Commit | Line | Data |
---|---|---|
2874c5fd | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
1da177e4 LT |
2 | /* |
3 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
4 | * operating system. INET is implemented using the BSD Socket | |
5 | * interface as the means of communication with the user level. | |
6 | * | |
7 | * PF_INET protocol family socket handler. | |
8 | * | |
02c30a84 | 9 | * Authors: Ross Biro |
1da177e4 LT |
10 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
11 | * Florian La Roche, <flla@stud.uni-sb.de> | |
12 | * Alan Cox, <A.Cox@swansea.ac.uk> | |
13 | * | |
14 | * Changes (see also sock.c) | |
15 | * | |
16 | * piggy, | |
17 | * Karl Knutson : Socket protocol table | |
18 | * A.N.Kuznetsov : Socket death error in accept(). | |
19 | * John Richardson : Fix non blocking error in connect() | |
20 | * so sockets that fail to connect | |
21 | * don't return -EINPROGRESS. | |
22 | * Alan Cox : Asynchronous I/O support | |
23 | * Alan Cox : Keep correct socket pointer on sock | |
24 | * structures | |
25 | * when accept() ed | |
26 | * Alan Cox : Semantics of SO_LINGER aren't state | |
27 | * moved to close when you look carefully. | |
28 | * With this fixed and the accept bug fixed | |
29 | * some RPC stuff seems happier. | |
30 | * Niibe Yutaka : 4.4BSD style write async I/O | |
31 | * Alan Cox, | |
32 | * Tony Gale : Fixed reuse semantics. | |
33 | * Alan Cox : bind() shouldn't abort existing but dead | |
34 | * sockets. Stops FTP netin:.. I hope. | |
35 | * Alan Cox : bind() works correctly for RAW sockets. | |
36 | * Note that FreeBSD at least was broken | |
37 | * in this respect so be careful with | |
38 | * compatibility tests... | |
39 | * Alan Cox : routing cache support | |
40 | * Alan Cox : memzero the socket structure for | |
41 | * compactness. | |
42 | * Matt Day : nonblock connect error handler | |
43 | * Alan Cox : Allow large numbers of pending sockets | |
44 | * (eg for big web sites), but only if | |
45 | * specifically application requested. | |
46 | * Alan Cox : New buffering throughout IP. Used | |
47 | * dumbly. | |
48 | * Alan Cox : New buffering now used smartly. | |
49 | * Alan Cox : BSD rather than common sense | |
50 | * interpretation of listen. | |
51 | * Germano Caronni : Assorted small races. | |
52 | * Alan Cox : sendmsg/recvmsg basic support. | |
53 | * Alan Cox : Only sendmsg/recvmsg now supported. | |
54 | * Alan Cox : Locked down bind (see security list). | |
55 | * Alan Cox : Loosened bind a little. | |
56 | * Mike McLagan : ADD/DEL DLCI Ioctls | |
57 | * Willy Konynenberg : Transparent proxying support. | |
58 | * David S. Miller : New socket lookup architecture. | |
59 | * Some other random speedups. | |
60 | * Cyrus Durgin : Cleaned up file for kmod hacks. | |
61 | * Andi Kleen : Fix inet_stream_connect TCP race. | |
1da177e4 LT |
62 | */ |
63 | ||
afd46503 JP |
64 | #define pr_fmt(fmt) "IPv4: " fmt |
65 | ||
f4c50d99 | 66 | #include <linux/err.h> |
1da177e4 LT |
67 | #include <linux/errno.h> |
68 | #include <linux/types.h> | |
69 | #include <linux/socket.h> | |
70 | #include <linux/in.h> | |
71 | #include <linux/kernel.h> | |
d3fc0353 | 72 | #include <linux/kmod.h> |
1da177e4 LT |
73 | #include <linux/sched.h> |
74 | #include <linux/timer.h> | |
75 | #include <linux/string.h> | |
76 | #include <linux/sockios.h> | |
77 | #include <linux/net.h> | |
4fc268d2 | 78 | #include <linux/capability.h> |
1da177e4 LT |
79 | #include <linux/fcntl.h> |
80 | #include <linux/mm.h> | |
81 | #include <linux/interrupt.h> | |
82 | #include <linux/stat.h> | |
83 | #include <linux/init.h> | |
84 | #include <linux/poll.h> | |
85 | #include <linux/netfilter_ipv4.h> | |
b3da2cf3 | 86 | #include <linux/random.h> |
5a0e3ad6 | 87 | #include <linux/slab.h> |
1da177e4 | 88 | |
7c0f6ba6 | 89 | #include <linux/uaccess.h> |
1da177e4 | 90 | |
1da177e4 LT |
91 | #include <linux/inet.h> |
92 | #include <linux/igmp.h> | |
14c85021 | 93 | #include <linux/inetdevice.h> |
1da177e4 | 94 | #include <linux/netdevice.h> |
73cc19f1 | 95 | #include <net/checksum.h> |
1da177e4 LT |
96 | #include <net/ip.h> |
97 | #include <net/protocol.h> | |
98 | #include <net/arp.h> | |
99 | #include <net/route.h> | |
100 | #include <net/ip_fib.h> | |
295f7324 | 101 | #include <net/inet_connection_sock.h> |
1da177e4 LT |
102 | #include <net/tcp.h> |
103 | #include <net/udp.h> | |
ba4e58ec | 104 | #include <net/udplite.h> |
c319b4d7 | 105 | #include <net/ping.h> |
1da177e4 LT |
106 | #include <linux/skbuff.h> |
107 | #include <net/sock.h> | |
108 | #include <net/raw.h> | |
109 | #include <net/icmp.h> | |
1da177e4 | 110 | #include <net/inet_common.h> |
045a0fa0 | 111 | #include <net/ip_tunnels.h> |
1da177e4 | 112 | #include <net/xfrm.h> |
9b4661bd | 113 | #include <net/net_namespace.h> |
aebda156 | 114 | #include <net/secure_seq.h> |
1da177e4 LT |
115 | #ifdef CONFIG_IP_MROUTE |
116 | #include <linux/mroute.h> | |
117 | #endif | |
3236b004 | 118 | #include <net/l3mdev.h> |
1da177e4 | 119 | |
563e0bb0 | 120 | #include <trace/events/sock.h> |
1da177e4 LT |
121 | |
122 | /* The inetsw table contains everything that inet_create needs to | |
123 | * build a new socket. | |
124 | */ | |
125 | static struct list_head inetsw[SOCK_MAX]; | |
126 | static DEFINE_SPINLOCK(inetsw_lock); | |
127 | ||
128 | /* New destruction routine */ | |
129 | ||
130 | void inet_sock_destruct(struct sock *sk) | |
131 | { | |
132 | struct inet_sock *inet = inet_sk(sk); | |
133 | ||
134 | __skb_queue_purge(&sk->sk_receive_queue); | |
8b27dae5 ED |
135 | if (sk->sk_rx_skb_cache) { |
136 | __kfree_skb(sk->sk_rx_skb_cache); | |
137 | sk->sk_rx_skb_cache = NULL; | |
138 | } | |
1da177e4 LT |
139 | __skb_queue_purge(&sk->sk_error_queue); |
140 | ||
95766fff HA |
141 | sk_mem_reclaim(sk); |
142 | ||
1da177e4 | 143 | if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) { |
3d1427f8 | 144 | pr_err("Attempt to release TCP socket in state %d %p\n", |
1da177e4 LT |
145 | sk->sk_state, sk); |
146 | return; | |
147 | } | |
148 | if (!sock_flag(sk, SOCK_DEAD)) { | |
3d1427f8 | 149 | pr_err("Attempt to release alive inet socket %p\n", sk); |
1da177e4 LT |
150 | return; |
151 | } | |
152 | ||
547b792c | 153 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); |
14afee4b | 154 | WARN_ON(refcount_read(&sk->sk_wmem_alloc)); |
547b792c IJ |
155 | WARN_ON(sk->sk_wmem_queued); |
156 | WARN_ON(sk->sk_forward_alloc); | |
1da177e4 | 157 | |
f6d8bd05 | 158 | kfree(rcu_dereference_protected(inet->inet_opt, 1)); |
5869b8fa | 159 | dst_release(rcu_dereference_protected(sk->sk_dst_cache, 1)); |
41063e9d | 160 | dst_release(sk->sk_rx_dst); |
e6848976 | 161 | sk_refcnt_debug_dec(sk); |
1da177e4 | 162 | } |
3d1427f8 | 163 | EXPORT_SYMBOL(inet_sock_destruct); |
1da177e4 LT |
164 | |
165 | /* | |
166 | * The routines beyond this point handle the behaviour of an AF_INET | |
167 | * socket object. Mostly it punts to the subprotocols of IP to do | |
168 | * the work. | |
169 | */ | |
170 | ||
171 | /* | |
172 | * Automatically bind an unbound socket. | |
173 | */ | |
174 | ||
175 | static int inet_autobind(struct sock *sk) | |
176 | { | |
177 | struct inet_sock *inet; | |
178 | /* We may need to bind the socket. */ | |
179 | lock_sock(sk); | |
180 | inet = inet_sk(sk); | |
c720c7e8 | 181 | if (!inet->inet_num) { |
1da177e4 LT |
182 | if (sk->sk_prot->get_port(sk, 0)) { |
183 | release_sock(sk); | |
184 | return -EAGAIN; | |
185 | } | |
c720c7e8 | 186 | inet->inet_sport = htons(inet->inet_num); |
1da177e4 LT |
187 | } |
188 | release_sock(sk); | |
189 | return 0; | |
190 | } | |
191 | ||
192 | /* | |
193 | * Move a socket into listening state. | |
194 | */ | |
195 | int inet_listen(struct socket *sock, int backlog) | |
196 | { | |
197 | struct sock *sk = sock->sk; | |
198 | unsigned char old_state; | |
e1cfcbe8 | 199 | int err, tcp_fastopen; |
1da177e4 LT |
200 | |
201 | lock_sock(sk); | |
202 | ||
203 | err = -EINVAL; | |
204 | if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM) | |
205 | goto out; | |
206 | ||
207 | old_state = sk->sk_state; | |
208 | if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN))) | |
209 | goto out; | |
210 | ||
1295e2cf | 211 | sk->sk_max_ack_backlog = backlog; |
1da177e4 LT |
212 | /* Really, if the socket is already in listen state |
213 | * we can only allow the backlog to be adjusted. | |
214 | */ | |
215 | if (old_state != TCP_LISTEN) { | |
cebc5cba | 216 | /* Enable TFO w/o requiring TCP_FASTOPEN socket option. |
8336886f | 217 | * Note that only TCP sockets (SOCK_STREAM) will reach here. |
cebc5cba YC |
218 | * Also fastopen backlog may already been set via the option |
219 | * because the socket was in TCP_LISTEN state previously but | |
220 | * was shutdown() rather than close(). | |
8336886f | 221 | */ |
e1cfcbe8 HY |
222 | tcp_fastopen = sock_net(sk)->ipv4.sysctl_tcp_fastopen; |
223 | if ((tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) && | |
224 | (tcp_fastopen & TFO_SERVER_ENABLE) && | |
0536fcc0 | 225 | !inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) { |
cebc5cba | 226 | fastopen_queue_tune(sk, backlog); |
43713848 | 227 | tcp_fastopen_init_key_once(sock_net(sk)); |
8336886f | 228 | } |
cebc5cba | 229 | |
72a3effa | 230 | err = inet_csk_listen_start(sk, backlog); |
1da177e4 LT |
231 | if (err) |
232 | goto out; | |
f333ee0c | 233 | tcp_call_bpf(sk, BPF_SOCK_OPS_TCP_LISTEN_CB, 0, NULL); |
1da177e4 | 234 | } |
1da177e4 LT |
235 | err = 0; |
236 | ||
237 | out: | |
238 | release_sock(sk); | |
239 | return err; | |
240 | } | |
3d1427f8 | 241 | EXPORT_SYMBOL(inet_listen); |
1da177e4 LT |
242 | |
243 | /* | |
244 | * Create an inet socket. | |
245 | */ | |
246 | ||
3f378b68 EP |
247 | static int inet_create(struct net *net, struct socket *sock, int protocol, |
