |
858 | err = arp_ioctl(net, cmd, (void __user *)arg); | |
859 | break; | |
860 | case SIOCGIFADDR: | |
861 | case SIOCSIFADDR: | |
862 | case SIOCGIFBRDADDR: | |
863 | case SIOCSIFBRDADDR: | |
864 | case SIOCGIFNETMASK: | |
865 | case SIOCSIFNETMASK: | |
866 | case SIOCGIFDSTADDR: | |
867 | case SIOCSIFDSTADDR: | |
868 | case SIOCSIFPFLAGS: | |
869 | case SIOCGIFPFLAGS: | |
870 | case SIOCSIFFLAGS: | |
871 | err = devinet_ioctl(net, cmd, (void __user *)arg); | |
872 | break; | |
873 | default: | |
874 | if (sk->sk_prot->ioctl) | |
875 | err = sk->sk_prot->ioctl(sk, cmd, arg); | |
876 | else | |
877 | err = -ENOIOCTLCMD; | |
878 | break; | |
1da177e4 LT |
879 | } |
880 | return err; | |
881 | } | |
3d1427f8 | 882 | EXPORT_SYMBOL(inet_ioctl); |
1da177e4 | 883 | |
709b46e8 | 884 | #ifdef CONFIG_COMPAT |
686dc6b6 | 885 | static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
709b46e8 EB |
886 | { |
887 | struct sock *sk = sock->sk; | |
888 | int err = -ENOIOCTLCMD; | |
889 | ||
890 | if (sk->sk_prot->compat_ioctl) | |
891 | err = sk->sk_prot->compat_ioctl(sk, cmd, arg); | |
892 | ||
893 | return err; | |
894 | } | |
895 | #endif | |
896 | ||
90ddc4f0 | 897 | const struct proto_ops inet_stream_ops = { |
543d9cfe ACM |
898 | .family = PF_INET, |
899 | .owner = THIS_MODULE, | |
900 | .release = inet_release, | |
901 | .bind = inet_bind, | |
902 | .connect = inet_stream_connect, | |
903 | .socketpair = sock_no_socketpair, | |
904 | .accept = inet_accept, | |
905 | .getname = inet_getname, | |
906 | .poll = tcp_poll, | |
907 | .ioctl = inet_ioctl, | |
908 | .listen = inet_listen, | |
909 | .shutdown = inet_shutdown, | |
910 | .setsockopt = sock_common_setsockopt, | |
911 | .getsockopt = sock_common_getsockopt, | |
7ba42910 | 912 | .sendmsg = inet_sendmsg, |
fec5e652 | 913 | .recvmsg = inet_recvmsg, |
543d9cfe | 914 | .mmap = sock_no_mmap, |
7ba42910 | 915 | .sendpage = inet_sendpage, |
9c55e01c | 916 | .splice_read = tcp_splice_read, |
3fdadf7d | 917 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
918 | .compat_setsockopt = compat_sock_common_setsockopt, |
919 | .compat_getsockopt = compat_sock_common_getsockopt, | |
709b46e8 | 920 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 921 | #endif |
1da177e4 | 922 | }; |
3d1427f8 | 923 | EXPORT_SYMBOL(inet_stream_ops); |
1da177e4 | 924 | |
90ddc4f0 | 925 | const struct proto_ops inet_dgram_ops = { |
543d9cfe ACM |
926 | .family = PF_INET, |
927 | .owner = THIS_MODULE, | |
928 | .release = inet_release, | |
929 | .bind = inet_bind, | |
930 | .connect = inet_dgram_connect, | |
931 | .socketpair = sock_no_socketpair, | |
932 | .accept = sock_no_accept, | |
933 | .getname = inet_getname, | |
934 | .poll = udp_poll, | |
935 | .ioctl = inet_ioctl, | |
936 | .listen = sock_no_listen, | |
937 | .shutdown = inet_shutdown, | |
938 | .setsockopt = sock_common_setsockopt, | |
939 | .getsockopt = sock_common_getsockopt, | |
940 | .sendmsg = inet_sendmsg, | |
fec5e652 | 941 | .recvmsg = inet_recvmsg, |
543d9cfe ACM |
942 | .mmap = sock_no_mmap, |
943 | .sendpage = inet_sendpage, | |
3fdadf7d | 944 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
945 | .compat_setsockopt = compat_sock_common_setsockopt, |
946 | .compat_getsockopt = compat_sock_common_getsockopt, | |
709b46e8 | 947 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 948 | #endif |
1da177e4 | 949 | }; |
3d1427f8 | 950 | EXPORT_SYMBOL(inet_dgram_ops); |
1da177e4 LT |
951 | |
952 | /* | |
953 | * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without | |
954 | * udp_poll | |
955 | */ | |
90ddc4f0 | 956 | static const struct proto_ops inet_sockraw_ops = { |
543d9cfe ACM |
957 | .family = PF_INET, |
958 | .owner = THIS_MODULE, | |
959 | .release = inet_release, | |
960 | .bind = inet_bind, | |
961 | .connect = inet_dgram_connect, | |
962 | .socketpair = sock_no_socketpair, | |
963 | .accept = sock_no_accept, | |
964 | .getname = inet_getname, | |
965 | .poll = datagram_poll, | |
966 | .ioctl = inet_ioctl, | |
967 | .listen = sock_no_listen, | |
968 | .shutdown = inet_shutdown, | |
969 | .setsockopt = sock_common_setsockopt, | |
970 | .getsockopt = sock_common_getsockopt, | |
971 | .sendmsg = inet_sendmsg, | |
fec5e652 | 972 | .recvmsg = inet_recvmsg, |
543d9cfe ACM |
973 | .mmap = sock_no_mmap, |
974 | .sendpage = inet_sendpage, | |
3fdadf7d | 975 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
976 | .compat_setsockopt = compat_sock_common_setsockopt, |
977 | .compat_getsockopt = compat_sock_common_getsockopt, | |
709b46e8 | 978 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 979 | #endif |
1da177e4 LT |
980 | }; |
