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> |
4721031c | 102 | #include <net/gro.h> |
d457a0e3 | 103 | #include <net/gso.h> |
1da177e4 LT |
104 | #include <net/tcp.h> |
105 | #include <net/udp.h> | |
ba4e58ec | 106 | #include <net/udplite.h> |
c319b4d7 | 107 | #include <net/ping.h> |
1da177e4 LT |
108 | #include <linux/skbuff.h> |
109 | #include <net/sock.h> | |
110 | #include <net/raw.h> | |
111 | #include <net/icmp.h> | |
1da177e4 | 112 | #include <net/inet_common.h> |
045a0fa0 | 113 | #include <net/ip_tunnels.h> |
1da177e4 | 114 | #include <net/xfrm.h> |
9b4661bd | 115 | #include <net/net_namespace.h> |
aebda156 | 116 | #include <net/secure_seq.h> |
1da177e4 LT |
117 | #ifdef CONFIG_IP_MROUTE |
118 | #include <linux/mroute.h> | |
119 | #endif | |
3236b004 | 120 | #include <net/l3mdev.h> |
dc13c876 | 121 | #include <net/compat.h> |
1da177e4 | 122 | |
563e0bb0 | 123 | #include <trace/events/sock.h> |
1da177e4 LT |
124 | |
125 | /* The inetsw table contains everything that inet_create needs to | |
126 | * build a new socket. | |
127 | */ | |
128 | static struct list_head inetsw[SOCK_MAX]; | |
129 | static DEFINE_SPINLOCK(inetsw_lock); | |
130 | ||
131 | /* New destruction routine */ | |
132 | ||
133 | void inet_sock_destruct(struct sock *sk) | |
134 | { | |
135 | struct inet_sock *inet = inet_sk(sk); | |
136 | ||
137 | __skb_queue_purge(&sk->sk_receive_queue); | |
138 | __skb_queue_purge(&sk->sk_error_queue); | |
139 | ||
2bb2f5fb | 140 | sk_mem_reclaim_final(sk); |
95766fff | 141 | |
1da177e4 | 142 | if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) { |
3d1427f8 | 143 | pr_err("Attempt to release TCP socket in state %d %p\n", |
1da177e4 LT |
144 | sk->sk_state, sk); |
145 | return; | |
146 | } | |
147 | if (!sock_flag(sk, SOCK_DEAD)) { | |
3d1427f8 | 148 | pr_err("Attempt to release alive inet socket %p\n", sk); |
1da177e4 LT |
149 | return; |
150 | } | |
151 | ||
3e7f2b8d ED |
152 | WARN_ON_ONCE(atomic_read(&sk->sk_rmem_alloc)); |
153 | WARN_ON_ONCE(refcount_read(&sk->sk_wmem_alloc)); | |
154 | WARN_ON_ONCE(sk->sk_wmem_queued); | |
155 | WARN_ON_ONCE(sk_forward_alloc_get(sk)); | |
1da177e4 | 156 | |
f6d8bd05 | 157 | kfree(rcu_dereference_protected(inet->inet_opt, 1)); |
5869b8fa | 158 | dst_release(rcu_dereference_protected(sk->sk_dst_cache, 1)); |
8f905c0e | 159 | dst_release(rcu_dereference_protected(sk->sk_rx_dst, 1)); |
1da177e4 | 160 | } |
3d1427f8 | 161 | EXPORT_SYMBOL(inet_sock_destruct); |
1da177e4 LT |
162 | |
163 | /* | |
164 | * The routines beyond this point handle the behaviour of an AF_INET | |
165 | * socket object. Mostly it punts to the subprotocols of IP to do | |
166 | * the work. | |
167 | */ | |
168 | ||
169 | /* | |
170 | * Automatically bind an unbound socket. | |
171 | */ | |
172 | ||
173 | static int inet_autobind(struct sock *sk) | |
174 | { | |
175 | struct inet_sock *inet; | |
176 | /* We may need to bind the socket. */ | |
177 | lock_sock(sk); | |
178 | inet = inet_sk(sk); | |
c720c7e8 | 179 | if (!inet->inet_num) { |
1da177e4 LT |
180 | if (sk->sk_prot->get_port(sk, 0)) { |
181 | release_sock(sk); | |
182 | return -EAGAIN; | |
183 | } | |
c720c7e8 | 184 | inet->inet_sport = htons(inet->inet_num); |
1da177e4 LT |
185 | } |
186 | release_sock(sk); | |
187 | return 0; | |
188 | } | |
189 | ||
71a9a874 | 190 | int __inet_listen_sk(struct sock *sk, int backlog) |
1da177e4 | 191 | { |
71a9a874 | 192 | unsigned char old_state = sk->sk_state; |
e1cfcbe8 | 193 | int err, tcp_fastopen; |
1da177e4 | 194 | |
1da177e4 | 195 | if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN))) |
71a9a874 | 196 | return -EINVAL; |
1da177e4 | 197 | |
099ecf59 | 198 | WRITE_ONCE(sk->sk_max_ack_backlog, backlog); |
1da177e4 LT |
199 | /* Really, if the socket is already in listen state |
200 | * we can only allow the backlog to be adjusted. | |
201 | */ | |
202 | if (old_state != TCP_LISTEN) { | |
cebc5cba | 203 | /* Enable TFO w/o requiring TCP_FASTOPEN socket option. |
8336886f | 204 | * Note that only TCP sockets (SOCK_STREAM) will reach here. |
cebc5cba YC |
205 | * Also fastopen backlog may already been set via the option |
206 | * because the socket was in TCP_LISTEN state previously but | |
207 | * was shutdown() rather than close(). | |
8336886f | 208 | */ |
5a542133 | 209 | tcp_fastopen = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_fastopen); |
e1cfcbe8 HY |
210 | if ((tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) && |
211 | (tcp_fastopen & TFO_SERVER_ENABLE) && | |
0536fcc0 | 212 | !inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) { |
cebc5cba | 213 | fastopen_queue_tune(sk, backlog); |
43713848 | 214 | tcp_fastopen_init_key_once(sock_net(sk)); |
8336886f | 215 | } |
cebc5cba | 216 | |
e7049395 | 217 | err = inet_csk_listen_start(sk); |
1da177e4 | 218 | if (err) |
71a9a874 PA |
219 | return err; |
220 | ||
f333ee0c | 221 | tcp_call_bpf(sk, BPF_SOCK_OPS_TCP_LISTEN_CB, 0, NULL); |
1da177e4 | 222 | } |
71a9a874 PA |
223 | return 0; |
224 | } | |
225 | ||
226 | /* | |
227 | * Move a socket into listening state. | |
228 | */ | |
229 | int inet_listen(struct socket *sock, int backlog) | |
230 | { | |
231 | struct sock *sk = sock->sk; | |
232 | int err = -EINVAL; | |
233 | ||
234 | lock_sock(sk); | |
235 | ||
236 | if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM) | |
237 | goto out; | |
238 | ||
239 | err = __inet_listen_sk(sk, backlog); | |
1da177e4 LT |
240 | |
241 | out: | |
242 | release_sock(sk); | |
243 | return err; | |
244 | } | |
3d1427f8 | 245 | EXPORT_SYMBOL(inet_listen); |
1da177e4 LT |
246 | |
247 | /* | |
248 | * Create an inet socket. | |
249 | */ | |
250 | ||
3f378b68 EP |
251 | static int inet_create(struct net *net, struct socket *sock, int protocol, |
252 | int kern) | |
1da177e4 LT |
253 | { |
254 | struct sock *sk; | |
1da177e4 LT |
255 | struct inet_protosw *answer; |
256 | struct inet_sock *inet; | |
257 | struct proto *answer_prot; | |
258 | unsigned char answer_flags; | |
bb97d31f | 259 | int try_loading_module = 0; |
86c8f9d1 | 260 | int err; |
1da177e4 | 261 | |
79462ad0 HFS |
262 | if (protocol < 0 || protocol >= IPPROTO_MAX) |
263 | return -EINVAL; | |
264 | ||
1da177e4 LT |
265 | sock->state = SS_UNCONNECTED; |
266 | ||
267 | /* Look for the requested type/protocol pair. */ | |
bb97d31f | 268 | lookup_protocol: |
86c8f9d1 | 269 | err = -ESOCKTNOSUPPORT; |
1da177e4 | 270 | rcu_read_lock(); |
696adfe8 | 271 | list_for_each_entry_rcu(answer, &inetsw[sock->type], list) { |
1da177e4 | 272 | |
696adfe8 | 273 | err = 0; |
1da177e4 LT |
274 | /* Check the non-wild match. */ |
275 | if (protocol == answer->protocol) { | |
276 | if (protocol != IPPROTO_IP) | |
277 | break; | |
278 | } else { | |
279 | /* Check for the two wild cases. */ | |
280 | if (IPPROTO_IP == protocol) { | |
281 | protocol = answer->protocol; | |
282 | break; | |
283 | } | |
284 | if (IPPROTO_IP == answer->protocol) | |
285 | break; | |
286 | } | |
86c8f9d1 | 287 | err = -EPROTONOSUPPORT; |
1da177e4 LT |
288 | } |
289 | ||
696adfe8 | 290 | if (unlikely(err)) { |
bb97d31f ACM |
291 | if (try_loading_module < 2) { |
292 | rcu_read_unlock(); | |
293 | /* | |
294 | * Be more specific, e.g. net-pf-2-proto-132-type-1 | |
295 | * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM) | |
296 | */ | |
297 | if (++try_loading_module == 1) | |
298 | request_module("net-pf-%d-proto-%d-type-%d", | |
299 | PF_INET, protocol, sock->type); | |
300 | /* | |
301 | * Fall back to generic, e.g. net-pf-2-proto-132 | |
302 | * (net-pf-PF_INET-proto-IPPROTO_SCTP) | |
303 | */ | |
304 | else | |
305 | request_module("net-pf-%d-proto-%d", | |
306 | PF_INET, protocol); | |
307 | goto lookup_protocol; | |
308 | } else | |
309 | goto out_rcu_unlock; | |
310 | } | |
311 | ||
1da177e4 | 312 | err = -EPERM; |
52e804c6 EB |
313 | if (sock->type == SOCK_RAW && !kern && |
314 | !ns_capable(net->user_ns, CAP_NET_RAW)) | |
1da177e4 | 315 | goto out_rcu_unlock; |
1da177e4 LT |
316 | |
317 | sock->ops = answer->ops; | |
318 | answer_prot = answer->prot; | |
1da177e4 LT |
319 | answer_flags = answer->flags; |
320 | rcu_read_unlock(); | |
321 | ||
51456b29 | 322 | WARN_ON(!answer_prot->slab); |
1da177e4 | 323 | |
ca746c55 | 324 | err = -ENOMEM; |
11aa9c28 | 325 | sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot, kern); |
51456b29 | 326 | if (!sk) |
1da177e4 LT |
327 | goto out; |
328 | ||
329 | err = 0; | |
1da177e4 | 330 | if (INET_PROTOSW_REUSE & answer_flags) |
4a17fd52 | 331 | sk->sk_reuse = SK_CAN_REUSE; |
1da177e4 LT |
332 | |
333 | inet = inet_sk(sk); | |
b1c0356a | 334 | inet_assign_bit(IS_ICSK, sk, INET_PROTOSW_ICSK & answer_flags); |
1da177e4 | 335 | |
f04b8d34 | 336 | inet_clear_bit(NODEFRAG, sk); |
7b2ff18e | 337 | |
1da177e4 | 338 | if (SOCK_RAW == sock->type) { |
c720c7e8 | 339 | inet->inet_num = protocol; |
1da177e4 | 340 | if (IPPROTO_RAW == protocol) |
cafbe182 | 341 | inet_set_bit(HDRINCL, sk); |
1da177e4 LT |
342 | } |
343 | ||
0968d2a4 | 344 | if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc)) |
1da177e4 LT |
345 | inet->pmtudisc = IP_PMTUDISC_DONT; |
346 | else | |
347 | inet->pmtudisc = IP_PMTUDISC_WANT; | |
348 | ||
f866fbc8 | 349 | atomic_set(&inet->inet_id, 0); |
1da177e4 LT |
350 | |
351 | sock_init_data(sock, sk); | |
352 | ||
353 | sk->sk_destruct = inet_sock_destruct; | |
1da177e4 LT |
354 | sk->sk_protocol = protocol; |
355 | sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv; | |
c11204c7 | 356 | sk->sk_txrehash = READ_ONCE(net->core.sysctl_txrehash); |