248 | int kern) | |
1da177e4 LT |
249 | { |
250 | struct sock *sk; | |
1da177e4 LT |
251 | struct inet_protosw *answer; |
252 | struct inet_sock *inet; | |
253 | struct proto *answer_prot; | |
254 | unsigned char answer_flags; | |
bb97d31f | 255 | int try_loading_module = 0; |
86c8f9d1 | 256 | int err; |
1da177e4 | 257 | |
79462ad0 HFS |
258 | if (protocol < 0 || protocol >= IPPROTO_MAX) |
259 | return -EINVAL; | |
260 | ||
1da177e4 LT |
261 | sock->state = SS_UNCONNECTED; |
262 | ||
263 | /* Look for the requested type/protocol pair. */ | |
bb97d31f | 264 | lookup_protocol: |
86c8f9d1 | 265 | err = -ESOCKTNOSUPPORT; |
1da177e4 | 266 | rcu_read_lock(); |
696adfe8 | 267 | list_for_each_entry_rcu(answer, &inetsw[sock->type], list) { |
1da177e4 | 268 | |
696adfe8 | 269 | err = 0; |
1da177e4 LT |
270 | /* Check the non-wild match. */ |
271 | if (protocol == answer->protocol) { | |
272 | if (protocol != IPPROTO_IP) | |
273 | break; | |
274 | } else { | |
275 | /* Check for the two wild cases. */ | |
276 | if (IPPROTO_IP == protocol) { | |
277 | protocol = answer->protocol; | |
278 | break; | |
279 | } | |
280 | if (IPPROTO_IP == answer->protocol) | |
281 | break; | |
282 | } | |
86c8f9d1 | 283 | err = -EPROTONOSUPPORT; |
1da177e4 LT |
284 | } |
285 | ||
696adfe8 | 286 | if (unlikely(err)) { |
bb97d31f ACM |
287 | if (try_loading_module < 2) { |
288 | rcu_read_unlock(); | |
289 | /* | |
290 | * Be more specific, e.g. net-pf-2-proto-132-type-1 | |
291 | * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM) | |
292 | */ | |
293 | if (++try_loading_module == 1) | |
294 | request_module("net-pf-%d-proto-%d-type-%d", | |
295 | PF_INET, protocol, sock->type); | |
296 | /* | |
297 | * Fall back to generic, e.g. net-pf-2-proto-132 | |
298 | * (net-pf-PF_INET-proto-IPPROTO_SCTP) | |
299 | */ | |
300 | else | |
301 | request_module("net-pf-%d-proto-%d", | |
302 | PF_INET, protocol); | |
303 | goto lookup_protocol; | |
304 | } else | |
305 | goto out_rcu_unlock; | |
306 | } | |
307 | ||
1da177e4 | 308 | err = -EPERM; |
52e804c6 EB |
309 | if (sock->type == SOCK_RAW && !kern && |
310 | !ns_capable(net->user_ns, CAP_NET_RAW)) | |
1da177e4 | 311 | goto out_rcu_unlock; |
1da177e4 LT |
312 | |
313 | sock->ops = answer->ops; | |
314 | answer_prot = answer->prot; | |
1da177e4 LT |
315 | answer_flags = answer->flags; |
316 | rcu_read_unlock(); | |
317 | ||
51456b29 | 318 | WARN_ON(!answer_prot->slab); |
1da177e4 LT |
319 | |
320 | err = -ENOBUFS; | |
11aa9c28 | 321 | sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot, kern); |
51456b29 | 322 | if (!sk) |
1da177e4 LT |
323 | goto out; |
324 | ||
325 | err = 0; | |
1da177e4 | 326 | if (INET_PROTOSW_REUSE & answer_flags) |
4a17fd52 | 327 | sk->sk_reuse = SK_CAN_REUSE; |
1da177e4 LT |
328 | |
329 | inet = inet_sk(sk); | |
469de9b9 | 330 | inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; |
1da177e4 | 331 | |
7b2ff18e JO |
332 | inet->nodefrag = 0; |
333 | ||
1da177e4 | 334 | if (SOCK_RAW == sock->type) { |
c720c7e8 | 335 | inet->inet_num = protocol; |
1da177e4 LT |
336 | if (IPPROTO_RAW == protocol) |
337 | inet->hdrincl = 1; | |
338 | } | |
339 | ||
974eda11 | 340 | if (net->ipv4.sysctl_ip_no_pmtu_disc) |
1da177e4 LT |
341 | inet->pmtudisc = IP_PMTUDISC_DONT; |
342 | else | |
343 | inet->pmtudisc = IP_PMTUDISC_WANT; | |
344 | ||
c720c7e8 | 345 | inet->inet_id = 0; |
1da177e4 LT |
346 | |
347 | sock_init_data(sock, sk); | |
348 | ||
349 | sk->sk_destruct = inet_sock_destruct; | |
1da177e4 LT |
350 | sk->sk_protocol = protocol; |
351 | sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv; | |
352 | ||
353 | inet->uc_ttl = -1; | |
354 | inet->mc_loop = 1; | |
355 | inet->mc_ttl = 1; | |
f771bef9 | 356 | inet->mc_all = 1; |
1da177e4 LT |
357 | inet->mc_index = 0; |
358 | inet->mc_list = NULL; | |
4c507d28 | 359 | inet->rcv_tos = 0; |
1da177e4 | 360 | |
e6848976 | 361 | sk_refcnt_debug_inc(sk); |
1da177e4 | 362 | |
c720c7e8 | 363 | if (inet->inet_num) { |
1da177e4 LT |
364 | /* It assumes that any protocol which allows |
365 | * the user to assign a number at socket | |
366 | * creation time automatically | |
367 | * shares. | |
368 | */ | |
c720c7e8 | 369 | inet->inet_sport = htons(inet->inet_num); |
1da177e4 | 370 | /* Add to protocol hash chains. */ |
086c653f CG |
371 | err = sk->sk_prot->hash(sk); |
372 | if (err) { | |
373 | sk_common_release(sk); | |
374 | goto out; | |
375 | } | |
1da177e4 LT |
376 | } |
377 | ||
378 | if (sk->sk_prot->init) { | |
379 | err = sk->sk_prot->init(sk); | |
61023658 DA |
380 | if (err) { |
381 | sk_common_release(sk); | |
382 | goto out; | |
383 | } | |
384 | } | |
385 | ||
386 | if (!kern) { | |
387 | err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk); | |
388 | if (err) { | |
1da177e4 | 389 | sk_common_release(sk); |
61023658 DA |
390 | goto out; |
391 | } | |
1da177e4 LT |
392 | } |
393 | out: | |
394 | return err; | |
395 | out_rcu_unlock: | |
396 | rcu_read_unlock(); | |
397 | goto out; | |
398 | } | |
399 | ||
400 | ||
401 | /* | |
402 | * The peer socket should always be NULL (or else). When we call this | |
403 | * function we are destroying the object and from then on nobody | |
404 | * should refer to it. | |
405 | */ | |
406 | int inet_release(struct socket *sock) | |
407 | { | |
408 | struct sock *sk = sock->sk; | |
409 | ||
410 | if (sk) { | |
411 | long timeout; | |
412 | ||
413 | /* Applications forget to leave groups before exiting */ | |
414 | ip_mc_drop_socket(sk); | |
415 | ||
416 | /* If linger is set, we don't return until the close | |
417 | * is complete. Otherwise we return immediately. The | |
418 | * actually closing is done the same either way. | |
419 | * | |
420 | * If the close is due to the process exiting, we never | |
421 | * linger.. | |
422 | */ | |
423 | timeout = 0; | |
424 | if (sock_flag(sk, SOCK_LINGER) && | |
425 | !(current->flags & PF_EXITING)) | |
426 | timeout = sk->sk_lingertime; | |
427 | sock->sk = NULL; | |
428 | sk->sk_prot->close(sk, timeout); | |
429 | } | |
430 | return 0; | |
431 | } | |
3d1427f8 | 432 | EXPORT_SYMBOL(inet_release); |
1da177e4 | 433 | |
1da177e4 LT |
434 | int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) |
435 | { | |
1da177e4 | 436 | struct sock *sk = sock->sk; |
1da177e4 LT |
437 | int err; |
438 | ||
439 | /* If the socket has its own bind function then use it. (RAW) */ | |
440 | if (sk->sk_prot->bind) { | |
3679d585 | 441 | return sk->sk_prot->bind(sk, uaddr, addr_len); |
1da177e4 | 442 | } |
1da177e4 | 443 | if (addr_len < sizeof(struct sockaddr_in)) |
3679d585 | 444 | return -EINVAL; |
4fbac77d AI |
445 | |
446 | /* BPF prog is run before any checks are done so that if the prog | |
447 | * changes context in a wrong way it will be caught. | |
448 | */ | |
449 | err = BPF_CGROUP_RUN_PROG_INET4_BIND(sk, uaddr); | |
450 | if (err) | |
3679d585 AI |
451 | return err; |
452 | ||
453 | return __inet_bind(sk, uaddr, addr_len, false, true); | |
454 | } | |
455 | EXPORT_SYMBOL(inet_bind); | |
456 | ||
457 | int __inet_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len, | |
458 | bool force_bind_address_no_port, bool with_lock) | |
459 | { | |
460 | struct sockaddr_in *addr = (struct sockaddr_in *)uaddr; | |
461 | struct inet_sock *inet = inet_sk(sk); | |
462 | struct net *net = sock_net(sk); | |
463 | unsigned short snum; | |
464 | int chk_addr_ret; | |
465 | u32 tb_id = RT_TABLE_LOCAL; | |
466 | int err; | |
1da177e4 | 467 | |
c349a528 | 468 | if (addr->sin_family != AF_INET) { |
29c486df ED |
469 | /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET) |
470 | * only if s_addr is INADDR_ANY. | |
471 | */ | |
c349a528 | 472 | err = -EAFNOSUPPORT; |
29c486df ED |
473 | if (addr->sin_family != AF_UNSPEC || |
474 | addr->sin_addr.s_addr != htonl(INADDR_ANY)) | |
475 | goto out; | |
c349a528 | 476 | } |
d0733d2e | 477 | |
3236b004 | 478 | tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id; |
30bbaa19 | 479 | chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id); |
1da177e4 LT |
480 | |
481 | /* Not specified by any standard per-se, however it breaks too | |
482 | * many applications when removed. It is unfortunate since | |
483 | * allowing applications to make a non-local bind solves | |
484 | * several problems with systems using dynamic addressing. | |
485 | * (ie. your servers still start up even if your ISDN link | |
486 | * is temporarily down) | |
487 | */ | |
488 | err = -EADDRNOTAVAIL; | |
83ba4645 | 489 | if (!inet_can_nonlocal_bind(net, inet) && |
e6f1cebf | 490 | addr->sin_addr.s_addr != htonl(INADDR_ANY) && |
1da177e4 LT |
491 | chk_addr_ret != RTN_LOCAL && |
492 | chk_addr_ret != RTN_MULTICAST && | |
493 | chk_addr_ret != RTN_BROADCAST) | |
494 | goto out; | |
495 | ||
496 | snum = ntohs(addr->sin_port); | |
497 | err = -EACCES; | |
4548b683 | 498 | if (snum && snum < inet_prot_sock(net) && |
3594698a | 499 | !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) |
1da177e4 LT |
500 | goto out; |
501 | ||
502 | /* We keep a pair of addresses. rcv_saddr is the one | |
503 | * used by hash lookups, and saddr is used for transmit. | |
504 | * | |
505 | * In the BSD API these are the same except where it | |
506 | * would be illegal to use them (multicast/broadcast) in | |
507 | * which case the sending device address is used. | |
508 | */ | |
3679d585 AI |
509 | if (with_lock) |
510 | lock_sock(sk); | |
1da177e4 LT |
511 | |
512 | /* Check these errors (active socket, double bind). */ | |
513 | err = -EINVAL; | |
c720c7e8 | 514 | if (sk->sk_state != TCP_CLOSE || inet->inet_num) |
1da177e4 LT |
515 | goto out_release_sock; |
516 | ||
c720c7e8 | 517 | inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; |
1da177e4 | 518 | if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) |
