12874c5fdSThomas Gleixner // SPDX-License-Identifier: GPL-2.0-or-later 21da177e4SLinus Torvalds /* 31da177e4SLinus Torvalds * INET An implementation of the TCP/IP protocol suite for the LINUX 41da177e4SLinus Torvalds * operating system. INET is implemented using the BSD Socket 51da177e4SLinus Torvalds * interface as the means of communication with the user level. 61da177e4SLinus Torvalds * 71da177e4SLinus Torvalds * PF_INET protocol family socket handler. 81da177e4SLinus Torvalds * 902c30a84SJesper Juhl * Authors: Ross Biro 101da177e4SLinus Torvalds * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 111da177e4SLinus Torvalds * Florian La Roche, <flla@stud.uni-sb.de> 121da177e4SLinus Torvalds * Alan Cox, <A.Cox@swansea.ac.uk> 131da177e4SLinus Torvalds * 141da177e4SLinus Torvalds * Changes (see also sock.c) 151da177e4SLinus Torvalds * 161da177e4SLinus Torvalds * piggy, 171da177e4SLinus Torvalds * Karl Knutson : Socket protocol table 181da177e4SLinus Torvalds * A.N.Kuznetsov : Socket death error in accept(). 191da177e4SLinus Torvalds * John Richardson : Fix non blocking error in connect() 201da177e4SLinus Torvalds * so sockets that fail to connect 211da177e4SLinus Torvalds * don't return -EINPROGRESS. 221da177e4SLinus Torvalds * Alan Cox : Asynchronous I/O support 231da177e4SLinus Torvalds * Alan Cox : Keep correct socket pointer on sock 241da177e4SLinus Torvalds * structures 251da177e4SLinus Torvalds * when accept() ed 261da177e4SLinus Torvalds * Alan Cox : Semantics of SO_LINGER aren't state 271da177e4SLinus Torvalds * moved to close when you look carefully. 281da177e4SLinus Torvalds * With this fixed and the accept bug fixed 291da177e4SLinus Torvalds * some RPC stuff seems happier. 301da177e4SLinus Torvalds * Niibe Yutaka : 4.4BSD style write async I/O 311da177e4SLinus Torvalds * Alan Cox, 321da177e4SLinus Torvalds * Tony Gale : Fixed reuse semantics. 331da177e4SLinus Torvalds * Alan Cox : bind() shouldn't abort existing but dead 341da177e4SLinus Torvalds * sockets. Stops FTP netin:.. I hope. 351da177e4SLinus Torvalds * Alan Cox : bind() works correctly for RAW sockets. 361da177e4SLinus Torvalds * Note that FreeBSD at least was broken 371da177e4SLinus Torvalds * in this respect so be careful with 381da177e4SLinus Torvalds * compatibility tests... 391da177e4SLinus Torvalds * Alan Cox : routing cache support 401da177e4SLinus Torvalds * Alan Cox : memzero the socket structure for 411da177e4SLinus Torvalds * compactness. 421da177e4SLinus Torvalds * Matt Day : nonblock connect error handler 431da177e4SLinus Torvalds * Alan Cox : Allow large numbers of pending sockets 441da177e4SLinus Torvalds * (eg for big web sites), but only if 451da177e4SLinus Torvalds * specifically application requested. 461da177e4SLinus Torvalds * Alan Cox : New buffering throughout IP. Used 471da177e4SLinus Torvalds * dumbly. 481da177e4SLinus Torvalds * Alan Cox : New buffering now used smartly. 491da177e4SLinus Torvalds * Alan Cox : BSD rather than common sense 501da177e4SLinus Torvalds * interpretation of listen. 511da177e4SLinus Torvalds * Germano Caronni : Assorted small races. 521da177e4SLinus Torvalds * Alan Cox : sendmsg/recvmsg basic support. 531da177e4SLinus Torvalds * Alan Cox : Only sendmsg/recvmsg now supported. 541da177e4SLinus Torvalds * Alan Cox : Locked down bind (see security list). 551da177e4SLinus Torvalds * Alan Cox : Loosened bind a little. 561da177e4SLinus Torvalds * Mike McLagan : ADD/DEL DLCI Ioctls 571da177e4SLinus Torvalds * Willy Konynenberg : Transparent proxying support. 581da177e4SLinus Torvalds * David S. Miller : New socket lookup architecture. 591da177e4SLinus Torvalds * Some other random speedups. 601da177e4SLinus Torvalds * Cyrus Durgin : Cleaned up file for kmod hacks. 611da177e4SLinus Torvalds * Andi Kleen : Fix inet_stream_connect TCP race. 621da177e4SLinus Torvalds */ 631da177e4SLinus Torvalds 64afd46503SJoe Perches #define pr_fmt(fmt) "IPv4: " fmt 65afd46503SJoe Perches 66f4c50d99SHerbert Xu #include <linux/err.h> 671da177e4SLinus Torvalds #include <linux/errno.h> 681da177e4SLinus Torvalds #include <linux/types.h> 691da177e4SLinus Torvalds #include <linux/socket.h> 701da177e4SLinus Torvalds #include <linux/in.h> 711da177e4SLinus Torvalds #include <linux/kernel.h> 72d3fc0353SPaul Gortmaker #include <linux/kmod.h> 731da177e4SLinus Torvalds #include <linux/sched.h> 741da177e4SLinus Torvalds #include <linux/timer.h> 751da177e4SLinus Torvalds #include <linux/string.h> 761da177e4SLinus Torvalds #include <linux/sockios.h> 771da177e4SLinus Torvalds #include <linux/net.h> 784fc268d2SRandy Dunlap #include <linux/capability.h> 791da177e4SLinus Torvalds #include <linux/fcntl.h> 801da177e4SLinus Torvalds #include <linux/mm.h> 811da177e4SLinus Torvalds #include <linux/interrupt.h> 821da177e4SLinus Torvalds #include <linux/stat.h> 831da177e4SLinus Torvalds #include <linux/init.h> 841da177e4SLinus Torvalds #include <linux/poll.h> 851da177e4SLinus Torvalds #include <linux/netfilter_ipv4.h> 86b3da2cf3SDavid S. Miller #include <linux/random.h> 875a0e3ad6STejun Heo #include <linux/slab.h> 881da177e4SLinus Torvalds 897c0f6ba6SLinus Torvalds #include <linux/uaccess.h> 901da177e4SLinus Torvalds 911da177e4SLinus Torvalds #include <linux/inet.h> 921da177e4SLinus Torvalds #include <linux/igmp.h> 9314c85021SArnaldo Carvalho de Melo #include <linux/inetdevice.h> 941da177e4SLinus Torvalds #include <linux/netdevice.h> 9573cc19f1SHerbert Xu #include <net/checksum.h> 961da177e4SLinus Torvalds #include <net/ip.h> 971da177e4SLinus Torvalds #include <net/protocol.h> 981da177e4SLinus Torvalds #include <net/arp.h> 991da177e4SLinus Torvalds #include <net/route.h> 1001da177e4SLinus Torvalds #include <net/ip_fib.h> 101295f7324SArnaldo Carvalho de Melo #include <net/inet_connection_sock.h> 1024721031cSEric Dumazet #include <net/gro.h> 103d457a0e3SEric Dumazet #include <net/gso.h> 1041da177e4SLinus Torvalds #include <net/tcp.h> 1051da177e4SLinus Torvalds #include <net/udp.h> 106ba4e58ecSGerrit Renker #include <net/udplite.h> 107c319b4d7SVasiliy Kulikov #include <net/ping.h> 1081da177e4SLinus Torvalds #include <linux/skbuff.h> 1091da177e4SLinus Torvalds #include <net/sock.h> 1101da177e4SLinus Torvalds #include <net/raw.h> 1111da177e4SLinus Torvalds #include <net/icmp.h> 1121da177e4SLinus Torvalds #include <net/inet_common.h> 113045a0fa0SThomas Graf #include <net/ip_tunnels.h> 1141da177e4SLinus Torvalds #include <net/xfrm.h> 1159b4661bdSPavel Emelyanov #include <net/net_namespace.h> 116aebda156SEric Dumazet #include <net/secure_seq.h> 1171da177e4SLinus Torvalds #ifdef CONFIG_IP_MROUTE 1181da177e4SLinus Torvalds #include <linux/mroute.h> 1191da177e4SLinus Torvalds #endif 1203236b004SDavid Ahern #include <net/l3mdev.h> 121dc13c876SChristoph Hellwig #include <net/compat.h> 1221da177e4SLinus Torvalds 123563e0bb0SYafang Shao #include <trace/events/sock.h> 1241da177e4SLinus Torvalds 1251da177e4SLinus Torvalds /* The inetsw table contains everything that inet_create needs to 1261da177e4SLinus Torvalds * build a new socket. 1271da177e4SLinus Torvalds */ 1281da177e4SLinus Torvalds static struct list_head inetsw[SOCK_MAX]; 1291da177e4SLinus Torvalds static DEFINE_SPINLOCK(inetsw_lock); 1301da177e4SLinus Torvalds 1311da177e4SLinus Torvalds /* New destruction routine */ 1321da177e4SLinus Torvalds 1331da177e4SLinus Torvalds void inet_sock_destruct(struct sock *sk) 1341da177e4SLinus Torvalds { 1351da177e4SLinus Torvalds struct inet_sock *inet = inet_sk(sk); 1361da177e4SLinus Torvalds 1371da177e4SLinus Torvalds __skb_queue_purge(&sk->sk_receive_queue); 1381da177e4SLinus Torvalds __skb_queue_purge(&sk->sk_error_queue); 1391da177e4SLinus Torvalds 1402bb2f5fbSWei Wang sk_mem_reclaim_final(sk); 14195766fffSHideo Aoki 1421da177e4SLinus Torvalds if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) { 1433d1427f8SEric Dumazet pr_err("Attempt to release TCP socket in state %d %p\n", 1441da177e4SLinus Torvalds sk->sk_state, sk); 1451da177e4SLinus Torvalds return; 1461da177e4SLinus Torvalds } 1471da177e4SLinus Torvalds if (!sock_flag(sk, SOCK_DEAD)) { 1483d1427f8SEric Dumazet pr_err("Attempt to release alive inet socket %p\n", sk); 1491da177e4SLinus Torvalds return; 1501da177e4SLinus Torvalds } 1511da177e4SLinus Torvalds 1523e7f2b8dSEric Dumazet WARN_ON_ONCE(atomic_read(&sk->sk_rmem_alloc)); 1533e7f2b8dSEric Dumazet WARN_ON_ONCE(refcount_read(&sk->sk_wmem_alloc)); 1543e7f2b8dSEric Dumazet WARN_ON_ONCE(sk->sk_wmem_queued); 1553e7f2b8dSEric Dumazet WARN_ON_ONCE(sk_forward_alloc_get(sk)); 1561da177e4SLinus Torvalds 157f6d8bd05SEric Dumazet kfree(rcu_dereference_protected(inet->inet_opt, 1)); 1585869b8faSXin Long dst_release(rcu_dereference_protected(sk->sk_dst_cache, 1)); 1598f905c0eSEric Dumazet dst_release(rcu_dereference_protected(sk->sk_rx_dst, 1)); 1601da177e4SLinus Torvalds } 1613d1427f8SEric Dumazet EXPORT_SYMBOL(inet_sock_destruct); 1621da177e4SLinus Torvalds 1631da177e4SLinus Torvalds /* 1641da177e4SLinus Torvalds * The routines beyond this point handle the behaviour of an AF_INET 1651da177e4SLinus Torvalds * socket object. Mostly it punts to the subprotocols of IP to do 1661da177e4SLinus Torvalds * the work. 1671da177e4SLinus Torvalds */ 1681da177e4SLinus Torvalds 1691da177e4SLinus Torvalds /* 1701da177e4SLinus Torvalds * Automatically bind an unbound socket. 1711da177e4SLinus Torvalds */ 1721da177e4SLinus Torvalds 1731da177e4SLinus Torvalds static int inet_autobind(struct sock *sk) 1741da177e4SLinus Torvalds { 1751da177e4SLinus Torvalds struct inet_sock *inet; 1761da177e4SLinus Torvalds /* We may need to bind the socket. */ 1771da177e4SLinus Torvalds lock_sock(sk); 1781da177e4SLinus Torvalds inet = inet_sk(sk); 179c720c7e8SEric Dumazet if (!inet->inet_num) { 1801da177e4SLinus Torvalds if (sk->sk_prot->get_port(sk, 0)) { 1811da177e4SLinus Torvalds release_sock(sk); 1821da177e4SLinus Torvalds return -EAGAIN; 1831da177e4SLinus Torvalds } 184c720c7e8SEric Dumazet inet->inet_sport = htons(inet->inet_num); 1851da177e4SLinus Torvalds } 1861da177e4SLinus Torvalds release_sock(sk); 1871da177e4SLinus Torvalds return 0; 1881da177e4SLinus Torvalds } 1891da177e4SLinus Torvalds 1901da177e4SLinus Torvalds /* 1911da177e4SLinus Torvalds * Move a socket into listening state. 1921da177e4SLinus Torvalds */ 1931da177e4SLinus Torvalds int inet_listen(struct socket *sock, int backlog) 1941da177e4SLinus Torvalds { 1951da177e4SLinus Torvalds struct sock *sk = sock->sk; 1961da177e4SLinus Torvalds unsigned char old_state; 197e1cfcbe8SHaishuang Yan int err, tcp_fastopen; 1981da177e4SLinus Torvalds 1991da177e4SLinus Torvalds lock_sock(sk); 2001da177e4SLinus Torvalds 2011da177e4SLinus Torvalds err = -EINVAL; 2021da177e4SLinus Torvalds if (sock->state != SS_UNCONNECTED || sock->type != SOCK_STREAM) 2031da177e4SLinus Torvalds goto out; 2041da177e4SLinus Torvalds 2051da177e4SLinus Torvalds old_state = sk->sk_state; 2061da177e4SLinus Torvalds if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN))) 2071da177e4SLinus Torvalds goto out; 2081da177e4SLinus Torvalds 209099ecf59SEric Dumazet WRITE_ONCE(sk->sk_max_ack_backlog, backlog); 2101da177e4SLinus Torvalds /* Really, if the socket is already in listen state 2111da177e4SLinus Torvalds * we can only allow the backlog to be adjusted. 2121da177e4SLinus Torvalds */ 2131da177e4SLinus Torvalds if (old_state != TCP_LISTEN) { 214cebc5cbaSYuchung Cheng /* Enable TFO w/o requiring TCP_FASTOPEN socket option. 