]>
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> |
dc13c876 | 119 | #include <net/compat.h> |
1da177e4 | 120 | |
563e0bb0 | 121 | #include <trace/events/sock.h> |
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); | |
8b27dae5 ED |
136 | if (sk->sk_rx_skb_cache) { |
137 | __kfree_skb(sk->sk_rx_skb_cache); | |
138 | sk->sk_rx_skb_cache = NULL; | |
139 | } | |
1da177e4 LT |
140 | __skb_queue_purge(&sk->sk_error_queue); |
141 | ||
95766fff HA |
142 | sk_mem_reclaim(sk); |
143 | ||
1da177e4 | 144 | if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) { |
3d1427f8 | 145 | pr_err("Attempt to release TCP socket in state %d %p\n", |
1da177e4 LT |
146 | sk->sk_state, sk); |
147 | return; | |
148 | } | |
149 | if (!sock_flag(sk, SOCK_DEAD)) { | |
3d1427f8 | 150 | pr_err("Attempt to release alive inet socket %p\n", sk); |
1da177e4 LT |
151 | return; |
152 | } | |
153 | ||
547b792c | 154 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); |
14afee4b | 155 | WARN_ON(refcount_read(&sk->sk_wmem_alloc)); |
547b792c IJ |
156 | WARN_ON(sk->sk_wmem_queued); |
157 | WARN_ON(sk->sk_forward_alloc); | |
1da177e4 | 158 | |
f6d8bd05 | 159 | kfree(rcu_dereference_protected(inet->inet_opt, 1)); |
5869b8fa | 160 | dst_release(rcu_dereference_protected(sk->sk_dst_cache, 1)); |
41063e9d | 161 | dst_release(sk->sk_rx_dst); |
e6848976 | 162 | sk_refcnt_debug_dec(sk); |
1da177e4 | 163 | } |
3d1427f8 | 164 | EXPORT_SYMBOL(inet_sock_destruct); |
1da177e4 LT |
165 | |
166 | /* | |
167 | * The routines beyond this point handle the behaviour of an AF_INET | |
168 | * socket object. Mostly it punts to the subprotocols of IP to do | |
169 | * the work. | |
170 | */ | |
171 | ||
172 | /* | |
173 | * Automatically bind an unbound socket. | |
174 | */ | |
175 | ||
176 | static int inet_autobind(struct sock *sk) | |
177 | { | |
178 | struct inet_sock *inet; | |
179 | /* We may need to bind the socket. */ | |
180 | lock_sock(sk); | |
181 | inet = inet_sk(sk); | |
c720c7e8 | 182 | if (!inet->inet_num) { |
1da177e4 LT |
183 | if (sk->sk_prot->get_port(sk, 0)) { |
184 | release_sock(sk); | |
185 | return -EAGAIN; | |
186 | } | |
c720c7e8 | 187 | inet->inet_sport = htons(inet->inet_num); |
1da177e4 LT |
188 | } |
189 | release_sock(sk); | |
190 | return 0; | |
191 | } | |
192 | ||
193 | /* | |
194 | * Move a socket into listening state. | |
195 | */ | |
196 | int inet_listen(struct socket *sock, int backlog) | |
197 | { | |
198 | struct sock *sk = sock->sk; | |
199 | unsigned char old_state; | |
e1cfcbe8 | 200 | int err, tcp_fastopen; |
1da177e4 LT |
201 | |
202 | lock_sock(sk); | |
203 | ||
204 | err = -EINVAL; | |
205 | if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM) | |
206 | goto out; | |
207 | ||
208 | old_state = sk->sk_state; | |
209 | if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN))) | |
210 | goto out; | |
211 | ||
099ecf59 | 212 | WRITE_ONCE(sk->sk_max_ack_backlog, backlog); |
1da177e4 LT |
213 | /* Really, if the socket is already in listen state |
214 | * we can only allow the backlog to be adjusted. | |
215 | */ | |
216 | if (old_state != TCP_LISTEN) { | |
cebc5cba | 217 | /* Enable TFO w/o requiring TCP_FASTOPEN socket option. |
8336886f | 218 | * Note that only TCP sockets (SOCK_STREAM) will reach here. |
cebc5cba YC |
219 | * Also fastopen backlog may already been set via the option |
220 | * because the socket was in TCP_LISTEN state previously but | |
221 | * was shutdown() rather than close(). | |
8336886f | 222 | */ |
e1cfcbe8 HY |
223 | tcp_fastopen = sock_net(sk)->ipv4.sysctl_tcp_fastopen; |
224 | if ((tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) && | |
225 | (tcp_fastopen & TFO_SERVER_ENABLE) && | |
0536fcc0 | 226 | !inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) { |
cebc5cba | 227 | fastopen_queue_tune(sk, backlog); |
43713848 | 228 | tcp_fastopen_init_key_once(sock_net(sk)); |
8336886f | 229 | } |
cebc5cba | 230 | |
72a3effa | 231 | err = inet_csk_listen_start(sk, backlog); |
1da177e4 LT |
232 | if (err) |
233 | goto out; | |
f333ee0c | 234 | tcp_call_bpf(sk, BPF_SOCK_OPS_TCP_LISTEN_CB, 0, NULL); |
1da177e4 | 235 | } |
1da177e4 LT |
236 | err = 0; |
237 | ||
238 | out: | |
239 | release_sock(sk); | |
240 | return err; | |
241 | } | |
3d1427f8 | 242 | EXPORT_SYMBOL(inet_listen); |
1da177e4 LT |
243 | |
244 | /* | |
245 | * Create an inet socket. | |
246 | */ | |
247 | ||
3f378b68 EP |
248 | static int inet_create(struct net *net, struct socket *sock, int protocol, |
249 | int kern) | |
1da177e4 LT |
250 | { |
251 | struct sock *sk; | |
1da177e4 LT |
252 | struct inet_protosw *answer; |
253 | struct inet_sock *inet; | |
254 | struct proto *answer_prot; | |
255 | unsigned char answer_flags; | |
bb97d31f | 256 | int try_loading_module = 0; |
86c8f9d1 | 257 | int err; |
1da177e4 | 258 | |
79462ad0 HFS |
259 | if (protocol < 0 || protocol >= IPPROTO_MAX) |
260 | return -EINVAL; | |
261 | ||
1da177e4 LT |
262 | sock->state = SS_UNCONNECTED; |
263 | ||
264 | /* Look for the requested type/protocol pair. */ | |
bb97d31f | 265 | lookup_protocol: |
86c8f9d1 | 266 | err = -ESOCKTNOSUPPORT; |
1da177e4 | 267 | rcu_read_lock(); |
696adfe8 | 268 | list_for_each_entry_rcu(answer, &inetsw[sock->type], list) { |
1da177e4 | 269 | |
696adfe8 | 270 | err = 0; |
1da177e4 LT |
271 | /* Check the non-wild match. */ |
272 | if (protocol == answer->protocol) { | |
273 | if (protocol != IPPROTO_IP) | |
274 | break; | |
275 | } else { | |
276 | /* Check for the two wild cases. */ | |
277 | if (IPPROTO_IP == protocol) { | |
278 | protocol = answer->protocol; | |
279 | break; | |
280 | } | |
281 | if (IPPROTO_IP == answer->protocol) | |
282 | break; | |
283 | } | |
86c8f9d1 | 284 | err = -EPROTONOSUPPORT; |
1da177e4 LT |
285 | } |
286 | ||
696adfe8 | 287 | if (unlikely(err)) { |
bb97d31f ACM |
288 | if (try_loading_module < 2) { |
289 | rcu_read_unlock(); | |
290 | /* | |
291 | * Be more specific, e.g. net-pf-2-proto-132-type-1 | |
292 | * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM) | |
293 | */ | |
294 | if (++try_loading_module == 1) | |
295 | request_module("net-pf-%d-proto-%d-type-%d", | |
296 | PF_INET, protocol, sock->type); | |
297 | /* | |
298 | * Fall back to generic, e.g. net-pf-2-proto-132 | |
299 | * (net-pf-PF_INET-proto-IPPROTO_SCTP) | |
300 | */ | |
301 | else | |
302 | request_module("net-pf-%d-proto-%d", | |
303 | PF_INET, protocol); | |
304 | goto lookup_protocol; | |
305 | } else | |
306 | goto out_rcu_unlock; | |
307 | } | |
308 | ||
1da177e4 | 309 | err = -EPERM; |
52e804c6 EB |
310 | if (sock->type == SOCK_RAW && !kern && |
311 | !ns_capable(net->user_ns, CAP_NET_RAW)) | |
1da177e4 | 312 | goto out_rcu_unlock; |
1da177e4 LT |
313 | |
314 | sock->ops = answer->ops; | |
315 | answer_prot = answer->prot; | |
1da177e4 LT |
316 | answer_flags = answer->flags; |
317 | rcu_read_unlock(); | |
318 | ||
51456b29 | 319 | WARN_ON(!answer_prot->slab); |
1da177e4 LT |
320 | |
321 | err = -ENOBUFS; | |
11aa9c28 | 322 | sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot, kern); |
51456b29 | 323 | if (!sk) |
1da177e4 LT |
324 | goto out; |
325 | ||
326 | err = 0; | |
1da177e4 | 327 | if (INET_PROTOSW_REUSE & answer_flags) |
4a17fd52 | 328 | sk->sk_reuse = SK_CAN_REUSE; |
1da177e4 LT |
329 | |
330 | inet = inet_sk(sk); | |
469de9b9 | 331 | inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; |
1da177e4 | 332 | |
7b2ff18e JO |
333 | inet->nodefrag = 0; |
334 | ||
1da177e4 | 335 | if (SOCK_RAW == sock->type) { |
c720c7e8 | 336 | inet->inet_num = protocol; |
1da177e4 LT |
337 | if (IPPROTO_RAW == protocol) |
338 | inet->hdrincl = 1; | |
339 | } | |
340 | ||
974eda11 | 341 | if (net->ipv4.sysctl_ip_no_pmtu_disc) |
1da177e4 LT |
342 | inet->pmtudisc = IP_PMTUDISC_DONT; |
343 | else | |
344 | inet->pmtudisc = IP_PMTUDISC_WANT; | |
345 | ||
c720c7e8 | 346 | inet->inet_id = 0; |
1da177e4 LT |
347 | |
348 | sock_init_data(sock, sk); | |
349 | ||
350 | sk->sk_destruct = inet_sock_destruct; | |
1da177e4 LT |
351 | sk->sk_protocol = protocol; |
352 | sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv; | |
353 | ||
354 | inet->uc_ttl = -1; | |
355 | inet->mc_loop = 1; | |
356 | inet->mc_ttl = 1; | |
f771bef9 | 357 | inet->mc_all = 1; |
1da177e4 LT |
358 | inet->mc_index = 0; |
359 | inet->mc_list = NULL; | |
4c507d28 | 360 | inet->rcv_tos = 0; |
1da177e4 | 361 | |
e6848976 | 362 | sk_refcnt_debug_inc(sk); |
1da177e4 | 363 | |
c720c7e8 | 364 | if (inet->inet_num) { |
1da177e4 LT |
365 | /* It assumes that any protocol which allows |
366 | * the user to assign a number at socket | |
367 | * creation time automatically | |
368 | * shares. | |
369 | */ | |
c720c7e8 | 370 | inet->inet_sport = htons(inet->inet_num); |
1da177e4 | 371 | /* Add to protocol hash chains. */ |
086c653f CG |
372 | err = sk->sk_prot->hash(sk); |
373 | if (err) { | |
374 | sk_common_release(sk); | |
375 | goto out; | |
376 | } | |
1da177e4 LT |
377 | } |
378 | ||
379 | if (sk->sk_prot->init) { | |
380 | err = sk->sk_prot->init(sk); | |
61023658 DA |
381 | if (err) { |
382 | sk_common_release(sk); | |
383 | goto out; | |
384 | } | |
385 | } | |
386 | ||
387 | if (!kern) { | |
388 | err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk); | |
389 | if (err) { | |