981 | ||
ec1b4cf7 | 982 | static const struct net_proto_family inet_family_ops = { |
1da177e4 LT |
983 | .family = PF_INET, |
984 | .create = inet_create, | |
985 | .owner = THIS_MODULE, | |
986 | }; | |
987 | ||
1da177e4 LT |
988 | /* Upon startup we insert all the elements in inetsw_array[] into |
989 | * the linked list inetsw. | |
990 | */ | |
991 | static struct inet_protosw inetsw_array[] = | |
992 | { | |
e905a9ed YH |
993 | { |
994 | .type = SOCK_STREAM, | |
995 | .protocol = IPPROTO_TCP, | |
996 | .prot = &tcp_prot, | |
997 | .ops = &inet_stream_ops, | |
e905a9ed | 998 | .flags = INET_PROTOSW_PERMANENT | |
d83d8461 | 999 | INET_PROTOSW_ICSK, |
e905a9ed YH |
1000 | }, |
1001 | ||
1002 | { | |
1003 | .type = SOCK_DGRAM, | |
1004 | .protocol = IPPROTO_UDP, | |
1005 | .prot = &udp_prot, | |
1006 | .ops = &inet_dgram_ops, | |
e905a9ed | 1007 | .flags = INET_PROTOSW_PERMANENT, |
1da177e4 | 1008 | }, |
e905a9ed | 1009 | |
c319b4d7 VK |
1010 | { |
1011 | .type = SOCK_DGRAM, | |
1012 | .protocol = IPPROTO_ICMP, | |
1013 | .prot = &ping_prot, | |
1014 | .ops = &inet_dgram_ops, | |
c319b4d7 VK |
1015 | .flags = INET_PROTOSW_REUSE, |
1016 | }, | |
1da177e4 LT |
1017 | |
1018 | { | |
e905a9ed YH |
1019 | .type = SOCK_RAW, |
1020 | .protocol = IPPROTO_IP, /* wild card */ | |
1021 | .prot = &raw_prot, | |
1022 | .ops = &inet_sockraw_ops, | |
e905a9ed | 1023 | .flags = INET_PROTOSW_REUSE, |
1da177e4 LT |
1024 | } |
1025 | }; | |
1026 | ||
c40f6fff | 1027 | #define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array) |
1da177e4 LT |
1028 | |
1029 | void inet_register_protosw(struct inet_protosw *p) | |
1030 | { | |
1031 | struct list_head *lh; | |
1032 | struct inet_protosw *answer; | |
1033 | int protocol = p->protocol; | |
1034 | struct list_head *last_perm; | |
1035 | ||
1036 | spin_lock_bh(&inetsw_lock); | |
1037 | ||
1038 | if (p->type >= SOCK_MAX) | |
1039 | goto out_illegal; | |
1040 | ||
1041 | /* If we are trying to override a permanent protocol, bail. */ | |
1042 | answer = NULL; | |
1043 | last_perm = &inetsw[p->type]; | |
1044 | list_for_each(lh, &inetsw[p->type]) { | |
1045 | answer = list_entry(lh, struct inet_protosw, list); | |
1046 | ||
1047 | /* Check only the non-wild match. */ | |
1048 | if (INET_PROTOSW_PERMANENT & answer->flags) { | |
1049 | if (protocol == answer->protocol) | |
1050 | break; | |
1051 | last_perm = lh; | |
1052 | } | |
1053 | ||
1054 | answer = NULL; | |
1055 | } | |
1056 | if (answer) | |
1057 | goto out_permanent; | |
1058 | ||
1059 | /* Add the new entry after the last permanent entry if any, so that | |
1060 | * the new entry does not override a permanent entry when matched with | |
1061 | * a wild-card protocol. But it is allowed to override any existing | |
e905a9ed | 1062 | * non-permanent entry. This means that when we remove this entry, the |
1da177e4 LT |
1063 | * system automatically returns to the old behavior. |
1064 | */ | |
1065 | list_add_rcu(&p->list, last_perm); | |
1066 | out: | |
1067 | spin_unlock_bh(&inetsw_lock); | |
1068 | ||
1da177e4 LT |
1069 | return; |
1070 | ||
1071 | out_permanent: | |
058bd4d2 | 1072 | pr_err("Attempt to override permanent protocol %d\n", protocol); |
1da177e4 LT |
1073 | goto out; |
1074 | ||
1075 | out_illegal: | |
058bd4d2 | 1076 | pr_err("Ignoring attempt to register invalid socket type %d\n", |
1da177e4 LT |
1077 | p->type); |
1078 | goto out; | |
1079 | } | |
3d1427f8 | 1080 | EXPORT_SYMBOL(inet_register_protosw); |
1da177e4 LT |
1081 | |
1082 | void inet_unregister_protosw(struct inet_protosw *p) | |
1083 | { | |
1084 | if (INET_PROTOSW_PERMANENT & p->flags) { | |
058bd4d2 | 1085 | pr_err("Attempt to unregister permanent protocol %d\n", |
1da177e4 LT |
1086 | p->protocol); |
1087 | } else { | |
1088 | spin_lock_bh(&inetsw_lock); | |
1089 | list_del_rcu(&p->list); | |
1090 | spin_unlock_bh(&inetsw_lock); | |
1091 | ||
1092 | synchronize_net(); | |
1093 | } | |
1094 | } | |
3d1427f8 | 1095 | EXPORT_SYMBOL(inet_unregister_protosw); |
1da177e4 | 1096 | |
32519f11 ACM |
1097 | /* |
1098 | * Shall we try to damage output packets if routing dev changes? | |
1099 | */ | |
1100 | ||
ab32ea5d | 1101 | int sysctl_ip_dynaddr __read_mostly; |
32519f11 ACM |
1102 | |
1103 | static int inet_sk_reselect_saddr(struct sock *sk) | |
1104 | { | |
1105 | struct inet_sock *inet = inet_sk(sk); | |
c720c7e8 | 1106 | __be32 old_saddr = inet->inet_saddr; |
c720c7e8 | 1107 | __be32 daddr = inet->inet_daddr; |
6e869138 | 1108 | struct flowi4 *fl4; |
b23dd4fe DM |
1109 | struct rtable *rt; |
1110 | __be32 new_saddr; | |
f6d8bd05 | 1111 | struct ip_options_rcu *inet_opt; |