2158336886fSJerry Chu * Note that only TCP sockets (SOCK_STREAM) will reach here. 216cebc5cbaSYuchung Cheng * Also fastopen backlog may already been set via the option 217cebc5cbaSYuchung Cheng * because the socket was in TCP_LISTEN state previously but 218cebc5cbaSYuchung Cheng * was shutdown() rather than close(). 2198336886fSJerry Chu */ 2205a542133SKuniyuki Iwashima tcp_fastopen = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_fastopen); 221e1cfcbe8SHaishuang Yan if ((tcp_fastopen & TFO_SERVER_WO_SOCKOPT1) && 222e1cfcbe8SHaishuang Yan (tcp_fastopen & TFO_SERVER_ENABLE) && 2230536fcc0SEric Dumazet !inet_csk(sk)->icsk_accept_queue.fastopenq.max_qlen) { 2240536fcc0SEric Dumazet fastopen_queue_tune(sk, backlog); 22543713848SHaishuang Yan tcp_fastopen_init_key_once(sock_net(sk)); 2268336886fSJerry Chu } 227cebc5cbaSYuchung Cheng 228e7049395SKuniyuki Iwashima err = inet_csk_listen_start(sk); 2291da177e4SLinus Torvalds if (err) 2301da177e4SLinus Torvalds goto out; 231f333ee0cSAndrey Ignatov tcp_call_bpf(sk, BPF_SOCK_OPS_TCP_LISTEN_CB, 0, NULL); 2321da177e4SLinus Torvalds } 2331da177e4SLinus Torvalds err = 0; 2341da177e4SLinus Torvalds 2351da177e4SLinus Torvalds out: 2361da177e4SLinus Torvalds release_sock(sk); 2371da177e4SLinus Torvalds return err; 2381da177e4SLinus Torvalds } 2393d1427f8SEric Dumazet EXPORT_SYMBOL(inet_listen); 2401da177e4SLinus Torvalds 2411da177e4SLinus Torvalds /* 2421da177e4SLinus Torvalds * Create an inet socket. 2431da177e4SLinus Torvalds */ 2441da177e4SLinus Torvalds 2453f378b68SEric Paris static int inet_create(struct net *net, struct socket *sock, int protocol, 2463f378b68SEric Paris int kern) 2471da177e4SLinus Torvalds { 2481da177e4SLinus Torvalds struct sock *sk; 2491da177e4SLinus Torvalds struct inet_protosw *answer; 2501da177e4SLinus Torvalds struct inet_sock *inet; 2511da177e4SLinus Torvalds struct proto *answer_prot; 2521da177e4SLinus Torvalds unsigned char answer_flags; 253bb97d31fSArnaldo Carvalho de Melo int try_loading_module = 0; 25486c8f9d1SHerbert Xu int err; 2551da177e4SLinus Torvalds 25679462ad0SHannes Frederic Sowa if (protocol < 0 || protocol >= IPPROTO_MAX) 25779462ad0SHannes Frederic Sowa return -EINVAL; 25879462ad0SHannes Frederic Sowa 2591da177e4SLinus Torvalds sock->state = SS_UNCONNECTED; 2601da177e4SLinus Torvalds 2611da177e4SLinus Torvalds /* Look for the requested type/protocol pair. */ 262bb97d31fSArnaldo Carvalho de Melo lookup_protocol: 26386c8f9d1SHerbert Xu err = -ESOCKTNOSUPPORT; 2641da177e4SLinus Torvalds rcu_read_lock(); 265696adfe8SPaul E. McKenney list_for_each_entry_rcu(answer, &inetsw[sock->type], list) { 2661da177e4SLinus Torvalds 267696adfe8SPaul E. McKenney err = 0; 2681da177e4SLinus Torvalds /* Check the non-wild match. */ 2691da177e4SLinus Torvalds if (protocol == answer->protocol) { 2701da177e4SLinus Torvalds if (protocol != IPPROTO_IP) 2711da177e4SLinus Torvalds break; 2721da177e4SLinus Torvalds } else { 2731da177e4SLinus Torvalds /* Check for the two wild cases. */ 2741da177e4SLinus Torvalds if (IPPROTO_IP == protocol) { 2751da177e4SLinus Torvalds protocol = answer->protocol; 2761da177e4SLinus Torvalds break; 2771da177e4SLinus Torvalds } 2781da177e4SLinus Torvalds if (IPPROTO_IP == answer->protocol) 2791da177e4SLinus Torvalds break; 2801da177e4SLinus Torvalds } 28186c8f9d1SHerbert Xu err = -EPROTONOSUPPORT; 2821da177e4SLinus Torvalds } 2831da177e4SLinus Torvalds 284696adfe8SPaul E. McKenney if (unlikely(err)) { 285bb97d31fSArnaldo Carvalho de Melo if (try_loading_module < 2) { 286bb97d31fSArnaldo Carvalho de Melo rcu_read_unlock(); 287bb97d31fSArnaldo Carvalho de Melo /* 288bb97d31fSArnaldo Carvalho de Melo * Be more specific, e.g. net-pf-2-proto-132-type-1 289bb97d31fSArnaldo Carvalho de Melo * (net-pf-PF_INET-proto-IPPROTO_SCTP-type-SOCK_STREAM) 290bb97d31fSArnaldo Carvalho de Melo */ 291bb97d31fSArnaldo Carvalho de Melo if (++try_loading_module == 1) 292bb97d31fSArnaldo Carvalho de Melo request_module("net-pf-%d-proto-%d-type-%d", 293bb97d31fSArnaldo Carvalho de Melo PF_INET, protocol, sock->type); 294bb97d31fSArnaldo Carvalho de Melo /* 295bb97d31fSArnaldo Carvalho de Melo * Fall back to generic, e.g. net-pf-2-proto-132 296bb97d31fSArnaldo Carvalho de Melo * (net-pf-PF_INET-proto-IPPROTO_SCTP) 297bb97d31fSArnaldo Carvalho de Melo */ 298bb97d31fSArnaldo Carvalho de Melo else 299bb97d31fSArnaldo Carvalho de Melo request_module("net-pf-%d-proto-%d", 300bb97d31fSArnaldo Carvalho de Melo PF_INET, protocol); 301bb97d31fSArnaldo Carvalho de Melo goto lookup_protocol; 302bb97d31fSArnaldo Carvalho de Melo } else 3031da177e4SLinus Torvalds goto out_rcu_unlock; 304bb97d31fSArnaldo Carvalho de Melo } 305bb97d31fSArnaldo Carvalho de Melo 3061da177e4SLinus Torvalds err = -EPERM; 30752e804c6SEric W. Biederman if (sock->type == SOCK_RAW && !kern && 30852e804c6SEric W. Biederman !ns_capable(net->user_ns, CAP_NET_RAW)) 3091da177e4SLinus Torvalds goto out_rcu_unlock; 3101da177e4SLinus Torvalds 3111da177e4SLinus Torvalds sock->ops = answer->ops; 3121da177e4SLinus Torvalds answer_prot = answer->prot; 3131da177e4SLinus Torvalds answer_flags = answer->flags; 3141da177e4SLinus Torvalds rcu_read_unlock(); 3151da177e4SLinus Torvalds 31651456b29SIan Morris WARN_ON(!answer_prot->slab); 3171da177e4SLinus Torvalds 318ca746c55SZheng Yongjun err = -ENOMEM; 31911aa9c28SEric W. Biederman sk = sk_alloc(net, PF_INET, GFP_KERNEL, answer_prot, kern); 32051456b29SIan Morris if (!sk) 3211da177e4SLinus Torvalds goto out; 3221da177e4SLinus Torvalds 3231da177e4SLinus Torvalds err = 0; 3241da177e4SLinus Torvalds if (INET_PROTOSW_REUSE & answer_flags) 3254a17fd52SPavel Emelyanov sk->sk_reuse = SK_CAN_REUSE; 3261da177e4SLinus Torvalds 3271da177e4SLinus Torvalds inet = inet_sk(sk); 328469de9b9SPaul Moore inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; 3291da177e4SLinus Torvalds 3307b2ff18eSJiri Olsa inet->nodefrag = 0; 3317b2ff18eSJiri Olsa 3321da177e4SLinus Torvalds if (SOCK_RAW == sock->type) { 333c720c7e8SEric Dumazet inet->inet_num = protocol; 3341da177e4SLinus Torvalds if (IPPROTO_RAW == protocol) 3351da177e4SLinus Torvalds inet->hdrincl = 1; 3361da177e4SLinus Torvalds } 3371da177e4SLinus Torvalds 3380968d2a4SKuniyuki Iwashima if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc)) 3391da177e4SLinus Torvalds inet->pmtudisc = IP_PMTUDISC_DONT; 3401da177e4SLinus Torvalds else 3411da177e4SLinus Torvalds inet->pmtudisc = IP_PMTUDISC_WANT; 3421da177e4SLinus Torvalds 343c720c7e8SEric Dumazet inet->inet_id = 0; 3441da177e4SLinus Torvalds 3451da177e4SLinus Torvalds sock_init_data(sock, sk); 3461da177e4SLinus Torvalds 3471da177e4SLinus Torvalds sk->sk_destruct = inet_sock_destruct; 3481da177e4SLinus Torvalds sk->sk_protocol = protocol; 3491da177e4SLinus Torvalds sk->sk_backlog_rcv = sk->sk_prot->backlog_rcv; 350c11204c7SKevin Yang sk->sk_txrehash = READ_ONCE(net->core.sysctl_txrehash); 3511da177e4SLinus Torvalds 3521da177e4SLinus Torvalds inet->uc_ttl = -1; 3531da177e4SLinus Torvalds inet->mc_loop = 1; 3541da177e4SLinus Torvalds inet->mc_ttl = 1; 355f771bef9SNivedita Singhvi inet->mc_all = 1; 3561da177e4SLinus Torvalds inet->mc_index = 0; 3571da177e4SLinus Torvalds inet->mc_list = NULL; 3584c507d28SJiri Benc inet->rcv_tos = 0; 3591da177e4SLinus Torvalds 360c720c7e8SEric Dumazet if (inet->inet_num) { 3611da177e4SLinus Torvalds /* It assumes that any protocol which allows 3621da177e4SLinus Torvalds * the user to assign a number at socket 3631da177e4SLinus Torvalds * creation time automatically 3641da177e4SLinus Torvalds * shares. 3651da177e4SLinus Torvalds */ 366c720c7e8SEric Dumazet inet->inet_sport = htons(inet->inet_num); 3671da177e4SLinus Torvalds /* Add to protocol hash chains. */ 368086c653fSCraig Gallek err = sk->sk_prot->hash(sk); 369086c653fSCraig Gallek if (err) { 370086c653fSCraig Gallek sk_common_release(sk); 371086c653fSCraig Gallek goto out; 372086c653fSCraig Gallek } 3731da177e4SLinus Torvalds } 3741da177e4SLinus Torvalds 3751da177e4SLinus Torvalds if (sk->sk_prot->init) { 3761da177e4SLinus Torvalds err = sk->sk_prot->init(sk); 37761023658SDavid Ahern if (err) { 3781da177e4SLinus Torvalds sk_common_release(sk); 37961023658SDavid Ahern goto out; 38061023658SDavid Ahern } 38161023658SDavid Ahern } 38261023658SDavid Ahern 38361023658SDavid Ahern if (!kern) { 38461023658SDavid Ahern err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk); 38561023658SDavid Ahern if (err) { 38661023658SDavid Ahern sk_common_release(sk); 38761023658SDavid Ahern goto out; 38861023658SDavid Ahern } 3891da177e4SLinus Torvalds } 3901da177e4SLinus Torvalds out: 3911da177e4SLinus Torvalds return err; 3921da177e4SLinus Torvalds out_rcu_unlock: 3931da177e4SLinus Torvalds rcu_read_unlock(); 3941da177e4SLinus Torvalds goto out; 3951da177e4SLinus Torvalds } 3961da177e4SLinus Torvalds 3971da177e4SLinus Torvalds 3981da177e4SLinus Torvalds /* 3991da177e4SLinus Torvalds * The peer socket should always be NULL (or else). When we call this 4001da177e4SLinus Torvalds * function we are destroying the object and from then on nobody 4011da177e4SLinus Torvalds * should refer to it. 4021da177e4SLinus Torvalds */ 4031da177e4SLinus Torvalds int inet_release(struct socket *sock) 4041da177e4SLinus Torvalds { 4051da177e4SLinus Torvalds struct sock *sk = sock->sk; 4061da177e4SLinus Torvalds 4071da177e4SLinus Torvalds if (sk) { 4081da177e4SLinus Torvalds long timeout; 4091da177e4SLinus Torvalds 410f5836749SStanislav Fomichev if (!sk->sk_kern_sock) 411f5836749SStanislav Fomichev BPF_CGROUP_RUN_PROG_INET_SOCK_RELEASE(sk); 412f5836749SStanislav Fomichev 4131da177e4SLinus Torvalds /* Applications forget to leave groups before exiting */ 4141da177e4SLinus Torvalds ip_mc_drop_socket(sk); 4151da177e4SLinus Torvalds 4161da177e4SLinus Torvalds /* If linger is set, we don't return until the close 4171da177e4SLinus Torvalds * is complete. Otherwise we return immediately. The 4181da177e4SLinus Torvalds * actually closing is done the same either way. 4191da177e4SLinus Torvalds * 4201da177e4SLinus Torvalds * If the close is due to the process exiting, we never 4211da177e4SLinus Torvalds * linger.. 4221da177e4SLinus Torvalds */ 4231da177e4SLinus Torvalds timeout = 0; 4241da177e4SLinus Torvalds if (sock_flag(sk, SOCK_LINGER) && 4251da177e4SLinus Torvalds !(current->flags & PF_EXITING)) 4261da177e4SLinus Torvalds timeout = sk->sk_lingertime; 4271da177e4SLinus Torvalds sk->sk_prot->close(sk, timeout); 4282b81f816SJakub Kicinski sock->sk = NULL; 4291da177e4SLinus Torvalds } 4301da177e4SLinus Torvalds return 0; 4311da177e4SLinus Torvalds } 4323d1427f8SEric Dumazet EXPORT_SYMBOL(inet_release); 4331da177e4SLinus Torvalds 4341da177e4SLinus Torvalds int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) 4351da177e4SLinus Torvalds { 4361da177e4SLinus Torvalds struct sock *sk = sock->sk; 43777241217SStanislav Fomichev u32 flags = BIND_WITH_LOCK; 4381da177e4SLinus Torvalds int err; 4391da177e4SLinus Torvalds 4401da177e4SLinus Torvalds /* If the socket has its own bind function then use it. (RAW) */ 4411da177e4SLinus Torvalds if (sk->sk_prot->bind) { 4423679d585SAndrey Ignatov return sk->sk_prot->bind(sk, uaddr, addr_len); 4431da177e4SLinus Torvalds } 4441da177e4SLinus Torvalds if (addr_len < sizeof(struct sockaddr_in)) 4453679d585SAndrey Ignatov return -EINVAL; 4461da177e4SLinus Torvalds 4474fbac77dSAndrey Ignatov /* BPF prog is run before any checks are done so that if the prog 4484fbac77dSAndrey Ignatov * changes context in a wrong way it will be caught. 