1da177e4 | 390 | sk_common_release(sk); |
61023658 DA |
391 | goto out; |
392 | } | |
1da177e4 LT |
393 | } |
394 | out: | |
395 | return err; | |
396 | out_rcu_unlock: | |
397 | rcu_read_unlock(); | |
398 | goto out; | |
399 | } | |
400 | ||
401 | ||
402 | /* | |
403 | * The peer socket should always be NULL (or else). When we call this | |
404 | * function we are destroying the object and from then on nobody | |
405 | * should refer to it. | |
406 | */ | |
407 | int inet_release(struct socket *sock) | |
408 | { | |
409 | struct sock *sk = sock->sk; | |
410 | ||
411 | if (sk) { | |
412 | long timeout; | |
413 | ||
f5836749 SF |
414 | if (!sk->sk_kern_sock) |
415 | BPF_CGROUP_RUN_PROG_INET_SOCK_RELEASE(sk); | |
416 | ||
1da177e4 LT |
417 | /* Applications forget to leave groups before exiting */ |
418 | ip_mc_drop_socket(sk); | |
419 | ||
420 | /* If linger is set, we don't return until the close | |
421 | * is complete. Otherwise we return immediately. The | |
422 | * actually closing is done the same either way. | |
423 | * | |
424 | * If the close is due to the process exiting, we never | |
425 | * linger.. | |
426 | */ | |
427 | timeout = 0; | |
428 | if (sock_flag(sk, SOCK_LINGER) && | |
429 | !(current->flags & PF_EXITING)) | |
430 | timeout = sk->sk_lingertime; | |
1da177e4 | 431 | sk->sk_prot->close(sk, timeout); |
2b81f816 | 432 | sock->sk = NULL; |
1da177e4 LT |
433 | } |
434 | return 0; | |
435 | } | |
3d1427f8 | 436 | EXPORT_SYMBOL(inet_release); |
1da177e4 | 437 | |
1da177e4 LT |
438 | int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) |
439 | { | |
1da177e4 | 440 | struct sock *sk = sock->sk; |
1da177e4 LT |
441 | int err; |
442 | ||
443 | /* If the socket has its own bind function then use it. (RAW) */ | |
444 | if (sk->sk_prot->bind) { | |
3679d585 | 445 | return sk->sk_prot->bind(sk, uaddr, addr_len); |
1da177e4 | 446 | } |
1da177e4 | 447 | if (addr_len < sizeof(struct sockaddr_in)) |
3679d585 | 448 | return -EINVAL; |
4fbac77d AI |
449 | |
450 | /* BPF prog is run before any checks are done so that if the prog | |
451 | * changes context in a wrong way it will be caught. | |
452 | */ | |
427167c0 | 453 | err = BPF_CGROUP_RUN_PROG_INET4_BIND_LOCK(sk, uaddr); |
4fbac77d | 454 | if (err) |
3679d585 AI |
455 | return err; |
456 | ||
cb0721c7 | 457 | return __inet_bind(sk, uaddr, addr_len, BIND_WITH_LOCK); |
3679d585 AI |
458 | } |
459 | EXPORT_SYMBOL(inet_bind); | |
460 | ||
461 | int __inet_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len, | |
cb0721c7 | 462 | u32 flags) |
3679d585 AI |
463 | { |
464 | struct sockaddr_in *addr = (struct sockaddr_in *)uaddr; | |
465 | struct inet_sock *inet = inet_sk(sk); | |
466 | struct net *net = sock_net(sk); | |
467 | unsigned short snum; | |
468 | int chk_addr_ret; | |
469 | u32 tb_id = RT_TABLE_LOCAL; | |
470 | int err; | |
1da177e4 | 471 | |
c349a528 | 472 | if (addr->sin_family != AF_INET) { |
29c486df ED |
473 | /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET) |
474 | * only if s_addr is INADDR_ANY. | |
475 | */ | |
c349a528 | 476 | err = -EAFNOSUPPORT; |
29c486df ED |
477 | if (addr->sin_family != AF_UNSPEC || |
478 | addr->sin_addr.s_addr != htonl(INADDR_ANY)) | |
479 | goto out; | |
c349a528 | 480 | } |
d0733d2e | 481 | |
3236b004 | 482 | tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id; |
30bbaa19 | 483 | chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id); |
1da177e4 LT |
484 | |
485 | /* Not specified by any standard per-se, however it breaks too | |
486 | * many applications when removed. It is unfortunate since | |
487 | * allowing applications to make a non-local bind solves | |
488 | * several problems with systems using dynamic addressing. | |
489 | * (ie. your servers still start up even if your ISDN link | |
490 | * is temporarily down) | |
491 | */ | |
492 | err = -EADDRNOTAVAIL; | |
83ba4645 | 493 | if (!inet_can_nonlocal_bind(net, inet) && |
e6f1cebf | 494 | addr->sin_addr.s_addr != htonl(INADDR_ANY) && |
1da177e4 LT |
495 | chk_addr_ret != RTN_LOCAL && |
496 | chk_addr_ret != RTN_MULTICAST && | |
497 | chk_addr_ret != RTN_BROADCAST) | |
498 | goto out; | |
499 | ||
500 | snum = ntohs(addr->sin_port); | |
501 | err = -EACCES; | |
82f31ebf | 502 | if (snum && inet_port_requires_bind_service(net, snum) && |
3594698a | 503 | !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) |
1da177e4 LT |
504 | goto out; |
505 | ||
506 | /* We keep a pair of addresses. rcv_saddr is the one | |
507 | * used by hash lookups, and saddr is used for transmit. | |
508 | * | |
509 | * In the BSD API these are the same except where it | |
510 | * would be illegal to use them (multicast/broadcast) in | |
511 | * which case the sending device address is used. | |
512 | */ | |
cb0721c7 | 513 | if (flags & BIND_WITH_LOCK) |
3679d585 | 514 | lock_sock(sk); |
1da177e4 LT |
515 | |
516 | /* Check these errors (active socket, double bind). */ | |
517 | err = -EINVAL; | |
c720c7e8 | 518 | if (sk->sk_state != TCP_CLOSE || inet->inet_num) |
1da177e4 LT |
519 | goto out_release_sock; |
520 | ||
c720c7e8 | 521 | inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; |
1da177e4 | 522 | if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) |
c720c7e8 | 523 | inet->inet_saddr = 0; /* Use device */ |
1da177e4 LT |
524 | |
525 | /* Make sure we are allowed to bind here. */ | |
aac3fc32 | 526 | if (snum || !(inet->bind_address_no_port || |
cb0721c7 | 527 | (flags & BIND_FORCE_ADDRESS_NO_PORT))) { |
aac3fc32 AI |
528 | if (sk->sk_prot->get_port(sk, snum)) { |
529 | inet->inet_saddr = inet->inet_rcv_saddr = 0; | |
530 | err = -EADDRINUSE; | |
531 | goto out_release_sock; | |
532 | } | |
8086fbaf SF |
533 | if (!(flags & BIND_FROM_BPF)) { |
534 | err = BPF_CGROUP_RUN_PROG_INET4_POST_BIND(sk); | |
535 | if (err) { | |
536 | inet->inet_saddr = inet->inet_rcv_saddr = 0; | |
537 | goto out_release_sock; | |
538 | } | |
aac3fc32 | 539 | } |
1da177e4 LT |
540 | } |
541 | ||
c720c7e8 | 542 | if (inet->inet_rcv_saddr) |
1da177e4 LT |
543 | sk->sk_userlocks |= SOCK_BINDADDR_LOCK; |
544 | if (snum) | |
545 | sk->sk_userlocks |= SOCK_BINDPORT_LOCK; | |
c720c7e8 ED |
546 | inet->inet_sport = htons(inet->inet_num); |
547 | inet->inet_daddr = 0; | |
548 | inet->inet_dport = 0; | |
1da177e4 LT |
549 | sk_dst_reset(sk); |
550 | err = 0; | |
551 | out_release_sock: | |
cb0721c7 | 552 | if (flags & BIND_WITH_LOCK) |
3679d585 | 553 | release_sock(sk); |
1da177e4 LT |
554 | out: |
555 | return err; | |
556 | } | |
557 | ||
5e73ea1a | 558 | int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr, |
1da177e4 LT |
559 | int addr_len, int flags) |
560 | { | |
561 | struct sock *sk = sock->sk; | |
d74bad4e | 562 | int err; |
1da177e4 | 563 | |
6503d961 CG |
564 | if (addr_len < sizeof(uaddr->sa_family)) |
565 | return -EINVAL; | |
1da177e4 LT |
566 | if (uaddr->sa_family == AF_UNSPEC) |
567 | return sk->sk_prot->disconnect(sk, flags); | |
568 | ||
d74bad4e AI |
569 | if (BPF_CGROUP_PRE_CONNECT_ENABLED(sk)) { |
570 | err = sk->sk_prot->pre_connect(sk, uaddr, addr_len); | |
571 | if (err) | |
572 | return err; | |
573 | } | |
574 | ||
c720c7e8 | 575 | if (!inet_sk(sk)->inet_num && inet_autobind(sk)) |
1da177e4 | 576 | return -EAGAIN; |
e3192690 | 577 | return sk->sk_prot->connect(sk, uaddr, addr_len); |
1da177e4 | 578 | } |
3d1427f8 | 579 | EXPORT_SYMBOL(inet_dgram_connect); |
1da177e4 | 580 | |
783237e8 | 581 | static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias) |
1da177e4 | 582 | { |
14135f30 | 583 | DEFINE_WAIT_FUNC(wait, woken_wake_function); |
1da177e4 | 584 | |
14135f30 | 585 | add_wait_queue(sk_sleep(sk), &wait); |
783237e8 | 586 | sk->sk_write_pending += writebias; |
1da177e4 LT |
587 | |
588 | /* Basic assumption: if someone sets sk->sk_err, he _must_ | |
589 | * change state of the socket from TCP_SYN_*. | |
590 | * Connect() does not allow to get error notifications | |
591 | * without closing the socket. | |
592 | */ | |
593 | while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
594 | release_sock(sk); | |
14135f30 | 595 | timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); |
1da177e4 LT |
596 | lock_sock(sk); |
597 | if (signal_pending(current) || !timeo) | |
598 | break; | |
1da177e4 | 599 | } |
14135f30 | 600 | remove_wait_queue(sk_sleep(sk), &wait); |
783237e8 | 601 | sk->sk_write_pending -= writebias; |
1da177e4 LT |
602 | return timeo; |
603 | } | |
604 | ||
605 | /* | |
606 | * Connect to a remote host. There is regrettably still a little | |
607 | * TCP 'magic' in here. | |
608 | */ | |
cf60af03 | 609 | int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, |
3979ad7e | 610 | int addr_len, int flags, int is_sendmsg) |
1da177e4 LT |
611 | { |
612 | struct sock *sk = sock->sk; | |
613 | int err; | |
614 | long timeo; | |
615 | ||
19f6d3f3 WW |
616 | /* |
617 | * uaddr can be NULL and addr_len can be 0 if: | |
618 | * sk is a TCP fastopen active socket and | |
619 | * TCP_FASTOPEN_CONNECT sockopt is set and | |
620 | * we already have a valid cookie for this socket. | |
621 | * In this case, user can call write() after connect(). | |
622 | * write() will invoke tcp_sendmsg_fastopen() which calls | |
623 | * __inet_stream_connect(). | |
624 | */ | |
625 | if (uaddr) { | |