1da177e4 LT |
357 | |
358 | inet->uc_ttl = -1; | |
b09bde5c | 359 | inet_set_bit(MC_LOOP, sk); |
1da177e4 | 360 | inet->mc_ttl = 1; |
307b4ac6 | 361 | inet_set_bit(MC_ALL, sk); |
1da177e4 LT |
362 | inet->mc_index = 0; |
363 | inet->mc_list = NULL; | |
4c507d28 | 364 | inet->rcv_tos = 0; |
1da177e4 | 365 | |
c720c7e8 | 366 | if (inet->inet_num) { |
1da177e4 LT |
367 | /* It assumes that any protocol which allows |
368 | * the user to assign a number at socket | |
369 | * creation time automatically | |
370 | * shares. | |
371 | */ | |
c720c7e8 | 372 | inet->inet_sport = htons(inet->inet_num); |
1da177e4 | 373 | /* Add to protocol hash chains. */ |
086c653f CG |
374 | err = sk->sk_prot->hash(sk); |
375 | if (err) { | |
376 | sk_common_release(sk); | |
377 | goto out; | |
378 | } | |
1da177e4 LT |
379 | } |
380 | ||
381 | if (sk->sk_prot->init) { | |
382 | err = sk->sk_prot->init(sk); | |
61023658 DA |
383 | if (err) { |
384 | sk_common_release(sk); | |
385 | goto out; | |
386 | } | |
387 | } | |
388 | ||
389 | if (!kern) { | |
390 | err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk); | |
391 | if (err) { | |
1da177e4 | 392 | sk_common_release(sk); |
61023658 DA |
393 | goto out; |
394 | } | |
1da177e4 LT |
395 | } |
396 | out: | |
397 | return err; | |
398 | out_rcu_unlock: | |
399 | rcu_read_unlock(); | |
400 | goto out; | |
401 | } | |
402 | ||
403 | ||
404 | /* | |
405 | * The peer socket should always be NULL (or else). When we call this | |
406 | * function we are destroying the object and from then on nobody | |
407 | * should refer to it. | |
408 | */ | |
409 | int inet_release(struct socket *sock) | |
410 | { | |
411 | struct sock *sk = sock->sk; | |
412 | ||
413 | if (sk) { | |
414 | long timeout; | |
415 | ||
f5836749 SF |
416 | if (!sk->sk_kern_sock) |
417 | BPF_CGROUP_RUN_PROG_INET_SOCK_RELEASE(sk); | |
418 | ||
1da177e4 LT |
419 | /* Applications forget to leave groups before exiting */ |
420 | ip_mc_drop_socket(sk); | |
421 | ||
422 | /* If linger is set, we don't return until the close | |
423 | * is complete. Otherwise we return immediately. The | |
424 | * actually closing is done the same either way. | |
425 | * | |
426 | * If the close is due to the process exiting, we never | |
427 | * linger.. | |
428 | */ | |
429 | timeout = 0; | |
430 | if (sock_flag(sk, SOCK_LINGER) && | |
431 | !(current->flags & PF_EXITING)) | |
432 | timeout = sk->sk_lingertime; | |
1da177e4 | 433 | sk->sk_prot->close(sk, timeout); |
2b81f816 | 434 | sock->sk = NULL; |
1da177e4 LT |
435 | } |
436 | return 0; | |
437 | } | |
3d1427f8 | 438 | EXPORT_SYMBOL(inet_release); |
1da177e4 | 439 | |
e6d360ff | 440 | int inet_bind_sk(struct sock *sk, struct sockaddr *uaddr, int addr_len) |
1da177e4 | 441 | { |
77241217 | 442 | u32 flags = BIND_WITH_LOCK; |
1da177e4 LT |
443 | int err; |
444 | ||
445 | /* If the socket has its own bind function then use it. (RAW) */ | |
446 | if (sk->sk_prot->bind) { | |
3679d585 | 447 | return sk->sk_prot->bind(sk, uaddr, addr_len); |
1da177e4 | 448 | } |
1da177e4 | 449 | if (addr_len < sizeof(struct sockaddr_in)) |
3679d585 | 450 | return -EINVAL; |
4fbac77d AI |
451 | |
452 | /* BPF prog is run before any checks are done so that if the prog | |
453 | * changes context in a wrong way it will be caught. | |
454 | */ | |
77241217 | 455 | err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, |
6fc88c35 | 456 | CGROUP_INET4_BIND, &flags); |
4fbac77d | 457 | if (err) |
3679d585 AI |
458 | return err; |
459 | ||
77241217 | 460 | return __inet_bind(sk, uaddr, addr_len, flags); |
3679d585 | 461 | } |
e6d360ff PA |
462 | |
463 | int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
464 | { | |
465 | return inet_bind_sk(sock->sk, uaddr, addr_len); | |
466 | } | |
3679d585 AI |
467 | EXPORT_SYMBOL(inet_bind); |
468 | ||
469 | int __inet_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len, | |
cb0721c7 | 470 | u32 flags) |
3679d585 AI |
471 | { |
472 | struct sockaddr_in *addr = (struct sockaddr_in *)uaddr; | |
473 | struct inet_sock *inet = inet_sk(sk); | |
474 | struct net *net = sock_net(sk); | |
475 | unsigned short snum; | |
476 | int chk_addr_ret; | |
477 | u32 tb_id = RT_TABLE_LOCAL; | |
478 | int err; | |
1da177e4 | 479 | |
c349a528 | 480 | if (addr->sin_family != AF_INET) { |
29c486df ED |
481 | /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET) |
482 | * only if s_addr is INADDR_ANY. | |
483 | */ | |
c349a528 | 484 | err = -EAFNOSUPPORT; |
29c486df ED |
485 | if (addr->sin_family != AF_UNSPEC || |
486 | addr->sin_addr.s_addr != htonl(INADDR_ANY)) | |
487 | goto out; | |
c349a528 | 488 | } |
d0733d2e | 489 | |
3236b004 | 490 | tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id; |
30bbaa19 | 491 | chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id); |
1da177e4 LT |
492 | |
493 | /* Not specified by any standard per-se, however it breaks too | |
494 | * many applications when removed. It is unfortunate since | |
495 | * allowing applications to make a non-local bind solves | |
496 | * several problems with systems using dynamic addressing. | |
497 | * (ie. your servers still start up even if your ISDN link | |
498 | * is temporarily down) | |
499 | */ | |
500 | err = -EADDRNOTAVAIL; | |
8ff978b8 RPB |
501 | if (!inet_addr_valid_or_nonlocal(net, inet, addr->sin_addr.s_addr, |
502 | chk_addr_ret)) | |
1da177e4 LT |
503 | goto out; |
504 | ||
505 | snum = ntohs(addr->sin_port); | |
506 | err = -EACCES; | |
77241217 SF |
507 | if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) && |
508 | snum && inet_port_requires_bind_service(net, snum) && | |
3594698a | 509 | !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) |
1da177e4 LT |
510 | goto out; |
511 | ||
512 | /* We keep a pair of addresses. rcv_saddr is the one | |
513 | * used by hash lookups, and saddr is used for transmit. | |
514 | * | |
515 | * In the BSD API these are the same except where it | |
516 | * would be illegal to use them (multicast/broadcast) in | |
517 | * which case the sending device address is used. | |
518 | */ | |
cb0721c7 | 519 | if (flags & BIND_WITH_LOCK) |
3679d585 | 520 | lock_sock(sk); |
1da177e4 LT |
521 | |
522 | /* Check these errors (active socket, double bind). */ | |
523 | err = -EINVAL; | |
c720c7e8 | 524 | if (sk->sk_state != TCP_CLOSE || inet->inet_num) |
1da177e4 LT |
525 | goto out_release_sock; |
526 | ||
c720c7e8 | 527 | inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; |
1da177e4 | 528 | if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) |
c720c7e8 | 529 | inet->inet_saddr = 0; /* Use device */ |
1da177e4 LT |
530 | |
531 | /* Make sure we are allowed to bind here. */ | |
ca571e2e | 532 | if (snum || !(inet_test_bit(BIND_ADDRESS_NO_PORT, sk) || |
cb0721c7 | 533 | (flags & BIND_FORCE_ADDRESS_NO_PORT))) { |
7a7160ed KI |
534 | err = sk->sk_prot->get_port(sk, snum); |
535 | if (err) { | |
aac3fc32 | 536 | inet->inet_saddr = inet->inet_rcv_saddr = 0; |
aac3fc32 AI |
537 | goto out_release_sock; |
538 | } | |
8086fbaf SF |
539 | if (!(flags & BIND_FROM_BPF)) { |
540 | err = BPF_CGROUP_RUN_PROG_INET4_POST_BIND(sk); | |
541 | if (err) { | |
542 | inet->inet_saddr = inet->inet_rcv_saddr = 0; | |
91a760b2 MD |
543 | if (sk->sk_prot->put_port) |
544 | sk->sk_prot->put_port(sk); | |
8086fbaf SF |
545 | goto out_release_sock; |
546 | } | |
aac3fc32 | 547 | } |
1da177e4 LT |
548 | } |
549 | ||
c720c7e8 | 550 | if (inet->inet_rcv_saddr) |
1da177e4 LT |
551 | sk->sk_userlocks |= SOCK_BINDADDR_LOCK; |
552 | if (snum) | |
553 | sk->sk_userlocks |= SOCK_BINDPORT_LOCK; | |
c720c7e8 ED |
554 | inet->inet_sport = htons(inet->inet_num); |
555 | inet->inet_daddr = 0; | |
556 | inet->inet_dport = 0; | |
1da177e4 LT |
557 | sk_dst_reset(sk); |
558 | err = 0; | |
559 | out_release_sock: | |
cb0721c7 | 560 | if (flags & BIND_WITH_LOCK) |
3679d585 | 561 | release_sock(sk); |
1da177e4 LT |
562 | out: |
563 | return err; | |
564 | } | |
565 | ||
5e73ea1a | 566 | int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr, |
1da177e4 LT |
567 | int addr_len, int flags) |
568 | { | |
569 | struct sock *sk = sock->sk; | |
364f997b | 570 | const struct proto *prot; |
d74bad4e | 571 | int err; |
1da177e4 | 572 | |
6503d961 CG |
573 | if (addr_len < sizeof(uaddr->sa_family)) |
574 | return -EINVAL; | |
364f997b KI |
575 | |
576 | /* IPV6_ADDRFORM can change sk->sk_prot under us. */ | |
577 | prot = READ_ONCE(sk->sk_prot); | |
578 | ||
1da177e4 | 579 | if (uaddr->sa_family == AF_UNSPEC) |
364f997b | 580 | return prot->disconnect(sk, flags); |
1da177e4 | 581 | |
d74bad4e | 582 | if (BPF_CGROUP_PRE_CONNECT_ENABLED(sk)) { |
364f997b | 583 | err = prot->pre_connect(sk, uaddr, addr_len); |
d74bad4e AI |
584 | if (err) |
585 | return err; | |
586 | } | |
587 | ||
dcd01eea | 588 | if (data_race(!inet_sk(sk)->inet_num) && inet_autobind(sk)) |
1da177e4 | 589 | return -EAGAIN; |
364f997b | 590 | return prot->connect(sk, uaddr, addr_len); |
1da177e4 | 591 | } |
3d1427f8 | 592 | EXPORT_SYMBOL(inet_dgram_connect); |
1da177e4 | 593 | |
783237e8 | 594 | static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias) |
1da177e4 | 595 | { |
14135f30 | 596 | DEFINE_WAIT_FUNC(wait, woken_wake_function); |
1da177e4 | 597 | |
14135f30 | 598 | add_wait_queue(sk_sleep(sk), &wait); |
783237e8 | 599 | sk->sk_write_pending += writebias; |
4faeee0c | 600 | sk->sk_wait_pending++; |
1da177e4 LT |
601 | |
602 | /* Basic assumption: if someone sets sk->sk_err, he _must_ | |