4494fbac77dSAndrey Ignatov */ 45077241217SStanislav Fomichev err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, 4516fc88c35SDave Marchevsky CGROUP_INET4_BIND, &flags); 4524fbac77dSAndrey Ignatov if (err) 4533679d585SAndrey Ignatov return err; 4543679d585SAndrey Ignatov 45577241217SStanislav Fomichev return __inet_bind(sk, uaddr, addr_len, flags); 4563679d585SAndrey Ignatov } 4573679d585SAndrey Ignatov EXPORT_SYMBOL(inet_bind); 4583679d585SAndrey Ignatov 4593679d585SAndrey Ignatov int __inet_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len, 460cb0721c7SStanislav Fomichev u32 flags) 4613679d585SAndrey Ignatov { 4623679d585SAndrey Ignatov struct sockaddr_in *addr = (struct sockaddr_in *)uaddr; 4631da177e4SLinus Torvalds struct inet_sock *inet = inet_sk(sk); 4641da177e4SLinus Torvalds struct net *net = sock_net(sk); 4651da177e4SLinus Torvalds unsigned short snum; 4661da177e4SLinus Torvalds int chk_addr_ret; 4671da177e4SLinus Torvalds u32 tb_id = RT_TABLE_LOCAL; 4681da177e4SLinus Torvalds int err; 4691da177e4SLinus Torvalds 470c349a528SMarcus Meissner if (addr->sin_family != AF_INET) { 47129c486dfSEric Dumazet /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET) 47229c486dfSEric Dumazet * only if s_addr is INADDR_ANY. 47329c486dfSEric Dumazet */ 474c349a528SMarcus Meissner err = -EAFNOSUPPORT; 47529c486dfSEric Dumazet if (addr->sin_family != AF_UNSPEC || 47629c486dfSEric Dumazet addr->sin_addr.s_addr != htonl(INADDR_ANY)) 477d0733d2eSMarcus Meissner goto out; 478c349a528SMarcus Meissner } 479d0733d2eSMarcus Meissner 4803236b004SDavid Ahern tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id; 48130bbaa19SDavid Ahern chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id); 4821da177e4SLinus Torvalds 4831da177e4SLinus Torvalds /* Not specified by any standard per-se, however it breaks too 4841da177e4SLinus Torvalds * many applications when removed. It is unfortunate since 4851da177e4SLinus Torvalds * allowing applications to make a non-local bind solves 4861da177e4SLinus Torvalds * several problems with systems using dynamic addressing. 4871da177e4SLinus Torvalds * (ie. your servers still start up even if your ISDN link 4881da177e4SLinus Torvalds * is temporarily down) 4891da177e4SLinus Torvalds */ 4901da177e4SLinus Torvalds err = -EADDRNOTAVAIL; 4918ff978b8SRiccardo Paolo Bestetti if (!inet_addr_valid_or_nonlocal(net, inet, addr->sin_addr.s_addr, 4928ff978b8SRiccardo Paolo Bestetti chk_addr_ret)) 4931da177e4SLinus Torvalds goto out; 4941da177e4SLinus Torvalds 4951da177e4SLinus Torvalds snum = ntohs(addr->sin_port); 4961da177e4SLinus Torvalds err = -EACCES; 49777241217SStanislav Fomichev if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) && 49877241217SStanislav Fomichev snum && inet_port_requires_bind_service(net, snum) && 4993594698aSEric W. Biederman !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) 5001da177e4SLinus Torvalds goto out; 5011da177e4SLinus Torvalds 5021da177e4SLinus Torvalds /* We keep a pair of addresses. rcv_saddr is the one 5031da177e4SLinus Torvalds * used by hash lookups, and saddr is used for transmit. 5041da177e4SLinus Torvalds * 5051da177e4SLinus Torvalds * In the BSD API these are the same except where it 5061da177e4SLinus Torvalds * would be illegal to use them (multicast/broadcast) in 5071da177e4SLinus Torvalds * which case the sending device address is used. 5081da177e4SLinus Torvalds */ 509cb0721c7SStanislav Fomichev if (flags & BIND_WITH_LOCK) 5101da177e4SLinus Torvalds lock_sock(sk); 5111da177e4SLinus Torvalds 5121da177e4SLinus Torvalds /* Check these errors (active socket, double bind). */ 5131da177e4SLinus Torvalds err = -EINVAL; 514c720c7e8SEric Dumazet if (sk->sk_state != TCP_CLOSE || inet->inet_num) 5151da177e4SLinus Torvalds goto out_release_sock; 5161da177e4SLinus Torvalds 517c720c7e8SEric Dumazet inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; 5181da177e4SLinus Torvalds if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) 519c720c7e8SEric Dumazet inet->inet_saddr = 0; /* Use device */ 5201da177e4SLinus Torvalds 5211da177e4SLinus Torvalds /* Make sure we are allowed to bind here. */ 522aac3fc32SAndrey Ignatov if (snum || !(inet->bind_address_no_port || 523cb0721c7SStanislav Fomichev (flags & BIND_FORCE_ADDRESS_NO_PORT))) { 5247a7160edSKuniyuki Iwashima err = sk->sk_prot->get_port(sk, snum); 5257a7160edSKuniyuki Iwashima if (err) { 526c720c7e8SEric Dumazet inet->inet_saddr = inet->inet_rcv_saddr = 0; 5271da177e4SLinus Torvalds goto out_release_sock; 5281da177e4SLinus Torvalds } 5298086fbafSStanislav Fomichev if (!(flags & BIND_FROM_BPF)) { 530aac3fc32SAndrey Ignatov err = BPF_CGROUP_RUN_PROG_INET4_POST_BIND(sk); 531aac3fc32SAndrey Ignatov if (err) { 532aac3fc32SAndrey Ignatov inet->inet_saddr = inet->inet_rcv_saddr = 0; 53391a760b2SMenglong Dong if (sk->sk_prot->put_port) 53491a760b2SMenglong Dong sk->sk_prot->put_port(sk); 535aac3fc32SAndrey Ignatov goto out_release_sock; 536aac3fc32SAndrey Ignatov } 537aac3fc32SAndrey Ignatov } 5388086fbafSStanislav Fomichev } 5391da177e4SLinus Torvalds 540c720c7e8SEric Dumazet if (inet->inet_rcv_saddr) 5411da177e4SLinus Torvalds sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 5421da177e4SLinus Torvalds if (snum) 5431da177e4SLinus Torvalds sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 544c720c7e8SEric Dumazet inet->inet_sport = htons(inet->inet_num); 545c720c7e8SEric Dumazet inet->inet_daddr = 0; 546c720c7e8SEric Dumazet inet->inet_dport = 0; 5471da177e4SLinus Torvalds sk_dst_reset(sk); 5481da177e4SLinus Torvalds err = 0; 5491da177e4SLinus Torvalds out_release_sock: 550cb0721c7SStanislav Fomichev if (flags & BIND_WITH_LOCK) 5511da177e4SLinus Torvalds release_sock(sk); 5521da177e4SLinus Torvalds out: 5531da177e4SLinus Torvalds return err; 5541da177e4SLinus Torvalds } 5551da177e4SLinus Torvalds 5561da177e4SLinus Torvalds int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr, 5571da177e4SLinus Torvalds int addr_len, int flags) 5581da177e4SLinus Torvalds { 5591da177e4SLinus Torvalds struct sock *sk = sock->sk; 560364f997bSKuniyuki Iwashima const struct proto *prot; 561d74bad4eSAndrey Ignatov int err; 5621da177e4SLinus Torvalds 5636503d961SChangli Gao if (addr_len < sizeof(uaddr->sa_family)) 5646503d961SChangli Gao return -EINVAL; 565364f997bSKuniyuki Iwashima 566364f997bSKuniyuki Iwashima /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 567364f997bSKuniyuki Iwashima prot = READ_ONCE(sk->sk_prot); 568364f997bSKuniyuki Iwashima 5691da177e4SLinus Torvalds if (uaddr->sa_family == AF_UNSPEC) 570364f997bSKuniyuki Iwashima return prot->disconnect(sk, flags); 5711da177e4SLinus Torvalds 572d74bad4eSAndrey Ignatov if (BPF_CGROUP_PRE_CONNECT_ENABLED(sk)) { 573364f997bSKuniyuki Iwashima err = prot->pre_connect(sk, uaddr, addr_len); 574d74bad4eSAndrey Ignatov if (err) 575d74bad4eSAndrey Ignatov return err; 576d74bad4eSAndrey Ignatov } 577d74bad4eSAndrey Ignatov 578dcd01eeaSEric Dumazet if (data_race(!inet_sk(sk)->inet_num) && inet_autobind(sk)) 5791da177e4SLinus Torvalds return -EAGAIN; 580364f997bSKuniyuki Iwashima return prot->connect(sk, uaddr, addr_len); 5811da177e4SLinus Torvalds } 5823d1427f8SEric Dumazet EXPORT_SYMBOL(inet_dgram_connect); 5831da177e4SLinus Torvalds 584783237e8SYuchung Cheng static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias) 5851da177e4SLinus Torvalds { 58614135f30SWANG Cong DEFINE_WAIT_FUNC(wait, woken_wake_function); 5871da177e4SLinus Torvalds 58814135f30SWANG Cong add_wait_queue(sk_sleep(sk), &wait); 589783237e8SYuchung Cheng sk->sk_write_pending += writebias; 5904faeee0cSEric Dumazet sk->sk_wait_pending++; 5911da177e4SLinus Torvalds 5921da177e4SLinus Torvalds /* Basic assumption: if someone sets sk->sk_err, he _must_ 5931da177e4SLinus Torvalds * change state of the socket from TCP_SYN_*. 5941da177e4SLinus Torvalds * Connect() does not allow to get error notifications 5951da177e4SLinus Torvalds * without closing the socket. 5961da177e4SLinus Torvalds */ 5971da177e4SLinus Torvalds while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { 5981da177e4SLinus Torvalds release_sock(sk); 59914135f30SWANG Cong timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); 6001da177e4SLinus Torvalds lock_sock(sk); 6011da177e4SLinus Torvalds if (signal_pending(current) || !timeo) 6021da177e4SLinus Torvalds break; 6031da177e4SLinus Torvalds } 60414135f30SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 605783237e8SYuchung Cheng sk->sk_write_pending -= writebias; 6064faeee0cSEric Dumazet sk->sk_wait_pending--; 6071da177e4SLinus Torvalds return timeo; 6081da177e4SLinus Torvalds } 6091da177e4SLinus Torvalds 6101da177e4SLinus Torvalds /* 6111da177e4SLinus Torvalds * Connect to a remote host. There is regrettably still a little 6121da177e4SLinus Torvalds * TCP 'magic' in here. 6131da177e4SLinus Torvalds */ 614cf60af03SYuchung Cheng int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, 6153979ad7eSWilly Tarreau int addr_len, int flags, int is_sendmsg) 6161da177e4SLinus Torvalds { 6171da177e4SLinus Torvalds struct sock *sk = sock->sk; 6181da177e4SLinus Torvalds int err; 6191da177e4SLinus Torvalds long timeo; 6201da177e4SLinus Torvalds 62119f6d3f3SWei Wang /* 62219f6d3f3SWei Wang * uaddr can be NULL and addr_len can be 0 if: 62319f6d3f3SWei Wang * sk is a TCP fastopen active socket and 62419f6d3f3SWei Wang * TCP_FASTOPEN_CONNECT sockopt is set and 62519f6d3f3SWei Wang * we already have a valid cookie for this socket. 62619f6d3f3SWei Wang * In this case, user can call write() after connect(). 62719f6d3f3SWei Wang * write() will invoke tcp_sendmsg_fastopen() which calls 62819f6d3f3SWei Wang * __inet_stream_connect(). 62919f6d3f3SWei Wang */ 63019f6d3f3SWei Wang if (uaddr) { 6316503d961SChangli Gao if (addr_len < sizeof(uaddr->sa_family)) 6326503d961SChangli Gao return -EINVAL; 6336503d961SChangli Gao 6341da177e4SLinus Torvalds if (uaddr->sa_family == AF_UNSPEC) { 6351da177e4SLinus Torvalds err = sk->sk_prot->disconnect(sk, flags); 6361da177e4SLinus Torvalds sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; 6371da177e4SLinus Torvalds goto out; 6381da177e4SLinus Torvalds } 63919f6d3f3SWei Wang } 6401da177e4SLinus Torvalds 6411da177e4SLinus Torvalds switch (sock->state) { 6421da177e4SLinus Torvalds default: 6431da177e4SLinus Torvalds err = -EINVAL; 6441da177e4SLinus Torvalds goto out; 6451da177e4SLinus Torvalds case SS_CONNECTED: 6461da177e4SLinus Torvalds err = -EISCONN; 6471da177e4SLinus Torvalds goto out; 6481da177e4SLinus Torvalds case SS_CONNECTING: 64919f6d3f3SWei Wang if (inet_sk(sk)->defer_connect) 6503979ad7eSWilly Tarreau err = is_sendmsg ? -EINPROGRESS : -EISCONN; 65119f6d3f3SWei Wang else 6521da177e4SLinus Torvalds err = -EALREADY; 6531da177e4SLinus Torvalds /* Fall out of switch with err, set for this state */ 6541da177e4SLinus Torvalds break; 6551da177e4SLinus Torvalds case SS_UNCONNECTED: 6561da177e4SLinus Torvalds err = -EISCONN; 6571da177e4SLinus Torvalds if (sk->sk_state != TCP_CLOSE) 6581da177e4SLinus Torvalds goto out; 6591da177e4SLinus Torvalds 660d74bad4eSAndrey Ignatov if (BPF_CGROUP_PRE_CONNECT_ENABLED(sk)) { 661d74bad4eSAndrey Ignatov err = sk->sk_prot->pre_connect(sk, uaddr, addr_len); 662d74bad4eSAndrey Ignatov if (err) 663d74bad4eSAndrey Ignatov goto out; 664d74bad4eSAndrey Ignatov } 665d74bad4eSAndrey Ignatov 6661da177e4SLinus Torvalds err = sk->sk_prot->connect(sk, uaddr, addr_len); 6671da177e4SLinus Torvalds if (err < 0) 6681da177e4SLinus Torvalds goto out; 6691da177e4SLinus Torvalds 6701da177e4SLinus Torvalds sock->state = SS_CONNECTING; 6711da177e4SLinus Torvalds 67219f6d3f3SWei Wang if (!err && inet_sk(sk)->defer_connect) 67319f6d3f3SWei Wang goto out; 67419f6d3f3SWei Wang 6751da177e4SLinus Torvalds /* Just entered SS_CONNECTING state; the only 6761da177e4SLinus Torvalds * difference is that return value in non-blocking 6771da177e4SLinus Torvalds * case is EINPROGRESS, rather than EALREADY. 