626 | if (addr_len < sizeof(uaddr->sa_family)) | |
627 | return -EINVAL; | |
6503d961 | 628 | |
19f6d3f3 WW |
629 | if (uaddr->sa_family == AF_UNSPEC) { |
630 | err = sk->sk_prot->disconnect(sk, flags); | |
631 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
632 | goto out; | |
633 | } | |
1da177e4 LT |
634 | } |
635 | ||
636 | switch (sock->state) { | |
637 | default: | |
638 | err = -EINVAL; | |
639 | goto out; | |
640 | case SS_CONNECTED: | |
641 | err = -EISCONN; | |
642 | goto out; | |
643 | case SS_CONNECTING: | |
19f6d3f3 | 644 | if (inet_sk(sk)->defer_connect) |
3979ad7e | 645 | err = is_sendmsg ? -EINPROGRESS : -EISCONN; |
19f6d3f3 WW |
646 | else |
647 | err = -EALREADY; | |
1da177e4 LT |
648 | /* Fall out of switch with err, set for this state */ |
649 | break; | |
650 | case SS_UNCONNECTED: | |
651 | err = -EISCONN; | |
652 | if (sk->sk_state != TCP_CLOSE) | |
653 | goto out; | |
654 | ||
d74bad4e AI |
655 | if (BPF_CGROUP_PRE_CONNECT_ENABLED(sk)) { |
656 | err = sk->sk_prot->pre_connect(sk, uaddr, addr_len); | |
657 | if (err) | |
658 | goto out; | |
659 | } | |
660 | ||
1da177e4 LT |
661 | err = sk->sk_prot->connect(sk, uaddr, addr_len); |
662 | if (err < 0) | |
663 | goto out; | |
664 | ||
e905a9ed | 665 | sock->state = SS_CONNECTING; |
1da177e4 | 666 | |
19f6d3f3 WW |
667 | if (!err && inet_sk(sk)->defer_connect) |
668 | goto out; | |
669 | ||
1da177e4 LT |
670 | /* Just entered SS_CONNECTING state; the only |
671 | * difference is that return value in non-blocking | |
672 | * case is EINPROGRESS, rather than EALREADY. | |
673 | */ | |
674 | err = -EINPROGRESS; | |
675 | break; | |
676 | } | |
677 | ||
678 | timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); | |
679 | ||
680 | if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
783237e8 YC |
681 | int writebias = (sk->sk_protocol == IPPROTO_TCP) && |
682 | tcp_sk(sk)->fastopen_req && | |
683 | tcp_sk(sk)->fastopen_req->data ? 1 : 0; | |
684 | ||
1da177e4 | 685 | /* Error code is set above */ |
783237e8 | 686 | if (!timeo || !inet_wait_for_connect(sk, timeo, writebias)) |
1da177e4 LT |
687 | goto out; |
688 | ||
689 | err = sock_intr_errno(timeo); | |
690 | if (signal_pending(current)) | |
691 | goto out; | |
692 | } | |
693 | ||
694 | /* Connection was closed by RST, timeout, ICMP error | |
695 | * or another process disconnected us. | |
696 | */ | |
697 | if (sk->sk_state == TCP_CLOSE) | |
698 | goto sock_error; | |
699 | ||
700 | /* sk->sk_err may be not zero now, if RECVERR was ordered by user | |
701 | * and error was received after socket entered established state. | |
702 | * Hence, it is handled normally after connect() return successfully. | |
703 | */ | |
704 | ||
705 | sock->state = SS_CONNECTED; | |
706 | err = 0; | |
707 | out: | |
1da177e4 LT |
708 | return err; |
709 | ||
710 | sock_error: | |
711 | err = sock_error(sk) ? : -ECONNABORTED; | |
712 | sock->state = SS_UNCONNECTED; | |
713 | if (sk->sk_prot->disconnect(sk, flags)) | |
714 | sock->state = SS_DISCONNECTING; | |
715 | goto out; | |
716 | } | |
cf60af03 YC |
717 | EXPORT_SYMBOL(__inet_stream_connect); |
718 | ||
719 | int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, | |
720 | int addr_len, int flags) | |
721 | { | |
722 | int err; | |
723 | ||
724 | lock_sock(sock->sk); | |
3979ad7e | 725 | err = __inet_stream_connect(sock, uaddr, addr_len, flags, 0); |
cf60af03 YC |
726 | release_sock(sock->sk); |
727 | return err; | |
728 | } | |
3d1427f8 | 729 | EXPORT_SYMBOL(inet_stream_connect); |
1da177e4 LT |
730 | |
731 | /* | |
732 | * Accept a pending connection. The TCP layer now gives BSD semantics. | |
733 | */ | |
734 | ||
cdfbabfb DH |
735 | int inet_accept(struct socket *sock, struct socket *newsock, int flags, |
736 | bool kern) | |
1da177e4 LT |
737 | { |
738 | struct sock *sk1 = sock->sk; | |
739 | int err = -EINVAL; | |
cdfbabfb | 740 | struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err, kern); |
1da177e4 LT |
741 | |
742 | if (!sk2) | |
743 | goto do_err; | |
744 | ||
745 | lock_sock(sk2); | |
746 | ||
1eddcead | 747 | sock_rps_record_flow(sk2); |
547b792c | 748 | WARN_ON(!((1 << sk2->sk_state) & |
8336886f JC |
749 | (TCPF_ESTABLISHED | TCPF_SYN_RECV | |
750 | TCPF_CLOSE_WAIT | TCPF_CLOSE))); | |
1da177e4 LT |
751 | |
752 | sock_graft(sk2, newsock); | |
753 | ||
754 | newsock->state = SS_CONNECTED; | |
755 | err = 0; | |
756 | release_sock(sk2); | |
757 | do_err: | |
758 | return err; | |
759 | } | |
3d1427f8 | 760 | EXPORT_SYMBOL(inet_accept); |
1da177e4 | 761 | |
1da177e4 LT |
762 | /* |
763 | * This does both peername and sockname. | |
764 | */ | |
765 | int inet_getname(struct socket *sock, struct sockaddr *uaddr, | |
1b66d253 | 766 | int peer) |
1da177e4 LT |
767 | { |
768 | struct sock *sk = sock->sk; | |
769 | struct inet_sock *inet = inet_sk(sk); | |
38bfd8f5 | 770 | DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr); |
1da177e4 LT |
771 | |
772 | sin->sin_family = AF_INET; | |
773 | if (peer) { | |
c720c7e8 | 774 | if (!inet->inet_dport || |
1da177e4 LT |
775 | (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && |
776 | peer == 1)) | |
777 | return -ENOTCONN; | |
c720c7e8 ED |
778 | sin->sin_port = inet->inet_dport; |
779 | sin->sin_addr.s_addr = inet->inet_daddr; | |
1da177e4 | 780 | } else { |
c720c7e8 | 781 | __be32 addr = inet->inet_rcv_saddr; |
1da177e4 | 782 | if (!addr) |
c720c7e8 ED |
783 | addr = inet->inet_saddr; |
784 | sin->sin_port = inet->inet_sport; | |
1da177e4 LT |
785 | sin->sin_addr.s_addr = addr; |
786 | } | |
1b66d253 DB |
787 | if (cgroup_bpf_enabled) |
788 | BPF_CGROUP_RUN_SA_PROG_LOCK(sk, (struct sockaddr *)sin, | |
789 | peer ? BPF_CGROUP_INET4_GETPEERNAME : | |
790 | BPF_CGROUP_INET4_GETSOCKNAME, | |
791 | NULL); | |
1da177e4 | 792 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); |
9b2c45d4 | 793 | return sizeof(*sin); |
1da177e4 | 794 | } |
3d1427f8 | 795 | EXPORT_SYMBOL(inet_getname); |
1da177e4 | 796 | |
e4730936 | 797 | int inet_send_prepare(struct sock *sk) |
1da177e4 | 798 | { |
c58dc01b | 799 | sock_rps_record_flow(sk); |
fec5e652 | 800 | |
1da177e4 | 801 | /* We may need to bind the socket. */ |
7ba42910 CG |
802 | if (!inet_sk(sk)->inet_num && !sk->sk_prot->no_autobind && |
803 | inet_autobind(sk)) | |
1da177e4 LT |
804 | return -EAGAIN; |
805 | ||
e4730936 PA |
806 | return 0; |
807 | } | |
808 | EXPORT_SYMBOL_GPL(inet_send_prepare); | |
809 | ||
810 | int inet_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) | |
811 | { | |
812 | struct sock *sk = sock->sk; | |
813 | ||
814 | if (unlikely(inet_send_prepare(sk))) | |
815 | return -EAGAIN; | |
816 | ||
6f24080e PA |
817 | return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udp_sendmsg, |
818 | sk, msg, size); | |
1da177e4 | 819 | } |
3d1427f8 | 820 | EXPORT_SYMBOL(inet_sendmsg); |
1da177e4 | 821 | |
7ba42910 CG |
822 | ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, |
823 | size_t size, int flags) | |
1da177e4 LT |
824 | { |
825 | struct sock *sk = sock->sk; | |
826 | ||
e4730936 | 827 | if (unlikely(inet_send_prepare(sk))) |
1da177e4 LT |
828 | return -EAGAIN; |
829 | ||
830 | if (sk->sk_prot->sendpage) | |
831 | return sk->sk_prot->sendpage(sk, page, offset, size, flags); | |
832 | return sock_no_sendpage(sock, page, offset, size, flags); | |
833 | } | |
7ba42910 | 834 | EXPORT_SYMBOL(inet_sendpage); |
1da177e4 | 835 | |
6f24080e PA |
836 | INDIRECT_CALLABLE_DECLARE(int udp_recvmsg(struct sock *, struct msghdr *, |
837 | size_t, int, int, int *)); | |
1b784140 YX |
838 | int inet_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, |
839 | int flags) | |
fec5e652 TH |
840 | { |
841 | struct sock *sk = sock->sk; | |
842 | int addr_len = 0; | |
843 | int err; | |
844 | ||
e3f2c4a3 SHY |
845 | if (likely(!(flags & MSG_ERRQUEUE))) |
846 | sock_rps_record_flow(sk); | |
fec5e652 | 847 | |
6f24080e PA |
848 | err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udp_recvmsg, |
849 | sk, msg, size, flags & MSG_DONTWAIT, | |
850 | flags & ~MSG_DONTWAIT, &addr_len); | |
fec5e652 TH |
851 | if (err >= 0) |
852 | msg->msg_namelen = addr_len; | |
853 | return err; | |
854 | } | |
855 | EXPORT_SYMBOL(inet_recvmsg); | |
1da177e4 LT |
856 | |
857 | int inet_shutdown(struct socket *sock, int how) | |
858 | { | |
859 | struct sock *sk = sock->sk; | |
860 | int err = 0; | |
861 | ||
862 | /* This should really check to make sure | |
863 | * the socket is a TCP socket. (WHY AC...) | |
864 | */ | |
865 | how++; /* maps 0->1 has the advantage of making bit 1 rcvs and | |
866 | 1->2 bit 2 snds. | |
867 | 2->3 */ | |
868 | if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */ | |
869 | return -EINVAL; | |
870 | ||
871 | lock_sock(sk); | |
872 | if (sock->state == SS_CONNECTING) { | |
873 | if ((1 << sk->sk_state) & | |
874 | (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE)) | |
875 | sock->state = SS_DISCONNECTING; | |
876 | else | |
877 | sock->state = SS_CONNECTED; | |
878 | } | |
879 | ||
880 | switch (sk->sk_state) { | |
881 | case TCP_CLOSE: | |
882 | err = -ENOTCONN; | |
883 | /* Hack to wake up other listeners, who can poll for | |
a9a08845 | 884 | EPOLLHUP, even on eg. unconnected UDP sockets -- RR */ |
a8eceea8 | 885 | fallthrough; |
1da177e4 LT |
886 | default: |