32519f11 | 1112 | |
f6d8bd05 ED |
1113 | inet_opt = rcu_dereference_protected(inet->inet_opt, |
1114 | sock_owned_by_user(sk)); | |
1115 | if (inet_opt && inet_opt->opt.srr) | |
1116 | daddr = inet_opt->opt.faddr; | |
32519f11 ACM |
1117 | |
1118 | /* Query new route. */ | |
6e869138 DM |
1119 | fl4 = &inet->cork.fl.u.ip4; |
1120 | rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk), | |
b23dd4fe | 1121 | sk->sk_bound_dev_if, sk->sk_protocol, |
0e0d44ab | 1122 | inet->inet_sport, inet->inet_dport, sk); |
b23dd4fe DM |
1123 | if (IS_ERR(rt)) |
1124 | return PTR_ERR(rt); | |
32519f11 | 1125 | |
d8d1f30b | 1126 | sk_setup_caps(sk, &rt->dst); |
32519f11 | 1127 | |
6e869138 | 1128 | new_saddr = fl4->saddr; |
32519f11 ACM |
1129 | |
1130 | if (new_saddr == old_saddr) | |
1131 | return 0; | |
1132 | ||
1133 | if (sysctl_ip_dynaddr > 1) { | |
058bd4d2 JP |
1134 | pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n", |
1135 | __func__, &old_saddr, &new_saddr); | |
32519f11 ACM |
1136 | } |
1137 | ||
c720c7e8 | 1138 | inet->inet_saddr = inet->inet_rcv_saddr = new_saddr; |
32519f11 ACM |
1139 | |
1140 | /* | |
1141 | * XXX The only one ugly spot where we need to | |
1142 | * XXX really change the sockets identity after | |
1143 | * XXX it has entered the hashes. -DaveM | |
1144 | * | |
1145 | * Besides that, it does not check for connection | |
1146 | * uniqueness. Wait for troubles. | |
1147 | */ | |
1148 | __sk_prot_rehash(sk); | |
1149 | return 0; | |
1150 | } | |
1151 | ||
1152 | int inet_sk_rebuild_header(struct sock *sk) | |
1153 | { | |
1154 | struct inet_sock *inet = inet_sk(sk); | |
1155 | struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0); | |
3ca3c68e | 1156 | __be32 daddr; |
f6d8bd05 | 1157 | struct ip_options_rcu *inet_opt; |
6e869138 | 1158 | struct flowi4 *fl4; |
32519f11 ACM |
1159 | int err; |
1160 | ||
1161 | /* Route is OK, nothing to do. */ | |
1162 | if (rt) | |
1163 | return 0; | |
1164 | ||
1165 | /* Reroute. */ | |
f6d8bd05 ED |
1166 | rcu_read_lock(); |
1167 | inet_opt = rcu_dereference(inet->inet_opt); | |
c720c7e8 | 1168 | daddr = inet->inet_daddr; |
f6d8bd05 ED |
1169 | if (inet_opt && inet_opt->opt.srr) |
1170 | daddr = inet_opt->opt.faddr; | |
1171 | rcu_read_unlock(); | |
6e869138 DM |
1172 | fl4 = &inet->cork.fl.u.ip4; |
1173 | rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr, | |
78fbfd8a DM |
1174 | inet->inet_dport, inet->inet_sport, |
1175 | sk->sk_protocol, RT_CONN_FLAGS(sk), | |
1176 | sk->sk_bound_dev_if); | |
b23dd4fe DM |
1177 | if (!IS_ERR(rt)) { |
1178 | err = 0; | |
d8d1f30b | 1179 | sk_setup_caps(sk, &rt->dst); |
b23dd4fe DM |
1180 | } else { |
1181 | err = PTR_ERR(rt); | |
1182 | ||
32519f11 ACM |
1183 | /* Routing failed... */ |
1184 | sk->sk_route_caps = 0; | |
1185 | /* | |
1186 | * Other protocols have to map its equivalent state to TCP_SYN_SENT. | |
1187 | * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme | |
1188 | */ | |
1189 | if (!sysctl_ip_dynaddr || | |
1190 | sk->sk_state != TCP_SYN_SENT || | |
1191 | (sk->sk_userlocks & SOCK_BINDADDR_LOCK) || | |
1192 | (err = inet_sk_reselect_saddr(sk)) != 0) | |
1193 | sk->sk_err_soft = -err; | |
1194 | } | |
1195 | ||
1196 | return err; | |
1197 | } | |
32519f11 ACM |
1198 | EXPORT_SYMBOL(inet_sk_rebuild_header); |
1199 | ||
a430a43d HX |
1200 | static int inet_gso_send_check(struct sk_buff *skb) |
1201 | { | |
bca49f84 | 1202 | const struct net_offload *ops; |
f9242b6b | 1203 | const struct iphdr *iph; |
a430a43d HX |
1204 | int proto; |
1205 | int ihl; | |
1206 | int err = -EINVAL; | |
1207 | ||
1208 | if (unlikely(!pskb_may_pull(skb, sizeof(*iph)))) | |
1209 | goto out; | |
1210 | ||
eddc9ec5 | 1211 | iph = ip_hdr(skb); |
a430a43d HX |
1212 | ihl = iph->ihl * 4; |
1213 | if (ihl < sizeof(*iph)) | |
1214 | goto out; | |
1215 | ||
47d27aad ED |
1216 | proto = iph->protocol; |
1217 | ||
1218 | /* Warning: after this point, iph might be no longer valid */ | |
a430a43d HX |
1219 | if (unlikely(!pskb_may_pull(skb, ihl))) |
1220 | goto out; | |
badff6d0 | 1221 | __skb_pull(skb, ihl); |
47d27aad | 1222 | |
badff6d0 | 1223 | skb_reset_transport_header(skb); |
a430a43d HX |
1224 | err = -EPROTONOSUPPORT; |
1225 | ||
bca49f84 | 1226 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 VY |
1227 | if (likely(ops && ops->callbacks.gso_send_check)) |
1228 | err = ops->callbacks.gso_send_check(skb); | |
a430a43d HX |
1229 | |
1230 | out: | |
1231 | return err; | |
1232 | } | |
1233 | ||
c8f44aff | 1234 | static struct sk_buff *inet_gso_segment(struct sk_buff *skb, |
3347c960 | 1235 | netdev_features_t features) |
f4c50d99 HX |
1236 | { |