c720c7e8 | 519 | inet->inet_saddr = 0; /* Use device */ |
1da177e4 LT |
520 | |
521 | /* Make sure we are allowed to bind here. */ | |
aac3fc32 AI |
522 | if (snum || !(inet->bind_address_no_port || |
523 | force_bind_address_no_port)) { | |
524 | if (sk->sk_prot->get_port(sk, snum)) { | |
525 | inet->inet_saddr = inet->inet_rcv_saddr = 0; | |
526 | err = -EADDRINUSE; | |
527 | goto out_release_sock; | |
528 | } | |
529 | err = BPF_CGROUP_RUN_PROG_INET4_POST_BIND(sk); | |
530 | if (err) { | |
531 | inet->inet_saddr = inet->inet_rcv_saddr = 0; | |
532 | goto out_release_sock; | |
533 | } | |
1da177e4 LT |
534 | } |
535 | ||
c720c7e8 | 536 | if (inet->inet_rcv_saddr) |
1da177e4 LT |
537 | sk->sk_userlocks |= SOCK_BINDADDR_LOCK; |
538 | if (snum) | |
539 | sk->sk_userlocks |= SOCK_BINDPORT_LOCK; | |
c720c7e8 ED |
540 | inet->inet_sport = htons(inet->inet_num); |
541 | inet->inet_daddr = 0; | |
542 | inet->inet_dport = 0; | |
1da177e4 LT |
543 | sk_dst_reset(sk); |
544 | err = 0; | |
545 | out_release_sock: | |
3679d585 AI |
546 | if (with_lock) |
547 | release_sock(sk); | |
1da177e4 LT |
548 | out: |
549 | return err; | |
550 | } | |
551 | ||
5e73ea1a | 552 | int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr, |
1da177e4 LT |
553 | int addr_len, int flags) |
554 | { | |
555 | struct sock *sk = sock->sk; | |
d74bad4e | 556 | int err; |
1da177e4 | 557 | |
6503d961 CG |
558 | if (addr_len < sizeof(uaddr->sa_family)) |
559 | return -EINVAL; | |
1da177e4 LT |
560 | if (uaddr->sa_family == AF_UNSPEC) |
561 | return sk->sk_prot->disconnect(sk, flags); | |
562 | ||
d74bad4e AI |
563 | if (BPF_CGROUP_PRE_CONNECT_ENABLED(sk)) { |
564 | err = sk->sk_prot->pre_connect(sk, uaddr, addr_len); | |
565 | if (err) | |
566 | return err; | |
567 | } | |
568 | ||
c720c7e8 | 569 | if (!inet_sk(sk)->inet_num && inet_autobind(sk)) |
1da177e4 | 570 | return -EAGAIN; |
e3192690 | 571 | return sk->sk_prot->connect(sk, uaddr, addr_len); |
1da177e4 | 572 | } |
3d1427f8 | 573 | EXPORT_SYMBOL(inet_dgram_connect); |
1da177e4 | 574 | |
783237e8 | 575 | static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias) |
1da177e4 | 576 | { |
14135f30 | 577 | DEFINE_WAIT_FUNC(wait, woken_wake_function); |
1da177e4 | 578 | |
14135f30 | 579 | add_wait_queue(sk_sleep(sk), &wait); |
783237e8 | 580 | sk->sk_write_pending += writebias; |
1da177e4 LT |
581 | |
582 | /* Basic assumption: if someone sets sk->sk_err, he _must_ | |
583 | * change state of the socket from TCP_SYN_*. | |
584 | * Connect() does not allow to get error notifications | |
585 | * without closing the socket. | |
586 | */ | |
587 | while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
588 | release_sock(sk); | |
14135f30 | 589 | timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); |
1da177e4 LT |
590 | lock_sock(sk); |
591 | if (signal_pending(current) || !timeo) | |
592 | break; | |
1da177e4 | 593 | } |
14135f30 | 594 | remove_wait_queue(sk_sleep(sk), &wait); |
783237e8 | 595 | sk->sk_write_pending -= writebias; |
1da177e4 LT |
596 | return timeo; |
597 | } | |
598 | ||
599 | /* | |
600 | * Connect to a remote host. There is regrettably still a little | |
601 | * TCP 'magic' in here. | |
602 | */ | |
cf60af03 | 603 | int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, |
3979ad7e | 604 | int addr_len, int flags, int is_sendmsg) |
1da177e4 LT |
605 | { |
606 | struct sock *sk = sock->sk; | |
607 | int err; | |
608 | long timeo; | |
609 | ||
19f6d3f3 WW |
610 | /* |
611 | * uaddr can be NULL and addr_len can be 0 if: | |
612 | * sk is a TCP fastopen active socket and | |
613 | * TCP_FASTOPEN_CONNECT sockopt is set and | |
614 | * we already have a valid cookie for this socket. | |
615 | * In this case, user can call write() after connect(). | |
616 | * write() will invoke tcp_sendmsg_fastopen() which calls | |
617 | * __inet_stream_connect(). | |
618 | */ | |
619 | if (uaddr) { | |
620 | if (addr_len < sizeof(uaddr->sa_family)) | |
621 | return -EINVAL; | |
6503d961 | 622 | |
19f6d3f3 WW |
623 | if (uaddr->sa_family == AF_UNSPEC) { |
624 | err = sk->sk_prot->disconnect(sk, flags); | |
625 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
626 | goto out; | |
627 | } | |
1da177e4 LT |
628 | } |
629 | ||
630 | switch (sock->state) { | |
631 | default: | |
632 | err = -EINVAL; | |
633 | goto out; | |
634 | case SS_CONNECTED: | |
635 | err = -EISCONN; | |
636 | goto out; | |
637 | case SS_CONNECTING: | |
19f6d3f3 | 638 | if (inet_sk(sk)->defer_connect) |
3979ad7e | 639 | err = is_sendmsg ? -EINPROGRESS : -EISCONN; |
19f6d3f3 WW |
640 | else |
641 | err = -EALREADY; | |
1da177e4 LT |
642 | /* Fall out of switch with err, set for this state */ |
643 | break; | |
644 | case SS_UNCONNECTED: | |
645 | err = -EISCONN; | |
646 | if (sk->sk_state != TCP_CLOSE) | |
647 | goto out; | |
648 | ||
d74bad4e AI |
649 | if (BPF_CGROUP_PRE_CONNECT_ENABLED(sk)) { |
650 | err = sk->sk_prot->pre_connect(sk, uaddr, addr_len); | |
651 | if (err) | |
652 | goto out; | |
653 | } | |
654 | ||
1da177e4 LT |
655 | err = sk->sk_prot->connect(sk, uaddr, addr_len); |
656 | if (err < 0) | |
657 | goto out; | |
658 | ||
e905a9ed | 659 | sock->state = SS_CONNECTING; |
1da177e4 | 660 | |
19f6d3f3 WW |
661 | if (!err && inet_sk(sk)->defer_connect) |
662 | goto out; | |
663 | ||
1da177e4 LT |
664 | /* Just entered SS_CONNECTING state; the only |
665 | * difference is that return value in non-blocking | |
666 | * case is EINPROGRESS, rather than EALREADY. | |
667 | */ | |
668 | err = -EINPROGRESS; | |
669 | break; | |
670 | } | |
671 | ||
672 | timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); | |
673 | ||
674 | if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
783237e8 YC |
675 | int writebias = (sk->sk_protocol == IPPROTO_TCP) && |
676 | tcp_sk(sk)->fastopen_req && | |
677 | tcp_sk(sk)->fastopen_req->data ? 1 : 0; | |
678 | ||
1da177e4 | 679 | /* Error code is set above */ |
783237e8 | 680 | if (!timeo || !inet_wait_for_connect(sk, timeo, writebias)) |
1da177e4 LT |
681 | goto out; |
682 | ||
683 | err = sock_intr_errno(timeo); | |
684 | if (signal_pending(current)) | |
685 | goto out; | |
686 | } | |
687 | ||
688 | /* Connection was closed by RST, timeout, ICMP error | |
689 | * or another process disconnected us. | |
690 | */ | |
691 | if (sk->sk_state == TCP_CLOSE) | |
692 | goto sock_error; | |
693 | ||
694 | /* sk->sk_err may be not zero now, if RECVERR was ordered by user | |
695 | * and error was received after socket entered established state. | |
696 | * Hence, it is handled normally after connect() return successfully. | |
697 | */ | |
698 | ||
699 | sock->state = SS_CONNECTED; | |
700 | err = 0; | |
701 | out: | |
1da177e4 LT |
702 | return err; |
703 | ||
704 | sock_error: | |
705 | err = sock_error(sk) ? : -ECONNABORTED; | |
706 | sock->state = SS_UNCONNECTED; | |
707 | if (sk->sk_prot->disconnect(sk, flags)) | |
708 | sock->state = SS_DISCONNECTING; | |
709 | goto out; | |
710 | } | |
cf60af03 YC |
711 | EXPORT_SYMBOL(__inet_stream_connect); |
712 | ||
713 | int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, | |
714 | int addr_len, int flags) | |
715 | { | |
716 | int err; | |
717 | ||
718 | lock_sock(sock->sk); | |
3979ad7e | 719 | err = __inet_stream_connect(sock, uaddr, addr_len, flags, 0); |
cf60af03 YC |
720 | release_sock(sock->sk); |
721 | return err; | |
722 | } | |
3d1427f8 | 723 | EXPORT_SYMBOL(inet_stream_connect); |
1da177e4 LT |
724 | |
725 | /* | |
726 | * Accept a pending connection. The TCP layer now gives BSD semantics. | |
727 | */ | |
728 | ||
cdfbabfb DH |
729 | int inet_accept(struct socket *sock, struct socket *newsock, int flags, |
730 | bool kern) | |
1da177e4 LT |
731 | { |
732 | struct sock *sk1 = sock->sk; | |
733 | int err = -EINVAL; | |
cdfbabfb | 734 | struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err, kern); |
1da177e4 LT |
735 | |
736 | if (!sk2) | |
737 | goto do_err; | |
738 | ||
739 | lock_sock(sk2); | |
740 | ||
1eddcead | 741 | sock_rps_record_flow(sk2); |
547b792c | 742 | WARN_ON(!((1 << sk2->sk_state) & |
8336886f JC |
743 | (TCPF_ESTABLISHED | TCPF_SYN_RECV | |
744 | TCPF_CLOSE_WAIT | TCPF_CLOSE))); | |
1da177e4 LT |
745 | |
746 | sock_graft(sk2, newsock); | |
747 | ||
748 | newsock->state = SS_CONNECTED; | |
749 | err = 0; | |
750 | release_sock(sk2); | |
751 | do_err: | |
752 | return err; | |
753 | } | |
3d1427f8 | 754 | EXPORT_SYMBOL(inet_accept); |
1da177e4 LT |
755 | |
756 | ||
757 | /* | |
758 | * This does both peername and sockname. | |
759 | */ | |
760 | int inet_getname(struct socket *sock, struct sockaddr *uaddr, | |
9b2c45d4 | 761 | int peer) |
1da177e4 LT |
762 | { |
763 | struct sock *sk = sock->sk; | |
764 | struct inet_sock *inet = inet_sk(sk); | |
38bfd8f5 | 765 | DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr); |
1da177e4 LT |
766 | |
767 | sin->sin_family = AF_INET; | |
768 | if (peer) { | |
c720c7e8 | 769 | if (!inet->inet_dport || |
1da177e4 LT |
770 | (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && |
771 | peer == 1)) | |
772 | return -ENOTCONN; | |
c720c7e8 ED |
773 | sin->sin_port = inet->inet_dport; |
774 | sin->sin_addr.s_addr = inet->inet_daddr; | |
1da177e4 | 775 | } else { |
c720c7e8 | 776 | __be32 addr = inet->inet_rcv_saddr; |
1da177e4 | 777 | if (!addr) |
c720c7e8 ED |
778 | addr = inet->inet_saddr; |
779 | sin->sin_port = inet->inet_sport; | |
1da177e4 LT |
780 | sin->sin_addr.s_addr = addr; |
781 | } | |
782 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); | |
9b2c45d4 | 783 | return sizeof(*sin); |
1da177e4 | 784 | } |
3d1427f8 | 785 | EXPORT_SYMBOL(inet_getname); |
1da177e4 | 786 | |
1b784140 | 787 | int inet_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) |
1da177e4 LT |
788 | { |