887 | sk->sk_shutdown |= how; | |
888 | if (sk->sk_prot->shutdown) | |
889 | sk->sk_prot->shutdown(sk, how); | |
890 | break; | |
891 | ||
892 | /* Remaining two branches are temporary solution for missing | |
893 | * close() in multithreaded environment. It is _not_ a good idea, | |
894 | * but we have no choice until close() is repaired at VFS level. | |
895 | */ | |
896 | case TCP_LISTEN: | |
897 | if (!(how & RCV_SHUTDOWN)) | |
898 | break; | |
a8eceea8 | 899 | fallthrough; |
1da177e4 LT |
900 | case TCP_SYN_SENT: |
901 | err = sk->sk_prot->disconnect(sk, O_NONBLOCK); | |
902 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
903 | break; | |
904 | } | |
905 | ||
906 | /* Wake up anyone sleeping in poll. */ | |
907 | sk->sk_state_change(sk); | |
908 | release_sock(sk); | |
909 | return err; | |
910 | } | |
3d1427f8 | 911 | EXPORT_SYMBOL(inet_shutdown); |
1da177e4 LT |
912 | |
913 | /* | |
914 | * ioctl() calls you can issue on an INET socket. Most of these are | |
915 | * device configuration and stuff and very rarely used. Some ioctls | |
916 | * pass on to the socket itself. | |
917 | * | |
918 | * NOTE: I like the idea of a module for the config stuff. ie ifconfig | |
919 | * loads the devconfigure module does its configuring and unloads it. | |
920 | * There's a good 20K of config code hanging around the kernel. | |
921 | */ | |
922 | ||
923 | int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) | |
924 | { | |
925 | struct sock *sk = sock->sk; | |
926 | int err = 0; | |
3b1e0a65 | 927 | struct net *net = sock_net(sk); |
03aef17b AV |
928 | void __user *p = (void __user *)arg; |
929 | struct ifreq ifr; | |
ca25c300 | 930 | struct rtentry rt; |
1da177e4 LT |
931 | |
932 | switch (cmd) { | |
3d1427f8 ED |
933 | case SIOCADDRT: |
934 | case SIOCDELRT: | |
ca25c300 AV |
935 | if (copy_from_user(&rt, p, sizeof(struct rtentry))) |
936 | return -EFAULT; | |
937 | err = ip_rt_ioctl(net, cmd, &rt); | |
938 | break; | |
3d1427f8 | 939 | case SIOCRTMSG: |
ca25c300 | 940 | err = -EINVAL; |
3d1427f8 ED |
941 | break; |
942 | case SIOCDARP: | |
943 | case SIOCGARP: | |
944 | case SIOCSARP: | |
945 | err = arp_ioctl(net, cmd, (void __user *)arg); | |
946 | break; | |
947 | case SIOCGIFADDR: | |
3d1427f8 | 948 | case SIOCGIFBRDADDR: |
3d1427f8 | 949 | case SIOCGIFNETMASK: |
3d1427f8 | 950 | case SIOCGIFDSTADDR: |
03aef17b AV |
951 | case SIOCGIFPFLAGS: |
952 | if (copy_from_user(&ifr, p, sizeof(struct ifreq))) | |
953 | return -EFAULT; | |
954 | err = devinet_ioctl(net, cmd, &ifr); | |
955 | if (!err && copy_to_user(p, &ifr, sizeof(struct ifreq))) | |
956 | err = -EFAULT; | |
957 | break; | |
958 | ||
959 | case SIOCSIFADDR: | |
960 | case SIOCSIFBRDADDR: | |
961 | case SIOCSIFNETMASK: | |
3d1427f8 ED |
962 | case SIOCSIFDSTADDR: |
963 | case SIOCSIFPFLAGS: | |
3d1427f8 | 964 | case SIOCSIFFLAGS: |
03aef17b AV |
965 | if (copy_from_user(&ifr, p, sizeof(struct ifreq))) |
966 | return -EFAULT; | |
967 | err = devinet_ioctl(net, cmd, &ifr); | |
3d1427f8 ED |
968 | break; |
969 | default: | |
970 | if (sk->sk_prot->ioctl) | |
971 | err = sk->sk_prot->ioctl(sk, cmd, arg); | |
972 | else | |
973 | err = -ENOIOCTLCMD; | |
974 | break; | |
1da177e4 LT |
975 | } |
976 | return err; | |
977 | } | |
3d1427f8 | 978 | EXPORT_SYMBOL(inet_ioctl); |
1da177e4 | 979 | |
709b46e8 | 980 | #ifdef CONFIG_COMPAT |
dc13c876 CH |
981 | static int inet_compat_routing_ioctl(struct sock *sk, unsigned int cmd, |
982 | struct compat_rtentry __user *ur) | |
983 | { | |
984 | compat_uptr_t rtdev; | |
985 | struct rtentry rt; | |
986 | ||
987 | if (copy_from_user(&rt.rt_dst, &ur->rt_dst, | |
988 | 3 * sizeof(struct sockaddr)) || | |
989 | get_user(rt.rt_flags, &ur->rt_flags) || | |
990 | get_user(rt.rt_metric, &ur->rt_metric) || | |
991 | get_user(rt.rt_mtu, &ur->rt_mtu) || | |
992 | get_user(rt.rt_window, &ur->rt_window) || | |
993 | get_user(rt.rt_irtt, &ur->rt_irtt) || | |
994 | get_user(rtdev, &ur->rt_dev)) | |
995 | return -EFAULT; | |
996 | ||
997 | rt.rt_dev = compat_ptr(rtdev); | |
998 | return ip_rt_ioctl(sock_net(sk), cmd, &rt); | |
999 | } | |
1000 | ||
686dc6b6 | 1001 | static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
709b46e8 | 1002 | { |
dc13c876 | 1003 | void __user *argp = compat_ptr(arg); |
709b46e8 | 1004 | struct sock *sk = sock->sk; |
709b46e8 | 1005 | |
dc13c876 CH |
1006 | switch (cmd) { |
1007 | case SIOCADDRT: | |
1008 | case SIOCDELRT: | |
1009 | return inet_compat_routing_ioctl(sk, cmd, argp); | |
1010 | default: | |
1011 | if (!sk->sk_prot->compat_ioctl) | |
1012 | return -ENOIOCTLCMD; | |
1013 | return sk->sk_prot->compat_ioctl(sk, cmd, arg); | |
1014 | } | |
709b46e8 | 1015 | } |
dc13c876 | 1016 | #endif /* CONFIG_COMPAT */ |
709b46e8 | 1017 | |
90ddc4f0 | 1018 | const struct proto_ops inet_stream_ops = { |
543d9cfe | 1019 | .family = PF_INET, |
583bbf06 | 1020 | .flags = PROTO_CMSG_DATA_ONLY, |
543d9cfe ACM |
1021 | .owner = THIS_MODULE, |
1022 | .release = inet_release, | |
1023 | .bind = inet_bind, | |
1024 | .connect = inet_stream_connect, | |
1025 | .socketpair = sock_no_socketpair, | |
1026 | .accept = inet_accept, | |
1027 | .getname = inet_getname, | |
a11e1d43 | 1028 | .poll = tcp_poll, |
543d9cfe | 1029 | .ioctl = inet_ioctl, |
c7cbdbf2 | 1030 | .gettstamp = sock_gettstamp, |
543d9cfe ACM |
1031 | .listen = inet_listen, |
1032 | .shutdown = inet_shutdown, | |
1033 | .setsockopt = sock_common_setsockopt, | |
1034 | .getsockopt = sock_common_getsockopt, | |
7ba42910 | 1035 | .sendmsg = inet_sendmsg, |
fec5e652 | 1036 | .recvmsg = inet_recvmsg, |
05255b82 | 1037 | #ifdef CONFIG_MMU |
93ab6cc6 | 1038 | .mmap = tcp_mmap, |
05255b82 | 1039 | #endif |
7ba42910 | 1040 | .sendpage = inet_sendpage, |
9c55e01c | 1041 | .splice_read = tcp_splice_read, |
32035585 | 1042 | .read_sock = tcp_read_sock, |
306b13eb TH |
1043 | .sendmsg_locked = tcp_sendmsg_locked, |
1044 | .sendpage_locked = tcp_sendpage_locked, | |
32035585 | 1045 | .peek_len = tcp_peek_len, |
3fdadf7d | 1046 | #ifdef CONFIG_COMPAT |
709b46e8 | 1047 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 1048 | #endif |
d1361840 | 1049 | .set_rcvlowat = tcp_set_rcvlowat, |
1da177e4 | 1050 | }; |
3d1427f8 | 1051 | EXPORT_SYMBOL(inet_stream_ops); |
1da177e4 | 1052 | |
90ddc4f0 | 1053 | const struct proto_ops inet_dgram_ops = { |
543d9cfe ACM |
1054 | .family = PF_INET, |
1055 | .owner = THIS_MODULE, | |
1056 | .release = inet_release, | |
1057 | .bind = inet_bind, | |
1058 | .connect = inet_dgram_connect, | |
1059 | .socketpair = sock_no_socketpair, | |
1060 | .accept = sock_no_accept, | |
1061 | .getname = inet_getname, | |
a11e1d43 | 1062 | .poll = udp_poll, |
543d9cfe | 1063 | .ioctl = inet_ioctl, |
c7cbdbf2 | 1064 | .gettstamp = sock_gettstamp, |
543d9cfe ACM |
1065 | .listen = sock_no_listen, |
1066 | .shutdown = inet_shutdown, | |
1067 | .setsockopt = sock_common_setsockopt, | |
1068 | .getsockopt = sock_common_getsockopt, | |
1069 | .sendmsg = inet_sendmsg, | |
fec5e652 | 1070 | .recvmsg = inet_recvmsg, |
543d9cfe ACM |
1071 | .mmap = sock_no_mmap, |
1072 | .sendpage = inet_sendpage, | |
627d2d6b | 1073 | .set_peek_off = sk_set_peek_off, |
3fdadf7d | 1074 | #ifdef CONFIG_COMPAT |
709b46e8 | 1075 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 1076 | #endif |
1da177e4 | 1077 | }; |
3d1427f8 | 1078 | EXPORT_SYMBOL(inet_dgram_ops); |
1da177e4 LT |
1079 | |
1080 | /* | |
1081 | * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without | |
a11e1d43 | 1082 | * udp_poll |
1da177e4 | 1083 | */ |
90ddc4f0 | 1084 | static const struct proto_ops inet_sockraw_ops = { |
543d9cfe ACM |
1085 | .family = PF_INET, |
1086 | .owner = THIS_MODULE, | |
1087 | .release = inet_release, | |
1088 | .bind = inet_bind, | |
1089 | .connect = inet_dgram_connect, | |
1090 | .socketpair = sock_no_socketpair, | |
1091 | .accept = sock_no_accept, | |
1092 | .getname = inet_getname, | |
a11e1d43 | 1093 | .poll = datagram_poll, |
543d9cfe | 1094 | .ioctl = inet_ioctl, |
c7cbdbf2 | 1095 | .gettstamp = sock_gettstamp, |
543d9cfe ACM |
1096 | .listen = sock_no_listen, |
1097 | .shutdown = inet_shutdown, | |
1098 | .setsockopt = sock_common_setsockopt, | |
1099 | .getsockopt = sock_common_getsockopt, | |
1100 | .sendmsg = inet_sendmsg, | |
fec5e652 | 1101 | .recvmsg = inet_recvmsg, |
543d9cfe ACM |
1102 | .mmap = sock_no_mmap, |
1103 | .sendpage = inet_sendpage, | |
3fdadf7d | 1104 | #ifdef CONFIG_COMPAT |
709b46e8 | 1105 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 1106 | #endif |
1da177e4 LT |
1107 | }; |
1108 | ||
ec1b4cf7 | 1109 | static const struct net_proto_family inet_family_ops = { |
1da177e4 LT |
1110 | .family = PF_INET, |
1111 | .create = inet_create, | |
1112 | .owner = THIS_MODULE, | |
1113 | }; | |
1114 | ||
1da177e4 LT |
1115 | /* Upon startup we insert all the elements in inetsw_array[] into |
1116 | * the linked list inetsw. | |
1117 | */ | |
1118 | static struct inet_protosw inetsw_array[] = | |
1119 | { | |
e905a9ed YH |
1120 | { |
1121 | .type = SOCK_STREAM, | |
1122 | .protocol = IPPROTO_TCP, | |
1123 | .prot = &tcp_prot, | |
1124 | .ops = &inet_stream_ops, | |
e905a9ed | 1125 | .flags = INET_PROTOSW_PERMANENT | |
d83d8461 | 1126 | INET_PROTOSW_ICSK, |
e905a9ed YH |
1127 | }, |
1128 | ||
1129 | { | |