603 | * change state of the socket from TCP_SYN_*. | |
604 | * Connect() does not allow to get error notifications | |
605 | * without closing the socket. | |
606 | */ | |
607 | while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
608 | release_sock(sk); | |
14135f30 | 609 | timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); |
1da177e4 LT |
610 | lock_sock(sk); |
611 | if (signal_pending(current) || !timeo) | |
612 | break; | |
1da177e4 | 613 | } |
14135f30 | 614 | remove_wait_queue(sk_sleep(sk), &wait); |
783237e8 | 615 | sk->sk_write_pending -= writebias; |
4faeee0c | 616 | sk->sk_wait_pending--; |
1da177e4 LT |
617 | return timeo; |
618 | } | |
619 | ||
620 | /* | |
621 | * Connect to a remote host. There is regrettably still a little | |
622 | * TCP 'magic' in here. | |
623 | */ | |
cf60af03 | 624 | int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, |
3979ad7e | 625 | int addr_len, int flags, int is_sendmsg) |
1da177e4 LT |
626 | { |
627 | struct sock *sk = sock->sk; | |
628 | int err; | |
629 | long timeo; | |
630 | ||
19f6d3f3 WW |
631 | /* |
632 | * uaddr can be NULL and addr_len can be 0 if: | |
633 | * sk is a TCP fastopen active socket and | |
634 | * TCP_FASTOPEN_CONNECT sockopt is set and | |
635 | * we already have a valid cookie for this socket. | |
636 | * In this case, user can call write() after connect(). | |
637 | * write() will invoke tcp_sendmsg_fastopen() which calls | |
638 | * __inet_stream_connect(). | |
639 | */ | |
640 | if (uaddr) { | |
641 | if (addr_len < sizeof(uaddr->sa_family)) | |
642 | return -EINVAL; | |
6503d961 | 643 | |
19f6d3f3 WW |
644 | if (uaddr->sa_family == AF_UNSPEC) { |
645 | err = sk->sk_prot->disconnect(sk, flags); | |
646 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
647 | goto out; | |
648 | } | |
1da177e4 LT |
649 | } |
650 | ||
651 | switch (sock->state) { | |
652 | default: | |
653 | err = -EINVAL; | |
654 | goto out; | |
655 | case SS_CONNECTED: | |
656 | err = -EISCONN; | |
657 | goto out; | |
658 | case SS_CONNECTING: | |
08e39c0d | 659 | if (inet_test_bit(DEFER_CONNECT, sk)) |
3979ad7e | 660 | err = is_sendmsg ? -EINPROGRESS : -EISCONN; |
19f6d3f3 WW |
661 | else |
662 | err = -EALREADY; | |
1da177e4 LT |
663 | /* Fall out of switch with err, set for this state */ |
664 | break; | |
665 | case SS_UNCONNECTED: | |
666 | err = -EISCONN; | |
667 | if (sk->sk_state != TCP_CLOSE) | |
668 | goto out; | |
669 | ||
d74bad4e AI |
670 | if (BPF_CGROUP_PRE_CONNECT_ENABLED(sk)) { |
671 | err = sk->sk_prot->pre_connect(sk, uaddr, addr_len); | |
672 | if (err) | |
673 | goto out; | |
674 | } | |
675 | ||
1da177e4 LT |
676 | err = sk->sk_prot->connect(sk, uaddr, addr_len); |
677 | if (err < 0) | |
678 | goto out; | |
679 | ||
e905a9ed | 680 | sock->state = SS_CONNECTING; |
1da177e4 | 681 | |
08e39c0d | 682 | if (!err && inet_test_bit(DEFER_CONNECT, sk)) |
19f6d3f3 WW |
683 | goto out; |
684 | ||
1da177e4 LT |
685 | /* Just entered SS_CONNECTING state; the only |
686 | * difference is that return value in non-blocking | |
687 | * case is EINPROGRESS, rather than EALREADY. | |
688 | */ | |
689 | err = -EINPROGRESS; | |
690 | break; | |
691 | } | |
692 | ||
693 | timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); | |
694 | ||
695 | if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { | |
783237e8 YC |
696 | int writebias = (sk->sk_protocol == IPPROTO_TCP) && |
697 | tcp_sk(sk)->fastopen_req && | |
698 | tcp_sk(sk)->fastopen_req->data ? 1 : 0; | |
699 | ||
1da177e4 | 700 | /* Error code is set above */ |
783237e8 | 701 | if (!timeo || !inet_wait_for_connect(sk, timeo, writebias)) |
1da177e4 LT |
702 | goto out; |
703 | ||
704 | err = sock_intr_errno(timeo); | |
705 | if (signal_pending(current)) | |
706 | goto out; | |
707 | } | |
708 | ||
709 | /* Connection was closed by RST, timeout, ICMP error | |
710 | * or another process disconnected us. | |
711 | */ | |
712 | if (sk->sk_state == TCP_CLOSE) | |
713 | goto sock_error; | |
714 | ||
715 | /* sk->sk_err may be not zero now, if RECVERR was ordered by user | |
716 | * and error was received after socket entered established state. | |
717 | * Hence, it is handled normally after connect() return successfully. | |
718 | */ | |
719 | ||
720 | sock->state = SS_CONNECTED; | |
721 | err = 0; | |
722 | out: | |
1da177e4 LT |
723 | return err; |
724 | ||
725 | sock_error: | |
726 | err = sock_error(sk) ? : -ECONNABORTED; | |
727 | sock->state = SS_UNCONNECTED; | |
728 | if (sk->sk_prot->disconnect(sk, flags)) | |
729 | sock->state = SS_DISCONNECTING; | |
730 | goto out; | |
731 | } | |
cf60af03 YC |
732 | EXPORT_SYMBOL(__inet_stream_connect); |
733 | ||
734 | int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, | |
735 | int addr_len, int flags) | |
736 | { | |
737 | int err; | |
738 | ||
739 | lock_sock(sock->sk); | |
3979ad7e | 740 | err = __inet_stream_connect(sock, uaddr, addr_len, flags, 0); |
cf60af03 YC |
741 | release_sock(sock->sk); |
742 | return err; | |
743 | } | |
3d1427f8 | 744 | EXPORT_SYMBOL(inet_stream_connect); |
1da177e4 | 745 | |
711bdd51 PA |
746 | void __inet_accept(struct socket *sock, struct socket *newsock, struct sock *newsk) |
747 | { | |
748 | sock_rps_record_flow(newsk); | |
749 | WARN_ON(!((1 << newsk->sk_state) & | |
750 | (TCPF_ESTABLISHED | TCPF_SYN_RECV | | |
751 | TCPF_CLOSE_WAIT | TCPF_CLOSE))); | |
752 | ||
753 | if (test_bit(SOCK_SUPPORT_ZC, &sock->flags)) | |
754 | set_bit(SOCK_SUPPORT_ZC, &newsock->flags); | |
755 | sock_graft(newsk, newsock); | |
756 | ||
757 | newsock->state = SS_CONNECTED; | |
758 | } | |
759 | ||
1da177e4 LT |
760 | /* |
761 | * Accept a pending connection. The TCP layer now gives BSD semantics. | |
762 | */ | |
763 | ||
cdfbabfb DH |
764 | int inet_accept(struct socket *sock, struct socket *newsock, int flags, |
765 | bool kern) | |
1da177e4 | 766 | { |
364f997b | 767 | struct sock *sk1 = sock->sk, *sk2; |
1da177e4 | 768 | int err = -EINVAL; |
1da177e4 | 769 | |
364f997b KI |
770 | /* IPV6_ADDRFORM can change sk->sk_prot under us. */ |
771 | sk2 = READ_ONCE(sk1->sk_prot)->accept(sk1, flags, &err, kern); | |
1da177e4 | 772 | if (!sk2) |
711bdd51 | 773 | return err; |
1da177e4 LT |
774 | |
775 | lock_sock(sk2); | |
711bdd51 | 776 | __inet_accept(sock, newsock, sk2); |
1da177e4 | 777 | release_sock(sk2); |
711bdd51 | 778 | return 0; |
1da177e4 | 779 | } |
3d1427f8 | 780 | EXPORT_SYMBOL(inet_accept); |
1da177e4 | 781 | |
1da177e4 LT |
782 | /* |
783 | * This does both peername and sockname. | |
784 | */ | |
785 | int inet_getname(struct socket *sock, struct sockaddr *uaddr, | |
1b66d253 | 786 | int peer) |
1da177e4 LT |
787 | { |
788 | struct sock *sk = sock->sk; | |
789 | struct inet_sock *inet = inet_sk(sk); | |
38bfd8f5 | 790 | DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr); |
1da177e4 LT |
791 | |
792 | sin->sin_family = AF_INET; | |
9dfc685e | 793 | lock_sock(sk); |
1da177e4 | 794 | if (peer) { |
c720c7e8 | 795 | if (!inet->inet_dport || |
1da177e4 | 796 | (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && |
9dfc685e ED |
797 | peer == 1)) { |
798 | release_sock(sk); | |
1da177e4 | 799 | return -ENOTCONN; |
9dfc685e | 800 | } |
c720c7e8 ED |
801 | sin->sin_port = inet->inet_dport; |
802 | sin->sin_addr.s_addr = inet->inet_daddr; | |
9dfc685e ED |
803 | BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, |
804 | CGROUP_INET4_GETPEERNAME); | |
1da177e4 | 805 | } else { |
c720c7e8 | 806 | __be32 addr = inet->inet_rcv_saddr; |
1da177e4 | 807 | if (!addr) |
c720c7e8 ED |
808 | addr = inet->inet_saddr; |
809 | sin->sin_port = inet->inet_sport; | |
1da177e4 | 810 | sin->sin_addr.s_addr = addr; |
9dfc685e ED |
811 | BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, |
812 | CGROUP_INET4_GETSOCKNAME); | |
a9ed15da | 813 | } |
9dfc685e | 814 | release_sock(sk); |
1da177e4 | 815 | memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); |
9b2c45d4 | 816 | return sizeof(*sin); |
1da177e4 | 817 | } |
3d1427f8 | 818 | EXPORT_SYMBOL(inet_getname); |
1da177e4 | 819 | |
e4730936 | 820 | int inet_send_prepare(struct sock *sk) |
1da177e4 | 821 | { |
c58dc01b | 822 | sock_rps_record_flow(sk); |
fec5e652 | 823 | |
1da177e4 | 824 | /* We may need to bind the socket. */ |
dcd01eea | 825 | if (data_race(!inet_sk(sk)->inet_num) && !sk->sk_prot->no_autobind && |
7ba42910 | 826 | inet_autobind(sk)) |
1da177e4 LT |
827 | return -EAGAIN; |
828 | ||
e4730936 PA |
829 | return 0; |
830 | } | |
831 | EXPORT_SYMBOL_GPL(inet_send_prepare); | |
832 | ||
833 | int inet_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) | |
834 | { | |
835 | struct sock *sk = sock->sk; | |
836 | ||
837 | if (unlikely(inet_send_prepare(sk))) | |
838 | return -EAGAIN; | |
839 | ||
6f24080e PA |
840 | return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udp_sendmsg, |
841 | sk, msg, size); | |
1da177e4 | 842 | } |
3d1427f8 | 843 | EXPORT_SYMBOL(inet_sendmsg); |
1da177e4 | 844 | |
1d7e4538 DH |
845 | void inet_splice_eof(struct socket *sock) |
846 | { | |
847 | const struct proto *prot; | |
848 | struct sock *sk = sock->sk; | |
849 | ||
850 | if (unlikely(inet_send_prepare(sk))) | |
851 | return; | |
852 | ||
853 | /* IPV6_ADDRFORM can change sk->sk_prot under us. */ | |
854 | prot = READ_ONCE(sk->sk_prot); | |
855 | if (prot->splice_eof) | |
856 | prot->splice_eof(sock); | |
857 | } | |
858 | EXPORT_SYMBOL_GPL(inet_splice_eof); | |
859 | ||
6f24080e | 860 | INDIRECT_CALLABLE_DECLARE(int udp_recvmsg(struct sock *, struct msghdr *, |