789 | struct sock *sk = sock->sk; | |
790 | ||
c58dc01b | 791 | sock_rps_record_flow(sk); |
fec5e652 | 792 | |
1da177e4 | 793 | /* We may need to bind the socket. */ |
7ba42910 CG |
794 | if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind && |
795 | inet_autobind(sk)) | |
1da177e4 LT |
796 | return -EAGAIN; |
797 | ||
1b784140 | 798 | return sk->sk_prot->sendmsg(sk, msg, size); |
1da177e4 | 799 | } |
3d1427f8 | 800 | EXPORT_SYMBOL(inet_sendmsg); |
1da177e4 | 801 | |
7ba42910 CG |
802 | ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, |
803 | size_t size, int flags) | |
1da177e4 LT |
804 | { |
805 | struct sock *sk = sock->sk; | |
806 | ||
c58dc01b | 807 | sock_rps_record_flow(sk); |
fec5e652 | 808 | |
1da177e4 | 809 | /* We may need to bind the socket. */ |
7ba42910 CG |
810 | if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind && |
811 | inet_autobind(sk)) | |
1da177e4 LT |
812 | return -EAGAIN; |
813 | ||
814 | if (sk->sk_prot->sendpage) | |
815 | return sk->sk_prot->sendpage(sk, page, offset, size, flags); | |
816 | return sock_no_sendpage(sock, page, offset, size, flags); | |
817 | } | |
7ba42910 | 818 | EXPORT_SYMBOL(inet_sendpage); |
1da177e4 | 819 | |
1b784140 YX |
820 | int inet_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, |
821 | int flags) | |
fec5e652 TH |
822 | { |
823 | struct sock *sk = sock->sk; | |
824 | int addr_len = 0; | |
825 | int err; | |
826 | ||
e3f2c4a3 SHY |
827 | if (likely(!(flags & MSG_ERRQUEUE))) |
828 | sock_rps_record_flow(sk); | |
fec5e652 | 829 | |
1b784140 | 830 | err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT, |
fec5e652 TH |
831 | flags & ~MSG_DONTWAIT, &addr_len); |
832 | if (err >= 0) | |
833 | msg->msg_namelen = addr_len; | |
834 | return err; | |
835 | } | |
836 | EXPORT_SYMBOL(inet_recvmsg); | |
1da177e4 LT |
837 | |
838 | int inet_shutdown(struct socket *sock, int how) | |
839 | { | |
840 | struct sock *sk = sock->sk; | |
841 | int err = 0; | |
842 | ||
843 | /* This should really check to make sure | |
844 | * the socket is a TCP socket. (WHY AC...) | |
845 | */ | |
846 | how++; /* maps 0->1 has the advantage of making bit 1 rcvs and | |
847 | 1->2 bit 2 snds. | |
848 | 2->3 */ | |
849 | if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */ | |
850 | return -EINVAL; | |
851 | ||
852 | lock_sock(sk); | |
853 | if (sock->state == SS_CONNECTING) { | |
854 | if ((1 << sk->sk_state) & | |
855 | (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE)) | |
856 | sock->state = SS_DISCONNECTING; | |
857 | else | |
858 | sock->state = SS_CONNECTED; | |
859 | } | |
860 | ||
861 | switch (sk->sk_state) { | |
862 | case TCP_CLOSE: | |
863 | err = -ENOTCONN; | |
864 | /* Hack to wake up other listeners, who can poll for | |
a9a08845 | 865 | EPOLLHUP, even on eg. unconnected UDP sockets -- RR */ |
fcfd6dfa | 866 | /* fall through */ |
1da177e4 LT |
867 | default: |
868 | sk->sk_shutdown |= how; | |
869 | if (sk->sk_prot->shutdown) | |
870 | sk->sk_prot->shutdown(sk, how); | |
871 | break; | |
872 | ||
873 | /* Remaining two branches are temporary solution for missing | |
874 | * close() in multithreaded environment. It is _not_ a good idea, | |
875 | * but we have no choice until close() is repaired at VFS level. | |
876 | */ | |
877 | case TCP_LISTEN: | |
878 | if (!(how & RCV_SHUTDOWN)) | |
879 | break; | |
fcfd6dfa | 880 | /* fall through */ |
1da177e4 LT |
881 | case TCP_SYN_SENT: |
882 | err = sk->sk_prot->disconnect(sk, O_NONBLOCK); | |
883 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
884 | break; | |
885 | } | |
886 | ||
887 | /* Wake up anyone sleeping in poll. */ | |
888 | sk->sk_state_change(sk); | |
889 | release_sock(sk); | |
890 | return err; | |
891 | } | |
3d1427f8 | 892 | EXPORT_SYMBOL(inet_shutdown); |
1da177e4 LT |
893 | |
894 | /* | |
895 | * ioctl() calls you can issue on an INET socket. Most of these are | |
896 | * device configuration and stuff and very rarely used. Some ioctls | |
897 | * pass on to the socket itself. | |
898 | * | |
899 | * NOTE: I like the idea of a module for the config stuff. ie ifconfig | |
900 | * loads the devconfigure module does its configuring and unloads it. | |
901 | * There's a good 20K of config code hanging around the kernel. | |
902 | */ | |
903 | ||
904 | int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) | |
905 | { | |
906 | struct sock *sk = sock->sk; | |
907 | int err = 0; | |
3b1e0a65 | 908 | struct net *net = sock_net(sk); |
03aef17b AV |
909 | void __user *p = (void __user *)arg; |
910 | struct ifreq ifr; | |
ca25c300 | 911 | struct rtentry rt; |
1da177e4 LT |
912 | |
913 | switch (cmd) { | |
3d1427f8 ED |
914 | case SIOCADDRT: |
915 | case SIOCDELRT: | |
ca25c300 AV |
916 | if (copy_from_user(&rt, p, sizeof(struct rtentry))) |
917 | return -EFAULT; | |
918 | err = ip_rt_ioctl(net, cmd, &rt); | |
919 | break; | |
3d1427f8 | 920 | case SIOCRTMSG: |
ca25c300 | 921 | err = -EINVAL; |
3d1427f8 ED |
922 | break; |
923 | case SIOCDARP: | |
924 | case SIOCGARP: | |
925 | case SIOCSARP: | |
926 | err = arp_ioctl(net, cmd, (void __user *)arg); | |
927 | break; | |
928 | case SIOCGIFADDR: | |
3d1427f8 | 929 | case SIOCGIFBRDADDR: |
3d1427f8 | 930 | case SIOCGIFNETMASK: |
3d1427f8 | 931 | case SIOCGIFDSTADDR: |
03aef17b AV |
932 | case SIOCGIFPFLAGS: |
933 | if (copy_from_user(&ifr, p, sizeof(struct ifreq))) | |
934 | return -EFAULT; | |
935 | err = devinet_ioctl(net, cmd, &ifr); | |
936 | if (!err && copy_to_user(p, &ifr, sizeof(struct ifreq))) | |
937 | err = -EFAULT; | |
938 | break; | |
939 | ||
940 | case SIOCSIFADDR: | |
941 | case SIOCSIFBRDADDR: | |
942 | case SIOCSIFNETMASK: | |
3d1427f8 ED |
943 | case SIOCSIFDSTADDR: |
944 | case SIOCSIFPFLAGS: | |
3d1427f8 | 945 | case SIOCSIFFLAGS: |
03aef17b AV |
946 | if (copy_from_user(&ifr, p, sizeof(struct ifreq))) |
947 | return -EFAULT; | |
948 | err = devinet_ioctl(net, cmd, &ifr); | |
3d1427f8 ED |
949 | break; |
950 | default: | |
951 | if (sk->sk_prot->ioctl) | |
952 | err = sk->sk_prot->ioctl(sk, cmd, arg); | |
953 | else | |
954 | err = -ENOIOCTLCMD; | |
955 | break; | |
1da177e4 LT |
956 | } |
957 | return err; | |
958 | } | |
3d1427f8 | 959 | EXPORT_SYMBOL(inet_ioctl); |
1da177e4 | 960 | |
709b46e8 | 961 | #ifdef CONFIG_COMPAT |
686dc6b6 | 962 | static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
709b46e8 EB |
963 | { |
964 | struct sock *sk = sock->sk; | |
965 | int err = -ENOIOCTLCMD; | |
966 | ||
967 | if (sk->sk_prot->compat_ioctl) | |
968 | err = sk->sk_prot->compat_ioctl(sk, cmd, arg); | |
969 | ||
970 | return err; | |
971 | } | |
972 | #endif | |
973 | ||
90ddc4f0 | 974 | const struct proto_ops inet_stream_ops = { |
543d9cfe ACM |
975 | .family = PF_INET, |
976 | .owner = THIS_MODULE, | |
977 | .release = inet_release, | |
978 | .bind = inet_bind, | |
979 | .connect = inet_stream_connect, | |
980 | .socketpair = sock_no_socketpair, | |
981 | .accept = inet_accept, | |
982 | .getname = inet_getname, | |
a11e1d43 | 983 | .poll = tcp_poll, |
543d9cfe | 984 | .ioctl = inet_ioctl, |
c7cbdbf2 | 985 | .gettstamp = sock_gettstamp, |
543d9cfe ACM |
986 | .listen = inet_listen, |
987 | .shutdown = inet_shutdown, | |
988 | .setsockopt = sock_common_setsockopt, | |
989 | .getsockopt = sock_common_getsockopt, | |
7ba42910 | 990 | .sendmsg = inet_sendmsg, |
fec5e652 | 991 | .recvmsg = inet_recvmsg, |
05255b82 | 992 | #ifdef CONFIG_MMU |
93ab6cc6 | 993 | .mmap = tcp_mmap, |
05255b82 | 994 | #endif |
7ba42910 | 995 | .sendpage = inet_sendpage, |
9c55e01c | 996 | .splice_read = tcp_splice_read, |
32035585 | 997 | .read_sock = tcp_read_sock, |
306b13eb TH |
998 | .sendmsg_locked = tcp_sendmsg_locked, |
999 | .sendpage_locked = tcp_sendpage_locked, | |
32035585 | 1000 | .peek_len = tcp_peek_len, |
3fdadf7d | 1001 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1002 | .compat_setsockopt = compat_sock_common_setsockopt, |
1003 | .compat_getsockopt = compat_sock_common_getsockopt, | |
709b46e8 | 1004 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 1005 | #endif |
d1361840 | 1006 | .set_rcvlowat = tcp_set_rcvlowat, |
1da177e4 | 1007 | }; |
3d1427f8 | 1008 | EXPORT_SYMBOL(inet_stream_ops); |
1da177e4 | 1009 | |
90ddc4f0 | 1010 | const struct proto_ops inet_dgram_ops = { |
543d9cfe ACM |
1011 | .family = PF_INET, |
1012 | .owner = THIS_MODULE, | |
1013 | .release = inet_release, | |
1014 | .bind = inet_bind, | |
1015 | .connect = inet_dgram_connect, | |
1016 | .socketpair = sock_no_socketpair, | |
1017 | .accept = sock_no_accept, | |
1018 | .getname = inet_getname, | |
a11e1d43 | 1019 | .poll = udp_poll, |
543d9cfe | 1020 | .ioctl = inet_ioctl, |
c7cbdbf2 | 1021 | .gettstamp = sock_gettstamp, |
543d9cfe ACM |
1022 | .listen = sock_no_listen, |
1023 | .shutdown = inet_shutdown, | |
1024 | .setsockopt = sock_common_setsockopt, | |
1025 | .getsockopt = sock_common_getsockopt, | |
1026 | .sendmsg = inet_sendmsg, | |
fec5e652 | 1027 | .recvmsg = inet_recvmsg, |
543d9cfe ACM |
1028 | .mmap = sock_no_mmap, |
1029 | .sendpage = inet_sendpage, | |
627d2d6b | 1030 | .set_peek_off = sk_set_peek_off, |
3fdadf7d | 1031 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1032 | .compat_setsockopt = compat_sock_common_setsockopt, |
1033 | .compat_getsockopt = compat_sock_common_getsockopt, | |
709b46e8 | 1034 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 1035 | #endif |
1da177e4 | 1036 | }; |
3d1427f8 | 1037 | EXPORT_SYMBOL(inet_dgram_ops); |
1da177e4 LT |
1038 | |
1039 | /* | |
1040 | * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without | |
a11e1d43 | 1041 | * udp_poll |
1da177e4 | 1042 | */ |
90ddc4f0 | 1043 | static const struct proto_ops inet_sockraw_ops = { |
543d9cfe ACM |
1044 | .family = PF_INET, |
1045 | .owner = THIS_MODULE, | |
1046 | .release = inet_release, | |