1130 | .type = SOCK_DGRAM, | |
1131 | .protocol = IPPROTO_UDP, | |
1132 | .prot = &udp_prot, | |
1133 | .ops = &inet_dgram_ops, | |
e905a9ed | 1134 | .flags = INET_PROTOSW_PERMANENT, |
1da177e4 | 1135 | }, |
e905a9ed | 1136 | |
c319b4d7 VK |
1137 | { |
1138 | .type = SOCK_DGRAM, | |
1139 | .protocol = IPPROTO_ICMP, | |
1140 | .prot = &ping_prot, | |
77d4b1d3 | 1141 | .ops = &inet_sockraw_ops, |
c319b4d7 VK |
1142 | .flags = INET_PROTOSW_REUSE, |
1143 | }, | |
1da177e4 LT |
1144 | |
1145 | { | |
e905a9ed YH |
1146 | .type = SOCK_RAW, |
1147 | .protocol = IPPROTO_IP, /* wild card */ | |
1148 | .prot = &raw_prot, | |
1149 | .ops = &inet_sockraw_ops, | |
e905a9ed | 1150 | .flags = INET_PROTOSW_REUSE, |
1da177e4 LT |
1151 | } |
1152 | }; | |
1153 | ||
c40f6fff | 1154 | #define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array) |
1da177e4 LT |
1155 | |
1156 | void inet_register_protosw(struct inet_protosw *p) | |
1157 | { | |
1158 | struct list_head *lh; | |
1159 | struct inet_protosw *answer; | |
1160 | int protocol = p->protocol; | |
1161 | struct list_head *last_perm; | |
1162 | ||
1163 | spin_lock_bh(&inetsw_lock); | |
1164 | ||
1165 | if (p->type >= SOCK_MAX) | |
1166 | goto out_illegal; | |
1167 | ||
1168 | /* If we are trying to override a permanent protocol, bail. */ | |
1da177e4 LT |
1169 | last_perm = &inetsw[p->type]; |
1170 | list_for_each(lh, &inetsw[p->type]) { | |
1171 | answer = list_entry(lh, struct inet_protosw, list); | |
1da177e4 | 1172 | /* Check only the non-wild match. */ |
f6c53334 JZ |
1173 | if ((INET_PROTOSW_PERMANENT & answer->flags) == 0) |
1174 | break; | |
1175 | if (protocol == answer->protocol) | |
1176 | goto out_permanent; | |
1177 | last_perm = lh; | |
1da177e4 | 1178 | } |
1da177e4 LT |
1179 | |
1180 | /* Add the new entry after the last permanent entry if any, so that | |
1181 | * the new entry does not override a permanent entry when matched with | |
1182 | * a wild-card protocol. But it is allowed to override any existing | |
e905a9ed | 1183 | * non-permanent entry. This means that when we remove this entry, the |
1da177e4 LT |
1184 | * system automatically returns to the old behavior. |
1185 | */ | |
1186 | list_add_rcu(&p->list, last_perm); | |
1187 | out: | |
1188 | spin_unlock_bh(&inetsw_lock); | |
1189 | ||
1da177e4 LT |
1190 | return; |
1191 | ||
1192 | out_permanent: | |
058bd4d2 | 1193 | pr_err("Attempt to override permanent protocol %d\n", protocol); |
1da177e4 LT |
1194 | goto out; |
1195 | ||
1196 | out_illegal: | |
058bd4d2 | 1197 | pr_err("Ignoring attempt to register invalid socket type %d\n", |
1da177e4 LT |
1198 | p->type); |
1199 | goto out; | |
1200 | } | |
3d1427f8 | 1201 | EXPORT_SYMBOL(inet_register_protosw); |
1da177e4 LT |
1202 | |
1203 | void inet_unregister_protosw(struct inet_protosw *p) | |
1204 | { | |
1205 | if (INET_PROTOSW_PERMANENT & p->flags) { | |
058bd4d2 | 1206 | pr_err("Attempt to unregister permanent protocol %d\n", |
1da177e4 LT |
1207 | p->protocol); |
1208 | } else { | |
1209 | spin_lock_bh(&inetsw_lock); | |
1210 | list_del_rcu(&p->list); | |
1211 | spin_unlock_bh(&inetsw_lock); | |
1212 | ||
1213 | synchronize_net(); | |
1214 | } | |
1215 | } | |
3d1427f8 | 1216 | EXPORT_SYMBOL(inet_unregister_protosw); |
1da177e4 | 1217 | |
32519f11 ACM |
1218 | static int inet_sk_reselect_saddr(struct sock *sk) |
1219 | { | |
1220 | struct inet_sock *inet = inet_sk(sk); | |
c720c7e8 | 1221 | __be32 old_saddr = inet->inet_saddr; |
c720c7e8 | 1222 | __be32 daddr = inet->inet_daddr; |
6e869138 | 1223 | struct flowi4 *fl4; |
b23dd4fe DM |
1224 | struct rtable *rt; |
1225 | __be32 new_saddr; | |
f6d8bd05 | 1226 | struct ip_options_rcu *inet_opt; |
32519f11 | 1227 | |
f6d8bd05 | 1228 | inet_opt = rcu_dereference_protected(inet->inet_opt, |
1e1d04e6 | 1229 | lockdep_sock_is_held(sk)); |
f6d8bd05 ED |
1230 | if (inet_opt && inet_opt->opt.srr) |
1231 | daddr = inet_opt->opt.faddr; | |
32519f11 ACM |
1232 | |
1233 | /* Query new route. */ | |
6e869138 DM |
1234 | fl4 = &inet->cork.fl.u.ip4; |
1235 | rt = ip_route_connect(fl4, daddr, 0, RT_CONN_FLAGS(sk), | |
b23dd4fe | 1236 | sk->sk_bound_dev_if, sk->sk_protocol, |
0e0d44ab | 1237 | inet->inet_sport, inet->inet_dport, sk); |
b23dd4fe DM |
1238 | if (IS_ERR(rt)) |
1239 | return PTR_ERR(rt); | |
32519f11 | 1240 | |
d8d1f30b | 1241 | sk_setup_caps(sk, &rt->dst); |
32519f11 | 1242 | |
6e869138 | 1243 | new_saddr = fl4->saddr; |
32519f11 ACM |
1244 | |
1245 | if (new_saddr == old_saddr) | |
1246 | return 0; | |
1247 | ||
287b7f38 | 1248 | if (sock_net(sk)->ipv4.sysctl_ip_dynaddr > 1) { |
058bd4d2 JP |
1249 | pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n", |
1250 | __func__, &old_saddr, &new_saddr); | |
32519f11 ACM |
1251 | } |
1252 | ||
c720c7e8 | 1253 | inet->inet_saddr = inet->inet_rcv_saddr = new_saddr; |
32519f11 ACM |
1254 | |
1255 | /* | |
1256 | * XXX The only one ugly spot where we need to | |
1257 | * XXX really change the sockets identity after | |
1258 | * XXX it has entered the hashes. -DaveM | |
1259 | * | |
1260 | * Besides that, it does not check for connection | |
1261 | * uniqueness. Wait for troubles. | |
1262 | */ | |
086c653f | 1263 | return __sk_prot_rehash(sk); |
32519f11 ACM |
1264 | } |
1265 | ||
1266 | int inet_sk_rebuild_header(struct sock *sk) | |
1267 | { | |
1268 | struct inet_sock *inet = inet_sk(sk); | |
1269 | struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0); | |
3ca3c68e | 1270 | __be32 daddr; |
f6d8bd05 | 1271 | struct ip_options_rcu *inet_opt; |
6e869138 | 1272 | struct flowi4 *fl4; |
32519f11 ACM |
1273 | int err; |
1274 | ||
1275 | /* Route is OK, nothing to do. */ | |
1276 | if (rt) | |
1277 | return 0; | |
1278 | ||
1279 | /* Reroute. */ | |
f6d8bd05 ED |
1280 | rcu_read_lock(); |
1281 | inet_opt = rcu_dereference(inet->inet_opt); | |
c720c7e8 | 1282 | daddr = inet->inet_daddr; |
f6d8bd05 ED |
1283 | if (inet_opt && inet_opt->opt.srr) |
1284 | daddr = inet_opt->opt.faddr; | |
1285 | rcu_read_unlock(); | |
6e869138 DM |
1286 | fl4 = &inet->cork.fl.u.ip4; |
1287 | rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr, | |
78fbfd8a DM |
1288 | inet->inet_dport, inet->inet_sport, |
1289 | sk->sk_protocol, RT_CONN_FLAGS(sk), | |
1290 | sk->sk_bound_dev_if); | |
b23dd4fe DM |
1291 | if (!IS_ERR(rt)) { |
1292 | err = 0; | |
d8d1f30b | 1293 | sk_setup_caps(sk, &rt->dst); |
b23dd4fe DM |
1294 | } else { |
1295 | err = PTR_ERR(rt); | |
1296 | ||
32519f11 ACM |
1297 | /* Routing failed... */ |
1298 | sk->sk_route_caps = 0; | |
1299 | /* | |
1300 | * Other protocols have to map its equivalent state to TCP_SYN_SENT. | |
1301 | * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme | |
1302 | */ | |
287b7f38 | 1303 | if (!sock_net(sk)->ipv4.sysctl_ip_dynaddr || |
32519f11 ACM |
1304 | sk->sk_state != TCP_SYN_SENT || |
1305 | (sk->sk_userlocks & SOCK_BINDADDR_LOCK) || | |
1306 | (err = inet_sk_reselect_saddr(sk)) != 0) | |
1307 | sk->sk_err_soft = -err; | |
1308 | } | |
1309 | ||
1310 | return err; | |
1311 | } | |
32519f11 ACM |
1312 | EXPORT_SYMBOL(inet_sk_rebuild_header); |
1313 | ||
563e0bb0 YS |
1314 | void inet_sk_set_state(struct sock *sk, int state) |
1315 | { | |
1316 | trace_inet_sock_set_state(sk, sk->sk_state, state); | |
1317 | sk->sk_state = state; | |
1318 | } | |
1319 | EXPORT_SYMBOL(inet_sk_set_state); | |
1320 | ||
1321 | void inet_sk_state_store(struct sock *sk, int newstate) | |
1322 | { | |
1323 | trace_inet_sock_set_state(sk, sk->sk_state, newstate); | |
1324 | smp_store_release(&sk->sk_state, newstate); | |
1325 | } | |
1326 | ||
b8921ca8 TH |
1327 | struct sk_buff *inet_gso_segment(struct sk_buff *skb, |
1328 | netdev_features_t features) | |
f4c50d99 | 1329 | { |
0c19f846 | 1330 | bool udpfrag = false, fixedid = false, gso_partial, encap; |
f4c50d99 | 1331 | struct sk_buff *segs = ERR_PTR(-EINVAL); |
bca49f84 | 1332 | const struct net_offload *ops; |
0c19f846 | 1333 | unsigned int offset = 0; |
f9242b6b | 1334 | struct iphdr *iph; |
802ab55a | 1335 | int proto, tot_len; |
3347c960 | 1336 | int nhoff; |
f4c50d99 HX |
1337 | int ihl; |
1338 | int id; | |
1339 | ||
3347c960 ED |
1340 | skb_reset_network_header(skb); |
1341 | nhoff = skb_network_header(skb) - skb_mac_header(skb); | |
bbcf467d | 1342 | if (unlikely(!pskb_may_pull(skb, sizeof(*iph)))) |
f4c50d99 HX |
1343 | goto out; |
1344 | ||
eddc9ec5 | 1345 | iph = ip_hdr(skb); |
f4c50d99 HX |
1346 | ihl = iph->ihl * 4; |
1347 | if (ihl < sizeof(*iph)) | |
1348 | goto out; | |
1349 | ||
47d27aad ED |
1350 | id = ntohs(iph->id); |
1351 | proto = iph->protocol; | |
1352 | ||
1353 | /* Warning: after this point, iph might be no longer valid */ | |
bbcf467d | 1354 | if (unlikely(!pskb_may_pull(skb, ihl))) |
f4c50d99 | 1355 | goto out; |
47d27aad | 1356 | __skb_pull(skb, ihl); |
f4c50d99 | 1357 | |
8c3a897b ED |
1358 | encap = SKB_GSO_CB(skb)->encap_level > 0; |
1359 | if (encap) | |
1e16aa3d | 1360 | features &= skb->dev->hw_enc_features; |
3347c960 | 1361 | SKB_GSO_CB(skb)->encap_level += ihl; |
73136267 | 1362 | |
badff6d0 | 1363 | skb_reset_transport_header(skb); |
47d27aad | 1364 | |
f4c50d99 HX |
1365 | segs = ERR_PTR(-EPROTONOSUPPORT); |
1366 | ||
cbc53e08 | 1367 | if (!skb->encapsulation || encap) { |
0c19f846 | 1368 | udpfrag = !!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP); |