1237 | struct sk_buff *segs = ERR_PTR(-EINVAL); | |
bca49f84 | 1238 | const struct net_offload *ops; |
3347c960 | 1239 | unsigned int offset = 0; |
8c3a897b | 1240 | bool udpfrag, encap; |
f9242b6b | 1241 | struct iphdr *iph; |
f4c50d99 | 1242 | int proto; |
3347c960 | 1243 | int nhoff; |
f4c50d99 HX |
1244 | int ihl; |
1245 | int id; | |
1246 | ||
bbcf467d HX |
1247 | if (unlikely(skb_shinfo(skb)->gso_type & |
1248 | ~(SKB_GSO_TCPV4 | | |
1249 | SKB_GSO_UDP | | |
1250 | SKB_GSO_DODGY | | |
1251 | SKB_GSO_TCP_ECN | | |
68c33163 | 1252 | SKB_GSO_GRE | |
4749c09c | 1253 | SKB_GSO_GRE_CSUM | |
cb32f511 | 1254 | SKB_GSO_IPIP | |
61c1db7f | 1255 | SKB_GSO_SIT | |
9b3eb5ed | 1256 | SKB_GSO_TCPV6 | |
73136267 | 1257 | SKB_GSO_UDP_TUNNEL | |
0f4f4ffa | 1258 | SKB_GSO_UDP_TUNNEL_CSUM | |
0d89d203 | 1259 | SKB_GSO_MPLS | |
bbcf467d HX |
1260 | 0))) |
1261 | goto out; | |
1262 | ||
3347c960 ED |
1263 | skb_reset_network_header(skb); |
1264 | nhoff = skb_network_header(skb) - skb_mac_header(skb); | |
bbcf467d | 1265 | if (unlikely(!pskb_may_pull(skb, sizeof(*iph)))) |
f4c50d99 HX |
1266 | goto out; |
1267 | ||
eddc9ec5 | 1268 | iph = ip_hdr(skb); |
f4c50d99 HX |
1269 | ihl = iph->ihl * 4; |
1270 | if (ihl < sizeof(*iph)) | |
1271 | goto out; | |
1272 | ||
47d27aad ED |
1273 | id = ntohs(iph->id); |
1274 | proto = iph->protocol; | |
1275 | ||
1276 | /* Warning: after this point, iph might be no longer valid */ | |
bbcf467d | 1277 | if (unlikely(!pskb_may_pull(skb, ihl))) |
f4c50d99 | 1278 | goto out; |
47d27aad | 1279 | __skb_pull(skb, ihl); |
f4c50d99 | 1280 | |
8c3a897b ED |
1281 | encap = SKB_GSO_CB(skb)->encap_level > 0; |
1282 | if (encap) | |
3347c960 ED |
1283 | features = skb->dev->hw_enc_features & netif_skb_features(skb); |
1284 | SKB_GSO_CB(skb)->encap_level += ihl; | |
73136267 | 1285 | |
badff6d0 | 1286 | skb_reset_transport_header(skb); |
47d27aad | 1287 | |
f4c50d99 HX |
1288 | segs = ERR_PTR(-EPROTONOSUPPORT); |
1289 | ||
91a48a2e HFS |
1290 | if (skb->encapsulation && |
1291 | skb_shinfo(skb)->gso_type & (SKB_GSO_SIT|SKB_GSO_IPIP)) | |
1292 | udpfrag = proto == IPPROTO_UDP && encap; | |
1293 | else | |
1294 | udpfrag = proto == IPPROTO_UDP && !skb->encapsulation; | |
dcd60771 | 1295 | |
bca49f84 | 1296 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 VY |
1297 | if (likely(ops && ops->callbacks.gso_segment)) |
1298 | segs = ops->callbacks.gso_segment(skb, features); | |
f4c50d99 | 1299 | |
50c3a487 | 1300 | if (IS_ERR_OR_NULL(segs)) |
f4c50d99 HX |
1301 | goto out; |
1302 | ||
1303 | skb = segs; | |
1304 | do { | |
3347c960 | 1305 | iph = (struct iphdr *)(skb_mac_header(skb) + nhoff); |
8c3a897b | 1306 | if (udpfrag) { |
d7ca4cc0 SS |
1307 | iph->id = htons(id); |
1308 | iph->frag_off = htons(offset >> 3); | |
1309 | if (skb->next != NULL) | |
1310 | iph->frag_off |= htons(IP_MF); | |
3347c960 | 1311 | offset += skb->len - nhoff - ihl; |
8c3a897b | 1312 | } else { |
25c7704d | 1313 | iph->id = htons(id++); |
490ab081 | 1314 | } |
3347c960 | 1315 | iph->tot_len = htons(skb->len - nhoff); |
47d27aad | 1316 | ip_send_check(iph); |
8c3a897b | 1317 | if (encap) |
3347c960 | 1318 | skb_reset_inner_headers(skb); |
3347c960 | 1319 | skb->network_header = (u8 *)iph - skb->head; |
f4c50d99 HX |
1320 | } while ((skb = skb->next)); |
1321 | ||
1322 | out: | |
1323 | return segs; | |
1324 | } | |
1325 | ||
73cc19f1 HX |
1326 | static struct sk_buff **inet_gro_receive(struct sk_buff **head, |
1327 | struct sk_buff *skb) | |
1328 | { | |
bca49f84 | 1329 | const struct net_offload *ops; |
73cc19f1 HX |
1330 | struct sk_buff **pp = NULL; |
1331 | struct sk_buff *p; | |
b71d1d42 | 1332 | const struct iphdr *iph; |
a5b1cf28 HX |
1333 | unsigned int hlen; |
1334 | unsigned int off; | |
1075f3f6 | 1335 | unsigned int id; |
73cc19f1 HX |
1336 | int flush = 1; |
1337 | int proto; | |
73cc19f1 | 1338 | |
a5b1cf28 HX |
1339 | off = skb_gro_offset(skb); |
1340 | hlen = off + sizeof(*iph); | |
1341 | iph = skb_gro_header_fast(skb, off); | |
1342 | if (skb_gro_header_hard(skb, hlen)) { | |
1343 | iph = skb_gro_header_slow(skb, hlen, off); | |
1344 | if (unlikely(!iph)) | |
1345 | goto out; | |
1346 | } | |
73cc19f1 | 1347 | |
f9242b6b | 1348 | proto = iph->protocol; |
73cc19f1 HX |
1349 | |
1350 | rcu_read_lock(); | |
bca49f84 | 1351 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 | 1352 | if (!ops || !ops->callbacks.gro_receive) |
73cc19f1 HX |
1353 | goto out_unlock; |
1354 | ||
a5ad24be | 1355 | if (*(u8 *)iph != 0x45) |
73cc19f1 HX |
1356 | goto out_unlock; |
1357 | ||
a9e050f4 | 1358 | if (unlikely(ip_fast_csum((u8 *)iph, 5))) |
73cc19f1 HX |