ec095263 | 861 | size_t, int, int *)); |
1b784140 YX |
862 | int inet_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, |
863 | int flags) | |
fec5e652 TH |
864 | { |
865 | struct sock *sk = sock->sk; | |
866 | int addr_len = 0; | |
867 | int err; | |
868 | ||
e3f2c4a3 SHY |
869 | if (likely(!(flags & MSG_ERRQUEUE))) |
870 | sock_rps_record_flow(sk); | |
fec5e652 | 871 | |
6f24080e | 872 | err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udp_recvmsg, |
ec095263 | 873 | sk, msg, size, flags, &addr_len); |
fec5e652 TH |
874 | if (err >= 0) |
875 | msg->msg_namelen = addr_len; | |
876 | return err; | |
877 | } | |
878 | EXPORT_SYMBOL(inet_recvmsg); | |
1da177e4 LT |
879 | |
880 | int inet_shutdown(struct socket *sock, int how) | |
881 | { | |
882 | struct sock *sk = sock->sk; | |
883 | int err = 0; | |
884 | ||
885 | /* This should really check to make sure | |
886 | * the socket is a TCP socket. (WHY AC...) | |
887 | */ | |
888 | how++; /* maps 0->1 has the advantage of making bit 1 rcvs and | |
889 | 1->2 bit 2 snds. | |
890 | 2->3 */ | |
891 | if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */ | |
892 | return -EINVAL; | |
893 | ||
894 | lock_sock(sk); | |
895 | if (sock->state == SS_CONNECTING) { | |
896 | if ((1 << sk->sk_state) & | |
897 | (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE)) | |
898 | sock->state = SS_DISCONNECTING; | |
899 | else | |
900 | sock->state = SS_CONNECTED; | |
901 | } | |
902 | ||
903 | switch (sk->sk_state) { | |
904 | case TCP_CLOSE: | |
905 | err = -ENOTCONN; | |
906 | /* Hack to wake up other listeners, who can poll for | |
a9a08845 | 907 | EPOLLHUP, even on eg. unconnected UDP sockets -- RR */ |
a8eceea8 | 908 | fallthrough; |
1da177e4 | 909 | default: |
e14cadfd | 910 | WRITE_ONCE(sk->sk_shutdown, sk->sk_shutdown | how); |
1da177e4 LT |
911 | if (sk->sk_prot->shutdown) |
912 | sk->sk_prot->shutdown(sk, how); | |
913 | break; | |
914 | ||
915 | /* Remaining two branches are temporary solution for missing | |
916 | * close() in multithreaded environment. It is _not_ a good idea, | |
917 | * but we have no choice until close() is repaired at VFS level. | |
918 | */ | |
919 | case TCP_LISTEN: | |
920 | if (!(how & RCV_SHUTDOWN)) | |
921 | break; | |
a8eceea8 | 922 | fallthrough; |
1da177e4 LT |
923 | case TCP_SYN_SENT: |
924 | err = sk->sk_prot->disconnect(sk, O_NONBLOCK); | |
925 | sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; | |
926 | break; | |
927 | } | |
928 | ||
929 | /* Wake up anyone sleeping in poll. */ | |
930 | sk->sk_state_change(sk); | |
931 | release_sock(sk); | |
932 | return err; | |
933 | } | |
3d1427f8 | 934 | EXPORT_SYMBOL(inet_shutdown); |
1da177e4 LT |
935 | |
936 | /* | |
937 | * ioctl() calls you can issue on an INET socket. Most of these are | |
938 | * device configuration and stuff and very rarely used. Some ioctls | |
939 | * pass on to the socket itself. | |
940 | * | |
941 | * NOTE: I like the idea of a module for the config stuff. ie ifconfig | |
942 | * loads the devconfigure module does its configuring and unloads it. | |
943 | * There's a good 20K of config code hanging around the kernel. | |
944 | */ | |
945 | ||
946 | int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) | |
947 | { | |
948 | struct sock *sk = sock->sk; | |
949 | int err = 0; | |
3b1e0a65 | 950 | struct net *net = sock_net(sk); |
03aef17b AV |
951 | void __user *p = (void __user *)arg; |
952 | struct ifreq ifr; | |
ca25c300 | 953 | struct rtentry rt; |
1da177e4 LT |
954 | |
955 | switch (cmd) { | |
3d1427f8 ED |
956 | case SIOCADDRT: |
957 | case SIOCDELRT: | |
ca25c300 AV |
958 | if (copy_from_user(&rt, p, sizeof(struct rtentry))) |
959 | return -EFAULT; | |
960 | err = ip_rt_ioctl(net, cmd, &rt); | |
961 | break; | |
3d1427f8 | 962 | case SIOCRTMSG: |
ca25c300 | 963 | err = -EINVAL; |
3d1427f8 ED |
964 | break; |
965 | case SIOCDARP: | |
966 | case SIOCGARP: | |
967 | case SIOCSARP: | |
968 | err = arp_ioctl(net, cmd, (void __user *)arg); | |
969 | break; | |
970 | case SIOCGIFADDR: | |
3d1427f8 | 971 | case SIOCGIFBRDADDR: |
3d1427f8 | 972 | case SIOCGIFNETMASK: |
3d1427f8 | 973 | case SIOCGIFDSTADDR: |
03aef17b | 974 | case SIOCGIFPFLAGS: |
29c49648 | 975 | if (get_user_ifreq(&ifr, NULL, p)) |
03aef17b AV |
976 | return -EFAULT; |
977 | err = devinet_ioctl(net, cmd, &ifr); | |
29c49648 | 978 | if (!err && put_user_ifreq(&ifr, p)) |
03aef17b AV |
979 | err = -EFAULT; |
980 | break; | |
981 | ||
982 | case SIOCSIFADDR: | |
983 | case SIOCSIFBRDADDR: | |
984 | case SIOCSIFNETMASK: | |
3d1427f8 ED |
985 | case SIOCSIFDSTADDR: |
986 | case SIOCSIFPFLAGS: | |
3d1427f8 | 987 | case SIOCSIFFLAGS: |
29c49648 | 988 | if (get_user_ifreq(&ifr, NULL, p)) |
03aef17b AV |
989 | return -EFAULT; |
990 | err = devinet_ioctl(net, cmd, &ifr); | |
3d1427f8 ED |
991 | break; |
992 | default: | |
993 | if (sk->sk_prot->ioctl) | |
e1d001fa | 994 | err = sk_ioctl(sk, cmd, (void __user *)arg); |
3d1427f8 ED |
995 | else |
996 | err = -ENOIOCTLCMD; | |
997 | break; | |
1da177e4 LT |
998 | } |
999 | return err; | |
1000 | } | |
3d1427f8 | 1001 | EXPORT_SYMBOL(inet_ioctl); |
1da177e4 | 1002 | |
709b46e8 | 1003 | #ifdef CONFIG_COMPAT |
dc13c876 CH |
1004 | static int inet_compat_routing_ioctl(struct sock *sk, unsigned int cmd, |
1005 | struct compat_rtentry __user *ur) | |
1006 | { | |
1007 | compat_uptr_t rtdev; | |
1008 | struct rtentry rt; | |
1009 | ||
1010 | if (copy_from_user(&rt.rt_dst, &ur->rt_dst, | |
1011 | 3 * sizeof(struct sockaddr)) || | |
1012 | get_user(rt.rt_flags, &ur->rt_flags) || | |
1013 | get_user(rt.rt_metric, &ur->rt_metric) || | |
1014 | get_user(rt.rt_mtu, &ur->rt_mtu) || | |
1015 | get_user(rt.rt_window, &ur->rt_window) || | |
1016 | get_user(rt.rt_irtt, &ur->rt_irtt) || | |
1017 | get_user(rtdev, &ur->rt_dev)) | |
1018 | return -EFAULT; | |
1019 | ||
1020 | rt.rt_dev = compat_ptr(rtdev); | |
1021 | return ip_rt_ioctl(sock_net(sk), cmd, &rt); | |
1022 | } | |
1023 | ||
686dc6b6 | 1024 | static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
709b46e8 | 1025 | { |
dc13c876 | 1026 | void __user *argp = compat_ptr(arg); |
709b46e8 | 1027 | struct sock *sk = sock->sk; |
709b46e8 | 1028 | |
dc13c876 CH |
1029 | switch (cmd) { |
1030 | case SIOCADDRT: | |
1031 | case SIOCDELRT: | |
1032 | return inet_compat_routing_ioctl(sk, cmd, argp); | |
1033 | default: | |
1034 | if (!sk->sk_prot->compat_ioctl) | |
1035 | return -ENOIOCTLCMD; | |
1036 | return sk->sk_prot->compat_ioctl(sk, cmd, arg); | |
1037 | } | |
709b46e8 | 1038 | } |
dc13c876 | 1039 | #endif /* CONFIG_COMPAT */ |
709b46e8 | 1040 | |
90ddc4f0 | 1041 | const struct proto_ops inet_stream_ops = { |
543d9cfe ACM |
1042 | .family = PF_INET, |
1043 | .owner = THIS_MODULE, | |
1044 | .release = inet_release, | |
1045 | .bind = inet_bind, | |
1046 | .connect = inet_stream_connect, | |
1047 | .socketpair = sock_no_socketpair, | |
1048 | .accept = inet_accept, | |
1049 | .getname = inet_getname, | |
a11e1d43 | 1050 | .poll = tcp_poll, |
543d9cfe | 1051 | .ioctl = inet_ioctl, |
c7cbdbf2 | 1052 | .gettstamp = sock_gettstamp, |
543d9cfe ACM |
1053 | .listen = inet_listen, |
1054 | .shutdown = inet_shutdown, | |
1055 | .setsockopt = sock_common_setsockopt, | |
1056 | .getsockopt = sock_common_getsockopt, | |
7ba42910 | 1057 | .sendmsg = inet_sendmsg, |
fec5e652 | 1058 | .recvmsg = inet_recvmsg, |
05255b82 | 1059 | #ifdef CONFIG_MMU |
93ab6cc6 | 1060 | .mmap = tcp_mmap, |
05255b82 | 1061 | #endif |
1d7e4538 | 1062 | .splice_eof = inet_splice_eof, |
9c55e01c | 1063 | .splice_read = tcp_splice_read, |
32035585 | 1064 | .read_sock = tcp_read_sock, |
965b57b4 | 1065 | .read_skb = tcp_read_skb, |
306b13eb | 1066 | .sendmsg_locked = tcp_sendmsg_locked, |
32035585 | 1067 | .peek_len = tcp_peek_len, |
3fdadf7d | 1068 | #ifdef CONFIG_COMPAT |
709b46e8 | 1069 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 1070 | #endif |
d1361840 | 1071 | .set_rcvlowat = tcp_set_rcvlowat, |
1da177e4 | 1072 | }; |
3d1427f8 | 1073 | EXPORT_SYMBOL(inet_stream_ops); |
1da177e4 | 1074 | |
90ddc4f0 | 1075 | const struct proto_ops inet_dgram_ops = { |
543d9cfe ACM |
1076 | .family = PF_INET, |
1077 | .owner = THIS_MODULE, | |
1078 | .release = inet_release, | |
1079 | .bind = inet_bind, | |
1080 | .connect = inet_dgram_connect, | |
1081 | .socketpair = sock_no_socketpair, | |
1082 | .accept = sock_no_accept, | |
1083 | .getname = inet_getname, | |
a11e1d43 | 1084 | .poll = udp_poll, |
543d9cfe | 1085 | .ioctl = inet_ioctl, |
c7cbdbf2 | 1086 | .gettstamp = sock_gettstamp, |
543d9cfe ACM |
1087 | .listen = sock_no_listen, |
1088 | .shutdown = inet_shutdown, | |
1089 | .setsockopt = sock_common_setsockopt, | |
1090 | .getsockopt = sock_common_getsockopt, | |
1091 | .sendmsg = inet_sendmsg, | |
965b57b4 | 1092 | .read_skb = udp_read_skb, |
fec5e652 | 1093 | .recvmsg = inet_recvmsg, |
543d9cfe | 1094 | .mmap = sock_no_mmap, |
1d7e4538 | 1095 | .splice_eof = inet_splice_eof, |
627d2d6b | 1096 | .set_peek_off = sk_set_peek_off, |
3fdadf7d | 1097 | #ifdef CONFIG_COMPAT |
709b46e8 | 1098 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 1099 | #endif |
1da177e4 | 1100 | }; |
3d1427f8 | 1101 | EXPORT_SYMBOL(inet_dgram_ops); |
1da177e4 LT |
1102 | |
1103 | /* | |
1104 | * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without | |
a11e1d43 | 1105 | * udp_poll |
1da177e4 | 1106 | */ |
90ddc4f0 | 1107 | static const struct proto_ops inet_sockraw_ops = { |
543d9cfe ACM |
1108 | .family = PF_INET, |
1109 | .owner = THIS_MODULE, | |