cbc53e08 AD |
1369 | fixedid = !!(skb_shinfo(skb)->gso_type & SKB_GSO_TCP_FIXEDID); |
1370 | ||
1371 | /* fixed ID is invalid if DF bit is not set */ | |
a5108878 | 1372 | if (fixedid && !(ip_hdr(skb)->frag_off & htons(IP_DF))) |
cbc53e08 AD |
1373 | goto out; |
1374 | } | |
dcd60771 | 1375 | |
bca49f84 | 1376 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 VY |
1377 | if (likely(ops && ops->callbacks.gso_segment)) |
1378 | segs = ops->callbacks.gso_segment(skb, features); | |
f4c50d99 | 1379 | |
50c3a487 | 1380 | if (IS_ERR_OR_NULL(segs)) |
f4c50d99 HX |
1381 | goto out; |
1382 | ||
07b26c94 SK |
1383 | gso_partial = !!(skb_shinfo(segs)->gso_type & SKB_GSO_PARTIAL); |
1384 | ||
f4c50d99 HX |
1385 | skb = segs; |
1386 | do { | |
3347c960 | 1387 | iph = (struct iphdr *)(skb_mac_header(skb) + nhoff); |
0c19f846 WB |
1388 | if (udpfrag) { |
1389 | iph->frag_off = htons(offset >> 3); | |
1390 | if (skb->next) | |
1391 | iph->frag_off |= htons(IP_MF); | |
1392 | offset += skb->len - nhoff - ihl; | |
1393 | tot_len = skb->len - nhoff; | |
1394 | } else if (skb_is_gso(skb)) { | |
802ab55a AD |
1395 | if (!fixedid) { |
1396 | iph->id = htons(id); | |
1397 | id += skb_shinfo(skb)->gso_segs; | |
1398 | } | |
07b26c94 SK |
1399 | |
1400 | if (gso_partial) | |
1401 | tot_len = skb_shinfo(skb)->gso_size + | |
1402 | SKB_GSO_CB(skb)->data_offset + | |
1403 | skb->head - (unsigned char *)iph; | |
1404 | else | |
1405 | tot_len = skb->len - nhoff; | |
802ab55a AD |
1406 | } else { |
1407 | if (!fixedid) | |
1408 | iph->id = htons(id++); | |
1409 | tot_len = skb->len - nhoff; | |
490ab081 | 1410 | } |
802ab55a | 1411 | iph->tot_len = htons(tot_len); |
47d27aad | 1412 | ip_send_check(iph); |
8c3a897b | 1413 | if (encap) |
3347c960 | 1414 | skb_reset_inner_headers(skb); |
3347c960 | 1415 | skb->network_header = (u8 *)iph - skb->head; |
c56cae23 | 1416 | skb_reset_mac_len(skb); |
f4c50d99 HX |
1417 | } while ((skb = skb->next)); |
1418 | ||
1419 | out: | |
1420 | return segs; | |
1421 | } | |
b8921ca8 | 1422 | EXPORT_SYMBOL(inet_gso_segment); |
f4c50d99 | 1423 | |
418e897e WB |
1424 | static struct sk_buff *ipip_gso_segment(struct sk_buff *skb, |
1425 | netdev_features_t features) | |
1426 | { | |
1427 | if (!(skb_shinfo(skb)->gso_type & SKB_GSO_IPXIP4)) | |
1428 | return ERR_PTR(-EINVAL); | |
1429 | ||
1430 | return inet_gso_segment(skb, features); | |
1431 | } | |
1432 | ||
d4546c25 | 1433 | struct sk_buff *inet_gro_receive(struct list_head *head, struct sk_buff *skb) |
73cc19f1 | 1434 | { |
bca49f84 | 1435 | const struct net_offload *ops; |
d4546c25 | 1436 | struct sk_buff *pp = NULL; |
b71d1d42 | 1437 | const struct iphdr *iph; |
d4546c25 | 1438 | struct sk_buff *p; |
a5b1cf28 HX |
1439 | unsigned int hlen; |
1440 | unsigned int off; | |
1075f3f6 | 1441 | unsigned int id; |
73cc19f1 HX |
1442 | int flush = 1; |
1443 | int proto; | |
73cc19f1 | 1444 | |
a5b1cf28 HX |
1445 | off = skb_gro_offset(skb); |
1446 | hlen = off + sizeof(*iph); | |
1447 | iph = skb_gro_header_fast(skb, off); | |
1448 | if (skb_gro_header_hard(skb, hlen)) { | |
1449 | iph = skb_gro_header_slow(skb, hlen, off); | |
1450 | if (unlikely(!iph)) | |
1451 | goto out; | |
1452 | } | |
73cc19f1 | 1453 | |
f9242b6b | 1454 | proto = iph->protocol; |
73cc19f1 HX |
1455 | |
1456 | rcu_read_lock(); | |
bca49f84 | 1457 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 | 1458 | if (!ops || !ops->callbacks.gro_receive) |
73cc19f1 HX |
1459 | goto out_unlock; |
1460 | ||
a5ad24be | 1461 | if (*(u8 *)iph != 0x45) |
73cc19f1 HX |
1462 | goto out_unlock; |
1463 | ||
9b83e031 SK |
1464 | if (ip_is_fragment(iph)) |
1465 | goto out_unlock; | |
1466 | ||
a9e050f4 | 1467 | if (unlikely(ip_fast_csum((u8 *)iph, 5))) |
73cc19f1 HX |
1468 | goto out_unlock; |
1469 | ||
0eae88f3 | 1470 | id = ntohl(*(__be32 *)&iph->id); |
db8caf3d | 1471 | flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF)); |
1075f3f6 | 1472 | id >>= 16; |
73cc19f1 | 1473 | |
d4546c25 | 1474 | list_for_each_entry(p, head, list) { |
73cc19f1 | 1475 | struct iphdr *iph2; |
1530545e | 1476 | u16 flush_id; |
73cc19f1 HX |
1477 | |
1478 | if (!NAPI_GRO_CB(p)->same_flow) | |
1479 | continue; | |
1480 | ||
299603e8 JC |
1481 | iph2 = (struct iphdr *)(p->data + off); |
1482 | /* The above works because, with the exception of the top | |
1483 | * (inner most) layer, we only aggregate pkts with the same | |
1484 | * hdr length so all the hdrs we'll need to verify will start | |
1485 | * at the same offset. | |
1486 | */ | |
a5ad24be | 1487 | if ((iph->protocol ^ iph2->protocol) | |
0eae88f3 ED |
1488 | ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) | |
1489 | ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) { | |
73cc19f1 HX |
1490 | NAPI_GRO_CB(p)->same_flow = 0; |
1491 | continue; | |
1492 | } | |
1493 | ||
1494 | /* All fields must match except length and checksum. */ | |
1495 | NAPI_GRO_CB(p)->flush |= | |
a5ad24be | 1496 | (iph->ttl ^ iph2->ttl) | |
a9e050f4 | 1497 | (iph->tos ^ iph2->tos) | |
bf5a755f | 1498 | ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF)); |
73cc19f1 HX |
1499 | |
1500 | NAPI_GRO_CB(p)->flush |= flush; | |
1530545e AD |
1501 | |
1502 | /* We need to store of the IP ID check to be included later | |
1503 | * when we can verify that this packet does in fact belong | |
1504 | * to a given flow. | |
1505 | */ | |
1506 | flush_id = (u16)(id - ntohs(iph2->id)); | |
1507 | ||
1508 | /* This bit of code makes it much easier for us to identify | |
1509 | * the cases where we are doing atomic vs non-atomic IP ID | |
1510 | * checks. Specifically an atomic check can return IP ID | |
1511 | * values 0 - 0xFFFF, while a non-atomic check can only | |
1512 | * return 0 or 0xFFFF. | |
1513 | */ | |
1514 | if (!NAPI_GRO_CB(p)->is_atomic || | |
1515 | !(iph->frag_off & htons(IP_DF))) { | |
1516 | flush_id ^= NAPI_GRO_CB(p)->count; | |
1517 | flush_id = flush_id ? 0xFFFF : 0; | |
1518 | } | |
1519 | ||
1520 | /* If the previous IP ID value was based on an atomic | |
1521 | * datagram we can overwrite the value and ignore it. | |
1522 | */ | |
1523 | if (NAPI_GRO_CB(skb)->is_atomic) | |
1524 | NAPI_GRO_CB(p)->flush_id = flush_id; | |
1525 | else | |
1526 | NAPI_GRO_CB(p)->flush_id |= flush_id; | |
73cc19f1 HX |
1527 | } |
1528 | ||
1530545e | 1529 | NAPI_GRO_CB(skb)->is_atomic = !!(iph->frag_off & htons(IP_DF)); |
73cc19f1 | 1530 | NAPI_GRO_CB(skb)->flush |= flush; |
299603e8 JC |
1531 | skb_set_network_header(skb, off); |
1532 | /* The above will be needed by the transport layer if there is one | |
1533 | * immediately following this IP hdr. | |
1534 | */ | |
1535 | ||
2c804d0f ED |
1536 | /* Note : No need to call skb_gro_postpull_rcsum() here, |
1537 | * as we already checked checksum over ipv4 header was 0 | |
1538 | */ | |
86911732 HX |
1539 | skb_gro_pull(skb, sizeof(*iph)); |
1540 | skb_set_transport_header(skb, skb_gro_offset(skb)); | |
73cc19f1 | 1541 | |
028e0a47 PA |
1542 | pp = indirect_call_gro_receive(tcp4_gro_receive, udp4_gro_receive, |
1543 | ops->callbacks.gro_receive, head, skb); | |
73cc19f1 HX |
1544 | |
1545 | out_unlock: | |
1546 | rcu_read_unlock(); | |
1547 | ||
1548 | out: | |
5f114163 | 1549 | skb_gro_flush_final(skb, pp, flush); |
73cc19f1 HX |
1550 | |
1551 | return pp; | |
1552 | } | |
b8921ca8 | 1553 | EXPORT_SYMBOL(inet_gro_receive); |
73cc19f1 | 1554 | |
d4546c25 DM |
1555 | static struct sk_buff *ipip_gro_receive(struct list_head *head, |
1556 | struct sk_buff *skb) | |
fac8e0f5 JG |
1557 | { |
1558 | if (NAPI_GRO_CB(skb)->encap_mark) { | |
1559 | NAPI_GRO_CB(skb)->flush = 1; | |
1560 | return NULL; | |
1561 | } | |
1562 | ||
1563 | NAPI_GRO_CB(skb)->encap_mark = 1; | |
1564 | ||
1565 | return inet_gro_receive(head, skb); | |
1566 | } | |
1567 | ||
822c8685 DD |
1568 | #define SECONDS_PER_DAY 86400 |
1569 | ||
1570 | /* inet_current_timestamp - Return IP network timestamp | |
1571 | * | |
1572 | * Return milliseconds since midnight in network byte order. | |
1573 | */ | |
1574 | __be32 inet_current_timestamp(void) | |
1575 | { | |
1576 | u32 secs; | |
1577 | u32 msecs; | |
1578 | struct timespec64 ts; | |
1579 | ||
1580 | ktime_get_real_ts64(&ts); | |
1581 | ||
1582 | /* Get secs since midnight. */ | |
1583 | (void)div_u64_rem(ts.tv_sec, SECONDS_PER_DAY, &secs); | |
1584 | /* Convert to msecs. */ | |
1585 | msecs = secs * MSEC_PER_SEC; | |
1586 | /* Convert nsec to msec. */ | |
1587 | msecs += (u32)ts.tv_nsec / NSEC_PER_MSEC; | |
1588 | ||
1589 | /* Convert to network byte order. */ | |
3ba9d300 | 1590 | return htonl(msecs); |
822c8685 DD |
1591 | } |
1592 | EXPORT_SYMBOL(inet_current_timestamp); | |
1593 | ||
f4713a3d WB |
1594 | int inet_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len) |
1595 | { | |
1596 | if (sk->sk_family == AF_INET) | |
1597 | return ip_recv_error(sk, msg, len, addr_len); | |
1598 | #if IS_ENABLED(CONFIG_IPV6) | |
1599 | if (sk->sk_family == AF_INET6) | |
1600 | return pingv6_ops.ipv6_recv_error(sk, msg, len, addr_len); | |
1601 | #endif | |
1602 | return -EINVAL; | |
1603 | } | |
1604 | ||
b8921ca8 | 1605 | int inet_gro_complete(struct sk_buff *skb, int nhoff) |
73cc19f1 | 1606 | { |
299603e8 JC |
1607 | __be16 newlen = htons(skb->len - nhoff); |