1047 | .bind = inet_bind, | |
1048 | .connect = inet_dgram_connect, | |
1049 | .socketpair = sock_no_socketpair, | |
1050 | .accept = sock_no_accept, | |
1051 | .getname = inet_getname, | |
a11e1d43 | 1052 | .poll = datagram_poll, |
543d9cfe | 1053 | .ioctl = inet_ioctl, |
c7cbdbf2 | 1054 | .gettstamp = sock_gettstamp, |
543d9cfe ACM |
1055 | .listen = sock_no_listen, |
1056 | .shutdown = inet_shutdown, | |
1057 | .setsockopt = sock_common_setsockopt, | |
1058 | .getsockopt = sock_common_getsockopt, | |
1059 | .sendmsg = inet_sendmsg, | |
fec5e652 | 1060 | .recvmsg = inet_recvmsg, |
543d9cfe ACM |
1061 | .mmap = sock_no_mmap, |
1062 | .sendpage = inet_sendpage, | |
3fdadf7d | 1063 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1064 | .compat_setsockopt = compat_sock_common_setsockopt, |
1065 | .compat_getsockopt = compat_sock_common_getsockopt, | |
709b46e8 | 1066 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 1067 | #endif |
1da177e4 LT |
1068 | }; |
1069 | ||
ec1b4cf7 | 1070 | static const struct net_proto_family inet_family_ops = { |
1da177e4 LT |
1071 | .family = PF_INET, |
1072 | .create = inet_create, | |
1073 | .owner = THIS_MODULE, | |
1074 | }; | |
1075 | ||
1da177e4 LT |
1076 | /* Upon startup we insert all the elements in inetsw_array[] into |
1077 | * the linked list inetsw. | |
1078 | */ | |
1079 | static struct inet_protosw inetsw_array[] = | |
1080 | { | |
e905a9ed YH |
1081 | { |
1082 | .type = SOCK_STREAM, | |
1083 | .protocol = IPPROTO_TCP, | |
1084 | .prot = &tcp_prot, | |
1085 | .ops = &inet_stream_ops, | |
e905a9ed | 1086 | .flags = INET_PROTOSW_PERMANENT | |
d83d8461 | 1087 | INET_PROTOSW_ICSK, |
e905a9ed YH |
1088 | }, |
1089 | ||
1090 | { | |
1091 | .type = SOCK_DGRAM, | |
1092 | .protocol = IPPROTO_UDP, | |
1093 | .prot = &udp_prot, | |
1094 | .ops = &inet_dgram_ops, | |
e905a9ed | 1095 | .flags = INET_PROTOSW_PERMANENT, |
1da177e4 | 1096 | }, |
e905a9ed | 1097 | |
c319b4d7 VK |
1098 | { |
1099 | .type = SOCK_DGRAM, | |
1100 | .protocol = IPPROTO_ICMP, | |
1101 | .prot = &ping_prot, | |
77d4b1d3 | 1102 | .ops = &inet_sockraw_ops, |
c319b4d7 VK |
1103 | .flags = INET_PROTOSW_REUSE, |
1104 | }, | |
1da177e4 LT |
1105 | |
1106 | { | |
e905a9ed YH |
1107 | .type = SOCK_RAW, |
1108 | .protocol = IPPROTO_IP, /* wild card */ | |
1109 | .prot = &raw_prot, | |
1110 | .ops = &inet_sockraw_ops, | |
e905a9ed | 1111 | .flags = INET_PROTOSW_REUSE, |
1da177e4 LT |
1112 | } |
1113 | }; | |
1114 | ||
c40f6fff | 1115 | #define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array) |
1da177e4 LT |
1116 | |
1117 | void inet_register_protosw(struct inet_protosw *p) | |
1118 | { | |
1119 | struct list_head *lh; | |
1120 | struct inet_protosw *answer; | |
1121 | int protocol = p->protocol; | |
1122 | struct list_head *last_perm; | |
1123 | ||
1124 | spin_lock_bh(&inetsw_lock); | |
1125 | ||
1126 | if (p->type >= SOCK_MAX) | |
1127 | goto out_illegal; | |
1128 | ||
1129 | /* If we are trying to override a permanent protocol, bail. */ | |
1da177e4 LT |
1130 | last_perm = &inetsw[p->type]; |
1131 | list_for_each(lh, &inetsw[p->type]) { | |
1132 | answer = list_entry(lh, struct inet_protosw, list); | |
1da177e4 | 1133 | /* Check only the non-wild match. */ |
f6c53334 JZ |
1134 | if ((INET_PROTOSW_PERMANENT & answer->flags) == 0) |
1135 | break; | |
1136 | if (protocol == answer->protocol) | |
1137 | goto out_permanent; | |
1138 | last_perm = lh; | |
1da177e4 | 1139 | } |
1da177e4 LT |
1140 | |
1141 | /* Add the new entry after the last permanent entry if any, so that | |
1142 | * the new entry does not override a permanent entry when matched with | |
1143 | * a wild-card protocol. But it is allowed to override any existing | |
e905a9ed | 1144 | * non-permanent entry. This means that when we remove this entry, the |
1da177e4 LT |
1145 | * system automatically returns to the old behavior. |
1146 | */ | |
1147 | list_add_rcu(&p->list, last_perm); | |
1148 | out: | |
1149 | spin_unlock_bh(&inetsw_lock); | |
1150 | ||
1da177e4 LT |
1151 | return; |
1152 | ||
1153 | out_permanent: | |
058bd4d2 | 1154 | pr_err("Attempt to override permanent protocol %d\n", protocol); |
1da177e4 LT |
1155 | goto out; |
1156 | ||
1157 | out_illegal: | |
058bd4d2 | 1158 | pr_err("Ignoring attempt to register invalid socket type %d\n", |
1da177e4 LT |
1159 | p->type); |
1160 | goto out; | |
1161 | } | |
3d1427f8 | 1162 | EXPORT_SYMBOL(inet_register_protosw); |
1da177e4 LT |
1163 | |
1164 | void inet_unregister_protosw(struct inet_protosw *p) | |
1165 | { | |
1166 | if (INET_PROTOSW_PERMANENT & p->flags) { | |
058bd4d2 | 1167 | pr_err("Attempt to unregister permanent protocol %d\n", |
1da177e4 LT |
1168 | p->protocol); |
1169 | } else { | |
1170 | spin_lock_bh(&inetsw_lock); | |
1171 | list_del_rcu(&p->list); | |
1172 | spin_unlock_bh(&inetsw_lock); | |
1173 | ||
1174 | synchronize_net(); | |
1175 | } | |
1176 | } | |
3d1427f8 | 1177 | EXPORT_SYMBOL(inet_unregister_protosw); |
1da177e4 | 1178 | |
32519f11 ACM |
1179 | static int inet_sk_reselect_saddr(struct sock *sk) |
1180 | { | |
1181 | struct inet_sock *inet = inet_sk(sk); | |
c720c7e8 | 1182 | __be32 old_saddr = inet->inet_saddr; |
c720c7e8 | 1183 | __be32 daddr = inet->inet_daddr; |
6e869138 | 1184 | struct flowi4 *fl4; |
b23dd4fe DM |
1185 | struct rtable *rt; |
1186 | __be32 new_saddr; | |
f6d8bd05 | 1187 | struct ip_options_rcu *inet_opt; |
32519f11 | 1188 | |
f6d8bd05 | 1189 | inet_opt = rcu_dereference_protected(inet->inet_opt, |
1e1d04e6 | 1190 | lockdep_sock_is_held(sk)); |
f6d8bd05 ED |
1191 | if (inet_opt && inet_opt->opt.srr) |
1192 | daddr = inet_opt->opt.faddr; | |
32519f11 ACM |
1193 | |
1194 | /* Query new route. */ | |
6e869138 DM |
1195 | fl4 = &inet->cork.fl.u.ip4; |
1196 | rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk), | |
b23dd4fe | 1197 | sk->sk_bound_dev_if, sk->sk_protocol, |
0e0d44ab | 1198 | inet->inet_sport, inet->inet_dport, sk); |
b23dd4fe DM |
1199 | if (IS_ERR(rt)) |
1200 | return PTR_ERR(rt); | |
32519f11 | 1201 | |
d8d1f30b | 1202 | sk_setup_caps(sk, &rt->dst); |
32519f11 | 1203 | |
6e869138 | 1204 | new_saddr = fl4->saddr; |
32519f11 ACM |
1205 | |
1206 | if (new_saddr == old_saddr) | |
1207 | return 0; | |
1208 | ||
287b7f38 | 1209 | if (sock_net(sk)->ipv4.sysctl_ip_dynaddr > 1) { |
058bd4d2 JP |
1210 | pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n", |
1211 | __func__, &old_saddr, &new_saddr); | |
32519f11 ACM |
1212 | } |
1213 | ||
c720c7e8 | 1214 | inet->inet_saddr = inet->inet_rcv_saddr = new_saddr; |
32519f11 ACM |
1215 | |
1216 | /* | |
1217 | * XXX The only one ugly spot where we need to | |
1218 | * XXX really change the sockets identity after | |
1219 | * XXX it has entered the hashes. -DaveM | |
1220 | * | |
1221 | * Besides that, it does not check for connection | |
1222 | * uniqueness. Wait for troubles. | |
1223 | */ | |
086c653f | 1224 | return __sk_prot_rehash(sk); |
32519f11 ACM |
1225 | } |
1226 | ||
1227 | int inet_sk_rebuild_header(struct sock *sk) | |
1228 | { | |
1229 | struct inet_sock *inet = inet_sk(sk); | |
1230 | struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0); | |
3ca3c68e | 1231 | __be32 daddr; |
f6d8bd05 | 1232 | struct ip_options_rcu *inet_opt; |
6e869138 | 1233 | struct flowi4 *fl4; |
32519f11 ACM |
1234 | int err; |
1235 | ||
1236 | /* Route is OK, nothing to do. */ | |
1237 | if (rt) | |
1238 | return 0; | |
1239 | ||
1240 | /* Reroute. */ | |
f6d8bd05 ED |
1241 | rcu_read_lock(); |
1242 | inet_opt = rcu_dereference(inet->inet_opt); | |
c720c7e8 | 1243 | daddr = inet->inet_daddr; |
f6d8bd05 ED |
1244 | if (inet_opt && inet_opt->opt.srr) |
1245 | daddr = inet_opt->opt.faddr; | |
1246 | rcu_read_unlock(); | |
6e869138 DM |
1247 | fl4 = &inet->cork.fl.u.ip4; |
1248 | rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr, | |
78fbfd8a DM |
1249 | inet->inet_dport, inet->inet_sport, |
1250 | sk->sk_protocol, RT_CONN_FLAGS(sk), | |
1251 | sk->sk_bound_dev_if); | |
b23dd4fe DM |
1252 | if (!IS_ERR(rt)) { |
1253 | err = 0; | |
d8d1f30b | 1254 | sk_setup_caps(sk, &rt->dst); |
b23dd4fe DM |
1255 | } else { |
1256 | err = PTR_ERR(rt); | |
1257 | ||
32519f11 ACM |
1258 | /* Routing failed... */ |
1259 | sk->sk_route_caps = 0; | |
1260 | /* | |
1261 | * Other protocols have to map its equivalent state to TCP_SYN_SENT. | |
1262 | * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme | |
1263 | */ | |
287b7f38 | 1264 | if (!sock_net(sk)->ipv4.sysctl_ip_dynaddr || |
32519f11 ACM |
1265 | sk->sk_state != TCP_SYN_SENT || |
1266 | (sk->sk_userlocks & SOCK_BINDADDR_LOCK) || | |
1267 | (err = inet_sk_reselect_saddr(sk)) != 0) | |
1268 | sk->sk_err_soft = -err; | |
1269 | } | |
1270 | ||
1271 | return err; | |
1272 | } | |
32519f11 ACM |
1273 | EXPORT_SYMBOL(inet_sk_rebuild_header); |
1274 | ||
563e0bb0 YS |
1275 | void inet_sk_set_state(struct sock *sk, int state) |
1276 | { | |
1277 | trace_inet_sock_set_state(sk, sk->sk_state, state); | |
1278 | sk->sk_state = state; | |
1279 | } | |
1280 | EXPORT_SYMBOL(inet_sk_set_state); | |
1281 | ||
1282 | void inet_sk_state_store(struct sock *sk, int newstate) | |
1283 | { | |
1284 | trace_inet_sock_set_state(sk, sk->sk_state, newstate); | |
1285 | smp_store_release(&sk->sk_state, newstate); | |
1286 | } | |
1287 | ||
b8921ca8 TH |
1288 | struct sk_buff *inet_gso_segment(struct sk_buff *skb, |
1289 | netdev_features_t features) | |
f4c50d99 | 1290 | { |
0c19f846 | 1291 | bool udpfrag = false, fixedid = false, gso_partial, encap; |
f4c50d99 | 1292 | struct sk_buff *segs = ERR_PTR(-EINVAL); |
bca49f84 | 1293 | const struct net_offload *ops; |
0c19f846 | 1294 | unsigned int offset = 0; |
f9242b6b | 1295 | struct iphdr *iph; |
802ab55a | 1296 | int proto, tot_len; |
3347c960 | 1297 | int nhoff; |