1110 | .release = inet_release, | |
1111 | .bind = inet_bind, | |
1112 | .connect = inet_dgram_connect, | |
1113 | .socketpair = sock_no_socketpair, | |
1114 | .accept = sock_no_accept, | |
1115 | .getname = inet_getname, | |
a11e1d43 | 1116 | .poll = datagram_poll, |
543d9cfe | 1117 | .ioctl = inet_ioctl, |
c7cbdbf2 | 1118 | .gettstamp = sock_gettstamp, |
543d9cfe ACM |
1119 | .listen = sock_no_listen, |
1120 | .shutdown = inet_shutdown, | |
1121 | .setsockopt = sock_common_setsockopt, | |
1122 | .getsockopt = sock_common_getsockopt, | |
1123 | .sendmsg = inet_sendmsg, | |
fec5e652 | 1124 | .recvmsg = inet_recvmsg, |
543d9cfe | 1125 | .mmap = sock_no_mmap, |
1d7e4538 | 1126 | .splice_eof = inet_splice_eof, |
3fdadf7d | 1127 | #ifdef CONFIG_COMPAT |
709b46e8 | 1128 | .compat_ioctl = inet_compat_ioctl, |
3fdadf7d | 1129 | #endif |
1da177e4 LT |
1130 | }; |
1131 | ||
ec1b4cf7 | 1132 | static const struct net_proto_family inet_family_ops = { |
1da177e4 LT |
1133 | .family = PF_INET, |
1134 | .create = inet_create, | |
1135 | .owner = THIS_MODULE, | |
1136 | }; | |
1137 | ||
1da177e4 LT |
1138 | /* Upon startup we insert all the elements in inetsw_array[] into |
1139 | * the linked list inetsw. | |
1140 | */ | |
1141 | static struct inet_protosw inetsw_array[] = | |
1142 | { | |
e905a9ed YH |
1143 | { |
1144 | .type = SOCK_STREAM, | |
1145 | .protocol = IPPROTO_TCP, | |
1146 | .prot = &tcp_prot, | |
1147 | .ops = &inet_stream_ops, | |
e905a9ed | 1148 | .flags = INET_PROTOSW_PERMANENT | |
d83d8461 | 1149 | INET_PROTOSW_ICSK, |
e905a9ed YH |
1150 | }, |
1151 | ||
1152 | { | |
1153 | .type = SOCK_DGRAM, | |
1154 | .protocol = IPPROTO_UDP, | |
1155 | .prot = &udp_prot, | |
1156 | .ops = &inet_dgram_ops, | |
e905a9ed | 1157 | .flags = INET_PROTOSW_PERMANENT, |
1da177e4 | 1158 | }, |
e905a9ed | 1159 | |
c319b4d7 VK |
1160 | { |
1161 | .type = SOCK_DGRAM, | |
1162 | .protocol = IPPROTO_ICMP, | |
1163 | .prot = &ping_prot, | |
77d4b1d3 | 1164 | .ops = &inet_sockraw_ops, |
c319b4d7 VK |
1165 | .flags = INET_PROTOSW_REUSE, |
1166 | }, | |
1da177e4 LT |
1167 | |
1168 | { | |
e905a9ed YH |
1169 | .type = SOCK_RAW, |
1170 | .protocol = IPPROTO_IP, /* wild card */ | |
1171 | .prot = &raw_prot, | |
1172 | .ops = &inet_sockraw_ops, | |
e905a9ed | 1173 | .flags = INET_PROTOSW_REUSE, |
1da177e4 LT |
1174 | } |
1175 | }; | |
1176 | ||
c40f6fff | 1177 | #define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array) |
1da177e4 LT |
1178 | |
1179 | void inet_register_protosw(struct inet_protosw *p) | |
1180 | { | |
1181 | struct list_head *lh; | |
1182 | struct inet_protosw *answer; | |
1183 | int protocol = p->protocol; | |
1184 | struct list_head *last_perm; | |
1185 | ||
1186 | spin_lock_bh(&inetsw_lock); | |
1187 | ||
1188 | if (p->type >= SOCK_MAX) | |
1189 | goto out_illegal; | |
1190 | ||
1191 | /* If we are trying to override a permanent protocol, bail. */ | |
1da177e4 LT |
1192 | last_perm = &inetsw[p->type]; |
1193 | list_for_each(lh, &inetsw[p->type]) { | |
1194 | answer = list_entry(lh, struct inet_protosw, list); | |
1da177e4 | 1195 | /* Check only the non-wild match. */ |
f6c53334 JZ |
1196 | if ((INET_PROTOSW_PERMANENT & answer->flags) == 0) |
1197 | break; | |
1198 | if (protocol == answer->protocol) | |
1199 | goto out_permanent; | |
1200 | last_perm = lh; | |
1da177e4 | 1201 | } |
1da177e4 LT |
1202 | |
1203 | /* Add the new entry after the last permanent entry if any, so that | |
1204 | * the new entry does not override a permanent entry when matched with | |
1205 | * a wild-card protocol. But it is allowed to override any existing | |
e905a9ed | 1206 | * non-permanent entry. This means that when we remove this entry, the |
1da177e4 LT |
1207 | * system automatically returns to the old behavior. |
1208 | */ | |
1209 | list_add_rcu(&p->list, last_perm); | |
1210 | out: | |
1211 | spin_unlock_bh(&inetsw_lock); | |
1212 | ||
1da177e4 LT |
1213 | return; |
1214 | ||
1215 | out_permanent: | |
058bd4d2 | 1216 | pr_err("Attempt to override permanent protocol %d\n", protocol); |
1da177e4 LT |
1217 | goto out; |
1218 | ||
1219 | out_illegal: | |
058bd4d2 | 1220 | pr_err("Ignoring attempt to register invalid socket type %d\n", |
1da177e4 LT |
1221 | p->type); |
1222 | goto out; | |
1223 | } | |
3d1427f8 | 1224 | EXPORT_SYMBOL(inet_register_protosw); |
1da177e4 LT |
1225 | |
1226 | void inet_unregister_protosw(struct inet_protosw *p) | |
1227 | { | |
1228 | if (INET_PROTOSW_PERMANENT & p->flags) { | |
058bd4d2 | 1229 | pr_err("Attempt to unregister permanent protocol %d\n", |
1da177e4 LT |
1230 | p->protocol); |
1231 | } else { | |
1232 | spin_lock_bh(&inetsw_lock); | |
1233 | list_del_rcu(&p->list); | |
1234 | spin_unlock_bh(&inetsw_lock); | |
1235 | ||
1236 | synchronize_net(); | |
1237 | } | |
1238 | } | |
3d1427f8 | 1239 | EXPORT_SYMBOL(inet_unregister_protosw); |
1da177e4 | 1240 | |
32519f11 ACM |
1241 | static int inet_sk_reselect_saddr(struct sock *sk) |
1242 | { | |
1243 | struct inet_sock *inet = inet_sk(sk); | |
c720c7e8 | 1244 | __be32 old_saddr = inet->inet_saddr; |
c720c7e8 | 1245 | __be32 daddr = inet->inet_daddr; |
6e869138 | 1246 | struct flowi4 *fl4; |
b23dd4fe DM |
1247 | struct rtable *rt; |
1248 | __be32 new_saddr; | |
f6d8bd05 | 1249 | struct ip_options_rcu *inet_opt; |
28044fc1 | 1250 | int err; |
32519f11 | 1251 | |
f6d8bd05 | 1252 | inet_opt = rcu_dereference_protected(inet->inet_opt, |
1e1d04e6 | 1253 | lockdep_sock_is_held(sk)); |
f6d8bd05 ED |
1254 | if (inet_opt && inet_opt->opt.srr) |
1255 | daddr = inet_opt->opt.faddr; | |
32519f11 ACM |
1256 | |
1257 | /* Query new route. */ | |
6e869138 | 1258 | fl4 = &inet->cork.fl.u.ip4; |
67e1e2f4 GN |
1259 | rt = ip_route_connect(fl4, daddr, 0, sk->sk_bound_dev_if, |
1260 | sk->sk_protocol, inet->inet_sport, | |
1261 | inet->inet_dport, sk); | |
b23dd4fe DM |
1262 | if (IS_ERR(rt)) |
1263 | return PTR_ERR(rt); | |
32519f11 | 1264 | |
6e869138 | 1265 | new_saddr = fl4->saddr; |
32519f11 | 1266 | |
28044fc1 JK |
1267 | if (new_saddr == old_saddr) { |
1268 | sk_setup_caps(sk, &rt->dst); | |
32519f11 | 1269 | return 0; |
28044fc1 JK |
1270 | } |
1271 | ||
8c5dae4c | 1272 | err = inet_bhash2_update_saddr(sk, &new_saddr, AF_INET); |
28044fc1 | 1273 | if (err) { |
28044fc1 JK |
1274 | ip_rt_put(rt); |
1275 | return err; | |
1276 | } | |
1277 | ||
1278 | sk_setup_caps(sk, &rt->dst); | |
32519f11 | 1279 | |
e49e4aff | 1280 | if (READ_ONCE(sock_net(sk)->ipv4.sysctl_ip_dynaddr) > 1) { |
058bd4d2 JP |
1281 | pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n", |
1282 | __func__, &old_saddr, &new_saddr); | |
32519f11 ACM |
1283 | } |
1284 | ||
32519f11 ACM |
1285 | /* |
1286 | * XXX The only one ugly spot where we need to | |
1287 | * XXX really change the sockets identity after | |
1288 | * XXX it has entered the hashes. -DaveM | |
1289 | * | |
1290 | * Besides that, it does not check for connection | |
1291 | * uniqueness. Wait for troubles. | |
1292 | */ | |
086c653f | 1293 | return __sk_prot_rehash(sk); |
32519f11 ACM |
1294 | } |
1295 | ||
1296 | int inet_sk_rebuild_header(struct sock *sk) | |
1297 | { | |
1298 | struct inet_sock *inet = inet_sk(sk); | |
1299 | struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0); | |
3ca3c68e | 1300 | __be32 daddr; |
f6d8bd05 | 1301 | struct ip_options_rcu *inet_opt; |
6e869138 | 1302 | struct flowi4 *fl4; |
32519f11 ACM |
1303 | int err; |
1304 | ||
1305 | /* Route is OK, nothing to do. */ | |
1306 | if (rt) | |
1307 | return 0; | |
1308 | ||
1309 | /* Reroute. */ | |
f6d8bd05 ED |
1310 | rcu_read_lock(); |
1311 | inet_opt = rcu_dereference(inet->inet_opt); | |
c720c7e8 | 1312 | daddr = inet->inet_daddr; |
f6d8bd05 ED |
1313 | if (inet_opt && inet_opt->opt.srr) |
1314 | daddr = inet_opt->opt.faddr; | |
1315 | rcu_read_unlock(); | |
6e869138 DM |
1316 | fl4 = &inet->cork.fl.u.ip4; |
1317 | rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr, | |
78fbfd8a DM |
1318 | inet->inet_dport, inet->inet_sport, |
1319 | sk->sk_protocol, RT_CONN_FLAGS(sk), | |
1320 | sk->sk_bound_dev_if); | |
b23dd4fe DM |
1321 | if (!IS_ERR(rt)) { |
1322 | err = 0; | |
d8d1f30b | 1323 | sk_setup_caps(sk, &rt->dst); |
b23dd4fe DM |
1324 | } else { |
1325 | err = PTR_ERR(rt); | |
1326 | ||
32519f11 ACM |
1327 | /* Routing failed... */ |
1328 | sk->sk_route_caps = 0; | |
1329 | /* | |
1330 | * Other protocols have to map its equivalent state to TCP_SYN_SENT. | |
1331 | * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme | |
1332 | */ | |
e49e4aff | 1333 | if (!READ_ONCE(sock_net(sk)->ipv4.sysctl_ip_dynaddr) || |
32519f11 ACM |
1334 | sk->sk_state != TCP_SYN_SENT || |
1335 | (sk->sk_userlocks & SOCK_BINDADDR_LOCK) || | |
1336 | (err = inet_sk_reselect_saddr(sk)) != 0) | |
2f2d9972 | 1337 | WRITE_ONCE(sk->sk_err_soft, -err); |
32519f11 ACM |
1338 | } |
1339 | ||
1340 | return err; | |
1341 | } | |
32519f11 ACM |
1342 | EXPORT_SYMBOL(inet_sk_rebuild_header); |
1343 | ||
563e0bb0 YS |
1344 | void inet_sk_set_state(struct sock *sk, int state) |
1345 | { | |
1346 | trace_inet_sock_set_state(sk, sk->sk_state, state); | |
1347 | sk->sk_state = state; | |
1348 | } | |
1349 | EXPORT_SYMBOL(inet_sk_set_state); | |
1350 | ||
1351 | void inet_sk_state_store(struct sock *sk, int newstate) | |
1352 | { | |
1353 | trace_inet_sock_set_state(sk, sk->sk_state, newstate); | |
1354 | smp_store_release(&sk->sk_state, newstate); | |
1355 | } | |
1356 | ||
b8921ca8 TH |
1357 | struct sk_buff *inet_gso_segment(struct sk_buff *skb, |