1608 | struct iphdr *iph = (struct iphdr *)(skb->data + nhoff); | |
bca49f84 | 1609 | const struct net_offload *ops; |
f9242b6b | 1610 | int proto = iph->protocol; |
73cc19f1 | 1611 | int err = -ENOSYS; |
73cc19f1 | 1612 | |
294acf1c PA |
1613 | if (skb->encapsulation) { |
1614 | skb_set_inner_protocol(skb, cpu_to_be16(ETH_P_IP)); | |
c3caf119 | 1615 | skb_set_inner_network_header(skb, nhoff); |
294acf1c | 1616 | } |
c3caf119 | 1617 | |
73cc19f1 HX |
1618 | csum_replace2(&iph->check, iph->tot_len, newlen); |
1619 | iph->tot_len = newlen; | |
1620 | ||
1621 | rcu_read_lock(); | |
bca49f84 | 1622 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 | 1623 | if (WARN_ON(!ops || !ops->callbacks.gro_complete)) |
73cc19f1 HX |
1624 | goto out_unlock; |
1625 | ||
299603e8 JC |
1626 | /* Only need to add sizeof(*iph) to get to the next hdr below |
1627 | * because any hdr with option will have been flushed in | |
1628 | * inet_gro_receive(). | |
1629 | */ | |
028e0a47 PA |
1630 | err = INDIRECT_CALL_2(ops->callbacks.gro_complete, |
1631 | tcp4_gro_complete, udp4_gro_complete, | |
1632 | skb, nhoff + sizeof(*iph)); | |
73cc19f1 HX |
1633 | |
1634 | out_unlock: | |
1635 | rcu_read_unlock(); | |
1636 | ||
1637 | return err; | |
1638 | } | |
b8921ca8 | 1639 | EXPORT_SYMBOL(inet_gro_complete); |
73cc19f1 | 1640 | |
b8cba75b JG |
1641 | static int ipip_gro_complete(struct sk_buff *skb, int nhoff) |
1642 | { | |
1643 | skb->encapsulation = 1; | |
7e13318d | 1644 | skb_shinfo(skb)->gso_type |= SKB_GSO_IPXIP4; |
b8cba75b JG |
1645 | return inet_gro_complete(skb, nhoff); |
1646 | } | |
1647 | ||
eee4fe4d | 1648 | int inet_ctl_sock_create(struct sock **sk, unsigned short family, |
5677242f DL |
1649 | unsigned short type, unsigned char protocol, |
1650 | struct net *net) | |
3d58b5fa | 1651 | { |
eee4fe4d | 1652 | struct socket *sock; |
26abe143 | 1653 | int rc = sock_create_kern(net, family, type, protocol, &sock); |
3d58b5fa DL |
1654 | |
1655 | if (rc == 0) { | |
eee4fe4d DL |
1656 | *sk = sock->sk; |
1657 | (*sk)->sk_allocation = GFP_ATOMIC; | |
3d58b5fa DL |
1658 | /* |
1659 | * Unhash it so that IP input processing does not even see it, | |
1660 | * we do not wish this socket to see incoming packets. | |
1661 | */ | |
eee4fe4d | 1662 | (*sk)->sk_prot->unhash(*sk); |
3d58b5fa DL |
1663 | } |
1664 | return rc; | |
1665 | } | |
3d58b5fa DL |
1666 | EXPORT_SYMBOL_GPL(inet_ctl_sock_create); |
1667 | ||
c4c6bc31 R |
1668 | u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offt) |
1669 | { | |
1670 | return *(((unsigned long *)per_cpu_ptr(mib, cpu)) + offt); | |
1671 | } | |
1672 | EXPORT_SYMBOL_GPL(snmp_get_cpu_field); | |
1673 | ||
698365fa | 1674 | unsigned long snmp_fold_field(void __percpu *mib, int offt) |
5e0f0435 HX |
1675 | { |
1676 | unsigned long res = 0; | |
698365fa | 1677 | int i; |
5e0f0435 | 1678 | |
698365fa | 1679 | for_each_possible_cpu(i) |
c4c6bc31 | 1680 | res += snmp_get_cpu_field(mib, i, offt); |
5e0f0435 HX |
1681 | return res; |
1682 | } | |
1683 | EXPORT_SYMBOL_GPL(snmp_fold_field); | |
1684 | ||
4ce3c183 ED |
1685 | #if BITS_PER_LONG==32 |
1686 | ||
80ec1927 | 1687 | u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offt, |
c4c6bc31 R |
1688 | size_t syncp_offset) |
1689 | { | |
1690 | void *bhptr; | |
1691 | struct u64_stats_sync *syncp; | |
1692 | u64 v; | |
1693 | unsigned int start; | |
1694 | ||
1695 | bhptr = per_cpu_ptr(mib, cpu); | |
1696 | syncp = (struct u64_stats_sync *)(bhptr + syncp_offset); | |
1697 | do { | |
1698 | start = u64_stats_fetch_begin_irq(syncp); | |
1699 | v = *(((u64 *)bhptr) + offt); | |
1700 | } while (u64_stats_fetch_retry_irq(syncp, start)); | |
1701 | ||
1702 | return v; | |
1703 | } | |
1704 | EXPORT_SYMBOL_GPL(snmp_get_cpu_field64); | |
1705 | ||
698365fa | 1706 | u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset) |
4ce3c183 ED |
1707 | { |
1708 | u64 res = 0; | |
1709 | int cpu; | |
1710 | ||
1711 | for_each_possible_cpu(cpu) { | |
d1bfc625 | 1712 | res += snmp_get_cpu_field64(mib, cpu, offt, syncp_offset); |
4ce3c183 ED |
1713 | } |
1714 | return res; | |
1715 | } | |
1716 | EXPORT_SYMBOL_GPL(snmp_fold_field64); | |
1717 | #endif | |
1718 | ||
1da177e4 | 1719 | #ifdef CONFIG_IP_MULTICAST |
32613090 | 1720 | static const struct net_protocol igmp_protocol = { |
1da177e4 | 1721 | .handler = igmp_rcv, |
b4ee07df | 1722 | .netns_ok = 1, |
1da177e4 LT |
1723 | }; |
1724 | #endif | |
1725 | ||
a8e3bb34 DA |
1726 | /* thinking of making this const? Don't. |
1727 | * early_demux can change based on sysctl. | |
1728 | */ | |
03157937 | 1729 | static struct net_protocol tcp_protocol = { |
41063e9d | 1730 | .early_demux = tcp_v4_early_demux, |
dddb64bc | 1731 | .early_demux_handler = tcp_v4_early_demux, |
41063e9d DM |
1732 | .handler = tcp_v4_rcv, |
1733 | .err_handler = tcp_v4_err, | |
41063e9d DM |
1734 | .no_policy = 1, |
1735 | .netns_ok = 1, | |
8ed1dc44 | 1736 | .icmp_strict_tag_validation = 1, |
1da177e4 LT |
1737 | }; |
1738 | ||
a8e3bb34 DA |
1739 | /* thinking of making this const? Don't. |
1740 | * early_demux can change based on sysctl. | |
1741 | */ | |
03157937 | 1742 | static struct net_protocol udp_protocol = { |
421b3885 | 1743 | .early_demux = udp_v4_early_demux, |
dddb64bc | 1744 | .early_demux_handler = udp_v4_early_demux, |
1da177e4 LT |
1745 | .handler = udp_rcv, |
1746 | .err_handler = udp_err, | |
1747 | .no_policy = 1, | |
92f1fecb | 1748 | .netns_ok = 1, |
1da177e4 LT |
1749 | }; |
1750 | ||
32613090 | 1751 | static const struct net_protocol icmp_protocol = { |
1da177e4 | 1752 | .handler = icmp_rcv, |
5b052042 | 1753 | .err_handler = icmp_err, |
8b7817f3 | 1754 | .no_policy = 1, |
92f1fecb | 1755 | .netns_ok = 1, |
1da177e4 LT |
1756 | }; |
1757 | ||
9b4661bd PE |
1758 | static __net_init int ipv4_mib_init_net(struct net *net) |
1759 | { | |
827da44c JS |
1760 | int i; |
1761 | ||
698365fa WC |
1762 | net->mib.tcp_statistics = alloc_percpu(struct tcp_mib); |
1763 | if (!net->mib.tcp_statistics) | |
57ef42d5 | 1764 | goto err_tcp_mib; |
698365fa WC |
1765 | net->mib.ip_statistics = alloc_percpu(struct ipstats_mib); |
1766 | if (!net->mib.ip_statistics) | |
a20f5799 | 1767 | goto err_ip_mib; |
827da44c JS |
1768 | |
1769 | for_each_possible_cpu(i) { | |
1770 | struct ipstats_mib *af_inet_stats; | |
698365fa | 1771 | af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i); |
827da44c | 1772 | u64_stats_init(&af_inet_stats->syncp); |
827da44c JS |
1773 | } |
1774 | ||
698365fa WC |
1775 | net->mib.net_statistics = alloc_percpu(struct linux_mib); |
1776 | if (!net->mib.net_statistics) | |
61a7e260 | 1777 | goto err_net_mib; |
698365fa WC |
1778 | net->mib.udp_statistics = alloc_percpu(struct udp_mib); |
1779 | if (!net->mib.udp_statistics) | |
2f275f91 | 1780 | goto err_udp_mib; |
698365fa WC |
1781 | net->mib.udplite_statistics = alloc_percpu(struct udp_mib); |
1782 | if (!net->mib.udplite_statistics) | |
386019d3 | 1783 | goto err_udplite_mib; |
698365fa WC |
1784 | net->mib.icmp_statistics = alloc_percpu(struct icmp_mib); |
1785 | if (!net->mib.icmp_statistics) | |
b60538a0 | 1786 | goto err_icmp_mib; |
acb32ba3 ED |
1787 | net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib), |
1788 | GFP_KERNEL); | |
1789 | if (!net->mib.icmpmsg_statistics) | |
923c6586 | 1790 | goto err_icmpmsg_mib; |
57ef42d5 PE |
1791 | |
1792 | tcp_mib_init(net); | |
9b4661bd | 1793 | return 0; |
57ef42d5 | 1794 | |
923c6586 | 1795 | err_icmpmsg_mib: |
698365fa | 1796 | free_percpu(net->mib.icmp_statistics); |
b60538a0 | 1797 | err_icmp_mib: |
698365fa | 1798 | free_percpu(net->mib.udplite_statistics); |
386019d3 | 1799 | err_udplite_mib: |
698365fa | 1800 | free_percpu(net->mib.udp_statistics); |
2f275f91 | 1801 | err_udp_mib: |
698365fa | 1802 | free_percpu(net->mib.net_statistics); |
61a7e260 | 1803 | err_net_mib: |
698365fa | 1804 | free_percpu(net->mib.ip_statistics); |
a20f5799 | 1805 | err_ip_mib: |
698365fa | 1806 | free_percpu(net->mib.tcp_statistics); |
57ef42d5 PE |
1807 | err_tcp_mib: |
1808 | return -ENOMEM; | |
9b4661bd PE |
1809 | } |
1810 | ||
1811 | static __net_exit void ipv4_mib_exit_net(struct net *net) | |
1812 | { | |
acb32ba3 | 1813 | kfree(net->mib.icmpmsg_statistics); |
698365fa WC |
1814 | free_percpu(net->mib.icmp_statistics); |
1815 | free_percpu(net->mib.udplite_statistics); | |
1816 | free_percpu(net->mib.udp_statistics); | |
1817 | free_percpu(net->mib.net_statistics); | |
1818 | free_percpu(net->mib.ip_statistics); | |
1819 | free_percpu(net->mib.tcp_statistics); | |
fc518953 FW |
1820 | #ifdef CONFIG_MPTCP |
1821 | /* allocated on demand, see mptcp_init_sock() */ | |
1822 | free_percpu(net->mib.mptcp_statistics); | |
1823 | #endif | |
9b4661bd PE |
1824 | } |
1825 | ||
1826 | static __net_initdata struct pernet_operations ipv4_mib_ops = { | |
1827 | .init = ipv4_mib_init_net, | |
1828 | .exit = ipv4_mib_exit_net, | |
1829 | }; | |
1830 | ||
1da177e4 LT |
1831 | static int __init init_ipv4_mibs(void) |
1832 | { | |
d89cbbb1 | 1833 | return register_pernet_subsys(&ipv4_mib_ops); |
1da177e4 LT |
1834 | } |
1835 | ||
c9d8f1a6 CW |
1836 | static __net_init int inet_init_net(struct net *net) |
1837 | { | |
1838 | /* | |
1839 | * Set defaults for local port range | |
1840 | */ | |