f4c50d99 HX |
1298 | int ihl; |
1299 | int id; | |
1300 | ||
3347c960 ED |
1301 | skb_reset_network_header(skb); |
1302 | nhoff = skb_network_header(skb) - skb_mac_header(skb); | |
bbcf467d | 1303 | if (unlikely(!pskb_may_pull(skb, sizeof(*iph)))) |
f4c50d99 HX |
1304 | goto out; |
1305 | ||
eddc9ec5 | 1306 | iph = ip_hdr(skb); |
f4c50d99 HX |
1307 | ihl = iph->ihl * 4; |
1308 | if (ihl < sizeof(*iph)) | |
1309 | goto out; | |
1310 | ||
47d27aad ED |
1311 | id = ntohs(iph->id); |
1312 | proto = iph->protocol; | |
1313 | ||
1314 | /* Warning: after this point, iph might be no longer valid */ | |
bbcf467d | 1315 | if (unlikely(!pskb_may_pull(skb, ihl))) |
f4c50d99 | 1316 | goto out; |
47d27aad | 1317 | __skb_pull(skb, ihl); |
f4c50d99 | 1318 | |
8c3a897b ED |
1319 | encap = SKB_GSO_CB(skb)->encap_level > 0; |
1320 | if (encap) | |
1e16aa3d | 1321 | features &= skb->dev->hw_enc_features; |
3347c960 | 1322 | SKB_GSO_CB(skb)->encap_level += ihl; |
73136267 | 1323 | |
badff6d0 | 1324 | skb_reset_transport_header(skb); |
47d27aad | 1325 | |
f4c50d99 HX |
1326 | segs = ERR_PTR(-EPROTONOSUPPORT); |
1327 | ||
cbc53e08 | 1328 | if (!skb->encapsulation || encap) { |
0c19f846 | 1329 | udpfrag = !!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP); |
cbc53e08 AD |
1330 | fixedid = !!(skb_shinfo(skb)->gso_type & SKB_GSO_TCP_FIXEDID); |
1331 | ||
1332 | /* fixed ID is invalid if DF bit is not set */ | |
a5108878 | 1333 | if (fixedid && !(ip_hdr(skb)->frag_off & htons(IP_DF))) |
cbc53e08 AD |
1334 | goto out; |
1335 | } | |
dcd60771 | 1336 | |
bca49f84 | 1337 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 VY |
1338 | if (likely(ops && ops->callbacks.gso_segment)) |
1339 | segs = ops->callbacks.gso_segment(skb, features); | |
f4c50d99 | 1340 | |
50c3a487 | 1341 | if (IS_ERR_OR_NULL(segs)) |
f4c50d99 HX |
1342 | goto out; |
1343 | ||
07b26c94 SK |
1344 | gso_partial = !!(skb_shinfo(segs)->gso_type & SKB_GSO_PARTIAL); |
1345 | ||
f4c50d99 HX |
1346 | skb = segs; |
1347 | do { | |
3347c960 | 1348 | iph = (struct iphdr *)(skb_mac_header(skb) + nhoff); |
0c19f846 WB |
1349 | if (udpfrag) { |
1350 | iph->frag_off = htons(offset >> 3); | |
1351 | if (skb->next) | |
1352 | iph->frag_off |= htons(IP_MF); | |
1353 | offset += skb->len - nhoff - ihl; | |
1354 | tot_len = skb->len - nhoff; | |
1355 | } else if (skb_is_gso(skb)) { | |
802ab55a AD |
1356 | if (!fixedid) { |
1357 | iph->id = htons(id); | |
1358 | id += skb_shinfo(skb)->gso_segs; | |
1359 | } | |
07b26c94 SK |
1360 | |
1361 | if (gso_partial) | |
1362 | tot_len = skb_shinfo(skb)->gso_size + | |
1363 | SKB_GSO_CB(skb)->data_offset + | |
1364 | skb->head - (unsigned char *)iph; | |
1365 | else | |
1366 | tot_len = skb->len - nhoff; | |
802ab55a AD |
1367 | } else { |
1368 | if (!fixedid) | |
1369 | iph->id = htons(id++); | |
1370 | tot_len = skb->len - nhoff; | |
490ab081 | 1371 | } |
802ab55a | 1372 | iph->tot_len = htons(tot_len); |
47d27aad | 1373 | ip_send_check(iph); |
8c3a897b | 1374 | if (encap) |
3347c960 | 1375 | skb_reset_inner_headers(skb); |
3347c960 | 1376 | skb->network_header = (u8 *)iph - skb->head; |
c56cae23 | 1377 | skb_reset_mac_len(skb); |
f4c50d99 HX |
1378 | } while ((skb = skb->next)); |
1379 | ||
1380 | out: | |
1381 | return segs; | |
1382 | } | |
b8921ca8 | 1383 | EXPORT_SYMBOL(inet_gso_segment); |
f4c50d99 | 1384 | |
418e897e WB |
1385 | static struct sk_buff *ipip_gso_segment(struct sk_buff *skb, |
1386 | netdev_features_t features) | |
1387 | { | |
1388 | if (!(skb_shinfo(skb)->gso_type & SKB_GSO_IPXIP4)) | |
1389 | return ERR_PTR(-EINVAL); | |
1390 | ||
1391 | return inet_gso_segment(skb, features); | |
1392 | } | |
1393 | ||
028e0a47 PA |
1394 | INDIRECT_CALLABLE_DECLARE(struct sk_buff *tcp4_gro_receive(struct list_head *, |
1395 | struct sk_buff *)); | |
1396 | INDIRECT_CALLABLE_DECLARE(struct sk_buff *udp4_gro_receive(struct list_head *, | |
1397 | struct sk_buff *)); | |
d4546c25 | 1398 | struct sk_buff *inet_gro_receive(struct list_head *head, struct sk_buff *skb) |
73cc19f1 | 1399 | { |
bca49f84 | 1400 | const struct net_offload *ops; |
d4546c25 | 1401 | struct sk_buff *pp = NULL; |
b71d1d42 | 1402 | const struct iphdr *iph; |
d4546c25 | 1403 | struct sk_buff *p; |
a5b1cf28 HX |
1404 | unsigned int hlen; |
1405 | unsigned int off; | |
1075f3f6 | 1406 | unsigned int id; |
73cc19f1 HX |
1407 | int flush = 1; |
1408 | int proto; | |
73cc19f1 | 1409 | |
a5b1cf28 HX |
1410 | off = skb_gro_offset(skb); |
1411 | hlen = off + sizeof(*iph); | |
1412 | iph = skb_gro_header_fast(skb, off); | |
1413 | if (skb_gro_header_hard(skb, hlen)) { | |
1414 | iph = skb_gro_header_slow(skb, hlen, off); | |
1415 | if (unlikely(!iph)) | |
1416 | goto out; | |
1417 | } | |
73cc19f1 | 1418 | |
f9242b6b | 1419 | proto = iph->protocol; |
73cc19f1 HX |
1420 | |
1421 | rcu_read_lock(); | |
bca49f84 | 1422 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 | 1423 | if (!ops || !ops->callbacks.gro_receive) |
73cc19f1 HX |
1424 | goto out_unlock; |
1425 | ||
a5ad24be | 1426 | if (*(u8 *)iph != 0x45) |
73cc19f1 HX |
1427 | goto out_unlock; |
1428 | ||
9b83e031 SK |
1429 | if (ip_is_fragment(iph)) |
1430 | goto out_unlock; | |
1431 | ||
a9e050f4 | 1432 | if (unlikely(ip_fast_csum((u8 *)iph, 5))) |
73cc19f1 HX |
1433 | goto out_unlock; |
1434 | ||
0eae88f3 | 1435 | id = ntohl(*(__be32 *)&iph->id); |
db8caf3d | 1436 | flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF)); |
1075f3f6 | 1437 | id >>= 16; |
73cc19f1 | 1438 | |
d4546c25 | 1439 | list_for_each_entry(p, head, list) { |
73cc19f1 | 1440 | struct iphdr *iph2; |
1530545e | 1441 | u16 flush_id; |
73cc19f1 HX |
1442 | |
1443 | if (!NAPI_GRO_CB(p)->same_flow) | |
1444 | continue; | |
1445 | ||
299603e8 JC |
1446 | iph2 = (struct iphdr *)(p->data + off); |
1447 | /* The above works because, with the exception of the top | |
1448 | * (inner most) layer, we only aggregate pkts with the same | |
1449 | * hdr length so all the hdrs we'll need to verify will start | |
1450 | * at the same offset. | |
1451 | */ | |
a5ad24be | 1452 | if ((iph->protocol ^ iph2->protocol) | |
0eae88f3 ED |
1453 | ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) | |
1454 | ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) { | |
73cc19f1 HX |
1455 | NAPI_GRO_CB(p)->same_flow = 0; |
1456 | continue; | |
1457 | } | |
1458 | ||
1459 | /* All fields must match except length and checksum. */ | |
1460 | NAPI_GRO_CB(p)->flush |= | |
a5ad24be | 1461 | (iph->ttl ^ iph2->ttl) | |
a9e050f4 | 1462 | (iph->tos ^ iph2->tos) | |
bf5a755f | 1463 | ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF)); |
73cc19f1 HX |
1464 | |
1465 | NAPI_GRO_CB(p)->flush |= flush; | |
1530545e AD |
1466 | |
1467 | /* We need to store of the IP ID check to be included later | |
1468 | * when we can verify that this packet does in fact belong | |
1469 | * to a given flow. | |
1470 | */ | |
1471 | flush_id = (u16)(id - ntohs(iph2->id)); | |
1472 | ||
1473 | /* This bit of code makes it much easier for us to identify | |
1474 | * the cases where we are doing atomic vs non-atomic IP ID | |
1475 | * checks. Specifically an atomic check can return IP ID | |
1476 | * values 0 - 0xFFFF, while a non-atomic check can only | |
1477 | * return 0 or 0xFFFF. | |
1478 | */ | |
1479 | if (!NAPI_GRO_CB(p)->is_atomic || | |
1480 | !(iph->frag_off & htons(IP_DF))) { | |
1481 | flush_id ^= NAPI_GRO_CB(p)->count; | |
1482 | flush_id = flush_id ? 0xFFFF : 0; | |
1483 | } | |
1484 | ||
1485 | /* If the previous IP ID value was based on an atomic | |
1486 | * datagram we can overwrite the value and ignore it. | |
1487 | */ | |
1488 | if (NAPI_GRO_CB(skb)->is_atomic) | |
1489 | NAPI_GRO_CB(p)->flush_id = flush_id; | |
1490 | else | |
1491 | NAPI_GRO_CB(p)->flush_id |= flush_id; | |
73cc19f1 HX |
1492 | } |
1493 | ||
1530545e | 1494 | NAPI_GRO_CB(skb)->is_atomic = !!(iph->frag_off & htons(IP_DF)); |
73cc19f1 | 1495 | NAPI_GRO_CB(skb)->flush |= flush; |
299603e8 JC |
1496 | skb_set_network_header(skb, off); |
1497 | /* The above will be needed by the transport layer if there is one | |
1498 | * immediately following this IP hdr. | |
1499 | */ | |
1500 | ||
2c804d0f ED |
1501 | /* Note : No need to call skb_gro_postpull_rcsum() here, |
1502 | * as we already checked checksum over ipv4 header was 0 | |
1503 | */ | |
86911732 HX |
1504 | skb_gro_pull(skb, sizeof(*iph)); |
1505 | skb_set_transport_header(skb, skb_gro_offset(skb)); | |
73cc19f1 | 1506 | |
028e0a47 PA |
1507 | pp = indirect_call_gro_receive(tcp4_gro_receive, udp4_gro_receive, |
1508 | ops->callbacks.gro_receive, head, skb); | |
73cc19f1 HX |
1509 | |
1510 | out_unlock: | |
1511 | rcu_read_unlock(); | |
1512 | ||
1513 | out: | |
5f114163 | 1514 | skb_gro_flush_final(skb, pp, flush); |
73cc19f1 HX |
1515 | |
1516 | return pp; | |
1517 | } | |
b8921ca8 | 1518 | EXPORT_SYMBOL(inet_gro_receive); |
73cc19f1 | 1519 | |
d4546c25 DM |
1520 | static struct sk_buff *ipip_gro_receive(struct list_head *head, |
1521 | struct sk_buff *skb) | |
fac8e0f5 JG |
1522 | { |
1523 | if (NAPI_GRO_CB(skb)->encap_mark) { | |
1524 | NAPI_GRO_CB(skb)->flush = 1; | |
1525 | return NULL; | |
1526 | } | |
1527 | ||
1528 | NAPI_GRO_CB(skb)->encap_mark = 1; | |
1529 | ||
1530 | return inet_gro_receive(head, skb); | |
1531 | } | |
1532 | ||
822c8685 DD |
1533 | #define SECONDS_PER_DAY 86400 |
1534 | ||
1535 | /* inet_current_timestamp - Return IP network timestamp | |
1536 | * | |
1537 | * Return milliseconds since midnight in network byte order. | |
1538 | */ | |
1539 | __be32 inet_current_timestamp(void) | |
1540 | { | |
1541 | u32 secs; | |
1542 | u32 msecs; | |
1543 | struct timespec64 ts; | |
1544 | ||
1545 | ktime_get_real_ts64(&ts); | |
1546 | ||
1547 | /* Get secs since midnight. */ | |