1358 | netdev_features_t features) | |
f4c50d99 | 1359 | { |
0c19f846 | 1360 | bool udpfrag = false, fixedid = false, gso_partial, encap; |
f4c50d99 | 1361 | struct sk_buff *segs = ERR_PTR(-EINVAL); |
bca49f84 | 1362 | const struct net_offload *ops; |
0c19f846 | 1363 | unsigned int offset = 0; |
f9242b6b | 1364 | struct iphdr *iph; |
802ab55a | 1365 | int proto, tot_len; |
3347c960 | 1366 | int nhoff; |
f4c50d99 HX |
1367 | int ihl; |
1368 | int id; | |
1369 | ||
3347c960 ED |
1370 | skb_reset_network_header(skb); |
1371 | nhoff = skb_network_header(skb) - skb_mac_header(skb); | |
bbcf467d | 1372 | if (unlikely(!pskb_may_pull(skb, sizeof(*iph)))) |
f4c50d99 HX |
1373 | goto out; |
1374 | ||
eddc9ec5 | 1375 | iph = ip_hdr(skb); |
f4c50d99 HX |
1376 | ihl = iph->ihl * 4; |
1377 | if (ihl < sizeof(*iph)) | |
1378 | goto out; | |
1379 | ||
47d27aad ED |
1380 | id = ntohs(iph->id); |
1381 | proto = iph->protocol; | |
1382 | ||
1383 | /* Warning: after this point, iph might be no longer valid */ | |
bbcf467d | 1384 | if (unlikely(!pskb_may_pull(skb, ihl))) |
f4c50d99 | 1385 | goto out; |
47d27aad | 1386 | __skb_pull(skb, ihl); |
f4c50d99 | 1387 | |
8c3a897b ED |
1388 | encap = SKB_GSO_CB(skb)->encap_level > 0; |
1389 | if (encap) | |
1e16aa3d | 1390 | features &= skb->dev->hw_enc_features; |
3347c960 | 1391 | SKB_GSO_CB(skb)->encap_level += ihl; |
73136267 | 1392 | |
badff6d0 | 1393 | skb_reset_transport_header(skb); |
47d27aad | 1394 | |
f4c50d99 HX |
1395 | segs = ERR_PTR(-EPROTONOSUPPORT); |
1396 | ||
cbc53e08 | 1397 | if (!skb->encapsulation || encap) { |
0c19f846 | 1398 | udpfrag = !!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP); |
cbc53e08 AD |
1399 | fixedid = !!(skb_shinfo(skb)->gso_type & SKB_GSO_TCP_FIXEDID); |
1400 | ||
1401 | /* fixed ID is invalid if DF bit is not set */ | |
a5108878 | 1402 | if (fixedid && !(ip_hdr(skb)->frag_off & htons(IP_DF))) |
cbc53e08 AD |
1403 | goto out; |
1404 | } | |
dcd60771 | 1405 | |
bca49f84 | 1406 | ops = rcu_dereference(inet_offloads[proto]); |
cc20cced | 1407 | if (likely(ops && ops->callbacks.gso_segment)) { |
f191a1d1 | 1408 | segs = ops->callbacks.gso_segment(skb, features); |
cc20cced TL |
1409 | if (!segs) |
1410 | skb->network_header = skb_mac_header(skb) + nhoff - skb->head; | |
1411 | } | |
f4c50d99 | 1412 | |
50c3a487 | 1413 | if (IS_ERR_OR_NULL(segs)) |
f4c50d99 HX |
1414 | goto out; |
1415 | ||
07b26c94 SK |
1416 | gso_partial = !!(skb_shinfo(segs)->gso_type & SKB_GSO_PARTIAL); |
1417 | ||
f4c50d99 HX |
1418 | skb = segs; |
1419 | do { | |
3347c960 | 1420 | iph = (struct iphdr *)(skb_mac_header(skb) + nhoff); |
0c19f846 WB |
1421 | if (udpfrag) { |
1422 | iph->frag_off = htons(offset >> 3); | |
1423 | if (skb->next) | |
1424 | iph->frag_off |= htons(IP_MF); | |
1425 | offset += skb->len - nhoff - ihl; | |
1426 | tot_len = skb->len - nhoff; | |
1427 | } else if (skb_is_gso(skb)) { | |
802ab55a AD |
1428 | if (!fixedid) { |
1429 | iph->id = htons(id); | |
1430 | id += skb_shinfo(skb)->gso_segs; | |
1431 | } | |
07b26c94 SK |
1432 | |
1433 | if (gso_partial) | |
1434 | tot_len = skb_shinfo(skb)->gso_size + | |
1435 | SKB_GSO_CB(skb)->data_offset + | |
1436 | skb->head - (unsigned char *)iph; | |
1437 | else | |
1438 | tot_len = skb->len - nhoff; | |
802ab55a AD |
1439 | } else { |
1440 | if (!fixedid) | |
1441 | iph->id = htons(id++); | |
1442 | tot_len = skb->len - nhoff; | |
490ab081 | 1443 | } |
802ab55a | 1444 | iph->tot_len = htons(tot_len); |
47d27aad | 1445 | ip_send_check(iph); |
8c3a897b | 1446 | if (encap) |
3347c960 | 1447 | skb_reset_inner_headers(skb); |
3347c960 | 1448 | skb->network_header = (u8 *)iph - skb->head; |
c56cae23 | 1449 | skb_reset_mac_len(skb); |
f4c50d99 HX |
1450 | } while ((skb = skb->next)); |
1451 | ||
1452 | out: | |
1453 | return segs; | |
1454 | } | |
1455 | ||
418e897e WB |
1456 | static struct sk_buff *ipip_gso_segment(struct sk_buff *skb, |
1457 | netdev_features_t features) | |
1458 | { | |
1459 | if (!(skb_shinfo(skb)->gso_type & SKB_GSO_IPXIP4)) | |
1460 | return ERR_PTR(-EINVAL); | |
1461 | ||
1462 | return inet_gso_segment(skb, features); | |
1463 | } | |
1464 | ||
d4546c25 | 1465 | struct sk_buff *inet_gro_receive(struct list_head *head, struct sk_buff *skb) |
73cc19f1 | 1466 | { |
bca49f84 | 1467 | const struct net_offload *ops; |
d4546c25 | 1468 | struct sk_buff *pp = NULL; |
b71d1d42 | 1469 | const struct iphdr *iph; |
d4546c25 | 1470 | struct sk_buff *p; |
a5b1cf28 HX |
1471 | unsigned int hlen; |
1472 | unsigned int off; | |
1075f3f6 | 1473 | unsigned int id; |
73cc19f1 HX |
1474 | int flush = 1; |
1475 | int proto; | |
73cc19f1 | 1476 | |
a5b1cf28 HX |
1477 | off = skb_gro_offset(skb); |
1478 | hlen = off + sizeof(*iph); | |
35ffb665 RG |
1479 | iph = skb_gro_header(skb, hlen, off); |
1480 | if (unlikely(!iph)) | |
1481 | goto out; | |
73cc19f1 | 1482 | |
f9242b6b | 1483 | proto = iph->protocol; |
73cc19f1 | 1484 | |
bca49f84 | 1485 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 | 1486 | if (!ops || !ops->callbacks.gro_receive) |
fc1ca334 | 1487 | goto out; |
73cc19f1 | 1488 | |
a5ad24be | 1489 | if (*(u8 *)iph != 0x45) |
fc1ca334 | 1490 | goto out; |
73cc19f1 | 1491 | |
9b83e031 | 1492 | if (ip_is_fragment(iph)) |
fc1ca334 | 1493 | goto out; |
9b83e031 | 1494 | |
a9e050f4 | 1495 | if (unlikely(ip_fast_csum((u8 *)iph, 5))) |
fc1ca334 | 1496 | goto out; |
73cc19f1 | 1497 | |
b1a78b9b | 1498 | NAPI_GRO_CB(skb)->proto = proto; |
0eae88f3 | 1499 | id = ntohl(*(__be32 *)&iph->id); |
db8caf3d | 1500 | flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF)); |
1075f3f6 | 1501 | id >>= 16; |
73cc19f1 | 1502 | |
d4546c25 | 1503 | list_for_each_entry(p, head, list) { |
73cc19f1 | 1504 | struct iphdr *iph2; |
1530545e | 1505 | u16 flush_id; |
73cc19f1 HX |
1506 | |
1507 | if (!NAPI_GRO_CB(p)->same_flow) | |
1508 | continue; | |
1509 | ||
299603e8 JC |
1510 | iph2 = (struct iphdr *)(p->data + off); |
1511 | /* The above works because, with the exception of the top | |
1512 | * (inner most) layer, we only aggregate pkts with the same | |
1513 | * hdr length so all the hdrs we'll need to verify will start | |
1514 | * at the same offset. | |
1515 | */ | |
a5ad24be | 1516 | if ((iph->protocol ^ iph2->protocol) | |
0eae88f3 ED |
1517 | ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) | |
1518 | ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) { | |
73cc19f1 HX |
1519 | NAPI_GRO_CB(p)->same_flow = 0; |
1520 | continue; | |
1521 | } | |
1522 | ||
1523 | /* All fields must match except length and checksum. */ | |
1524 | NAPI_GRO_CB(p)->flush |= | |
a5ad24be | 1525 | (iph->ttl ^ iph2->ttl) | |
a9e050f4 | 1526 | (iph->tos ^ iph2->tos) | |
bf5a755f | 1527 | ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF)); |
73cc19f1 HX |
1528 | |
1529 | NAPI_GRO_CB(p)->flush |= flush; | |
1530545e AD |
1530 | |
1531 | /* We need to store of the IP ID check to be included later | |
1532 | * when we can verify that this packet does in fact belong | |
1533 | * to a given flow. | |
1534 | */ | |
1535 | flush_id = (u16)(id - ntohs(iph2->id)); | |
1536 | ||
1537 | /* This bit of code makes it much easier for us to identify | |
1538 | * the cases where we are doing atomic vs non-atomic IP ID | |
1539 | * checks. Specifically an atomic check can return IP ID | |
1540 | * values 0 - 0xFFFF, while a non-atomic check can only | |
1541 | * return 0 or 0xFFFF. | |
1542 | */ | |
1543 | if (!NAPI_GRO_CB(p)->is_atomic || | |
1544 | !(iph->frag_off & htons(IP_DF))) { | |
1545 | flush_id ^= NAPI_GRO_CB(p)->count; | |
1546 | flush_id = flush_id ? 0xFFFF : 0; | |
1547 | } | |
1548 | ||
1549 | /* If the previous IP ID value was based on an atomic | |
1550 | * datagram we can overwrite the value and ignore it. | |
1551 | */ | |
1552 | if (NAPI_GRO_CB(skb)->is_atomic) | |
1553 | NAPI_GRO_CB(p)->flush_id = flush_id; | |
1554 | else | |
1555 | NAPI_GRO_CB(p)->flush_id |= flush_id; | |
73cc19f1 HX |
1556 | } |
1557 | ||
1530545e | 1558 | NAPI_GRO_CB(skb)->is_atomic = !!(iph->frag_off & htons(IP_DF)); |
73cc19f1 | 1559 | NAPI_GRO_CB(skb)->flush |= flush; |
299603e8 JC |
1560 | skb_set_network_header(skb, off); |
1561 | /* The above will be needed by the transport layer if there is one | |
1562 | * immediately following this IP hdr. | |
1563 | */ | |
1564 | ||
2c804d0f ED |
1565 | /* Note : No need to call skb_gro_postpull_rcsum() here, |
1566 | * as we already checked checksum over ipv4 header was 0 | |
1567 | */ | |
86911732 HX |
1568 | skb_gro_pull(skb, sizeof(*iph)); |
1569 | skb_set_transport_header(skb, skb_gro_offset(skb)); | |
73cc19f1 | 1570 | |
028e0a47 PA |
1571 | pp = indirect_call_gro_receive(tcp4_gro_receive, udp4_gro_receive, |
1572 | ops->callbacks.gro_receive, head, skb); | |
73cc19f1 | 1573 | |
73cc19f1 | 1574 | out: |
5f114163 | 1575 | skb_gro_flush_final(skb, pp, flush); |
73cc19f1 HX |
1576 | |
1577 | return pp; | |
1578 | } | |
1579 | ||
d4546c25 DM |
1580 | static struct sk_buff *ipip_gro_receive(struct list_head *head, |
1581 | struct sk_buff *skb) | |
fac8e0f5 JG |
1582 | { |
1583 | if (NAPI_GRO_CB(skb)->encap_mark) { | |
1584 | NAPI_GRO_CB(skb)->flush = 1; | |
1585 | return NULL; | |
1586 | } | |
1587 | ||
1588 | NAPI_GRO_CB(skb)->encap_mark = 1; | |
1589 | ||
1590 | return inet_gro_receive(head, skb); | |
1591 | } | |
1592 | ||
822c8685 DD |
1593 | #define SECONDS_PER_DAY 86400 |
1594 | ||
1595 | /* inet_current_timestamp - Return IP network timestamp | |
1596 | * | |
1597 | * Return milliseconds since midnight in network byte order. | |