1359 | goto out_unlock; |
1360 | ||
0eae88f3 | 1361 | id = ntohl(*(__be32 *)&iph->id); |
db8caf3d | 1362 | flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF)); |
1075f3f6 | 1363 | id >>= 16; |
73cc19f1 HX |
1364 | |
1365 | for (p = *head; p; p = p->next) { | |
1366 | struct iphdr *iph2; | |
1367 | ||
1368 | if (!NAPI_GRO_CB(p)->same_flow) | |
1369 | continue; | |
1370 | ||
299603e8 JC |
1371 | iph2 = (struct iphdr *)(p->data + off); |
1372 | /* The above works because, with the exception of the top | |
1373 | * (inner most) layer, we only aggregate pkts with the same | |
1374 | * hdr length so all the hdrs we'll need to verify will start | |
1375 | * at the same offset. | |
1376 | */ | |
a5ad24be | 1377 | if ((iph->protocol ^ iph2->protocol) | |
0eae88f3 ED |
1378 | ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) | |
1379 | ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) { | |
73cc19f1 HX |
1380 | NAPI_GRO_CB(p)->same_flow = 0; |
1381 | continue; | |
1382 | } | |
1383 | ||
1384 | /* All fields must match except length and checksum. */ | |
1385 | NAPI_GRO_CB(p)->flush |= | |
a5ad24be | 1386 | (iph->ttl ^ iph2->ttl) | |
a9e050f4 | 1387 | (iph->tos ^ iph2->tos) | |
bf5a755f | 1388 | ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF)); |
73cc19f1 | 1389 | |
bf5a755f JC |
1390 | /* Save the IP ID check to be included later when we get to |
1391 | * the transport layer so only the inner most IP ID is checked. | |
1392 | * This is because some GSO/TSO implementations do not | |
1393 | * correctly increment the IP ID for the outer hdrs. | |
1394 | */ | |
1395 | NAPI_GRO_CB(p)->flush_id = | |
1396 | ((u16)(ntohs(iph2->id) + NAPI_GRO_CB(p)->count) ^ id); | |
73cc19f1 HX |
1397 | NAPI_GRO_CB(p)->flush |= flush; |
1398 | } | |
1399 | ||
1400 | NAPI_GRO_CB(skb)->flush |= flush; | |
299603e8 JC |
1401 | skb_set_network_header(skb, off); |
1402 | /* The above will be needed by the transport layer if there is one | |
1403 | * immediately following this IP hdr. | |
1404 | */ | |
1405 | ||
86911732 HX |
1406 | skb_gro_pull(skb, sizeof(*iph)); |
1407 | skb_set_transport_header(skb, skb_gro_offset(skb)); | |
73cc19f1 | 1408 | |
f191a1d1 | 1409 | pp = ops->callbacks.gro_receive(head, skb); |
73cc19f1 HX |
1410 | |
1411 | out_unlock: | |
1412 | rcu_read_unlock(); | |
1413 | ||
1414 | out: | |
1415 | NAPI_GRO_CB(skb)->flush |= flush; | |
1416 | ||
1417 | return pp; | |
1418 | } | |
1419 | ||
299603e8 | 1420 | static int inet_gro_complete(struct sk_buff *skb, int nhoff) |
73cc19f1 | 1421 | { |
299603e8 JC |
1422 | __be16 newlen = htons(skb->len - nhoff); |
1423 | struct iphdr *iph = (struct iphdr *)(skb->data + nhoff); | |
bca49f84 | 1424 | const struct net_offload *ops; |
f9242b6b | 1425 | int proto = iph->protocol; |
73cc19f1 | 1426 | int err = -ENOSYS; |
73cc19f1 | 1427 | |
c3caf119 JC |
1428 | if (skb->encapsulation) |
1429 | skb_set_inner_network_header(skb, nhoff); | |
1430 | ||
73cc19f1 HX |
1431 | csum_replace2(&iph->check, iph->tot_len, newlen); |
1432 | iph->tot_len = newlen; | |
1433 | ||
1434 | rcu_read_lock(); | |
bca49f84 | 1435 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 | 1436 | if (WARN_ON(!ops || !ops->callbacks.gro_complete)) |
73cc19f1 HX |
1437 | goto out_unlock; |
1438 | ||
299603e8 JC |
1439 | /* Only need to add sizeof(*iph) to get to the next hdr below |
1440 | * because any hdr with option will have been flushed in | |
1441 | * inet_gro_receive(). | |
1442 | */ | |
1443 | err = ops->callbacks.gro_complete(skb, nhoff + sizeof(*iph)); | |
73cc19f1 HX |
1444 | |
1445 | out_unlock: | |
1446 | rcu_read_unlock(); | |
1447 | ||
1448 | return err; | |
1449 | } | |
1450 | ||
eee4fe4d | 1451 | int inet_ctl_sock_create(struct sock **sk, unsigned short family, |
5677242f DL |
1452 | unsigned short type, unsigned char protocol, |
1453 | struct net *net) | |
3d58b5fa | 1454 | { |
eee4fe4d DL |
1455 | struct socket *sock; |
1456 | int rc = sock_create_kern(family, type, protocol, &sock); | |
3d58b5fa DL |
1457 | |
1458 | if (rc == 0) { | |
eee4fe4d DL |
1459 | *sk = sock->sk; |
1460 | (*sk)->sk_allocation = GFP_ATOMIC; | |
3d58b5fa DL |
1461 | /* |
1462 | * Unhash it so that IP input processing does not even see it, | |
1463 | * we do not wish this socket to see incoming packets. | |
1464 | */ | |
eee4fe4d | 1465 | (*sk)->sk_prot->unhash(*sk); |
5677242f DL |
1466 | |
1467 | sk_change_net(*sk, net); | |
3d58b5fa DL |
1468 | } |
1469 | return rc; | |
1470 | } | |
3d58b5fa DL |
1471 | EXPORT_SYMBOL_GPL(inet_ctl_sock_create); |
1472 | ||
698365fa | 1473 | unsigned long snmp_fold_field(void __percpu *mib, int offt) |
5e0f0435 HX |