1548 | (void)div_u64_rem(ts.tv_sec, SECONDS_PER_DAY, &secs); | |
1549 | /* Convert to msecs. */ | |
1550 | msecs = secs * MSEC_PER_SEC; | |
1551 | /* Convert nsec to msec. */ | |
1552 | msecs += (u32)ts.tv_nsec / NSEC_PER_MSEC; | |
1553 | ||
1554 | /* Convert to network byte order. */ | |
3ba9d300 | 1555 | return htonl(msecs); |
822c8685 DD |
1556 | } |
1557 | EXPORT_SYMBOL(inet_current_timestamp); | |
1558 | ||
f4713a3d WB |
1559 | int inet_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len) |
1560 | { | |
1561 | if (sk->sk_family == AF_INET) | |
1562 | return ip_recv_error(sk, msg, len, addr_len); | |
1563 | #if IS_ENABLED(CONFIG_IPV6) | |
1564 | if (sk->sk_family == AF_INET6) | |
1565 | return pingv6_ops.ipv6_recv_error(sk, msg, len, addr_len); | |
1566 | #endif | |
1567 | return -EINVAL; | |
1568 | } | |
1569 | ||
028e0a47 PA |
1570 | INDIRECT_CALLABLE_DECLARE(int tcp4_gro_complete(struct sk_buff *, int)); |
1571 | INDIRECT_CALLABLE_DECLARE(int udp4_gro_complete(struct sk_buff *, int)); | |
b8921ca8 | 1572 | int inet_gro_complete(struct sk_buff *skb, int nhoff) |
73cc19f1 | 1573 | { |
299603e8 JC |
1574 | __be16 newlen = htons(skb->len - nhoff); |
1575 | struct iphdr *iph = (struct iphdr *)(skb->data + nhoff); | |
bca49f84 | 1576 | const struct net_offload *ops; |
f9242b6b | 1577 | int proto = iph->protocol; |
73cc19f1 | 1578 | int err = -ENOSYS; |
73cc19f1 | 1579 | |
294acf1c PA |
1580 | if (skb->encapsulation) { |
1581 | skb_set_inner_protocol(skb, cpu_to_be16(ETH_P_IP)); | |
c3caf119 | 1582 | skb_set_inner_network_header(skb, nhoff); |
294acf1c | 1583 | } |
c3caf119 | 1584 | |
73cc19f1 HX |
1585 | csum_replace2(&iph->check, iph->tot_len, newlen); |
1586 | iph->tot_len = newlen; | |
1587 | ||
1588 | rcu_read_lock(); | |
bca49f84 | 1589 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 | 1590 | if (WARN_ON(!ops || !ops->callbacks.gro_complete)) |
73cc19f1 HX |
1591 | goto out_unlock; |
1592 | ||
299603e8 JC |
1593 | /* Only need to add sizeof(*iph) to get to the next hdr below |
1594 | * because any hdr with option will have been flushed in | |
1595 | * inet_gro_receive(). | |
1596 | */ | |
028e0a47 PA |
1597 | err = INDIRECT_CALL_2(ops->callbacks.gro_complete, |
1598 | tcp4_gro_complete, udp4_gro_complete, | |
1599 | skb, nhoff + sizeof(*iph)); | |
73cc19f1 HX |
1600 | |
1601 | out_unlock: | |
1602 | rcu_read_unlock(); | |
1603 | ||
1604 | return err; | |
1605 | } | |
b8921ca8 | 1606 | EXPORT_SYMBOL(inet_gro_complete); |
73cc19f1 | 1607 | |
b8cba75b JG |
1608 | static int ipip_gro_complete(struct sk_buff *skb, int nhoff) |
1609 | { | |
1610 | skb->encapsulation = 1; | |
7e13318d | 1611 | skb_shinfo(skb)->gso_type |= SKB_GSO_IPXIP4; |
b8cba75b JG |
1612 | return inet_gro_complete(skb, nhoff); |
1613 | } | |
1614 | ||
eee4fe4d | 1615 | int inet_ctl_sock_create(struct sock **sk, unsigned short family, |
5677242f DL |
1616 | unsigned short type, unsigned char protocol, |
1617 | struct net *net) | |
3d58b5fa | 1618 | { |
eee4fe4d | 1619 | struct socket *sock; |
26abe143 | 1620 | int rc = sock_create_kern(net, family, type, protocol, &sock); |
3d58b5fa DL |
1621 | |
1622 | if (rc == 0) { | |
eee4fe4d DL |
1623 | *sk = sock->sk; |
1624 | (*sk)->sk_allocation = GFP_ATOMIC; | |
3d58b5fa DL |
1625 | /* |
1626 | * Unhash it so that IP input processing does not even see it, | |
1627 | * we do not wish this socket to see incoming packets. | |
1628 | */ | |
eee4fe4d | 1629 | (*sk)->sk_prot->unhash(*sk); |
3d58b5fa DL |
1630 | } |
1631 | return rc; | |
1632 | } | |
3d58b5fa DL |
1633 | EXPORT_SYMBOL_GPL(inet_ctl_sock_create); |
1634 | ||
c4c6bc31 R |
1635 | u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offt) |
1636 | { | |
1637 | return *(((unsigned long *)per_cpu_ptr(mib, cpu)) + offt); | |
1638 | } | |
1639 | EXPORT_SYMBOL_GPL(snmp_get_cpu_field); | |
1640 | ||
698365fa | 1641 | unsigned long snmp_fold_field(void __percpu *mib, int offt) |
5e0f0435 HX |
1642 | { |
1643 | unsigned long res = 0; | |
698365fa | 1644 | int i; |
5e0f0435 | 1645 | |
698365fa | 1646 | for_each_possible_cpu(i) |
c4c6bc31 | 1647 | res += snmp_get_cpu_field(mib, i, offt); |
5e0f0435 HX |
1648 | return res; |
1649 | } | |
1650 | EXPORT_SYMBOL_GPL(snmp_fold_field); | |
1651 | ||
4ce3c183 ED |
1652 | #if BITS_PER_LONG==32 |
1653 | ||
80ec1927 | 1654 | u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offt, |
c4c6bc31 R |
1655 | size_t syncp_offset) |
1656 | { | |
1657 | void *bhptr; | |
1658 | struct u64_stats_sync *syncp; | |
1659 | u64 v; | |
1660 | unsigned int start; | |
1661 | ||
1662 | bhptr = per_cpu_ptr(mib, cpu); | |
1663 | syncp = (struct u64_stats_sync *)(bhptr + syncp_offset); | |
1664 | do { | |
1665 | start = u64_stats_fetch_begin_irq(syncp); | |
1666 | v = *(((u64 *)bhptr) + offt); | |
1667 | } while (u64_stats_fetch_retry_irq(syncp, start)); | |
1668 | ||
1669 | return v; | |
1670 | } | |
1671 | EXPORT_SYMBOL_GPL(snmp_get_cpu_field64); | |
1672 | ||
698365fa | 1673 | u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset) |
4ce3c183 ED |
1674 | { |
1675 | u64 res = 0; | |
1676 | int cpu; | |
1677 | ||
1678 | for_each_possible_cpu(cpu) { | |
d1bfc625 | 1679 | res += snmp_get_cpu_field64(mib, cpu, offt, syncp_offset); |
4ce3c183 ED |
1680 | } |
1681 | return res; | |
1682 | } | |
1683 | EXPORT_SYMBOL_GPL(snmp_fold_field64); | |
1684 | #endif | |
1685 | ||
1da177e4 | 1686 | #ifdef CONFIG_IP_MULTICAST |
32613090 | 1687 | static const struct net_protocol igmp_protocol = { |
1da177e4 | 1688 | .handler = igmp_rcv, |
b4ee07df | 1689 | .netns_ok = 1, |
1da177e4 LT |
1690 | }; |
1691 | #endif | |
1692 | ||
a8e3bb34 DA |
1693 | /* thinking of making this const? Don't. |
1694 | * early_demux can change based on sysctl. | |
1695 | */ | |
03157937 | 1696 | static struct net_protocol tcp_protocol = { |
41063e9d | 1697 | .early_demux = tcp_v4_early_demux, |
dddb64bc | 1698 | .early_demux_handler = tcp_v4_early_demux, |
41063e9d DM |
1699 | .handler = tcp_v4_rcv, |
1700 | .err_handler = tcp_v4_err, | |
41063e9d DM |
1701 | .no_policy = 1, |
1702 | .netns_ok = 1, | |
8ed1dc44 | 1703 | .icmp_strict_tag_validation = 1, |
1da177e4 LT |
1704 | }; |
1705 | ||
a8e3bb34 DA |
1706 | /* thinking of making this const? Don't. |
1707 | * early_demux can change based on sysctl. | |
1708 | */ | |
03157937 | 1709 | static struct net_protocol udp_protocol = { |
421b3885 | 1710 | .early_demux = udp_v4_early_demux, |
dddb64bc | 1711 | .early_demux_handler = udp_v4_early_demux, |
1da177e4 LT |
1712 | .handler = udp_rcv, |
1713 | .err_handler = udp_err, | |
1714 | .no_policy = 1, | |
92f1fecb | 1715 | .netns_ok = 1, |
1da177e4 LT |
1716 | }; |
1717 | ||
32613090 | 1718 | static const struct net_protocol icmp_protocol = { |
1da177e4 | 1719 | .handler = icmp_rcv, |
5b052042 | 1720 | .err_handler = icmp_err, |
8b7817f3 | 1721 | .no_policy = 1, |
92f1fecb | 1722 | .netns_ok = 1, |
1da177e4 LT |
1723 | }; |
1724 | ||
9b4661bd PE |
1725 | static __net_init int ipv4_mib_init_net(struct net *net) |
1726 | { | |
827da44c JS |
1727 | int i; |
1728 | ||
698365fa WC |
1729 | net->mib.tcp_statistics = alloc_percpu(struct tcp_mib); |
1730 | if (!net->mib.tcp_statistics) | |
57ef42d5 | 1731 | goto err_tcp_mib; |
698365fa WC |
1732 | net->mib.ip_statistics = alloc_percpu(struct ipstats_mib); |
1733 | if (!net->mib.ip_statistics) | |
a20f5799 | 1734 | goto err_ip_mib; |
827da44c JS |
1735 | |
1736 | for_each_possible_cpu(i) { | |
1737 | struct ipstats_mib *af_inet_stats; | |
698365fa | 1738 | af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i); |
827da44c | 1739 | u64_stats_init(&af_inet_stats->syncp); |
827da44c JS |
1740 | } |
1741 | ||
698365fa WC |
1742 | net->mib.net_statistics = alloc_percpu(struct linux_mib); |
1743 | if (!net->mib.net_statistics) | |
61a7e260 | 1744 | goto err_net_mib; |
698365fa WC |
1745 | net->mib.udp_statistics = alloc_percpu(struct udp_mib); |
1746 | if (!net->mib.udp_statistics) | |
2f275f91 | 1747 | goto err_udp_mib; |
698365fa WC |
1748 | net->mib.udplite_statistics = alloc_percpu(struct udp_mib); |
1749 | if (!net->mib.udplite_statistics) | |
386019d3 | 1750 | goto err_udplite_mib; |
698365fa WC |
1751 | net->mib.icmp_statistics = alloc_percpu(struct icmp_mib); |
1752 | if (!net->mib.icmp_statistics) | |
b60538a0 | 1753 | goto err_icmp_mib; |
acb32ba3 ED |
1754 | net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib), |
1755 | GFP_KERNEL); | |
1756 | if (!net->mib.icmpmsg_statistics) | |
923c6586 | 1757 | goto err_icmpmsg_mib; |
57ef42d5 PE |
1758 | |
1759 | tcp_mib_init(net); | |
9b4661bd | 1760 | return 0; |
57ef42d5 | 1761 | |
923c6586 | 1762 | err_icmpmsg_mib: |
698365fa | 1763 | free_percpu(net->mib.icmp_statistics); |
b60538a0 | 1764 | err_icmp_mib: |
698365fa | 1765 | free_percpu(net->mib.udplite_statistics); |
386019d3 | 1766 | err_udplite_mib: |
698365fa | 1767 | free_percpu(net->mib.udp_statistics); |
2f275f91 | 1768 | err_udp_mib: |
698365fa | 1769 | free_percpu(net->mib.net_statistics); |
61a7e260 | 1770 | err_net_mib: |
698365fa | 1771 | free_percpu(net->mib.ip_statistics); |
a20f5799 | 1772 | err_ip_mib: |
698365fa | 1773 | free_percpu(net->mib.tcp_statistics); |
57ef42d5 PE |
1774 | err_tcp_mib: |
1775 | return -ENOMEM; | |
9b4661bd PE |
1776 | } |
1777 | ||
1778 | static __net_exit void ipv4_mib_exit_net(struct net *net) | |
1779 | { | |
acb32ba3 | 1780 | kfree(net->mib.icmpmsg_statistics); |
698365fa WC |
1781 | free_percpu(net->mib.icmp_statistics); |
1782 | free_percpu(net->mib.udplite_statistics); | |
1783 | free_percpu(net->mib.udp_statistics); | |
1784 | free_percpu(net->mib.net_statistics); | |
1785 | free_percpu(net->mib.ip_statistics); | |
1786 | free_percpu(net->mib.tcp_statistics); | |
9b4661bd PE |
1787 | } |
1788 | ||
1789 | static __net_initdata struct pernet_operations ipv4_mib_ops = { | |