6781da177e4SLinus Torvalds */ 6791da177e4SLinus Torvalds err = -EINPROGRESS; 6801da177e4SLinus Torvalds break; 6811da177e4SLinus Torvalds } 6821da177e4SLinus Torvalds 6831da177e4SLinus Torvalds timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); 6841da177e4SLinus Torvalds 6851da177e4SLinus Torvalds if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { 686783237e8SYuchung Cheng int writebias = (sk->sk_protocol == IPPROTO_TCP) && 687783237e8SYuchung Cheng tcp_sk(sk)->fastopen_req && 688783237e8SYuchung Cheng tcp_sk(sk)->fastopen_req->data ? 1 : 0; 689783237e8SYuchung Cheng 6901da177e4SLinus Torvalds /* Error code is set above */ 691783237e8SYuchung Cheng if (!timeo || !inet_wait_for_connect(sk, timeo, writebias)) 6921da177e4SLinus Torvalds goto out; 6931da177e4SLinus Torvalds 6941da177e4SLinus Torvalds err = sock_intr_errno(timeo); 6951da177e4SLinus Torvalds if (signal_pending(current)) 6961da177e4SLinus Torvalds goto out; 6971da177e4SLinus Torvalds } 6981da177e4SLinus Torvalds 6991da177e4SLinus Torvalds /* Connection was closed by RST, timeout, ICMP error 7001da177e4SLinus Torvalds * or another process disconnected us. 7011da177e4SLinus Torvalds */ 7021da177e4SLinus Torvalds if (sk->sk_state == TCP_CLOSE) 7031da177e4SLinus Torvalds goto sock_error; 7041da177e4SLinus Torvalds 7051da177e4SLinus Torvalds /* sk->sk_err may be not zero now, if RECVERR was ordered by user 7061da177e4SLinus Torvalds * and error was received after socket entered established state. 7071da177e4SLinus Torvalds * Hence, it is handled normally after connect() return successfully. 7081da177e4SLinus Torvalds */ 7091da177e4SLinus Torvalds 7101da177e4SLinus Torvalds sock->state = SS_CONNECTED; 7111da177e4SLinus Torvalds err = 0; 7121da177e4SLinus Torvalds out: 7131da177e4SLinus Torvalds return err; 7141da177e4SLinus Torvalds 7151da177e4SLinus Torvalds sock_error: 7161da177e4SLinus Torvalds err = sock_error(sk) ? : -ECONNABORTED; 7171da177e4SLinus Torvalds sock->state = SS_UNCONNECTED; 7181da177e4SLinus Torvalds if (sk->sk_prot->disconnect(sk, flags)) 7191da177e4SLinus Torvalds sock->state = SS_DISCONNECTING; 7201da177e4SLinus Torvalds goto out; 7211da177e4SLinus Torvalds } 722cf60af03SYuchung Cheng EXPORT_SYMBOL(__inet_stream_connect); 723cf60af03SYuchung Cheng 724cf60af03SYuchung Cheng int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, 725cf60af03SYuchung Cheng int addr_len, int flags) 726cf60af03SYuchung Cheng { 727cf60af03SYuchung Cheng int err; 728cf60af03SYuchung Cheng 729cf60af03SYuchung Cheng lock_sock(sock->sk); 7303979ad7eSWilly Tarreau err = __inet_stream_connect(sock, uaddr, addr_len, flags, 0); 731cf60af03SYuchung Cheng release_sock(sock->sk); 732cf60af03SYuchung Cheng return err; 733cf60af03SYuchung Cheng } 7343d1427f8SEric Dumazet EXPORT_SYMBOL(inet_stream_connect); 7351da177e4SLinus Torvalds 736711bdd51SPaolo Abeni void __inet_accept(struct socket *sock, struct socket *newsock, struct sock *newsk) 737711bdd51SPaolo Abeni { 738711bdd51SPaolo Abeni sock_rps_record_flow(newsk); 739711bdd51SPaolo Abeni WARN_ON(!((1 << newsk->sk_state) & 740711bdd51SPaolo Abeni (TCPF_ESTABLISHED | TCPF_SYN_RECV | 741711bdd51SPaolo Abeni TCPF_CLOSE_WAIT | TCPF_CLOSE))); 742711bdd51SPaolo Abeni 743711bdd51SPaolo Abeni if (test_bit(SOCK_SUPPORT_ZC, &sock->flags)) 744711bdd51SPaolo Abeni set_bit(SOCK_SUPPORT_ZC, &newsock->flags); 745711bdd51SPaolo Abeni sock_graft(newsk, newsock); 746711bdd51SPaolo Abeni 747711bdd51SPaolo Abeni newsock->state = SS_CONNECTED; 748711bdd51SPaolo Abeni } 749711bdd51SPaolo Abeni 7501da177e4SLinus Torvalds /* 7511da177e4SLinus Torvalds * Accept a pending connection. The TCP layer now gives BSD semantics. 7521da177e4SLinus Torvalds */ 7531da177e4SLinus Torvalds 754cdfbabfbSDavid Howells int inet_accept(struct socket *sock, struct socket *newsock, int flags, 755cdfbabfbSDavid Howells bool kern) 7561da177e4SLinus Torvalds { 757364f997bSKuniyuki Iwashima struct sock *sk1 = sock->sk, *sk2; 7581da177e4SLinus Torvalds int err = -EINVAL; 7591da177e4SLinus Torvalds 760364f997bSKuniyuki Iwashima /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 761364f997bSKuniyuki Iwashima sk2 = READ_ONCE(sk1->sk_prot)->accept(sk1, flags, &err, kern); 7621da177e4SLinus Torvalds if (!sk2) 763711bdd51SPaolo Abeni return err; 7641da177e4SLinus Torvalds 7651da177e4SLinus Torvalds lock_sock(sk2); 766711bdd51SPaolo Abeni __inet_accept(sock, newsock, sk2); 7671da177e4SLinus Torvalds release_sock(sk2); 768711bdd51SPaolo Abeni return 0; 7691da177e4SLinus Torvalds } 7703d1427f8SEric Dumazet EXPORT_SYMBOL(inet_accept); 7711da177e4SLinus Torvalds 7721da177e4SLinus Torvalds /* 7731da177e4SLinus Torvalds * This does both peername and sockname. 7741da177e4SLinus Torvalds */ 7751da177e4SLinus Torvalds int inet_getname(struct socket *sock, struct sockaddr *uaddr, 7769b2c45d4SDenys Vlasenko int peer) 7771da177e4SLinus Torvalds { 7781da177e4SLinus Torvalds struct sock *sk = sock->sk; 7791da177e4SLinus Torvalds struct inet_sock *inet = inet_sk(sk); 78038bfd8f5SCyrill Gorcunov DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr); 7811da177e4SLinus Torvalds 7821da177e4SLinus Torvalds sin->sin_family = AF_INET; 7839dfc685eSEric Dumazet lock_sock(sk); 7841da177e4SLinus Torvalds if (peer) { 785c720c7e8SEric Dumazet if (!inet->inet_dport || 7861da177e4SLinus Torvalds (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && 7879dfc685eSEric Dumazet peer == 1)) { 7889dfc685eSEric Dumazet release_sock(sk); 7891da177e4SLinus Torvalds return -ENOTCONN; 7909dfc685eSEric Dumazet } 791c720c7e8SEric Dumazet sin->sin_port = inet->inet_dport; 792c720c7e8SEric Dumazet sin->sin_addr.s_addr = inet->inet_daddr; 7939dfc685eSEric Dumazet BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 7949dfc685eSEric Dumazet CGROUP_INET4_GETPEERNAME); 7951da177e4SLinus Torvalds } else { 796c720c7e8SEric Dumazet __be32 addr = inet->inet_rcv_saddr; 7971da177e4SLinus Torvalds if (!addr) 798c720c7e8SEric Dumazet addr = inet->inet_saddr; 799c720c7e8SEric Dumazet sin->sin_port = inet->inet_sport; 8001da177e4SLinus Torvalds sin->sin_addr.s_addr = addr; 8019dfc685eSEric Dumazet BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 8029dfc685eSEric Dumazet CGROUP_INET4_GETSOCKNAME); 803a9ed15daSStanislav Fomichev } 8049dfc685eSEric Dumazet release_sock(sk); 8051da177e4SLinus Torvalds memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 8069b2c45d4SDenys Vlasenko return sizeof(*sin); 8071da177e4SLinus Torvalds } 8083d1427f8SEric Dumazet EXPORT_SYMBOL(inet_getname); 8091da177e4SLinus Torvalds 810e4730936SPaolo Abeni int inet_send_prepare(struct sock *sk) 8111da177e4SLinus Torvalds { 812c58dc01bSDavid S. Miller sock_rps_record_flow(sk); 813fec5e652STom Herbert 8141da177e4SLinus Torvalds /* We may need to bind the socket. */ 815dcd01eeaSEric Dumazet if (data_race(!inet_sk(sk)->inet_num) && !sk->sk_prot->no_autobind && 8167ba42910SChangli Gao inet_autobind(sk)) 8171da177e4SLinus Torvalds return -EAGAIN; 8181da177e4SLinus Torvalds 819e4730936SPaolo Abeni return 0; 820e4730936SPaolo Abeni } 821e4730936SPaolo Abeni EXPORT_SYMBOL_GPL(inet_send_prepare); 822e4730936SPaolo Abeni 823e4730936SPaolo Abeni int inet_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 824e4730936SPaolo Abeni { 825e4730936SPaolo Abeni struct sock *sk = sock->sk; 826e4730936SPaolo Abeni 827e4730936SPaolo Abeni if (unlikely(inet_send_prepare(sk))) 828e4730936SPaolo Abeni return -EAGAIN; 829e4730936SPaolo Abeni 8306f24080eSPaolo Abeni return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udp_sendmsg, 8316f24080eSPaolo Abeni sk, msg, size); 8321da177e4SLinus Torvalds } 8333d1427f8SEric Dumazet EXPORT_SYMBOL(inet_sendmsg); 8341da177e4SLinus Torvalds 8351d7e4538SDavid Howells void inet_splice_eof(struct socket *sock) 8361d7e4538SDavid Howells { 8371d7e4538SDavid Howells const struct proto *prot; 8381d7e4538SDavid Howells struct sock *sk = sock->sk; 8391d7e4538SDavid Howells 8401d7e4538SDavid Howells if (unlikely(inet_send_prepare(sk))) 8411d7e4538SDavid Howells return; 8421d7e4538SDavid Howells 8431d7e4538SDavid Howells /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 8441d7e4538SDavid Howells prot = READ_ONCE(sk->sk_prot); 8451d7e4538SDavid Howells if (prot->splice_eof) 8461d7e4538SDavid Howells prot->splice_eof(sock); 8471d7e4538SDavid Howells } 8481d7e4538SDavid Howells EXPORT_SYMBOL_GPL(inet_splice_eof); 8491d7e4538SDavid Howells 8507ba42910SChangli Gao ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, 8513d1427f8SEric Dumazet size_t size, int flags) 8521da177e4SLinus Torvalds { 8531da177e4SLinus Torvalds struct sock *sk = sock->sk; 854364f997bSKuniyuki Iwashima const struct proto *prot; 8551da177e4SLinus Torvalds 856e4730936SPaolo Abeni if (unlikely(inet_send_prepare(sk))) 8571da177e4SLinus Torvalds return -EAGAIN; 8581da177e4SLinus Torvalds 859364f997bSKuniyuki Iwashima /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 860364f997bSKuniyuki Iwashima prot = READ_ONCE(sk->sk_prot); 861364f997bSKuniyuki Iwashima if (prot->sendpage) 862364f997bSKuniyuki Iwashima return prot->sendpage(sk, page, offset, size, flags); 8631da177e4SLinus Torvalds return sock_no_sendpage(sock, page, offset, size, flags); 8641da177e4SLinus Torvalds } 8657ba42910SChangli Gao EXPORT_SYMBOL(inet_sendpage); 8661da177e4SLinus Torvalds 8676f24080eSPaolo Abeni INDIRECT_CALLABLE_DECLARE(int udp_recvmsg(struct sock *, struct msghdr *, 868ec095263SOliver Hartkopp size_t, int, int *)); 8691b784140SYing Xue int inet_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, 8701b784140SYing Xue int flags) 871fec5e652STom Herbert { 872fec5e652STom Herbert struct sock *sk = sock->sk; 873fec5e652STom Herbert int addr_len = 0; 874fec5e652STom Herbert int err; 875fec5e652STom Herbert 876e3f2c4a3SSoheil Hassas Yeganeh if (likely(!(flags & MSG_ERRQUEUE))) 877c58dc01bSDavid S. Miller sock_rps_record_flow(sk); 878fec5e652STom Herbert 8796f24080eSPaolo Abeni err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udp_recvmsg, 880ec095263SOliver Hartkopp sk, msg, size, flags, &addr_len); 881fec5e652STom Herbert if (err >= 0) 882fec5e652STom Herbert msg->msg_namelen = addr_len; 883fec5e652STom Herbert return err; 884fec5e652STom Herbert } 885fec5e652STom Herbert EXPORT_SYMBOL(inet_recvmsg); 8861da177e4SLinus Torvalds 8871da177e4SLinus Torvalds int inet_shutdown(struct socket *sock, int how) 8881da177e4SLinus Torvalds { 8891da177e4SLinus Torvalds struct sock *sk = sock->sk; 8901da177e4SLinus Torvalds int err = 0; 8911da177e4SLinus Torvalds 8921da177e4SLinus Torvalds /* This should really check to make sure 8931da177e4SLinus Torvalds * the socket is a TCP socket. (WHY AC...) 8941da177e4SLinus Torvalds */ 8951da177e4SLinus Torvalds how++; /* maps 0->1 has the advantage of making bit 1 rcvs and 8961da177e4SLinus Torvalds 1->2 bit 2 snds. 8971da177e4SLinus Torvalds 2->3 */ 8981da177e4SLinus Torvalds if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */ 8991da177e4SLinus Torvalds return -EINVAL; 9001da177e4SLinus Torvalds 9011da177e4SLinus Torvalds lock_sock(sk); 9021da177e4SLinus Torvalds if (sock->state == SS_CONNECTING) { 9031da177e4SLinus Torvalds if ((1 << sk->sk_state) & 9041da177e4SLinus Torvalds (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE)) 9051da177e4SLinus Torvalds sock->state = SS_DISCONNECTING; 9061da177e4SLinus Torvalds else 9071da177e4SLinus Torvalds sock->state = SS_CONNECTED; 9081da177e4SLinus Torvalds } 9091da177e4SLinus Torvalds 9101da177e4SLinus Torvalds switch (sk->sk_state) { 9111da177e4SLinus Torvalds case TCP_CLOSE: 9121da177e4SLinus Torvalds err = -ENOTCONN; 9131da177e4SLinus Torvalds /* Hack to wake up other listeners, who can poll for 914a9a08845SLinus Torvalds EPOLLHUP, even on eg. unconnected UDP sockets -- RR */ 915a8eceea8SJoe Perches fallthrough; 9161da177e4SLinus Torvalds default: 917e14cadfdSEric Dumazet WRITE_ONCE(sk->sk_shutdown, sk->sk_shutdown | how); 9181da177e4SLinus Torvalds if (sk->sk_prot->shutdown) 9191da177e4SLinus Torvalds sk->sk_prot->shutdown(sk, how); 9201da177e4SLinus Torvalds break; 9211da177e4SLinus Torvalds 9221da177e4SLinus Torvalds /* Remaining two branches are temporary solution for missing 9231da177e4SLinus Torvalds * close() in multithreaded environment. It is _not_ a good idea, 9241da177e4SLinus Torvalds * but we have no choice until close() is repaired at VFS level. 