1474 | { |
1475 | unsigned long res = 0; | |
698365fa | 1476 | int i; |
5e0f0435 | 1477 | |
698365fa WC |
1478 | for_each_possible_cpu(i) |
1479 | res += *(((unsigned long *) per_cpu_ptr(mib, i)) + offt); | |
5e0f0435 HX |
1480 | return res; |
1481 | } | |
1482 | EXPORT_SYMBOL_GPL(snmp_fold_field); | |
1483 | ||
4ce3c183 ED |
1484 | #if BITS_PER_LONG==32 |
1485 | ||
698365fa | 1486 | u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset) |
4ce3c183 ED |
1487 | { |
1488 | u64 res = 0; | |
1489 | int cpu; | |
1490 | ||
1491 | for_each_possible_cpu(cpu) { | |
8f0ea0fe | 1492 | void *bhptr; |
4ce3c183 | 1493 | struct u64_stats_sync *syncp; |
8f0ea0fe | 1494 | u64 v; |
4ce3c183 ED |
1495 | unsigned int start; |
1496 | ||
698365fa | 1497 | bhptr = per_cpu_ptr(mib, cpu); |
4ce3c183 ED |
1498 | syncp = (struct u64_stats_sync *)(bhptr + syncp_offset); |
1499 | do { | |
57a7744e | 1500 | start = u64_stats_fetch_begin_irq(syncp); |
8f0ea0fe | 1501 | v = *(((u64 *) bhptr) + offt); |
57a7744e | 1502 | } while (u64_stats_fetch_retry_irq(syncp, start)); |
4ce3c183 | 1503 | |
8f0ea0fe | 1504 | res += v; |
4ce3c183 ED |
1505 | } |
1506 | return res; | |
1507 | } | |
1508 | EXPORT_SYMBOL_GPL(snmp_fold_field64); | |
1509 | #endif | |
1510 | ||
1da177e4 | 1511 | #ifdef CONFIG_IP_MULTICAST |
32613090 | 1512 | static const struct net_protocol igmp_protocol = { |
1da177e4 | 1513 | .handler = igmp_rcv, |
b4ee07df | 1514 | .netns_ok = 1, |
1da177e4 LT |
1515 | }; |
1516 | #endif | |
1517 | ||
32613090 | 1518 | static const struct net_protocol tcp_protocol = { |
41063e9d DM |
1519 | .early_demux = tcp_v4_early_demux, |
1520 | .handler = tcp_v4_rcv, | |
1521 | .err_handler = tcp_v4_err, | |
41063e9d DM |
1522 | .no_policy = 1, |
1523 | .netns_ok = 1, | |
8ed1dc44 | 1524 | .icmp_strict_tag_validation = 1, |
1da177e4 LT |
1525 | }; |
1526 | ||
32613090 | 1527 | static const struct net_protocol udp_protocol = { |
421b3885 | 1528 | .early_demux = udp_v4_early_demux, |
1da177e4 LT |
1529 | .handler = udp_rcv, |
1530 | .err_handler = udp_err, | |
1531 | .no_policy = 1, | |
92f1fecb | 1532 | .netns_ok = 1, |
1da177e4 LT |
1533 | }; |
1534 | ||
32613090 | 1535 | static const struct net_protocol icmp_protocol = { |
1da177e4 | 1536 | .handler = icmp_rcv, |
5b052042 | 1537 | .err_handler = icmp_err, |
8b7817f3 | 1538 | .no_policy = 1, |
92f1fecb | 1539 | .netns_ok = 1, |
1da177e4 LT |
1540 | }; |
1541 | ||
9b4661bd PE |
1542 | static __net_init int ipv4_mib_init_net(struct net *net) |
1543 | { | |
827da44c JS |
1544 | int i; |
1545 | ||
698365fa WC |
1546 | net->mib.tcp_statistics = alloc_percpu(struct tcp_mib); |
1547 | if (!net->mib.tcp_statistics) | |
57ef42d5 | 1548 | goto err_tcp_mib; |
698365fa WC |
1549 | net->mib.ip_statistics = alloc_percpu(struct ipstats_mib); |
1550 | if (!net->mib.ip_statistics) | |
a20f5799 | 1551 | goto err_ip_mib; |
827da44c JS |
1552 | |
1553 | for_each_possible_cpu(i) { | |
1554 | struct ipstats_mib *af_inet_stats; | |
698365fa | 1555 | af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i); |
827da44c | 1556 | u64_stats_init(&af_inet_stats->syncp); |
827da44c JS |
1557 | } |
1558 | ||
698365fa WC |
1559 | net->mib.net_statistics = alloc_percpu(struct linux_mib); |
1560 | if (!net->mib.net_statistics) | |
61a7e260 | 1561 | goto err_net_mib; |
698365fa WC |
1562 | net->mib.udp_statistics = alloc_percpu(struct udp_mib); |
1563 | if (!net->mib.udp_statistics) | |
2f275f91 | 1564 | goto err_udp_mib; |
698365fa WC |
1565 | net->mib.udplite_statistics = alloc_percpu(struct udp_mib); |
1566 | if (!net->mib.udplite_statistics) | |
386019d3 | 1567 | goto err_udplite_mib; |
698365fa WC |
1568 | net->mib.icmp_statistics = alloc_percpu(struct icmp_mib); |
1569 | if (!net->mib.icmp_statistics) | |
b60538a0 | 1570 | goto err_icmp_mib; |
acb32ba3 ED |
1571 | net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib), |
1572 | GFP_KERNEL); | |
1573 | if (!net->mib.icmpmsg_statistics) | |
923c6586 | 1574 | goto err_icmpmsg_mib; |
57ef42d5 PE |
1575 | |
1576 | tcp_mib_init(net); | |
9b4661bd | 1577 | return 0; |
57ef42d5 | 1578 | |
923c6586 | 1579 | err_icmpmsg_mib: |
698365fa | 1580 | free_percpu(net->mib.icmp_statistics); |
b60538a0 | 1581 | err_icmp_mib: |
698365fa | 1582 | free_percpu(net->mib.udplite_statistics); |
386019d3 | 1583 | err_udplite_mib: |
698365fa | 1584 | free_percpu(net->mib.udp_statistics); |
2f275f91 | 1585 | err_udp_mib: |
698365fa | 1586 | free_percpu(net->mib.net_statistics); |
61a7e260 | 1587 | err_net_mib: |
698365fa | 1588 | free_percpu(net->mib.ip_statistics); |
a20f5799 | 1589 | err_ip_mib: |