1790 | .init = ipv4_mib_init_net, | |
1791 | .exit = ipv4_mib_exit_net, | |
1792 | }; | |
1793 | ||
1da177e4 LT |
1794 | static int __init init_ipv4_mibs(void) |
1795 | { | |
d89cbbb1 | 1796 | return register_pernet_subsys(&ipv4_mib_ops); |
1da177e4 LT |
1797 | } |
1798 | ||
c9d8f1a6 CW |
1799 | static __net_init int inet_init_net(struct net *net) |
1800 | { | |
1801 | /* | |
1802 | * Set defaults for local port range | |
1803 | */ | |
1804 | seqlock_init(&net->ipv4.ip_local_ports.lock); | |
1805 | net->ipv4.ip_local_ports.range[0] = 32768; | |
07f4c900 | 1806 | net->ipv4.ip_local_ports.range[1] = 60999; |
ba6b918a CW |
1807 | |
1808 | seqlock_init(&net->ipv4.ping_group_range.lock); | |
1809 | /* | |
1810 | * Sane defaults - nobody may create ping sockets. | |
1811 | * Boot scripts should set this to distro-specific group. | |
1812 | */ | |
1813 | net->ipv4.ping_group_range.range[0] = make_kgid(&init_user_ns, 1); | |
1814 | net->ipv4.ping_group_range.range[1] = make_kgid(&init_user_ns, 0); | |
049bbf58 EG |
1815 | |
1816 | /* Default values for sysctl-controlled parameters. | |
1817 | * We set them here, in case sysctl is not compiled. | |
1818 | */ | |
1819 | net->ipv4.sysctl_ip_default_ttl = IPDEFTTL; | |
432e05d3 | 1820 | net->ipv4.sysctl_ip_fwd_update_priority = 1; |
049bbf58 EG |
1821 | net->ipv4.sysctl_ip_dynaddr = 0; |
1822 | net->ipv4.sysctl_ip_early_demux = 1; | |
dddb64bc | 1823 | net->ipv4.sysctl_udp_early_demux = 1; |
1824 | net->ipv4.sysctl_tcp_early_demux = 1; | |
4548b683 KJ |
1825 | #ifdef CONFIG_SYSCTL |
1826 | net->ipv4.sysctl_ip_prot_sock = PROT_SOCK; | |
1827 | #endif | |
049bbf58 | 1828 | |
1714020e NB |
1829 | /* Some igmp sysctl, whose values are always used */ |
1830 | net->ipv4.sysctl_igmp_max_memberships = 20; | |
1831 | net->ipv4.sysctl_igmp_max_msf = 10; | |
1832 | /* IGMP reports for link-local multicast groups are enabled by default */ | |
1833 | net->ipv4.sysctl_igmp_llm_reports = 1; | |
1834 | net->ipv4.sysctl_igmp_qrv = 2; | |
1835 | ||
c9d8f1a6 CW |
1836 | return 0; |
1837 | } | |
1838 | ||
1839 | static __net_exit void inet_exit_net(struct net *net) | |
1840 | { | |
1841 | } | |
1842 | ||
1843 | static __net_initdata struct pernet_operations af_inet_ops = { | |
1844 | .init = inet_init_net, | |
1845 | .exit = inet_exit_net, | |
1846 | }; | |
1847 | ||
1848 | static int __init init_inet_pernet_ops(void) | |
1849 | { | |
1850 | return register_pernet_subsys(&af_inet_ops); | |
1851 | } | |
1852 | ||
1da177e4 | 1853 | static int ipv4_proc_init(void); |
1da177e4 | 1854 | |
30e224d7 HX |
1855 | /* |
1856 | * IP protocol layer initialiser | |
1857 | */ | |
1858 | ||
22061d80 VY |
1859 | static struct packet_offload ip_packet_offload __read_mostly = { |
1860 | .type = cpu_to_be16(ETH_P_IP), | |
f191a1d1 | 1861 | .callbacks = { |
f191a1d1 VY |
1862 | .gso_segment = inet_gso_segment, |
1863 | .gro_receive = inet_gro_receive, | |
1864 | .gro_complete = inet_gro_complete, | |
1865 | }, | |
30e224d7 HX |
1866 | }; |
1867 | ||
cb32f511 ED |
1868 | static const struct net_offload ipip_offload = { |
1869 | .callbacks = { | |
418e897e | 1870 | .gso_segment = ipip_gso_segment, |
fac8e0f5 | 1871 | .gro_receive = ipip_gro_receive, |
b8cba75b | 1872 | .gro_complete = ipip_gro_complete, |
cb32f511 ED |
1873 | }, |
1874 | }; | |
1875 | ||
93b1b31f TZ |
1876 | static int __init ipip_offload_init(void) |
1877 | { | |
1878 | return inet_add_offload(&ipip_offload, IPPROTO_IPIP); | |
1879 | } | |
1880 | ||
808a8f88 VY |
1881 | static int __init ipv4_offload_init(void) |
1882 | { | |
1883 | /* | |
1884 | * Add offloads | |
1885 | */ | |
da5bab07 | 1886 | if (udpv4_offload_init() < 0) |
808a8f88 | 1887 | pr_crit("%s: Cannot add UDP protocol offload\n", __func__); |
28850dc7 DB |
1888 | if (tcpv4_offload_init() < 0) |
1889 | pr_crit("%s: Cannot add TCP protocol offload\n", __func__); | |
93b1b31f TZ |
1890 | if (ipip_offload_init() < 0) |
1891 | pr_crit("%s: Cannot add IPIP protocol offload\n", __func__); | |
808a8f88 VY |
1892 | |
1893 | dev_add_offload(&ip_packet_offload); | |
1894 | return 0; | |
1895 | } | |
1896 | ||
1897 | fs_initcall(ipv4_offload_init); | |
1898 | ||
1899 | static struct packet_type ip_packet_type __read_mostly = { | |
1900 | .type = cpu_to_be16(ETH_P_IP), | |
1901 | .func = ip_rcv, | |
17266ee9 | 1902 | .list_func = ip_list_rcv, |
808a8f88 VY |
1903 | }; |
1904 | ||
1da177e4 LT |
1905 | static int __init inet_init(void) |
1906 | { | |
1da177e4 LT |
1907 | struct inet_protosw *q; |
1908 | struct list_head *r; | |
1909 | int rc = -EINVAL; | |
1910 | ||
b4772ef8 | 1911 | sock_skb_cb_check_size(sizeof(struct inet_skb_parm)); |
1da177e4 LT |
1912 | |
1913 | rc = proto_register(&tcp_prot, 1); | |
1914 | if (rc) | |
122ff243 | 1915 | goto out; |
1da177e4 LT |
1916 | |
1917 | rc = proto_register(&udp_prot, 1); | |
1918 | if (rc) | |
1919 | goto out_unregister_tcp_proto; | |
1920 | ||
1921 | rc = proto_register(&raw_prot, 1); | |
1922 | if (rc) | |
1923 | goto out_unregister_udp_proto; | |
1924 | ||
c319b4d7 VK |
1925 | rc = proto_register(&ping_prot, 1); |
1926 | if (rc) | |
1927 | goto out_unregister_raw_proto; | |
1928 | ||
1da177e4 | 1929 | /* |
e905a9ed | 1930 | * Tell SOCKET that we are alive... |
1da177e4 LT |
1931 | */ |
1932 | ||
e905a9ed | 1933 | (void)sock_register(&inet_family_ops); |
1da177e4 | 1934 | |
bd7b1533 AV |
1935 | #ifdef CONFIG_SYSCTL |
1936 | ip_static_sysctl_init(); | |
1937 | #endif | |
1938 | ||
1da177e4 LT |
1939 | /* |
1940 | * Add all the base protocols. | |
1941 | */ | |
1942 | ||
1943 | if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0) | |
058bd4d2 | 1944 | pr_crit("%s: Cannot add ICMP protocol\n", __func__); |
1da177e4 | 1945 | if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0) |
058bd4d2 | 1946 | pr_crit("%s: Cannot add UDP protocol\n", __func__); |
1da177e4 | 1947 | if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0) |
058bd4d2 | 1948 | pr_crit("%s: Cannot add TCP protocol\n", __func__); |
1da177e4 LT |
1949 | #ifdef CONFIG_IP_MULTICAST |
1950 | if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0) | |
058bd4d2 | 1951 | pr_crit("%s: Cannot add IGMP protocol\n", __func__); |
1da177e4 LT |
1952 | #endif |
1953 | ||
1954 | /* Register the socket-side information for inet_create. */ | |
1955 | for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r) | |
1956 | INIT_LIST_HEAD(r); | |
1957 | ||
1958 | for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q) | |
1959 | inet_register_protosw(q); | |
1960 | ||
1961 | /* | |
1962 | * Set the ARP module up | |
1963 | */ | |
1964 | ||
1965 | arp_init(); | |
1966 | ||
e905a9ed YH |
1967 | /* |
1968 | * Set the IP module up | |
1969 | */ | |
1da177e4 LT |
1970 | |
1971 | ip_init(); | |
1972 | ||
1da177e4 LT |
1973 | /* Setup TCP slab cache for open requests. */ |
1974 | tcp_init(); | |
1975 | ||
95766fff HA |
1976 | /* Setup UDP memory threshold */ |
1977 | udp_init(); | |
1978 | ||
ba4e58ec GR |
1979 | /* Add UDP-Lite (RFC 3828) */ |
1980 | udplite4_register(); | |
1da177e4 | 1981 | |
6897445f MM |
1982 | raw_init(); |
1983 | ||
c319b4d7 VK |
1984 | ping_init(); |
1985 | ||
1da177e4 LT |
1986 | /* |
1987 | * Set the ICMP layer up | |
1988 | */ | |
1989 | ||
a5710d65 DL |
1990 | if (icmp_init() < 0) |
1991 | panic("Failed to create the ICMP control socket.\n"); | |
1da177e4 LT |
1992 | |
1993 | /* | |
1994 | * Initialise the multicast router | |
1995 | */ | |
1996 | #if defined(CONFIG_IP_MROUTE) | |
03d2f897 | 1997 | if (ip_mr_init()) |
058bd4d2 | 1998 | pr_crit("%s: Cannot init ipv4 mroute\n", __func__); |
1da177e4 | 1999 | #endif |
c9d8f1a6 CW |
2000 | |
2001 | if (init_inet_pernet_ops()) | |
2002 | pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__); | |
1da177e4 LT |
2003 | /* |
2004 | * Initialise per-cpu ipv4 mibs | |
e905a9ed | 2005 | */ |
1da177e4 | 2006 | |
132adf54 | 2007 | if (init_ipv4_mibs()) |
058bd4d2 | 2008 | pr_crit("%s: Cannot init ipv4 mibs\n", __func__); |
e905a9ed | 2009 | |
1da177e4 LT |
2010 | ipv4_proc_init(); |
2011 | ||
2012 | ipfrag_init(); | |
2013 | ||
30e224d7 HX |
2014 | dev_add_pack(&ip_packet_type); |
2015 | ||
045a0fa0 TG |
2016 | ip_tunnel_core_init(); |
2017 | ||
1da177e4 LT |
2018 | rc = 0; |
2019 | out: | |
2020 | return rc; | |
c319b4d7 VK |
2021 | out_unregister_raw_proto: |
2022 | proto_unregister(&raw_prot); | |
1da177e4 LT |
2023 | out_unregister_udp_proto: |
2024 | proto_unregister(&udp_prot); | |
321efff7 OK |
2025 | out_unregister_tcp_proto: |
2026 | proto_unregister(&tcp_prot); | |
1da177e4 LT |
2027 | goto out; |
2028 | } | |
2029 | ||
a536e077 | 2030 | fs_initcall(inet_init); |
1da177e4 LT |
2031 | |
2032 | /* ------------------------------------------------------------------------ */ | |
2033 | ||
2034 | #ifdef CONFIG_PROC_FS | |
1da177e4 LT |
2035 | static int __init ipv4_proc_init(void) |
2036 | { | |
2037 | int rc = 0; | |
2038 | ||
2039 | if (raw_proc_init()) | |
2040 | goto out_raw; | |
2041 | if (tcp4_proc_init()) | |
2042 | goto out_tcp; | |
2043 | if (udp4_proc_init()) | |
2044 | goto out_udp; | |
c319b4d7 VK |
2045 | if (ping_proc_init()) |
2046 | goto out_ping; | |
1da177e4 LT |
2047 | if (ip_misc_proc_init()) |
2048 | goto out_misc; | |
2049 | out: | |
2050 | return rc; | |
2051 | out_misc: | |
c319b4d7 VK |
2052 | ping_proc_exit(); |
2053 | out_ping: | |
1da177e4 LT |
2054 | udp4_proc_exit(); |
2055 | out_udp: | |
2056 | tcp4_proc_exit(); | |
2057 | out_tcp: | |
2058 | raw_proc_exit(); | |
2059 | out_raw: | |
2060 | rc = -ENOMEM; | |
2061 | goto out; | |
2062 | } | |
2063 | ||
2064 | #else /* CONFIG_PROC_FS */ | |
2065 | static int __init ipv4_proc_init(void) | |
2066 | { | |
2067 | return 0; | |
2068 | } | |
2069 | #endif /* CONFIG_PROC_FS */ |