9251da177e4SLinus Torvalds */ 9261da177e4SLinus Torvalds case TCP_LISTEN: 9271da177e4SLinus Torvalds if (!(how & RCV_SHUTDOWN)) 9281da177e4SLinus Torvalds break; 929a8eceea8SJoe Perches fallthrough; 9301da177e4SLinus Torvalds case TCP_SYN_SENT: 9311da177e4SLinus Torvalds err = sk->sk_prot->disconnect(sk, O_NONBLOCK); 9321da177e4SLinus Torvalds sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; 9331da177e4SLinus Torvalds break; 9341da177e4SLinus Torvalds } 9351da177e4SLinus Torvalds 9361da177e4SLinus Torvalds /* Wake up anyone sleeping in poll. */ 9371da177e4SLinus Torvalds sk->sk_state_change(sk); 9381da177e4SLinus Torvalds release_sock(sk); 9391da177e4SLinus Torvalds return err; 9401da177e4SLinus Torvalds } 9413d1427f8SEric Dumazet EXPORT_SYMBOL(inet_shutdown); 9421da177e4SLinus Torvalds 9431da177e4SLinus Torvalds /* 9441da177e4SLinus Torvalds * ioctl() calls you can issue on an INET socket. Most of these are 9451da177e4SLinus Torvalds * device configuration and stuff and very rarely used. Some ioctls 9461da177e4SLinus Torvalds * pass on to the socket itself. 9471da177e4SLinus Torvalds * 9481da177e4SLinus Torvalds * NOTE: I like the idea of a module for the config stuff. ie ifconfig 9491da177e4SLinus Torvalds * loads the devconfigure module does its configuring and unloads it. 9501da177e4SLinus Torvalds * There's a good 20K of config code hanging around the kernel. 9511da177e4SLinus Torvalds */ 9521da177e4SLinus Torvalds 9531da177e4SLinus Torvalds int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 9541da177e4SLinus Torvalds { 9551da177e4SLinus Torvalds struct sock *sk = sock->sk; 9561da177e4SLinus Torvalds int err = 0; 9573b1e0a65SYOSHIFUJI Hideaki struct net *net = sock_net(sk); 95803aef17bSAl Viro void __user *p = (void __user *)arg; 95903aef17bSAl Viro struct ifreq ifr; 960ca25c300SAl Viro struct rtentry rt; 9611da177e4SLinus Torvalds 9621da177e4SLinus Torvalds switch (cmd) { 9631da177e4SLinus Torvalds case SIOCADDRT: 9641da177e4SLinus Torvalds case SIOCDELRT: 965ca25c300SAl Viro if (copy_from_user(&rt, p, sizeof(struct rtentry))) 966ca25c300SAl Viro return -EFAULT; 967ca25c300SAl Viro err = ip_rt_ioctl(net, cmd, &rt); 968ca25c300SAl Viro break; 9691da177e4SLinus Torvalds case SIOCRTMSG: 970ca25c300SAl Viro err = -EINVAL; 9711da177e4SLinus Torvalds break; 9721da177e4SLinus Torvalds case SIOCDARP: 9731da177e4SLinus Torvalds case SIOCGARP: 9741da177e4SLinus Torvalds case SIOCSARP: 975e5b13cb1SDenis V. Lunev err = arp_ioctl(net, cmd, (void __user *)arg); 9761da177e4SLinus Torvalds break; 9771da177e4SLinus Torvalds case SIOCGIFADDR: 9781da177e4SLinus Torvalds case SIOCGIFBRDADDR: 9791da177e4SLinus Torvalds case SIOCGIFNETMASK: 9801da177e4SLinus Torvalds case SIOCGIFDSTADDR: 98103aef17bSAl Viro case SIOCGIFPFLAGS: 98229c49648SArnd Bergmann if (get_user_ifreq(&ifr, NULL, p)) 98303aef17bSAl Viro return -EFAULT; 98403aef17bSAl Viro err = devinet_ioctl(net, cmd, &ifr); 98529c49648SArnd Bergmann if (!err && put_user_ifreq(&ifr, p)) 98603aef17bSAl Viro err = -EFAULT; 98703aef17bSAl Viro break; 98803aef17bSAl Viro 98903aef17bSAl Viro case SIOCSIFADDR: 99003aef17bSAl Viro case SIOCSIFBRDADDR: 99103aef17bSAl Viro case SIOCSIFNETMASK: 9921da177e4SLinus Torvalds case SIOCSIFDSTADDR: 9931da177e4SLinus Torvalds case SIOCSIFPFLAGS: 9941da177e4SLinus Torvalds case SIOCSIFFLAGS: 99529c49648SArnd Bergmann if (get_user_ifreq(&ifr, NULL, p)) 99603aef17bSAl Viro return -EFAULT; 99703aef17bSAl Viro err = devinet_ioctl(net, cmd, &ifr); 9981da177e4SLinus Torvalds break; 9991da177e4SLinus Torvalds default: 1000b5e5fa5eSChristoph Hellwig if (sk->sk_prot->ioctl) 1001*e1d001faSBreno Leitao err = sk_ioctl(sk, cmd, (void __user *)arg); 1002b5e5fa5eSChristoph Hellwig else 1003b5e5fa5eSChristoph Hellwig err = -ENOIOCTLCMD; 10041da177e4SLinus Torvalds break; 10051da177e4SLinus Torvalds } 10061da177e4SLinus Torvalds return err; 10071da177e4SLinus Torvalds } 10083d1427f8SEric Dumazet EXPORT_SYMBOL(inet_ioctl); 10091da177e4SLinus Torvalds 1010709b46e8SEric W. Biederman #ifdef CONFIG_COMPAT 1011dc13c876SChristoph Hellwig static int inet_compat_routing_ioctl(struct sock *sk, unsigned int cmd, 1012dc13c876SChristoph Hellwig struct compat_rtentry __user *ur) 1013dc13c876SChristoph Hellwig { 1014dc13c876SChristoph Hellwig compat_uptr_t rtdev; 1015dc13c876SChristoph Hellwig struct rtentry rt; 1016dc13c876SChristoph Hellwig 1017dc13c876SChristoph Hellwig if (copy_from_user(&rt.rt_dst, &ur->rt_dst, 1018dc13c876SChristoph Hellwig 3 * sizeof(struct sockaddr)) || 1019dc13c876SChristoph Hellwig get_user(rt.rt_flags, &ur->rt_flags) || 1020dc13c876SChristoph Hellwig get_user(rt.rt_metric, &ur->rt_metric) || 1021dc13c876SChristoph Hellwig get_user(rt.rt_mtu, &ur->rt_mtu) || 1022dc13c876SChristoph Hellwig get_user(rt.rt_window, &ur->rt_window) || 1023dc13c876SChristoph Hellwig get_user(rt.rt_irtt, &ur->rt_irtt) || 1024dc13c876SChristoph Hellwig get_user(rtdev, &ur->rt_dev)) 1025dc13c876SChristoph Hellwig return -EFAULT; 1026dc13c876SChristoph Hellwig 1027dc13c876SChristoph Hellwig rt.rt_dev = compat_ptr(rtdev); 1028dc13c876SChristoph Hellwig return ip_rt_ioctl(sock_net(sk), cmd, &rt); 1029dc13c876SChristoph Hellwig } 1030dc13c876SChristoph Hellwig 1031686dc6b6SGerrit Renker static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 1032709b46e8SEric W. Biederman { 1033dc13c876SChristoph Hellwig void __user *argp = compat_ptr(arg); 1034709b46e8SEric W. Biederman struct sock *sk = sock->sk; 1035709b46e8SEric W. Biederman 1036dc13c876SChristoph Hellwig switch (cmd) { 1037dc13c876SChristoph Hellwig case SIOCADDRT: 1038dc13c876SChristoph Hellwig case SIOCDELRT: 1039dc13c876SChristoph Hellwig return inet_compat_routing_ioctl(sk, cmd, argp); 1040dc13c876SChristoph Hellwig default: 1041dc13c876SChristoph Hellwig if (!sk->sk_prot->compat_ioctl) 1042dc13c876SChristoph Hellwig return -ENOIOCTLCMD; 1043dc13c876SChristoph Hellwig return sk->sk_prot->compat_ioctl(sk, cmd, arg); 1044709b46e8SEric W. Biederman } 1045dc13c876SChristoph Hellwig } 1046dc13c876SChristoph Hellwig #endif /* CONFIG_COMPAT */ 1047709b46e8SEric W. Biederman 104890ddc4f0SEric Dumazet const struct proto_ops inet_stream_ops = { 10491da177e4SLinus Torvalds .family = PF_INET, 10501da177e4SLinus Torvalds .owner = THIS_MODULE, 10511da177e4SLinus Torvalds .release = inet_release, 10521da177e4SLinus Torvalds .bind = inet_bind, 10531da177e4SLinus Torvalds .connect = inet_stream_connect, 10541da177e4SLinus Torvalds .socketpair = sock_no_socketpair, 10551da177e4SLinus Torvalds .accept = inet_accept, 10561da177e4SLinus Torvalds .getname = inet_getname, 1057a11e1d43SLinus Torvalds .poll = tcp_poll, 10581da177e4SLinus Torvalds .ioctl = inet_ioctl, 1059c7cbdbf2SArnd Bergmann .gettstamp = sock_gettstamp, 10601da177e4SLinus Torvalds .listen = inet_listen, 10611da177e4SLinus Torvalds .shutdown = inet_shutdown, 10621da177e4SLinus Torvalds .setsockopt = sock_common_setsockopt, 10631da177e4SLinus Torvalds .getsockopt = sock_common_getsockopt, 10647ba42910SChangli Gao .sendmsg = inet_sendmsg, 1065fec5e652STom Herbert .recvmsg = inet_recvmsg, 106605255b82SEric Dumazet #ifdef CONFIG_MMU 106793ab6cc6SEric Dumazet .mmap = tcp_mmap, 106805255b82SEric Dumazet #endif 10691d7e4538SDavid Howells .splice_eof = inet_splice_eof, 10707ba42910SChangli Gao .sendpage = inet_sendpage, 10719c55e01cSJens Axboe .splice_read = tcp_splice_read, 107232035585STom Herbert .read_sock = tcp_read_sock, 1073965b57b4SCong Wang .read_skb = tcp_read_skb, 1074306b13ebSTom Herbert .sendmsg_locked = tcp_sendmsg_locked, 1075306b13ebSTom Herbert .sendpage_locked = tcp_sendpage_locked, 107632035585STom Herbert .peek_len = tcp_peek_len, 10773fdadf7dSDmitry Mishin #ifdef CONFIG_COMPAT 1078709b46e8SEric W. Biederman .compat_ioctl = inet_compat_ioctl, 10793fdadf7dSDmitry Mishin #endif 1080d1361840SEric Dumazet .set_rcvlowat = tcp_set_rcvlowat, 10811da177e4SLinus Torvalds }; 10823d1427f8SEric Dumazet EXPORT_SYMBOL(inet_stream_ops); 10831da177e4SLinus Torvalds 108490ddc4f0SEric Dumazet const struct proto_ops inet_dgram_ops = { 10851da177e4SLinus Torvalds .family = PF_INET, 10861da177e4SLinus Torvalds .owner = THIS_MODULE, 10871da177e4SLinus Torvalds .release = inet_release, 10881da177e4SLinus Torvalds .bind = inet_bind, 10891da177e4SLinus Torvalds .connect = inet_dgram_connect, 10901da177e4SLinus Torvalds .socketpair = sock_no_socketpair, 10911da177e4SLinus Torvalds .accept = sock_no_accept, 10921da177e4SLinus Torvalds .getname = inet_getname, 1093a11e1d43SLinus Torvalds .poll = udp_poll, 10941da177e4SLinus Torvalds .ioctl = inet_ioctl, 1095c7cbdbf2SArnd Bergmann .gettstamp = sock_gettstamp, 10961da177e4SLinus Torvalds .listen = sock_no_listen, 10971da177e4SLinus Torvalds .shutdown = inet_shutdown, 10981da177e4SLinus Torvalds .setsockopt = sock_common_setsockopt, 10991da177e4SLinus Torvalds .getsockopt = sock_common_getsockopt, 11001da177e4SLinus Torvalds .sendmsg = inet_sendmsg, 1101965b57b4SCong Wang .read_skb = udp_read_skb, 1102fec5e652STom Herbert .recvmsg = inet_recvmsg, 11031da177e4SLinus Torvalds .mmap = sock_no_mmap, 11041d7e4538SDavid Howells .splice_eof = inet_splice_eof, 11051da177e4SLinus Torvalds .sendpage = inet_sendpage, 1106627d2d6bSsamanthakumar .set_peek_off = sk_set_peek_off, 1107543d9cfeSArnaldo Carvalho de Melo #ifdef CONFIG_COMPAT 1108709b46e8SEric W. Biederman .compat_ioctl = inet_compat_ioctl, 1109543d9cfeSArnaldo Carvalho de Melo #endif 11101da177e4SLinus Torvalds }; 11113d1427f8SEric Dumazet EXPORT_SYMBOL(inet_dgram_ops); 11121da177e4SLinus Torvalds 11131da177e4SLinus Torvalds /* 11141da177e4SLinus Torvalds * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without 1115a11e1d43SLinus Torvalds * udp_poll 11161da177e4SLinus Torvalds */ 111790ddc4f0SEric Dumazet static const struct proto_ops inet_sockraw_ops = { 11181da177e4SLinus Torvalds .family = PF_INET, 11191da177e4SLinus Torvalds .owner = THIS_MODULE, 11201da177e4SLinus Torvalds .release = inet_release, 11211da177e4SLinus Torvalds .bind = inet_bind, 11221da177e4SLinus Torvalds .connect = inet_dgram_connect, 11231da177e4SLinus Torvalds .socketpair = sock_no_socketpair, 11241da177e4SLinus Torvalds .accept = sock_no_accept, 11251da177e4SLinus Torvalds .getname = inet_getname, 1126a11e1d43SLinus Torvalds .poll = datagram_poll, 11271da177e4SLinus Torvalds .ioctl = inet_ioctl, 1128c7cbdbf2SArnd Bergmann .gettstamp = sock_gettstamp, 11291da177e4SLinus Torvalds .listen = sock_no_listen, 11301da177e4SLinus Torvalds .shutdown = inet_shutdown, 11311da177e4SLinus Torvalds .setsockopt = sock_common_setsockopt, 11321da177e4SLinus Torvalds .getsockopt = sock_common_getsockopt, 11331da177e4SLinus Torvalds .sendmsg = inet_sendmsg, 1134fec5e652STom Herbert .recvmsg = inet_recvmsg, 11351da177e4SLinus Torvalds .mmap = sock_no_mmap, 11361d7e4538SDavid Howells .splice_eof = inet_splice_eof, 11371da177e4SLinus Torvalds .sendpage = inet_sendpage, 1138543d9cfeSArnaldo Carvalho de Melo #ifdef CONFIG_COMPAT 1139709b46e8SEric W. Biederman .compat_ioctl = inet_compat_ioctl, 1140543d9cfeSArnaldo Carvalho de Melo #endif 11411da177e4SLinus Torvalds }; 11421da177e4SLinus Torvalds 1143ec1b4cf7SStephen Hemminger static const struct net_proto_family inet_family_ops = { 11441da177e4SLinus Torvalds .family = PF_INET, 11451da177e4SLinus Torvalds .create = inet_create, 11461da177e4SLinus Torvalds .owner = THIS_MODULE, 11471da177e4SLinus Torvalds }; 11481da177e4SLinus Torvalds 11491da177e4SLinus Torvalds /* Upon startup we insert all the elements in inetsw_array[] into 11501da177e4SLinus Torvalds * the linked list inetsw. 