1598 | */ | |
1599 | __be32 inet_current_timestamp(void) | |
1600 | { | |
1601 | u32 secs; | |
1602 | u32 msecs; | |
1603 | struct timespec64 ts; | |
1604 | ||
1605 | ktime_get_real_ts64(&ts); | |
1606 | ||
1607 | /* Get secs since midnight. */ | |
1608 | (void)div_u64_rem(ts.tv_sec, SECONDS_PER_DAY, &secs); | |
1609 | /* Convert to msecs. */ | |
1610 | msecs = secs * MSEC_PER_SEC; | |
1611 | /* Convert nsec to msec. */ | |
1612 | msecs += (u32)ts.tv_nsec / NSEC_PER_MSEC; | |
1613 | ||
1614 | /* Convert to network byte order. */ | |
3ba9d300 | 1615 | return htonl(msecs); |
822c8685 DD |
1616 | } |
1617 | EXPORT_SYMBOL(inet_current_timestamp); | |
1618 | ||
f4713a3d WB |
1619 | int inet_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len) |
1620 | { | |
1621 | if (sk->sk_family == AF_INET) | |
1622 | return ip_recv_error(sk, msg, len, addr_len); | |
1623 | #if IS_ENABLED(CONFIG_IPV6) | |
1624 | if (sk->sk_family == AF_INET6) | |
1625 | return pingv6_ops.ipv6_recv_error(sk, msg, len, addr_len); | |
1626 | #endif | |
1627 | return -EINVAL; | |
1628 | } | |
1629 | ||
b8921ca8 | 1630 | int inet_gro_complete(struct sk_buff *skb, int nhoff) |
73cc19f1 | 1631 | { |
299603e8 | 1632 | struct iphdr *iph = (struct iphdr *)(skb->data + nhoff); |
bca49f84 | 1633 | const struct net_offload *ops; |
b1a78b9b | 1634 | __be16 totlen = iph->tot_len; |
f9242b6b | 1635 | int proto = iph->protocol; |
73cc19f1 | 1636 | int err = -ENOSYS; |
73cc19f1 | 1637 | |
294acf1c PA |
1638 | if (skb->encapsulation) { |
1639 | skb_set_inner_protocol(skb, cpu_to_be16(ETH_P_IP)); | |
c3caf119 | 1640 | skb_set_inner_network_header(skb, nhoff); |
294acf1c | 1641 | } |
c3caf119 | 1642 | |
b1a78b9b XL |
1643 | iph_set_totlen(iph, skb->len - nhoff); |
1644 | csum_replace2(&iph->check, totlen, iph->tot_len); | |
73cc19f1 | 1645 | |
bca49f84 | 1646 | ops = rcu_dereference(inet_offloads[proto]); |
f191a1d1 | 1647 | if (WARN_ON(!ops || !ops->callbacks.gro_complete)) |
627b94f7 | 1648 | goto out; |
73cc19f1 | 1649 | |
299603e8 JC |
1650 | /* Only need to add sizeof(*iph) to get to the next hdr below |
1651 | * because any hdr with option will have been flushed in | |
1652 | * inet_gro_receive(). | |
1653 | */ | |
028e0a47 PA |
1654 | err = INDIRECT_CALL_2(ops->callbacks.gro_complete, |
1655 | tcp4_gro_complete, udp4_gro_complete, | |
1656 | skb, nhoff + sizeof(*iph)); | |
73cc19f1 | 1657 | |
627b94f7 | 1658 | out: |
73cc19f1 HX |
1659 | return err; |
1660 | } | |
1661 | ||
b8cba75b JG |
1662 | static int ipip_gro_complete(struct sk_buff *skb, int nhoff) |
1663 | { | |
1664 | skb->encapsulation = 1; | |
7e13318d | 1665 | skb_shinfo(skb)->gso_type |= SKB_GSO_IPXIP4; |
b8cba75b JG |
1666 | return inet_gro_complete(skb, nhoff); |
1667 | } | |
1668 | ||
eee4fe4d | 1669 | int inet_ctl_sock_create(struct sock **sk, unsigned short family, |
5677242f DL |
1670 | unsigned short type, unsigned char protocol, |
1671 | struct net *net) | |
3d58b5fa | 1672 | { |
eee4fe4d | 1673 | struct socket *sock; |
26abe143 | 1674 | int rc = sock_create_kern(net, family, type, protocol, &sock); |
3d58b5fa DL |
1675 | |
1676 | if (rc == 0) { | |
eee4fe4d DL |
1677 | *sk = sock->sk; |
1678 | (*sk)->sk_allocation = GFP_ATOMIC; | |
1ac88557 | 1679 | (*sk)->sk_use_task_frag = false; |
3d58b5fa DL |
1680 | /* |
1681 | * Unhash it so that IP input processing does not even see it, | |
1682 | * we do not wish this socket to see incoming packets. | |
1683 | */ | |
eee4fe4d | 1684 | (*sk)->sk_prot->unhash(*sk); |
3d58b5fa DL |
1685 | } |
1686 | return rc; | |
1687 | } | |
3d58b5fa DL |
1688 | EXPORT_SYMBOL_GPL(inet_ctl_sock_create); |
1689 | ||
698365fa | 1690 | unsigned long snmp_fold_field(void __percpu *mib, int offt) |
5e0f0435 HX |
1691 | { |
1692 | unsigned long res = 0; | |
698365fa | 1693 | int i; |
5e0f0435 | 1694 | |
698365fa | 1695 | for_each_possible_cpu(i) |
c4c6bc31 | 1696 | res += snmp_get_cpu_field(mib, i, offt); |
5e0f0435 HX |
1697 | return res; |
1698 | } | |
1699 | EXPORT_SYMBOL_GPL(snmp_fold_field); | |
1700 | ||
4ce3c183 ED |
1701 | #if BITS_PER_LONG==32 |
1702 | ||
80ec1927 | 1703 | u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offt, |
c4c6bc31 R |
1704 | size_t syncp_offset) |
1705 | { | |
1706 | void *bhptr; | |
1707 | struct u64_stats_sync *syncp; | |
1708 | u64 v; | |
1709 | unsigned int start; | |
1710 | ||
1711 | bhptr = per_cpu_ptr(mib, cpu); | |
1712 | syncp = (struct u64_stats_sync *)(bhptr + syncp_offset); | |
1713 | do { | |
d120d1a6 | 1714 | start = u64_stats_fetch_begin(syncp); |
c4c6bc31 | 1715 | v = *(((u64 *)bhptr) + offt); |
d120d1a6 | 1716 | } while (u64_stats_fetch_retry(syncp, start)); |
c4c6bc31 R |
1717 | |
1718 | return v; | |
1719 | } | |
1720 | EXPORT_SYMBOL_GPL(snmp_get_cpu_field64); | |
1721 | ||
698365fa | 1722 | u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset) |
4ce3c183 ED |
1723 | { |
1724 | u64 res = 0; | |
1725 | int cpu; | |
1726 | ||
1727 | for_each_possible_cpu(cpu) { | |
d1bfc625 | 1728 | res += snmp_get_cpu_field64(mib, cpu, offt, syncp_offset); |
4ce3c183 ED |
1729 | } |
1730 | return res; | |
1731 | } | |
1732 | EXPORT_SYMBOL_GPL(snmp_fold_field64); | |
1733 | #endif | |
1734 | ||
1da177e4 | 1735 | #ifdef CONFIG_IP_MULTICAST |
32613090 | 1736 | static const struct net_protocol igmp_protocol = { |
1da177e4 LT |
1737 | .handler = igmp_rcv, |
1738 | }; | |
1739 | #endif | |
1740 | ||
11052589 | 1741 | static const struct net_protocol tcp_protocol = { |
41063e9d DM |
1742 | .handler = tcp_v4_rcv, |
1743 | .err_handler = tcp_v4_err, | |
41063e9d | 1744 | .no_policy = 1, |
8ed1dc44 | 1745 | .icmp_strict_tag_validation = 1, |
1da177e4 LT |
1746 | }; |
1747 | ||
11052589 | 1748 | static const struct net_protocol udp_protocol = { |
1da177e4 LT |
1749 | .handler = udp_rcv, |
1750 | .err_handler = udp_err, | |
1751 | .no_policy = 1, | |
1752 | }; | |
1753 | ||
32613090 | 1754 | static const struct net_protocol icmp_protocol = { |
1da177e4 | 1755 | .handler = icmp_rcv, |
5b052042 | 1756 | .err_handler = icmp_err, |
8b7817f3 | 1757 | .no_policy = 1, |
1da177e4 LT |
1758 | }; |
1759 | ||
9b4661bd PE |
1760 | static __net_init int ipv4_mib_init_net(struct net *net) |
1761 | { | |
827da44c JS |
1762 | int i; |
1763 | ||
698365fa WC |
1764 | net->mib.tcp_statistics = alloc_percpu(struct tcp_mib); |
1765 | if (!net->mib.tcp_statistics) | |
57ef42d5 | 1766 | goto err_tcp_mib; |
698365fa WC |
1767 | net->mib.ip_statistics = alloc_percpu(struct ipstats_mib); |
1768 | if (!net->mib.ip_statistics) | |
a20f5799 | 1769 | goto err_ip_mib; |
827da44c JS |
1770 | |
1771 | for_each_possible_cpu(i) { | |
1772 | struct ipstats_mib *af_inet_stats; | |
698365fa | 1773 | af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i); |
827da44c | 1774 | u64_stats_init(&af_inet_stats->syncp); |
827da44c JS |
1775 | } |
1776 | ||
698365fa WC |
1777 | net->mib.net_statistics = alloc_percpu(struct linux_mib); |
1778 | if (!net->mib.net_statistics) | |
61a7e260 | 1779 | goto err_net_mib; |
698365fa WC |
1780 | net->mib.udp_statistics = alloc_percpu(struct udp_mib); |
1781 | if (!net->mib.udp_statistics) | |
2f275f91 | 1782 | goto err_udp_mib; |
698365fa WC |
1783 | net->mib.udplite_statistics = alloc_percpu(struct udp_mib); |
1784 | if (!net->mib.udplite_statistics) | |
386019d3 | 1785 | goto err_udplite_mib; |
698365fa WC |
1786 | net->mib.icmp_statistics = alloc_percpu(struct icmp_mib); |
1787 | if (!net->mib.icmp_statistics) | |
b60538a0 | 1788 | goto err_icmp_mib; |
acb32ba3 ED |
1789 | net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib), |
1790 | GFP_KERNEL); | |
1791 | if (!net->mib.icmpmsg_statistics) | |
923c6586 | 1792 | goto err_icmpmsg_mib; |
57ef42d5 PE |
1793 | |
1794 | tcp_mib_init(net); | |
9b4661bd | 1795 | return 0; |
57ef42d5 | 1796 | |
923c6586 | 1797 | err_icmpmsg_mib: |
698365fa | 1798 | free_percpu(net->mib.icmp_statistics); |
b60538a0 | 1799 | err_icmp_mib: |
698365fa | 1800 | free_percpu(net->mib.udplite_statistics); |
386019d3 | 1801 | err_udplite_mib: |
698365fa | 1802 | free_percpu(net->mib.udp_statistics); |
2f275f91 | 1803 | err_udp_mib: |
698365fa | 1804 | free_percpu(net->mib.net_statistics); |
61a7e260 | 1805 | err_net_mib: |
698365fa | 1806 | free_percpu(net->mib.ip_statistics); |
a20f5799 | 1807 | err_ip_mib: |
698365fa | 1808 | free_percpu(net->mib.tcp_statistics); |
57ef42d5 PE |
1809 | err_tcp_mib: |
1810 | return -ENOMEM; | |
9b4661bd PE |
1811 | } |
1812 | ||
1813 | static __net_exit void ipv4_mib_exit_net(struct net *net) | |
1814 | { | |
acb32ba3 | 1815 | kfree(net->mib.icmpmsg_statistics); |
698365fa WC |
1816 | free_percpu(net->mib.icmp_statistics); |
1817 | free_percpu(net->mib.udplite_statistics); | |
1818 | free_percpu(net->mib.udp_statistics); | |
1819 | free_percpu(net->mib.net_statistics); | |
1820 | free_percpu(net->mib.ip_statistics); | |
1821 | free_percpu(net->mib.tcp_statistics); | |
fc518953 FW |
1822 | #ifdef CONFIG_MPTCP |
1823 | /* allocated on demand, see mptcp_init_sock() */ | |
1824 | free_percpu(net->mib.mptcp_statistics); | |
1825 | #endif | |
9b4661bd PE |
1826 | } |
1827 | ||
1828 | static __net_initdata struct pernet_operations ipv4_mib_ops = { | |
1829 | .init = ipv4_mib_init_net, | |
1830 | .exit = ipv4_mib_exit_net, | |
1831 | }; | |
1832 | ||
1da177e4 LT |
1833 | static int __init init_ipv4_mibs(void) |
1834 | { | |
d89cbbb1 | 1835 | return register_pernet_subsys(&ipv4_mib_ops); |
1da177e4 LT |
1836 | } |
1837 | ||
c9d8f1a6 CW |