698365fa | 1590 | free_percpu(net->mib.tcp_statistics); |
57ef42d5 PE |
1591 | err_tcp_mib: |
1592 | return -ENOMEM; | |
9b4661bd PE |
1593 | } |
1594 | ||
1595 | static __net_exit void ipv4_mib_exit_net(struct net *net) | |
1596 | { | |
acb32ba3 | 1597 | kfree(net->mib.icmpmsg_statistics); |
698365fa WC |
1598 | free_percpu(net->mib.icmp_statistics); |
1599 | free_percpu(net->mib.udplite_statistics); | |
1600 | free_percpu(net->mib.udp_statistics); | |
1601 | free_percpu(net->mib.net_statistics); | |
1602 | free_percpu(net->mib.ip_statistics); | |
1603 | free_percpu(net->mib.tcp_statistics); | |
9b4661bd PE |
1604 | } |
1605 | ||
1606 | static __net_initdata struct pernet_operations ipv4_mib_ops = { | |
1607 | .init = ipv4_mib_init_net, | |
1608 | .exit = ipv4_mib_exit_net, | |
1609 | }; | |
1610 | ||
1da177e4 LT |
1611 | static int __init init_ipv4_mibs(void) |
1612 | { | |
d89cbbb1 | 1613 | return register_pernet_subsys(&ipv4_mib_ops); |
1da177e4 LT |
1614 | } |
1615 | ||
c9d8f1a6 CW |
1616 | static __net_init int inet_init_net(struct net *net) |
1617 | { | |
1618 | /* | |
1619 | * Set defaults for local port range | |
1620 | */ | |
1621 | seqlock_init(&net->ipv4.ip_local_ports.lock); | |
1622 | net->ipv4.ip_local_ports.range[0] = 32768; | |
1623 | net->ipv4.ip_local_ports.range[1] = 61000; | |
ba6b918a CW |
1624 | |
1625 | seqlock_init(&net->ipv4.ping_group_range.lock); | |
1626 | /* | |
1627 | * Sane defaults - nobody may create ping sockets. | |
1628 | * Boot scripts should set this to distro-specific group. | |
1629 | */ | |
1630 | net->ipv4.ping_group_range.range[0] = make_kgid(&init_user_ns, 1); | |
1631 | net->ipv4.ping_group_range.range[1] = make_kgid(&init_user_ns, 0); | |
c9d8f1a6 CW |
1632 | return 0; |
1633 | } | |
1634 | ||
1635 | static __net_exit void inet_exit_net(struct net *net) | |
1636 | { | |
1637 | } | |
1638 | ||
1639 | static __net_initdata struct pernet_operations af_inet_ops = { | |
1640 | .init = inet_init_net, | |
1641 | .exit = inet_exit_net, | |
1642 | }; | |
1643 | ||
1644 | static int __init init_inet_pernet_ops(void) | |
1645 | { | |
1646 | return register_pernet_subsys(&af_inet_ops); | |
1647 | } | |
1648 | ||
1da177e4 | 1649 | static int ipv4_proc_init(void); |
1da177e4 | 1650 | |
30e224d7 HX |
1651 | /* |
1652 | * IP protocol layer initialiser | |
1653 | */ | |
1654 | ||
22061d80 VY |
1655 | static struct packet_offload ip_packet_offload __read_mostly = { |
1656 | .type = cpu_to_be16(ETH_P_IP), | |
f191a1d1 VY |
1657 | .callbacks = { |
1658 | .gso_send_check = inet_gso_send_check, | |
1659 | .gso_segment = inet_gso_segment, | |
1660 | .gro_receive = inet_gro_receive, | |
1661 | .gro_complete = inet_gro_complete, | |
1662 | }, | |
30e224d7 HX |
1663 | }; |
1664 | ||
cb32f511 ED |
1665 | static const struct net_offload ipip_offload = { |
1666 | .callbacks = { | |
1667 | .gso_send_check = inet_gso_send_check, | |
1668 | .gso_segment = inet_gso_segment, | |
9667e9bb TH |
1669 | .gro_receive = inet_gro_receive, |
1670 | .gro_complete = inet_gro_complete, | |
cb32f511 ED |
1671 | }, |
1672 | }; | |
1673 | ||
808a8f88 VY |
1674 | static int __init ipv4_offload_init(void) |
1675 | { | |
1676 | /* | |
1677 | * Add offloads | |
1678 | */ | |
da5bab07 | 1679 | if (udpv4_offload_init() < 0) |
808a8f88 | 1680 | pr_crit("%s: Cannot add UDP protocol offload\n", __func__); |
28850dc7 DB |
1681 | if (tcpv4_offload_init() < 0) |
1682 | pr_crit("%s: Cannot add TCP protocol offload\n", __func__); | |
808a8f88 VY |
1683 | |
1684 | dev_add_offload(&ip_packet_offload); | |
cb32f511 | 1685 | inet_add_offload(&ipip_offload, IPPROTO_IPIP); |
808a8f88 VY |
1686 | return 0; |
1687 | } | |
1688 | ||
1689 | fs_initcall(ipv4_offload_init); | |
1690 | ||
1691 | static struct packet_type ip_packet_type __read_mostly = { | |
1692 | .type = cpu_to_be16(ETH_P_IP), | |
1693 | .func = ip_rcv, | |
1694 | }; | |
1695 | ||
1da177e4 LT |
1696 | static int __init inet_init(void) |
1697 | { | |
1da177e4 LT |
1698 | struct inet_protosw *q; |
1699 | struct list_head *r; | |
1700 | int rc = -EINVAL; | |
1701 | ||
95c7e0e4 | 1702 | BUILD_BUG_ON(sizeof(struct inet_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb)); |
1da177e4 LT |
1703 | |
1704 | rc = proto_register(&tcp_prot, 1); | |
1705 | if (rc) | |
122ff243 | 1706 | goto out; |
1da177e4 LT |
1707 | |
1708 | rc = proto_register(&udp_prot, 1); | |
1709 | if (rc) | |
1710 | goto out_unregister_tcp_proto; | |
1711 | ||
1712 | rc = proto_register(&raw_prot, 1); | |
1713 | if (rc) | |
1714 | goto out_unregister_udp_proto; | |
1715 | ||
c319b4d7 VK |
1716 | rc = proto_register(&ping_prot, 1); |
1717 | if (rc) | |