11511da177e4SLinus Torvalds */ 11521da177e4SLinus Torvalds static struct inet_protosw inetsw_array[] = 11531da177e4SLinus Torvalds { 11541da177e4SLinus Torvalds { 11551da177e4SLinus Torvalds .type = SOCK_STREAM, 11561da177e4SLinus Torvalds .protocol = IPPROTO_TCP, 11571da177e4SLinus Torvalds .prot = &tcp_prot, 11581da177e4SLinus Torvalds .ops = &inet_stream_ops, 1159d83d8461SArnaldo Carvalho de Melo .flags = INET_PROTOSW_PERMANENT | 1160d83d8461SArnaldo Carvalho de Melo INET_PROTOSW_ICSK, 11611da177e4SLinus Torvalds }, 11621da177e4SLinus Torvalds 11631da177e4SLinus Torvalds { 11641da177e4SLinus Torvalds .type = SOCK_DGRAM, 11651da177e4SLinus Torvalds .protocol = IPPROTO_UDP, 11661da177e4SLinus Torvalds .prot = &udp_prot, 11671da177e4SLinus Torvalds .ops = &inet_dgram_ops, 11681da177e4SLinus Torvalds .flags = INET_PROTOSW_PERMANENT, 11691da177e4SLinus Torvalds }, 11701da177e4SLinus Torvalds 1171c319b4d7SVasiliy Kulikov { 1172c319b4d7SVasiliy Kulikov .type = SOCK_DGRAM, 1173c319b4d7SVasiliy Kulikov .protocol = IPPROTO_ICMP, 1174c319b4d7SVasiliy Kulikov .prot = &ping_prot, 117577d4b1d3SEric Dumazet .ops = &inet_sockraw_ops, 1176c319b4d7SVasiliy Kulikov .flags = INET_PROTOSW_REUSE, 1177c319b4d7SVasiliy Kulikov }, 11781da177e4SLinus Torvalds 11791da177e4SLinus Torvalds { 11801da177e4SLinus Torvalds .type = SOCK_RAW, 11811da177e4SLinus Torvalds .protocol = IPPROTO_IP, /* wild card */ 11821da177e4SLinus Torvalds .prot = &raw_prot, 11831da177e4SLinus Torvalds .ops = &inet_sockraw_ops, 11841da177e4SLinus Torvalds .flags = INET_PROTOSW_REUSE, 11851da177e4SLinus Torvalds } 11861da177e4SLinus Torvalds }; 11871da177e4SLinus Torvalds 1188c40f6fffSDenis Cheng #define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array) 11891da177e4SLinus Torvalds 11901da177e4SLinus Torvalds void inet_register_protosw(struct inet_protosw *p) 11911da177e4SLinus Torvalds { 11921da177e4SLinus Torvalds struct list_head *lh; 11931da177e4SLinus Torvalds struct inet_protosw *answer; 11941da177e4SLinus Torvalds int protocol = p->protocol; 11951da177e4SLinus Torvalds struct list_head *last_perm; 11961da177e4SLinus Torvalds 11971da177e4SLinus Torvalds spin_lock_bh(&inetsw_lock); 11981da177e4SLinus Torvalds 11991da177e4SLinus Torvalds if (p->type >= SOCK_MAX) 12001da177e4SLinus Torvalds goto out_illegal; 12011da177e4SLinus Torvalds 12021da177e4SLinus Torvalds /* If we are trying to override a permanent protocol, bail. */ 12031da177e4SLinus Torvalds last_perm = &inetsw[p->type]; 12041da177e4SLinus Torvalds list_for_each(lh, &inetsw[p->type]) { 12051da177e4SLinus Torvalds answer = list_entry(lh, struct inet_protosw, list); 12061da177e4SLinus Torvalds /* Check only the non-wild match. */ 1207f6c53334SJunwei Zhang if ((INET_PROTOSW_PERMANENT & answer->flags) == 0) 12081da177e4SLinus Torvalds break; 1209f6c53334SJunwei Zhang if (protocol == answer->protocol) 1210f6c53334SJunwei Zhang goto out_permanent; 12111da177e4SLinus Torvalds last_perm = lh; 12121da177e4SLinus Torvalds } 12131da177e4SLinus Torvalds 12141da177e4SLinus Torvalds /* Add the new entry after the last permanent entry if any, so that 12151da177e4SLinus Torvalds * the new entry does not override a permanent entry when matched with 12161da177e4SLinus Torvalds * a wild-card protocol. But it is allowed to override any existing 12171da177e4SLinus Torvalds * non-permanent entry. This means that when we remove this entry, the 12181da177e4SLinus Torvalds * system automatically returns to the old behavior. 12191da177e4SLinus Torvalds */ 12201da177e4SLinus Torvalds list_add_rcu(&p->list, last_perm); 12211da177e4SLinus Torvalds out: 12221da177e4SLinus Torvalds spin_unlock_bh(&inetsw_lock); 12231da177e4SLinus Torvalds 12241da177e4SLinus Torvalds return; 12251da177e4SLinus Torvalds 12261da177e4SLinus Torvalds out_permanent: 1227058bd4d2SJoe Perches pr_err("Attempt to override permanent protocol %d\n", protocol); 12281da177e4SLinus Torvalds goto out; 12291da177e4SLinus Torvalds 12301da177e4SLinus Torvalds out_illegal: 1231058bd4d2SJoe Perches pr_err("Ignoring attempt to register invalid socket type %d\n", 12321da177e4SLinus Torvalds p->type); 12331da177e4SLinus Torvalds goto out; 12341da177e4SLinus Torvalds } 12353d1427f8SEric Dumazet EXPORT_SYMBOL(inet_register_protosw); 12361da177e4SLinus Torvalds 12371da177e4SLinus Torvalds void inet_unregister_protosw(struct inet_protosw *p) 12381da177e4SLinus Torvalds { 12391da177e4SLinus Torvalds if (INET_PROTOSW_PERMANENT & p->flags) { 1240058bd4d2SJoe Perches pr_err("Attempt to unregister permanent protocol %d\n", 12411da177e4SLinus Torvalds p->protocol); 12421da177e4SLinus Torvalds } else { 12431da177e4SLinus Torvalds spin_lock_bh(&inetsw_lock); 12441da177e4SLinus Torvalds list_del_rcu(&p->list); 12451da177e4SLinus Torvalds spin_unlock_bh(&inetsw_lock); 12461da177e4SLinus Torvalds 12471da177e4SLinus Torvalds synchronize_net(); 12481da177e4SLinus Torvalds } 12491da177e4SLinus Torvalds } 12503d1427f8SEric Dumazet EXPORT_SYMBOL(inet_unregister_protosw); 12511da177e4SLinus Torvalds 125232519f11SArnaldo Carvalho de Melo static int inet_sk_reselect_saddr(struct sock *sk) 125332519f11SArnaldo Carvalho de Melo { 125432519f11SArnaldo Carvalho de Melo struct inet_sock *inet = inet_sk(sk); 1255c720c7e8SEric Dumazet __be32 old_saddr = inet->inet_saddr; 1256c720c7e8SEric Dumazet __be32 daddr = inet->inet_daddr; 12576e869138SDavid S. Miller struct flowi4 *fl4; 1258b23dd4feSDavid S. Miller struct rtable *rt; 1259b23dd4feSDavid S. Miller __be32 new_saddr; 1260f6d8bd05SEric Dumazet struct ip_options_rcu *inet_opt; 126128044fc1SJoanne Koong int err; 126232519f11SArnaldo Carvalho de Melo 1263f6d8bd05SEric Dumazet inet_opt = rcu_dereference_protected(inet->inet_opt, 12641e1d04e6SHannes Frederic Sowa lockdep_sock_is_held(sk)); 1265f6d8bd05SEric Dumazet if (inet_opt && inet_opt->opt.srr) 1266f6d8bd05SEric Dumazet daddr = inet_opt->opt.faddr; 126732519f11SArnaldo Carvalho de Melo 126832519f11SArnaldo Carvalho de Melo /* Query new route. */ 12696e869138SDavid S. Miller fl4 = &inet->cork.fl.u.ip4; 127067e1e2f4SGuillaume Nault rt = ip_route_connect(fl4, daddr, 0, sk->sk_bound_dev_if, 127167e1e2f4SGuillaume Nault sk->sk_protocol, inet->inet_sport, 127267e1e2f4SGuillaume Nault inet->inet_dport, sk); 1273b23dd4feSDavid S. Miller if (IS_ERR(rt)) 1274b23dd4feSDavid S. Miller return PTR_ERR(rt); 127532519f11SArnaldo Carvalho de Melo 12766e869138SDavid S. Miller new_saddr = fl4->saddr; 127732519f11SArnaldo Carvalho de Melo 127828044fc1SJoanne Koong if (new_saddr == old_saddr) { 127928044fc1SJoanne Koong sk_setup_caps(sk, &rt->dst); 128032519f11SArnaldo Carvalho de Melo return 0; 128128044fc1SJoanne Koong } 128228044fc1SJoanne Koong 12838c5dae4cSKuniyuki Iwashima err = inet_bhash2_update_saddr(sk, &new_saddr, AF_INET); 128428044fc1SJoanne Koong if (err) { 128528044fc1SJoanne Koong ip_rt_put(rt); 128628044fc1SJoanne Koong return err; 128728044fc1SJoanne Koong } 128828044fc1SJoanne Koong 128928044fc1SJoanne Koong sk_setup_caps(sk, &rt->dst); 129032519f11SArnaldo Carvalho de Melo 1291e49e4affSKuniyuki Iwashima if (READ_ONCE(sock_net(sk)->ipv4.sysctl_ip_dynaddr) > 1) { 1292058bd4d2SJoe Perches pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n", 1293673d57e7SHarvey Harrison __func__, &old_saddr, &new_saddr); 129432519f11SArnaldo Carvalho de Melo } 129532519f11SArnaldo Carvalho de Melo 129632519f11SArnaldo Carvalho de Melo /* 129732519f11SArnaldo Carvalho de Melo * XXX The only one ugly spot where we need to 129832519f11SArnaldo Carvalho de Melo * XXX really change the sockets identity after 129932519f11SArnaldo Carvalho de Melo * XXX it has entered the hashes. -DaveM 130032519f11SArnaldo Carvalho de Melo * 130132519f11SArnaldo Carvalho de Melo * Besides that, it does not check for connection 130232519f11SArnaldo Carvalho de Melo * uniqueness. Wait for troubles. 130332519f11SArnaldo Carvalho de Melo */ 1304086c653fSCraig Gallek return __sk_prot_rehash(sk); 130532519f11SArnaldo Carvalho de Melo } 130632519f11SArnaldo Carvalho de Melo 130732519f11SArnaldo Carvalho de Melo int inet_sk_rebuild_header(struct sock *sk) 130832519f11SArnaldo Carvalho de Melo { 130932519f11SArnaldo Carvalho de Melo struct inet_sock *inet = inet_sk(sk); 131032519f11SArnaldo Carvalho de Melo struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0); 13113ca3c68eSAl Viro __be32 daddr; 1312f6d8bd05SEric Dumazet struct ip_options_rcu *inet_opt; 13136e869138SDavid S. Miller struct flowi4 *fl4; 131432519f11SArnaldo Carvalho de Melo int err; 131532519f11SArnaldo Carvalho de Melo 131632519f11SArnaldo Carvalho de Melo /* Route is OK, nothing to do. */ 131732519f11SArnaldo Carvalho de Melo if (rt) 131832519f11SArnaldo Carvalho de Melo return 0; 131932519f11SArnaldo Carvalho de Melo 132032519f11SArnaldo Carvalho de Melo /* Reroute. */ 1321f6d8bd05SEric Dumazet rcu_read_lock(); 1322f6d8bd05SEric Dumazet inet_opt = rcu_dereference(inet->inet_opt); 1323c720c7e8SEric Dumazet daddr = inet->inet_daddr; 1324f6d8bd05SEric Dumazet if (inet_opt && inet_opt->opt.srr) 1325f6d8bd05SEric Dumazet daddr = inet_opt->opt.faddr; 1326f6d8bd05SEric Dumazet rcu_read_unlock(); 13276e869138SDavid S. Miller fl4 = &inet->cork.fl.u.ip4; 13286e869138SDavid S. Miller rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr, 132978fbfd8aSDavid S. Miller inet->inet_dport, inet->inet_sport, 133078fbfd8aSDavid S. Miller sk->sk_protocol, RT_CONN_FLAGS(sk), 133178fbfd8aSDavid S. Miller sk->sk_bound_dev_if); 1332b23dd4feSDavid S. Miller if (!IS_ERR(rt)) { 1333b23dd4feSDavid S. Miller err = 0; 1334d8d1f30bSChangli Gao sk_setup_caps(sk, &rt->dst); 1335b23dd4feSDavid S. Miller } else { 1336b23dd4feSDavid S. Miller err = PTR_ERR(rt); 1337b23dd4feSDavid S. Miller 133832519f11SArnaldo Carvalho de Melo /* Routing failed... */ 133932519f11SArnaldo Carvalho de Melo sk->sk_route_caps = 0; 134032519f11SArnaldo Carvalho de Melo /* 134132519f11SArnaldo Carvalho de Melo * Other protocols have to map its equivalent state to TCP_SYN_SENT. 134232519f11SArnaldo Carvalho de Melo * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme 134332519f11SArnaldo Carvalho de Melo */ 1344e49e4affSKuniyuki Iwashima if (!READ_ONCE(sock_net(sk)->ipv4.sysctl_ip_dynaddr) || 134532519f11SArnaldo Carvalho de Melo sk->sk_state != TCP_SYN_SENT || 134632519f11SArnaldo Carvalho