1841 | seqlock_init(&net->ipv4.ip_local_ports.lock); | |
1842 | net->ipv4.ip_local_ports.range[0] = 32768; | |
07f4c900 | 1843 | net->ipv4.ip_local_ports.range[1] = 60999; |
ba6b918a CW |
1844 | |
1845 | seqlock_init(&net->ipv4.ping_group_range.lock); | |
1846 | /* | |
1847 | * Sane defaults - nobody may create ping sockets. | |
1848 | * Boot scripts should set this to distro-specific group. | |
1849 | */ | |
1850 | net->ipv4.ping_group_range.range[0] = make_kgid(&init_user_ns, 1); | |
1851 | net->ipv4.ping_group_range.range[1] = make_kgid(&init_user_ns, 0); | |
049bbf58 EG |
1852 | |
1853 | /* Default values for sysctl-controlled parameters. | |
1854 | * We set them here, in case sysctl is not compiled. | |
1855 | */ | |
1856 | net->ipv4.sysctl_ip_default_ttl = IPDEFTTL; | |
432e05d3 | 1857 | net->ipv4.sysctl_ip_fwd_update_priority = 1; |
049bbf58 EG |
1858 | net->ipv4.sysctl_ip_dynaddr = 0; |
1859 | net->ipv4.sysctl_ip_early_demux = 1; | |
dddb64bc SAK |
1860 | net->ipv4.sysctl_udp_early_demux = 1; |
1861 | net->ipv4.sysctl_tcp_early_demux = 1; | |
4f80116d | 1862 | net->ipv4.sysctl_nexthop_compat_mode = 1; |
4548b683 KJ |
1863 | #ifdef CONFIG_SYSCTL |
1864 | net->ipv4.sysctl_ip_prot_sock = PROT_SOCK; | |
1865 | #endif | |
049bbf58 | 1866 | |
1714020e NB |
1867 | /* Some igmp sysctl, whose values are always used */ |
1868 | net->ipv4.sysctl_igmp_max_memberships = 20; | |
1869 | net->ipv4.sysctl_igmp_max_msf = 10; | |
1870 | /* IGMP reports for link-local multicast groups are enabled by default */ | |
1871 | net->ipv4.sysctl_igmp_llm_reports = 1; | |
1872 | net->ipv4.sysctl_igmp_qrv = 2; | |
1873 | ||
c9d8f1a6 CW |
1874 | return 0; |
1875 | } | |
1876 | ||
c9d8f1a6 CW |
1877 | static __net_initdata struct pernet_operations af_inet_ops = { |
1878 | .init = inet_init_net, | |
c9d8f1a6 CW |
1879 | }; |
1880 | ||
1881 | static int __init init_inet_pernet_ops(void) | |
1882 | { | |
1883 | return register_pernet_subsys(&af_inet_ops); | |
1884 | } | |
1885 | ||
1da177e4 | 1886 | static int ipv4_proc_init(void); |
1da177e4 | 1887 | |
30e224d7 HX |
1888 | /* |
1889 | * IP protocol layer initialiser | |
1890 | */ | |
1891 | ||
22061d80 VY |
1892 | static struct packet_offload ip_packet_offload __read_mostly = { |
1893 | .type = cpu_to_be16(ETH_P_IP), | |
f191a1d1 | 1894 | .callbacks = { |
f191a1d1 VY |
1895 | .gso_segment = inet_gso_segment, |
1896 | .gro_receive = inet_gro_receive, | |
1897 | .gro_complete = inet_gro_complete, | |
1898 | }, | |
30e224d7 HX |
1899 | }; |
1900 | ||
cb32f511 ED |
1901 | static const struct net_offload ipip_offload = { |
1902 | .callbacks = { | |
418e897e | 1903 | .gso_segment = ipip_gso_segment, |
fac8e0f5 | 1904 | .gro_receive = ipip_gro_receive, |
b8cba75b | 1905 | .gro_complete = ipip_gro_complete, |
cb32f511 ED |
1906 | }, |
1907 | }; | |
1908 | ||
93b1b31f TZ |
1909 | static int __init ipip_offload_init(void) |
1910 | { | |
1911 | return inet_add_offload(&ipip_offload, IPPROTO_IPIP); | |
1912 | } | |
1913 | ||
808a8f88 VY |
1914 | static int __init ipv4_offload_init(void) |
1915 | { | |
1916 | /* | |
1917 | * Add offloads | |
1918 | */ | |
da5bab07 | 1919 | if (udpv4_offload_init() < 0) |
808a8f88 | 1920 | pr_crit("%s: Cannot add UDP protocol offload\n", __func__); |
28850dc7 DB |
1921 | if (tcpv4_offload_init() < 0) |
1922 | pr_crit("%s: Cannot add TCP protocol offload\n", __func__); | |
93b1b31f TZ |
1923 | if (ipip_offload_init() < 0) |
1924 | pr_crit("%s: Cannot add IPIP protocol offload\n", __func__); | |
808a8f88 VY |
1925 | |
1926 | dev_add_offload(&ip_packet_offload); | |
1927 | return 0; | |
1928 | } | |
1929 | ||
1930 | fs_initcall(ipv4_offload_init); | |
1931 | ||
1932 | static struct packet_type ip_packet_type __read_mostly = { | |
1933 | .type = cpu_to_be16(ETH_P_IP), | |
1934 | .func = ip_rcv, | |
17266ee9 | 1935 | .list_func = ip_list_rcv, |
808a8f88 VY |
1936 | }; |
1937 | ||
1da177e4 LT |
1938 | static int __init inet_init(void) |
1939 | { | |
1da177e4 LT |
1940 | struct inet_protosw *q; |
1941 | struct list_head *r; | |
b6246f4d | 1942 | int rc; |
1da177e4 | 1943 | |
b4772ef8 | 1944 | sock_skb_cb_check_size(sizeof(struct inet_skb_parm)); |
1da177e4 LT |
1945 | |
1946 | rc = proto_register(&tcp_prot, 1); | |
1947 | if (rc) | |
122ff243 | 1948 | goto out; |
1da177e4 LT |
1949 | |
1950 | rc = proto_register(&udp_prot, 1); | |
1951 | if (rc) | |
1952 | goto out_unregister_tcp_proto; | |
1953 | ||
1954 | rc = proto_register(&raw_prot, 1); | |
1955 | if (rc) | |
1956 | goto out_unregister_udp_proto; | |
1957 | ||
c319b4d7 VK |
1958 | rc = proto_register(&ping_prot, 1); |
1959 | if (rc) | |
1960 | goto out_unregister_raw_proto; | |
1961 | ||
1da177e4 | 1962 | /* |
e905a9ed | 1963 | * Tell SOCKET that we are alive... |
1da177e4 LT |
1964 | */ |
1965 | ||
e905a9ed | 1966 | (void)sock_register(&inet_family_ops); |
1da177e4 | 1967 | |
bd7b1533 AV |
1968 | #ifdef CONFIG_SYSCTL |
1969 | ip_static_sysctl_init(); | |
1970 | #endif | |
1971 | ||
1da177e4 LT |
1972 | /* |
1973 | * Add all the base protocols. | |
1974 | */ | |
1975 | ||
1976 | if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0) | |
058bd4d2 | 1977 | pr_crit("%s: Cannot add ICMP protocol\n", __func__); |
1da177e4 | 1978 | if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0) |
058bd4d2 | 1979 | pr_crit("%s: Cannot add UDP protocol\n", __func__); |
1da177e4 | 1980 | if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0) |
058bd4d2 | 1981 | pr_crit("%s: Cannot add TCP protocol\n", __func__); |
1da177e4 LT |
1982 | #ifdef CONFIG_IP_MULTICAST |
1983 | if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0) | |
058bd4d2 | 1984 | pr_crit("%s: Cannot add IGMP protocol\n", __func__); |
1da177e4 LT |
1985 | #endif |
1986 | ||
1987 | /* Register the socket-side information for inet_create. */ | |
1988 | for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r) | |
1989 | INIT_LIST_HEAD(r); | |
1990 | ||
1991 | for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q) | |
1992 | inet_register_protosw(q); | |
1993 | ||
1994 | /* | |
1995 | * Set the ARP module up | |
1996 | */ | |
1997 | ||
1998 | arp_init(); | |
1999 | ||
e905a9ed YH |
2000 | /* |
2001 | * Set the IP module up | |
2002 | */ | |
1da177e4 LT |
2003 | |
2004 | ip_init(); | |
2005 | ||
1da177e4 LT |
2006 | /* Setup TCP slab cache for open requests. */ |
2007 | tcp_init(); | |
2008 | ||
95766fff HA |
2009 | /* Setup UDP memory threshold */ |
2010 | udp_init(); | |
2011 | ||
ba4e58ec GR |
2012 | /* Add UDP-Lite (RFC 3828) */ |
2013 | udplite4_register(); | |
1da177e4 | 2014 | |
6897445f MM |
2015 | raw_init(); |
2016 | ||
c319b4d7 VK |
2017 | ping_init(); |
2018 | ||
1da177e4 LT |
2019 | /* |
2020 | * Set the ICMP layer up | |
2021 | */ | |
2022 | ||
a5710d65 DL |
2023 | if (icmp_init() < 0) |
2024 | panic("Failed to create the ICMP control socket.\n"); | |
1da177e4 LT |
2025 | |
2026 | /* | |
2027 | * Initialise the multicast router | |
2028 | */ | |
2029 | #if defined(CONFIG_IP_MROUTE) | |
03d2f897 | 2030 | if (ip_mr_init()) |
058bd4d2 | 2031 | pr_crit("%s: Cannot init ipv4 mroute\n", __func__); |
1da177e4 | 2032 | #endif |
c9d8f1a6 CW |
2033 | |
2034 | if (init_inet_pernet_ops()) | |
2035 | pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__); | |
1da177e4 LT |
2036 | /* |
2037 | * Initialise per-cpu ipv4 mibs | |
e905a9ed | 2038 | */ |
1da177e4 | 2039 | |
132adf54 | 2040 | if (init_ipv4_mibs()) |
058bd4d2 | 2041 | pr_crit("%s: Cannot init ipv4 mibs\n", __func__); |
e905a9ed | 2042 | |
1da177e4 LT |
2043 | ipv4_proc_init(); |
2044 | ||
2045 | ipfrag_init(); | |
2046 | ||
30e224d7 HX |
2047 | dev_add_pack(&ip_packet_type); |
2048 | ||
045a0fa0 TG |
2049 | ip_tunnel_core_init(); |
2050 | ||
1da177e4 LT |
2051 | rc = 0; |
2052 | out: | |
2053 | return rc; | |
c319b4d7 VK |
2054 | out_unregister_raw_proto: |
2055 | proto_unregister(&raw_prot); | |
1da177e4 LT |
2056 | out_unregister_udp_proto: |
2057 | proto_unregister(&udp_prot); | |
321efff7 OK |
2058 | out_unregister_tcp_proto: |
2059 | proto_unregister(&tcp_prot); | |
1da177e4 LT |
2060 | goto out; |
2061 | } | |
2062 | ||
a536e077 | 2063 | fs_initcall(inet_init); |
1da177e4 LT |
2064 | |
2065 | /* ------------------------------------------------------------------------ */ | |
2066 | ||
2067 | #ifdef CONFIG_PROC_FS | |
1da177e4 LT |
2068 | static int __init ipv4_proc_init(void) |
2069 | { | |
2070 | int rc = 0; | |
2071 | ||
2072 | if (raw_proc_init()) | |
2073 | goto out_raw; | |
2074 | if (tcp4_proc_init()) | |
2075 | goto out_tcp; | |
2076 | if (udp4_proc_init()) | |
2077 | goto out_udp; | |
c319b4d7 VK |
2078 | if (ping_proc_init()) |
2079 | goto out_ping; | |
1da177e4 LT |
2080 | if (ip_misc_proc_init()) |
2081 | goto out_misc; | |
2082 | out: | |
2083 | return rc; | |
2084 | out_misc: | |
c319b4d7 VK |
2085 | ping_proc_exit(); |
2086 | out_ping: | |
1da177e4 LT |
2087 | udp4_proc_exit(); |
2088 | out_udp: | |
2089 | tcp4_proc_exit(); | |
2090 | out_tcp: | |
2091 | raw_proc_exit(); | |
2092 | out_raw: | |
2093 | rc = -ENOMEM; | |
2094 | goto out; | |
2095 | } | |
2096 | ||
2097 | #else /* CONFIG_PROC_FS */ | |
2098 | static int __init ipv4_proc_init(void) | |
2099 | { | |
2100 | return 0; | |
2101 | } | |
2102 | #endif /* CONFIG_PROC_FS */ |