1718 | goto out_unregister_raw_proto; | |
1719 | ||
1da177e4 | 1720 | /* |
e905a9ed | 1721 | * Tell SOCKET that we are alive... |
1da177e4 LT |
1722 | */ |
1723 | ||
e905a9ed | 1724 | (void)sock_register(&inet_family_ops); |
1da177e4 | 1725 | |
bd7b1533 AV |
1726 | #ifdef CONFIG_SYSCTL |
1727 | ip_static_sysctl_init(); | |
1728 | #endif | |
1729 | ||
1da177e4 LT |
1730 | /* |
1731 | * Add all the base protocols. | |
1732 | */ | |
1733 | ||
1734 | if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0) | |
058bd4d2 | 1735 | pr_crit("%s: Cannot add ICMP protocol\n", __func__); |
1da177e4 | 1736 | if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0) |
058bd4d2 | 1737 | pr_crit("%s: Cannot add UDP protocol\n", __func__); |
1da177e4 | 1738 | if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0) |
058bd4d2 | 1739 | pr_crit("%s: Cannot add TCP protocol\n", __func__); |
1da177e4 LT |
1740 | #ifdef CONFIG_IP_MULTICAST |
1741 | if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0) | |
058bd4d2 | 1742 | pr_crit("%s: Cannot add IGMP protocol\n", __func__); |
1da177e4 LT |
1743 | #endif |
1744 | ||
1745 | /* Register the socket-side information for inet_create. */ | |
1746 | for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r) | |
1747 | INIT_LIST_HEAD(r); | |
1748 | ||
1749 | for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q) | |
1750 | inet_register_protosw(q); | |
1751 | ||
1752 | /* | |
1753 | * Set the ARP module up | |
1754 | */ | |
1755 | ||
1756 | arp_init(); | |
1757 | ||
e905a9ed YH |
1758 | /* |
1759 | * Set the IP module up | |
1760 | */ | |
1da177e4 LT |
1761 | |
1762 | ip_init(); | |
1763 | ||
9b0f976f | 1764 | tcp_v4_init(); |
1da177e4 LT |
1765 | |
1766 | /* Setup TCP slab cache for open requests. */ | |
1767 | tcp_init(); | |
1768 | ||
95766fff HA |
1769 | /* Setup UDP memory threshold */ |
1770 | udp_init(); | |
1771 | ||
ba4e58ec GR |
1772 | /* Add UDP-Lite (RFC 3828) */ |
1773 | udplite4_register(); | |
1da177e4 | 1774 | |
c319b4d7 VK |
1775 | ping_init(); |
1776 | ||
1da177e4 LT |
1777 | /* |
1778 | * Set the ICMP layer up | |
1779 | */ | |
1780 | ||
a5710d65 DL |
1781 | if (icmp_init() < 0) |
1782 | panic("Failed to create the ICMP control socket.\n"); | |
1da177e4 LT |
1783 | |
1784 | /* | |
1785 | * Initialise the multicast router | |
1786 | */ | |
1787 | #if defined(CONFIG_IP_MROUTE) | |
03d2f897 | 1788 | if (ip_mr_init()) |
058bd4d2 | 1789 | pr_crit("%s: Cannot init ipv4 mroute\n", __func__); |
1da177e4 | 1790 | #endif |
c9d8f1a6 CW |
1791 | |
1792 | if (init_inet_pernet_ops()) | |
1793 | pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__); | |
1da177e4 LT |
1794 | /* |
1795 | * Initialise per-cpu ipv4 mibs | |
e905a9ed | 1796 | */ |
1da177e4 | 1797 | |
132adf54 | 1798 | if (init_ipv4_mibs()) |
058bd4d2 | 1799 | pr_crit("%s: Cannot init ipv4 mibs\n", __func__); |
e905a9ed | 1800 | |
1da177e4 LT |
1801 | ipv4_proc_init(); |
1802 | ||
1803 | ipfrag_init(); | |
1804 | ||
30e224d7 HX |
1805 | dev_add_pack(&ip_packet_type); |
1806 | ||
1da177e4 LT |
1807 | rc = 0; |
1808 | out: | |
1809 | return rc; | |
c319b4d7 VK |
1810 | out_unregister_raw_proto: |
1811 | proto_unregister(&raw_prot); | |
1da177e4 LT |
1812 | out_unregister_udp_proto: |
1813 | proto_unregister(&udp_prot); | |
321efff7 OK |
1814 | out_unregister_tcp_proto: |
1815 | proto_unregister(&tcp_prot); | |
1da177e4 LT |
1816 | goto out; |
1817 | } | |
1818 | ||
a536e077 | 1819 | fs_initcall(inet_init); |
1da177e4 LT |
1820 | |
1821 | /* ------------------------------------------------------------------------ */ | |
1822 | ||
1823 | #ifdef CONFIG_PROC_FS | |
1da177e4 LT |
1824 | static int __init ipv4_proc_init(void) |
1825 | { | |
1826 | int rc = 0; | |
1827 | ||
1828 | if (raw_proc_init()) | |
1829 | goto out_raw; | |
1830 | if (tcp4_proc_init()) | |
1831 | goto out_tcp; | |
1832 | if (udp4_proc_init()) | |
1833 | goto out_udp; | |
c319b4d7 VK |
1834 | if (ping_proc_init()) |
1835 | goto out_ping; | |
1da177e4 LT |
1836 | if (ip_misc_proc_init()) |
1837 | goto out_misc; | |
1838 | out: | |
1839 | return rc; | |
1840 | out_misc: | |
c319b4d7 VK |
1841 | ping_proc_exit(); |
1842 | out_ping: | |
1da177e4 LT |
1843 | udp4_proc_exit(); |
1844 | out_udp: | |
1845 | tcp4_proc_exit(); | |
1846 | out_tcp: | |
1847 | raw_proc_exit(); | |
1848 | out_raw: | |
1849 | rc = -ENOMEM; | |
1850 | goto out; | |
1851 | } | |
1852 | ||
1853 | #else /* CONFIG_PROC_FS */ | |
1854 | static int __init ipv4_proc_init(void) | |
1855 | { | |
1856 | return 0; | |
1857 | } | |
1858 | #endif /* CONFIG_PROC_FS */ | |
1859 | ||
1860 | MODULE_ALIAS_NETPROTO(PF_INET); | |
1861 |