de Melo (sk->sk_userlocks & SOCK_BINDADDR_LOCK) || 134732519f11SArnaldo Carvalho de Melo (err = inet_sk_reselect_saddr(sk)) != 0) 13482f2d9972SEric Dumazet WRITE_ONCE(sk->sk_err_soft, -err); 134932519f11SArnaldo Carvalho de Melo } 135032519f11SArnaldo Carvalho de Melo 135132519f11SArnaldo Carvalho de Melo return err; 135232519f11SArnaldo Carvalho de Melo } 135332519f11SArnaldo Carvalho de Melo EXPORT_SYMBOL(inet_sk_rebuild_header); 135432519f11SArnaldo Carvalho de Melo 1355563e0bb0SYafang Shao void inet_sk_set_state(struct sock *sk, int state) 1356563e0bb0SYafang Shao { 1357563e0bb0SYafang Shao trace_inet_sock_set_state(sk, sk->sk_state, state); 1358563e0bb0SYafang Shao sk->sk_state = state; 1359563e0bb0SYafang Shao } 1360563e0bb0SYafang Shao EXPORT_SYMBOL(inet_sk_set_state); 1361563e0bb0SYafang Shao 1362563e0bb0SYafang Shao void inet_sk_state_store(struct sock *sk, int newstate) 1363563e0bb0SYafang Shao { 1364563e0bb0SYafang Shao trace_inet_sock_set_state(sk, sk->sk_state, newstate); 1365563e0bb0SYafang Shao smp_store_release(&sk->sk_state, newstate); 1366563e0bb0SYafang Shao } 1367563e0bb0SYafang Shao 1368b8921ca8STom Herbert struct sk_buff *inet_gso_segment(struct sk_buff *skb, 1369c8f44affSMichał Mirosław netdev_features_t features) 1370f4c50d99SHerbert Xu { 13710c19f846SWillem de Bruijn bool udpfrag = false, fixedid = false, gso_partial, encap; 1372f4c50d99SHerbert Xu struct sk_buff *segs = ERR_PTR(-EINVAL); 1373bca49f84SVlad Yasevich const struct net_offload *ops; 13740c19f846SWillem de Bruijn unsigned int offset = 0; 1375f9242b6bSDavid S. Miller struct iphdr *iph; 1376802ab55aSAlexander Duyck int proto, tot_len; 13773347c960SEric Dumazet int nhoff; 1378f4c50d99SHerbert Xu int ihl; 1379f4c50d99SHerbert Xu int id; 1380f4c50d99SHerbert Xu 13813347c960SEric Dumazet skb_reset_network_header(skb); 13823347c960SEric Dumazet nhoff = skb_network_header(skb) - skb_mac_header(skb); 1383bbcf467dSHerbert Xu if (unlikely(!pskb_may_pull(skb, sizeof(*iph)))) 1384f4c50d99SHerbert Xu goto out; 1385f4c50d99SHerbert Xu 1386eddc9ec5SArnaldo Carvalho de Melo iph = ip_hdr(skb); 1387f4c50d99SHerbert Xu ihl = iph->ihl * 4; 1388f4c50d99SHerbert Xu if (ihl < sizeof(*iph)) 1389f4c50d99SHerbert Xu goto out; 1390f4c50d99SHerbert Xu 139147d27aadSEric Dumazet id = ntohs(iph->id); 139247d27aadSEric Dumazet proto = iph->protocol; 139347d27aadSEric Dumazet 139447d27aadSEric Dumazet /* Warning: after this point, iph might be no longer valid */ 1395bbcf467dSHerbert Xu if (unlikely(!pskb_may_pull(skb, ihl))) 1396f4c50d99SHerbert Xu goto out; 139747d27aadSEric Dumazet __skb_pull(skb, ihl); 1398f4c50d99SHerbert Xu 13998c3a897bSEric Dumazet encap = SKB_GSO_CB(skb)->encap_level > 0; 14008c3a897bSEric Dumazet if (encap) 14011e16aa3dSFlorian Westphal features &= skb->dev->hw_enc_features; 14023347c960SEric Dumazet SKB_GSO_CB(skb)->encap_level += ihl; 140373136267SPravin B Shelar 1404badff6d0SArnaldo Carvalho de Melo skb_reset_transport_header(skb); 140547d27aadSEric Dumazet 1406f4c50d99SHerbert Xu segs = ERR_PTR(-EPROTONOSUPPORT); 1407f4c50d99SHerbert Xu 1408cbc53e08SAlexander Duyck if (!skb->encapsulation || encap) { 14090c19f846SWillem de Bruijn udpfrag = !!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP); 1410cbc53e08SAlexander Duyck fixedid = !!(skb_shinfo(skb)->gso_type & SKB_GSO_TCP_FIXEDID); 1411cbc53e08SAlexander Duyck 1412cbc53e08SAlexander Duyck /* fixed ID is invalid if DF bit is not set */ 1413a5108878SArnaldo Carvalho de Melo if (fixedid && !(ip_hdr(skb)->frag_off & htons(IP_DF))) 1414cbc53e08SAlexander Duyck goto out; 1415cbc53e08SAlexander Duyck } 1416dcd60771SEric Dumazet 1417bca49f84SVlad Yasevich ops = rcu_dereference(inet_offloads[proto]); 1418cc20ccedSTao Liu if (likely(ops && ops->callbacks.gso_segment)) { 1419f191a1d1SVlad Yasevich segs = ops->callbacks.gso_segment(skb, features); 1420cc20ccedSTao Liu if (!segs) 1421cc20ccedSTao Liu skb->network_header = skb_mac_header(skb) + nhoff - skb->head; 1422cc20ccedSTao Liu } 1423f4c50d99SHerbert Xu 142450c3a487SYOSHIFUJI Hideaki / 吉藤英明 if (IS_ERR_OR_NULL(segs)) 1425f4c50d99SHerbert Xu goto out; 1426f4c50d99SHerbert Xu 142707b26c94SSteffen Klassert gso_partial = !!(skb_shinfo(segs)->gso_type & SKB_GSO_PARTIAL); 142807b26c94SSteffen Klassert 1429f4c50d99SHerbert Xu skb = segs; 1430f4c50d99SHerbert Xu do { 14313347c960SEric Dumazet iph = (struct iphdr *)(skb_mac_header(skb) + nhoff); 14320c19f846SWillem de Bruijn if (udpfrag) { 14330c19f846SWillem de Bruijn iph->frag_off = htons(offset >> 3); 14340c19f846SWillem de Bruijn if (skb->next) 14350c19f846SWillem de Bruijn iph->frag_off |= htons(IP_MF); 14360c19f846SWillem de Bruijn offset += skb->len - nhoff - ihl; 14370c19f846SWillem de Bruijn tot_len = skb->len - nhoff; 14380c19f846SWillem de Bruijn } else if (skb_is_gso(skb)) { 1439802ab55aSAlexander Duyck if (!fixedid) { 1440802ab55aSAlexander Duyck iph->id = htons(id); 1441802ab55aSAlexander Duyck id += skb_shinfo(skb)->gso_segs; 1442490ab081SPravin B Shelar } 144307b26c94SSteffen Klassert 144407b26c94SSteffen Klassert if (gso_partial) 1445802ab55aSAlexander Duyck tot_len = skb_shinfo(skb)->gso_size + 1446802ab55aSAlexander Duyck SKB_GSO_CB(skb)->data_offset + 1447802ab55aSAlexander Duyck skb->head - (unsigned char *)iph; 144807b26c94SSteffen Klassert else 144907b26c94SSteffen Klassert tot_len = skb->len - nhoff; 1450802ab55aSAlexander Duyck } else { 1451802ab55aSAlexander Duyck if (!fixedid) 1452802ab55aSAlexander Duyck iph->id = htons(id++); 1453802ab55aSAlexander Duyck tot_len = skb->len - nhoff; 1454802ab55aSAlexander Duyck } 1455802ab55aSAlexander Duyck iph->tot_len = htons(tot_len); 145647d27aadSEric Dumazet ip_send_check(iph); 14578c3a897bSEric Dumazet if (encap) 14583347c960SEric Dumazet skb_reset_inner_headers(skb); 14593347c960SEric Dumazet skb->network_header = (u8 *)iph - skb->head; 1460c56cae23SToke Høiland-Jørgensen skb_reset_mac_len(skb); 1461f4c50d99SHerbert Xu } while ((skb = skb->next)); 1462f4c50d99SHerbert Xu 1463f4c50d99SHerbert Xu out: 1464f4c50d99SHerbert Xu return segs; 1465f4c50d99SHerbert Xu } 1466f4c50d99SHerbert Xu 1467418e897eSWillem de Bruijn static struct sk_buff *ipip_gso_segment(struct sk_buff *skb, 1468418e897eSWillem de Bruijn netdev_features_t features) 1469418e897eSWillem de Bruijn { 1470418e897eSWillem de Bruijn if (!(skb_shinfo(skb)->gso_type & SKB_GSO_IPXIP4)) 1471418e897eSWillem de Bruijn return ERR_PTR(-EINVAL); 1472418e897eSWillem de Bruijn 1473418e897eSWillem de Bruijn return inet_gso_segment(skb, features); 1474418e897eSWillem de Bruijn } 1475418e897eSWillem de Bruijn 1476d4546c25SDavid Miller struct sk_buff *inet_gro_receive(struct list_head *head, struct sk_buff *skb) 147773cc19f1SHerbert Xu { 1478bca49f84SVlad Yasevich const struct net_offload *ops; 1479d4546c25SDavid Miller struct sk_buff *pp = NULL; 1480b71d1d42SEric Dumazet const struct iphdr *iph; 1481d4546c25SDavid Miller struct sk_buff *p; 1482a5b1cf28SHerbert Xu unsigned int hlen; 1483a5b1cf28SHerbert Xu unsigned int off; 14841075f3f6SHerbert Xu unsigned int id; 148573cc19f1SHerbert Xu int flush = 1; 148673cc19f1SHerbert Xu int proto; 148773cc19f1SHerbert Xu 1488a5b1cf28SHerbert Xu off = skb_gro_offset(skb); 1489a5b1cf28SHerbert Xu hlen = off + sizeof(*iph); 149035ffb665SRichard Gobert iph = skb_gro_header(skb, hlen, off); 149186911732SHerbert Xu if (unlikely(!iph)) 149273cc19f1SHerbert Xu goto out; 149373cc19f1SHerbert Xu 1494f9242b6bSDavid S. Miller proto = iph->protocol; 149573cc19f1SHerbert Xu 1496bca49f84SVlad Yasevich ops = rcu_dereference(inet_offloads[proto]); 1497f191a1d1SVlad Yasevich if (!ops || !ops->callbacks.gro_receive) 1498fc1ca334SEric Dumazet goto out; 149973cc19f1SHerbert Xu 1500a5ad24beSHerbert Xu if (*(u8 *)iph != 0x45) 1501fc1ca334SEric Dumazet goto out; 150273cc19f1SHerbert Xu 15039b83e031SSteffen Klassert if (ip_is_fragment(iph)) 1504fc1ca334SEric Dumazet goto out; 15059b83e031SSteffen Klassert 1506a9e050f4SEric Dumazet if (unlikely(ip_fast_csum((u8 *)iph, 5))) 1507fc1ca334SEric Dumazet goto out; 150873cc19f1SHerbert Xu 1509b1a78b9bSXin Long NAPI_GRO_CB(skb)->proto = proto; 15100eae88f3SEric Dumazet id = ntohl(*(__be32 *)&iph->id); 1511db8caf3dSEric Dumazet flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF)); 15121075f3f6SHerbert Xu id >>= 16; 151373cc19f1SHerbert Xu 1514d4546c25SDavid Miller list_for_each_entry(p, head, list) { 151573cc19f1SHerbert Xu struct iphdr *iph2; 15161530545eSAlexander Duyck u16 flush_id; 151773cc19f1SHerbert Xu 151873cc19f1SHerbert Xu if (!NAPI_GRO_CB(p)->same_flow) 151973cc19f1SHerbert Xu continue; 152073cc19f1SHerbert Xu 1521299603e8SJerry Chu iph2 = (struct iphdr *)(p->data + off); 1522299603e8SJerry Chu /* The above works because, with the exception of the top 1523299603e8SJerry Chu * (inner most) layer, we only aggregate pkts with the same 1524299603e8SJerry Chu * hdr length so all the hdrs we'll need to verify will start 1525299603e8SJerry Chu * at the same offset. 1526299603e8SJerry Chu */ 1527a5ad24beSHerbert Xu if ((iph->protocol ^ iph2->protocol) | 15280eae88f3SEric Dumazet ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) | 15290eae88f3SEric Dumazet ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) { 153073cc19f1SHerbert Xu NAPI_GRO_CB(p)->same_flow = 0; 153173cc19f1SHerbert Xu continue; 153273cc19f1SHerbert Xu } 153373cc19f1SHerbert Xu 153473cc19f1SHerbert Xu /* All fields must match except length and checksum. */ 153573cc19f1SHerbert Xu NAPI_GRO_CB(p)->flush |= 1536a5ad24beSHerbert Xu (iph->ttl ^ iph2->ttl) | 1537a9e050f4SEric Dumazet (iph->tos ^ iph2->tos) | 1538bf5a755fSJerry Chu ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF)); 153973cc19f1SHerbert Xu 154073cc19f1SHerbert Xu NAPI_GRO_CB(p)->flush |= flush; 15411530545eSAlexander Duyck 15421530545eSAlexander Duyck /* We need to store of the IP ID check to be included later 15431530545eSAlexander Duyck * when we can verify that this packet does in fact belong 15441530545eSAlexander Duyck * to a given flow. 15451530545eSAlexander Duyck */ 15461530545eSAlexander Duyck flush_id = (u16)(id - ntohs(iph2->id)); 15471530545eSAlexander Duyck 15481530545eSAlexander Duyck /* This bit of code makes it much easier for us to identify 15491530545eSAlexander Duyck * the cases where we are doing atomic vs non-atomic IP ID 15501530545eSAlexander Duyck * checks. Specifically an atomic check can return IP ID 15511530545eSAlexander Duyck * values 0 - 0xFFFF, while a non-atomic check can only 15521530545eSAlexander Duyck * return 0 or 0xFFFF. 15531530545eSAlexander Duyck */ 15541530545eSAlexander Duyck if (!NAPI_GRO_CB(p)->is_atomic || 15551530545eSAlexander Duyck !(iph->frag_off & htons(IP_DF))) { 15561530545eSAlexander Duyck flush_id ^= NAPI_GRO_CB(p)->count; 15571530545eSAlexander Duyck flush_id = flush_id ? 0xFFFF : 0; 155873cc19f1SHerbert Xu } 155973cc19f1SHerbert Xu 15601530545eSAlexander Duyck /* If the previous IP ID value was based on an atomic 15611530545eSAlexander Duyck * datagram we can overwrite the value and ignore it. 15621530545eSAlexander Duyck */ 15631530545eSAlexander Duyck if (NAPI_GRO_CB(skb)->is_atomic) 15641530545eSAlexander Duyck NAPI_GRO_CB(p)->flush_id = flush_id; 15651530545eSAlexander Duyck else 15661530545eSAlexander Duyck NAPI_GRO_CB(p)->flush_id |= flush_id; 15671530545eSAlexander Duyck } 15681530545eSAlexander Duyck 15691530545eSAlexander Duyck NAPI_GRO_CB(skb)->is_atomic = !!