1838 | static __net_init int inet_init_net(struct net *net) |
1839 | { | |
1840 | /* | |
1841 | * Set defaults for local port range | |
1842 | */ | |
1843 | seqlock_init(&net->ipv4.ip_local_ports.lock); | |
1844 | net->ipv4.ip_local_ports.range[0] = 32768; | |
07f4c900 | 1845 | net->ipv4.ip_local_ports.range[1] = 60999; |
ba6b918a CW |
1846 | |
1847 | seqlock_init(&net->ipv4.ping_group_range.lock); | |
1848 | /* | |
1849 | * Sane defaults - nobody may create ping sockets. | |
1850 | * Boot scripts should set this to distro-specific group. | |
1851 | */ | |
1852 | net->ipv4.ping_group_range.range[0] = make_kgid(&init_user_ns, 1); | |
1853 | net->ipv4.ping_group_range.range[1] = make_kgid(&init_user_ns, 0); | |
049bbf58 EG |
1854 | |
1855 | /* Default values for sysctl-controlled parameters. | |
1856 | * We set them here, in case sysctl is not compiled. | |
1857 | */ | |
1858 | net->ipv4.sysctl_ip_default_ttl = IPDEFTTL; | |
432e05d3 | 1859 | net->ipv4.sysctl_ip_fwd_update_priority = 1; |
049bbf58 EG |
1860 | net->ipv4.sysctl_ip_dynaddr = 0; |
1861 | net->ipv4.sysctl_ip_early_demux = 1; | |
dddb64bc | 1862 | net->ipv4.sysctl_udp_early_demux = 1; |
1863 | net->ipv4.sysctl_tcp_early_demux = 1; | |
4f80116d | 1864 | net->ipv4.sysctl_nexthop_compat_mode = 1; |
4548b683 KJ |
1865 | #ifdef CONFIG_SYSCTL |
1866 | net->ipv4.sysctl_ip_prot_sock = PROT_SOCK; | |
1867 | #endif | |
049bbf58 | 1868 | |
1714020e NB |
1869 | /* Some igmp sysctl, whose values are always used */ |
1870 | net->ipv4.sysctl_igmp_max_memberships = 20; | |
1871 | net->ipv4.sysctl_igmp_max_msf = 10; | |
1872 | /* IGMP reports for link-local multicast groups are enabled by default */ | |
1873 | net->ipv4.sysctl_igmp_llm_reports = 1; | |
1874 | net->ipv4.sysctl_igmp_qrv = 2; | |
1875 | ||
680aea08 AC |
1876 | net->ipv4.sysctl_fib_notify_on_flag_change = 0; |
1877 | ||
c9d8f1a6 CW |
1878 | return 0; |
1879 | } | |
1880 | ||
c9d8f1a6 CW |
1881 | static __net_initdata struct pernet_operations af_inet_ops = { |
1882 | .init = inet_init_net, | |
c9d8f1a6 CW |
1883 | }; |
1884 | ||
1885 | static int __init init_inet_pernet_ops(void) | |
1886 | { | |
1887 | return register_pernet_subsys(&af_inet_ops); | |
1888 | } | |
1889 | ||
1da177e4 | 1890 | static int ipv4_proc_init(void); |
1da177e4 | 1891 | |
30e224d7 HX |
1892 | /* |
1893 | * IP protocol layer initialiser | |
1894 | */ | |
1895 | ||
22061d80 VY |
1896 | static struct packet_offload ip_packet_offload __read_mostly = { |
1897 | .type = cpu_to_be16(ETH_P_IP), | |
f191a1d1 | 1898 | .callbacks = { |
f191a1d1 VY |
1899 | .gso_segment = inet_gso_segment, |
1900 | .gro_receive = inet_gro_receive, | |
1901 | .gro_complete = inet_gro_complete, | |
1902 | }, | |
30e224d7 HX |
1903 | }; |
1904 | ||
cb32f511 ED |
1905 | static const struct net_offload ipip_offload = { |
1906 | .callbacks = { | |
418e897e | 1907 | .gso_segment = ipip_gso_segment, |
fac8e0f5 | 1908 | .gro_receive = ipip_gro_receive, |
b8cba75b | 1909 | .gro_complete = ipip_gro_complete, |
cb32f511 ED |
1910 | }, |
1911 | }; | |
1912 | ||
93b1b31f TZ |
1913 | static int __init ipip_offload_init(void) |
1914 | { | |
1915 | return inet_add_offload(&ipip_offload, IPPROTO_IPIP); | |
1916 | } | |
1917 | ||
808a8f88 VY |
1918 | static int __init ipv4_offload_init(void) |
1919 | { | |
1920 | /* | |
1921 | * Add offloads | |
1922 | */ | |
da5bab07 | 1923 | if (udpv4_offload_init() < 0) |
808a8f88 | 1924 | pr_crit("%s: Cannot add UDP protocol offload\n", __func__); |
28850dc7 DB |
1925 | if (tcpv4_offload_init() < 0) |
1926 | pr_crit("%s: Cannot add TCP protocol offload\n", __func__); | |
93b1b31f TZ |
1927 | if (ipip_offload_init() < 0) |
1928 | pr_crit("%s: Cannot add IPIP protocol offload\n", __func__); | |
808a8f88 VY |
1929 | |
1930 | dev_add_offload(&ip_packet_offload); | |
1931 | return 0; | |
1932 | } | |
1933 | ||
1934 | fs_initcall(ipv4_offload_init); | |
1935 | ||
1936 | static struct packet_type ip_packet_type __read_mostly = { | |
1937 | .type = cpu_to_be16(ETH_P_IP), | |
1938 | .func = ip_rcv, | |
17266ee9 | 1939 | .list_func = ip_list_rcv, |
808a8f88 VY |
1940 | }; |
1941 | ||
1da177e4 LT |
1942 | static int __init inet_init(void) |
1943 | { | |
1da177e4 LT |
1944 | struct inet_protosw *q; |
1945 | struct list_head *r; | |
b6246f4d | 1946 | int rc; |
1da177e4 | 1947 | |
b4772ef8 | 1948 | sock_skb_cb_check_size(sizeof(struct inet_skb_parm)); |
1da177e4 | 1949 | |
0daf07e5 ED |
1950 | raw_hashinfo_init(&raw_v4_hashinfo); |
1951 | ||
1da177e4 LT |
1952 | rc = proto_register(&tcp_prot, 1); |
1953 | if (rc) | |
122ff243 | 1954 | goto out; |
1da177e4 LT |
1955 | |
1956 | rc = proto_register(&udp_prot, 1); | |
1957 | if (rc) | |
1958 | goto out_unregister_tcp_proto; | |
1959 | ||
1960 | rc = proto_register(&raw_prot, 1); | |
1961 | if (rc) | |
1962 | goto out_unregister_udp_proto; | |
1963 | ||
c319b4d7 VK |
1964 | rc = proto_register(&ping_prot, 1); |
1965 | if (rc) | |
1966 | goto out_unregister_raw_proto; | |
1967 | ||
1da177e4 | 1968 | /* |
e905a9ed | 1969 | * Tell SOCKET that we are alive... |
1da177e4 LT |
1970 | */ |
1971 | ||
e905a9ed | 1972 | (void)sock_register(&inet_family_ops); |
1da177e4 | 1973 | |
bd7b1533 AV |
1974 | #ifdef CONFIG_SYSCTL |
1975 | ip_static_sysctl_init(); | |
1976 | #endif | |
1977 | ||
1da177e4 LT |
1978 | /* |
1979 | * Add all the base protocols. | |
1980 | */ | |
1981 | ||
1982 | if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0) | |
058bd4d2 | 1983 | pr_crit("%s: Cannot add ICMP protocol\n", __func__); |
1da177e4 | 1984 | if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0) |
058bd4d2 | 1985 | pr_crit("%s: Cannot add UDP protocol\n", __func__); |
1da177e4 | 1986 | if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0) |
058bd4d2 | 1987 | pr_crit("%s: Cannot add TCP protocol\n", __func__); |
1da177e4 LT |
1988 | #ifdef CONFIG_IP_MULTICAST |
1989 | if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0) | |
058bd4d2 | 1990 | pr_crit("%s: Cannot add IGMP protocol\n", __func__); |
1da177e4 LT |
1991 | #endif |
1992 | ||
1993 | /* Register the socket-side information for inet_create. */ | |
1994 | for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r) | |
1995 | INIT_LIST_HEAD(r); | |
1996 | ||
1997 | for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q) | |
1998 | inet_register_protosw(q); | |
1999 | ||
2000 | /* | |
2001 | * Set the ARP module up | |
2002 | */ | |
2003 | ||
2004 | arp_init(); | |
2005 | ||
e905a9ed YH |
2006 | /* |
2007 | * Set the IP module up | |
2008 | */ | |
1da177e4 LT |
2009 | |
2010 | ip_init(); | |
2011 | ||
e22e45fc MS |
2012 | /* Initialise per-cpu ipv4 mibs */ |
2013 | if (init_ipv4_mibs()) | |
2014 | panic("%s: Cannot init ipv4 mibs\n", __func__); | |
2015 | ||
1da177e4 LT |
2016 | /* Setup TCP slab cache for open requests. */ |
2017 | tcp_init(); | |
2018 | ||
95766fff HA |
2019 | /* Setup UDP memory threshold */ |
2020 | udp_init(); | |
2021 | ||
ba4e58ec GR |
2022 | /* Add UDP-Lite (RFC 3828) */ |
2023 | udplite4_register(); | |
1da177e4 | 2024 | |
6897445f MM |
2025 | raw_init(); |
2026 | ||
c319b4d7 VK |
2027 | ping_init(); |
2028 | ||
1da177e4 LT |
2029 | /* |
2030 | * Set the ICMP layer up | |
2031 | */ | |
2032 | ||
a5710d65 DL |
2033 | if (icmp_init() < 0) |
2034 | panic("Failed to create the ICMP control socket.\n"); | |
1da177e4 LT |
2035 | |
2036 | /* | |
2037 | * Initialise the multicast router | |
2038 | */ | |
2039 | #if defined(CONFIG_IP_MROUTE) | |
03d2f897 | 2040 | if (ip_mr_init()) |
058bd4d2 | 2041 | pr_crit("%s: Cannot init ipv4 mroute\n", __func__); |
1da177e4 | 2042 | #endif |
c9d8f1a6 CW |
2043 | |
2044 | if (init_inet_pernet_ops()) | |
2045 | pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__); | |
e905a9ed | 2046 | |
1da177e4 LT |
2047 | ipv4_proc_init(); |
2048 | ||
2049 | ipfrag_init(); | |
2050 | ||
30e224d7 HX |
2051 | dev_add_pack(&ip_packet_type); |
2052 | ||
045a0fa0 TG |
2053 | ip_tunnel_core_init(); |
2054 | ||
1da177e4 LT |
2055 | rc = 0; |
2056 | out: | |
2057 | return rc; | |
c319b4d7 VK |
2058 | out_unregister_raw_proto: |
2059 | proto_unregister(&raw_prot); | |
1da177e4 LT |
2060 | out_unregister_udp_proto: |
2061 | proto_unregister(&udp_prot); | |
321efff7 OK |
2062 | out_unregister_tcp_proto: |
2063 | proto_unregister(&tcp_prot); | |
1da177e4 LT |
2064 | goto out; |
2065 | } | |
2066 | ||
a536e077 | 2067 | fs_initcall(inet_init); |
1da177e4 LT |
2068 | |
2069 | /* ------------------------------------------------------------------------ */ | |
2070 | ||
2071 | #ifdef CONFIG_PROC_FS | |
1da177e4 LT |
2072 | static int __init ipv4_proc_init(void) |
2073 | { | |
2074 | int rc = 0; | |
2075 | ||
2076 | if (raw_proc_init()) | |
2077 | goto out_raw; | |
2078 | if (tcp4_proc_init()) | |
2079 | goto out_tcp; | |
2080 | if (udp4_proc_init()) | |
2081 | goto out_udp; | |
c319b4d7 VK |
2082 | if (ping_proc_init()) |
2083 | goto out_ping; | |
1da177e4 LT |
2084 | if (ip_misc_proc_init()) |
2085 | goto out_misc; | |
2086 | out: | |
2087 | return rc; | |
2088 | out_misc: | |
c319b4d7 VK |
2089 | ping_proc_exit(); |
2090 | out_ping: | |
1da177e4 LT |
2091 | udp4_proc_exit(); |
2092 | out_udp: | |
2093 | tcp4_proc_exit(); | |
2094 | out_tcp: | |
2095 | raw_proc_exit(); | |
2096 | out_raw: | |
2097 | rc = -ENOMEM; | |
2098 | goto out; | |
2099 | } | |
2100 | ||
2101 | #else /* CONFIG_PROC_FS */ | |
2102 | static int __init ipv4_proc_init(void) | |
2103 | { | |
2104 | return 0; | |
2105 | } | |
2106 | #endif /* CONFIG_PROC_FS */ |