(iph->frag_off & htons(IP_DF)); 157073cc19f1SHerbert Xu NAPI_GRO_CB(skb)->flush |= flush; 1571299603e8SJerry Chu skb_set_network_header(skb, off); 1572299603e8SJerry Chu /* The above will be needed by the transport layer if there is one 1573299603e8SJerry Chu * immediately following this IP hdr. 1574299603e8SJerry Chu */ 1575299603e8SJerry Chu 15762c804d0fSEric Dumazet /* Note : No need to call skb_gro_postpull_rcsum() here, 15772c804d0fSEric Dumazet * as we already checked checksum over ipv4 header was 0 15782c804d0fSEric Dumazet */ 157986911732SHerbert Xu skb_gro_pull(skb, sizeof(*iph)); 158086911732SHerbert Xu skb_set_transport_header(skb, skb_gro_offset(skb)); 158173cc19f1SHerbert Xu 1582028e0a47SPaolo Abeni pp = indirect_call_gro_receive(tcp4_gro_receive, udp4_gro_receive, 1583028e0a47SPaolo Abeni ops->callbacks.gro_receive, head, skb); 158473cc19f1SHerbert Xu 158573cc19f1SHerbert Xu out: 15865f114163SSteffen Klassert skb_gro_flush_final(skb, pp, flush); 158773cc19f1SHerbert Xu 158873cc19f1SHerbert Xu return pp; 158973cc19f1SHerbert Xu } 159073cc19f1SHerbert Xu 1591d4546c25SDavid Miller static struct sk_buff *ipip_gro_receive(struct list_head *head, 1592fac8e0f5SJesse Gross struct sk_buff *skb) 1593fac8e0f5SJesse Gross { 1594fac8e0f5SJesse Gross if (NAPI_GRO_CB(skb)->encap_mark) { 1595fac8e0f5SJesse Gross NAPI_GRO_CB(skb)->flush = 1; 1596fac8e0f5SJesse Gross return NULL; 1597fac8e0f5SJesse Gross } 1598fac8e0f5SJesse Gross 1599fac8e0f5SJesse Gross NAPI_GRO_CB(skb)->encap_mark = 1; 1600fac8e0f5SJesse Gross 1601fac8e0f5SJesse Gross return inet_gro_receive(head, skb); 1602fac8e0f5SJesse Gross } 1603fac8e0f5SJesse Gross 1604822c8685SDeepa Dinamani #define SECONDS_PER_DAY 86400 1605822c8685SDeepa Dinamani 1606822c8685SDeepa Dinamani /* inet_current_timestamp - Return IP network timestamp 1607822c8685SDeepa Dinamani * 1608822c8685SDeepa Dinamani * Return milliseconds since midnight in network byte order. 1609822c8685SDeepa Dinamani */ 1610822c8685SDeepa Dinamani __be32 inet_current_timestamp(void) 1611822c8685SDeepa Dinamani { 1612822c8685SDeepa Dinamani u32 secs; 1613822c8685SDeepa Dinamani u32 msecs; 1614822c8685SDeepa Dinamani struct timespec64 ts; 1615822c8685SDeepa Dinamani 1616822c8685SDeepa Dinamani ktime_get_real_ts64(&ts); 1617822c8685SDeepa Dinamani 1618822c8685SDeepa Dinamani /* Get secs since midnight. */ 1619822c8685SDeepa Dinamani (void)div_u64_rem(ts.tv_sec, SECONDS_PER_DAY, &secs); 1620822c8685SDeepa Dinamani /* Convert to msecs. */ 1621822c8685SDeepa Dinamani msecs = secs * MSEC_PER_SEC; 1622822c8685SDeepa Dinamani /* Convert nsec to msec. */ 1623822c8685SDeepa Dinamani msecs += (u32)ts.tv_nsec / NSEC_PER_MSEC; 1624822c8685SDeepa Dinamani 1625822c8685SDeepa Dinamani /* Convert to network byte order. */ 16263ba9d300SDeepa Dinamani return htonl(msecs); 1627822c8685SDeepa Dinamani } 1628822c8685SDeepa Dinamani EXPORT_SYMBOL(inet_current_timestamp); 1629822c8685SDeepa Dinamani 1630f4713a3dSWillem de Bruijn int inet_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len) 1631f4713a3dSWillem de Bruijn { 1632f4713a3dSWillem de Bruijn if (sk->sk_family == AF_INET) 1633f4713a3dSWillem de Bruijn return ip_recv_error(sk, msg, len, addr_len); 1634f4713a3dSWillem de Bruijn #if IS_ENABLED(CONFIG_IPV6) 1635f4713a3dSWillem de Bruijn if (sk->sk_family == AF_INET6) 1636f4713a3dSWillem de Bruijn return pingv6_ops.ipv6_recv_error(sk, msg, len, addr_len); 1637f4713a3dSWillem de Bruijn #endif 1638f4713a3dSWillem de Bruijn return -EINVAL; 1639f4713a3dSWillem de Bruijn } 1640f4713a3dSWillem de Bruijn 1641b8921ca8STom Herbert int inet_gro_complete(struct sk_buff *skb, int nhoff) 164273cc19f1SHerbert Xu { 1643299603e8SJerry Chu struct iphdr *iph = (struct iphdr *)(skb->data + nhoff); 1644bca49f84SVlad Yasevich const struct net_offload *ops; 1645b1a78b9bSXin Long __be16 totlen = iph->tot_len; 1646f9242b6bSDavid S. Miller int proto = iph->protocol; 1647f9242b6bSDavid S. Miller int err = -ENOSYS; 164873cc19f1SHerbert Xu 1649294acf1cSPaolo Abeni if (skb->encapsulation) { 1650294acf1cSPaolo Abeni skb_set_inner_protocol(skb, cpu_to_be16(ETH_P_IP)); 1651c3caf119SJerry Chu skb_set_inner_network_header(skb, nhoff); 1652294acf1cSPaolo Abeni } 1653c3caf119SJerry Chu 1654b1a78b9bSXin Long iph_set_totlen(iph, skb->len - nhoff); 1655b1a78b9bSXin Long csum_replace2(&iph->check, totlen, iph->tot_len); 165673cc19f1SHerbert Xu 1657bca49f84SVlad Yasevich ops = rcu_dereference(inet_offloads[proto]); 1658f191a1d1SVlad Yasevich if (WARN_ON(!ops || !ops->callbacks.gro_complete)) 1659627b94f7SEric Dumazet goto out; 166073cc19f1SHerbert Xu 1661299603e8SJerry Chu /* Only need to add sizeof(*iph) to get to the next hdr below 1662299603e8SJerry Chu * because any hdr with option will have been flushed in 1663299603e8SJerry Chu * inet_gro_receive(). 1664299603e8SJerry Chu */ 1665028e0a47SPaolo Abeni err = INDIRECT_CALL_2(ops->callbacks.gro_complete, 1666028e0a47SPaolo Abeni tcp4_gro_complete, udp4_gro_complete, 1667028e0a47SPaolo Abeni skb, nhoff + sizeof(*iph)); 166873cc19f1SHerbert Xu 1669627b94f7SEric Dumazet out: 167073cc19f1SHerbert Xu return err; 167173cc19f1SHerbert Xu } 167273cc19f1SHerbert Xu 1673b8cba75bSJesse Gross static int ipip_gro_complete(struct sk_buff *skb, int nhoff) 1674b8cba75bSJesse Gross { 1675b8cba75bSJesse Gross skb->encapsulation = 1; 16767e13318dSTom Herbert skb_shinfo(skb)->gso_type |= SKB_GSO_IPXIP4; 1677b8cba75bSJesse Gross return inet_gro_complete(skb, nhoff); 1678b8cba75bSJesse Gross } 1679b8cba75bSJesse Gross 1680eee4fe4dSDenis V. Lunev int inet_ctl_sock_create(struct sock **sk, unsigned short family, 16815677242fSDenis V. Lunev unsigned short type, unsigned char protocol, 16825677242fSDenis V. Lunev struct net *net) 16833d58b5faSDenis V. Lunev { 1684eee4fe4dSDenis V. Lunev struct socket *sock; 168526abe143SEric W. Biederman int rc = sock_create_kern(net, family, type, protocol, &sock); 16863d58b5faSDenis V. Lunev 16873d58b5faSDenis V. Lunev if (rc == 0) { 1688eee4fe4dSDenis V. Lunev *sk = sock->sk; 1689eee4fe4dSDenis V. Lunev (*sk)->sk_allocation = GFP_ATOMIC; 16901ac88557SEric Dumazet (*sk)->sk_use_task_frag = false; 16913d58b5faSDenis V. Lunev /* 16923d58b5faSDenis V. Lunev * Unhash it so that IP input processing does not even see it, 16933d58b5faSDenis V. Lunev * we do not wish this socket to see incoming packets. 16943d58b5faSDenis V. Lunev */ 1695eee4fe4dSDenis V. Lunev (*sk)->sk_prot->unhash(*sk); 16963d58b5faSDenis V. Lunev } 16973d58b5faSDenis V. Lunev return rc; 16983d58b5faSDenis V. Lunev } 16993d58b5faSDenis V. Lunev EXPORT_SYMBOL_GPL(inet_ctl_sock_create); 17003d58b5faSDenis V. Lunev 1701698365faSWANG Cong unsigned long snmp_fold_field(void __percpu *mib, int offt) 17025e0f0435SHerbert Xu { 17035e0f0435SHerbert Xu unsigned long res = 0; 1704698365faSWANG Cong int i; 17055e0f0435SHerbert Xu 1706698365faSWANG Cong for_each_possible_cpu(i) 1707c4c6bc31SRaghavendra K T res += snmp_get_cpu_field(mib, i, offt); 17085e0f0435SHerbert Xu return res; 17095e0f0435SHerbert Xu } 17105e0f0435SHerbert Xu EXPORT_SYMBOL_GPL(snmp_fold_field); 17115e0f0435SHerbert Xu 17124ce3c183SEric Dumazet #if BITS_PER_LONG==32 17134ce3c183SEric Dumazet 171480ec1927SDavid S. Miller u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offt, 1715c4c6bc31SRaghavendra K T size_t syncp_offset) 17164ce3c183SEric Dumazet { 17178f0ea0feSEric Dumazet void *bhptr; 17184ce3c183SEric Dumazet struct u64_stats_sync *syncp; 17198f0ea0feSEric Dumazet u64 v; 17204ce3c183SEric Dumazet unsigned int start; 17214ce3c183SEric Dumazet 1722698365faSWANG Cong bhptr = per_cpu_ptr(mib, cpu); 17234ce3c183SEric Dumazet syncp = (struct u64_stats_sync *)(bhptr + syncp_offset); 17244ce3c183SEric Dumazet do { 1725d120d1a6SThomas Gleixner start = u64_stats_fetch_begin(syncp); 17268f0ea0feSEric Dumazet v = *(((u64 *)bhptr) + offt); 1727d120d1a6SThomas Gleixner } while (u64_stats_fetch_retry(syncp, start)); 17284ce3c183SEric Dumazet 1729c4c6bc31SRaghavendra K T return v; 1730c4c6bc31SRaghavendra K T } 1731c4c6bc31SRaghavendra K T EXPORT_SYMBOL_GPL(snmp_get_cpu_field64); 1732c4c6bc31SRaghavendra K T 1733c4c6bc31SRaghavendra K T u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset) 1734c4c6bc31SRaghavendra K T { 1735c4c6bc31SRaghavendra K T u64 res = 0; 1736c4c6bc31SRaghavendra K T int cpu; 1737c4c6bc31SRaghavendra K T 1738c4c6bc31SRaghavendra K T for_each_possible_cpu(cpu) { 1739d1bfc625SMadalin Bucur res += snmp_get_cpu_field64(mib, cpu, offt, syncp_offset); 17404ce3c183SEric Dumazet } 17414ce3c183SEric Dumazet return res; 17424ce3c183SEric Dumazet } 17434ce3c183SEric Dumazet EXPORT_SYMBOL_GPL(snmp_fold_field64); 17444ce3c183SEric Dumazet #endif 17454ce3c183SEric Dumazet 17461da177e4SLinus Torvalds #ifdef CONFIG_IP_MULTICAST 174732613090SAlexey Dobriyan static const struct net_protocol igmp_protocol = { 17481da177e4SLinus Torvalds .handler = igmp_rcv, 17491da177e4SLinus Torvalds }; 17501da177e4SLinus Torvalds #endif 17511da177e4SLinus Torvalds 175211052589SKuniyuki Iwashima static const struct net_protocol tcp_protocol = { 17531da177e4SLinus Torvalds .handler = tcp_v4_rcv, 17541da177e4SLinus Torvalds .err_handler = tcp_v4_err, 17551da177e4SLinus Torvalds .no_policy = 1, 17568ed1dc44SHannes Frederic Sowa .icmp_strict_tag_validation = 1, 17571da177e4SLinus Torvalds }; 17581da177e4SLinus Torvalds 175911052589SKuniyuki Iwashima static const struct net_protocol udp_protocol = { 17601da177e4SLinus Torvalds .handler = udp_rcv, 17611da177e4SLinus Torvalds .err_handler = udp_err, 17621da177e4SLinus Torvalds .no_policy = 1, 17631da177e4SLinus Torvalds }; 17641da177e4SLinus Torvalds 176532613090SAlexey Dobriyan static const struct net_protocol icmp_protocol = { 17661da177e4SLinus Torvalds .handler = icmp_rcv, 17675b052042SLi Wei .err_handler = icmp_err, 17688b7817f3SHerbert Xu .no_policy = 1, 17691da177e4SLinus Torvalds }; 17701da177e4SLinus Torvalds 17719b4661bdSPavel Emelyanov static __net_init int ipv4_mib_init_net(struct net *net) 17729b4661bdSPavel Emelyanov { 1773827da44cSJohn Stultz int i; 1774827da44cSJohn Stultz 1775698365faSWANG Cong net->mib.tcp_statistics = alloc_percpu(struct tcp_mib); 1776698365faSWANG Cong if (!net->mib.tcp_statistics) 177757ef42d5SPavel Emelyanov goto err_tcp_mib; 1778698365faSWANG Cong net->mib.ip_statistics = alloc_percpu(struct ipstats_mib); 1779698365faSWANG Cong if (!net->mib.ip_statistics) 1780a20f5799SPavel Emelyanov goto err_ip_mib; 1781827da44cSJohn Stultz 1782827da44cSJohn Stultz for_each_possible_cpu(i) { 1783827da44cSJohn Stultz struct ipstats_mib *af_inet_stats; 1784698365faSWANG Cong af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i); 1785827da44cSJohn Stultz u64_stats_init(&af_inet_stats->syncp); 1786827da44cSJohn Stultz } 1787827da44cSJohn Stultz 1788698365faSWANG Cong net->mib.net_statistics = alloc_percpu(struct linux_mib); 1789698365faSWANG Cong if (!net->mib.net_statistics) 179061a7e260SPavel Emelyanov goto err_net_mib; 1791698365faSWANG Cong net->mib.udp_statistics = alloc_percpu(struct udp_mib); 1792698365faSWANG Cong if (!net->mib.udp_statistics) 17932f275f91SPavel Emelyanov goto err_udp_mib; 1794698365faSWANG Cong net->mib.udplite_statistics = alloc_percpu(struct udp_mib); 1795698365faSWANG Cong if (!net->mib.udplite_statistics) 1796386019d3SPavel Emelyanov goto err_udplite_mib; 1797698365faSWANG Cong net->mib.icmp_statistics = alloc_percpu(struct icmp_mib); 1798698365faSWANG Cong if (!net->mib.icmp_statistics) 1799b60538a0SPavel Emelyanov goto err_icmp_mib; 1800acb32ba3SEric Dumazet net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib), 1801acb32ba3SEric Dumazet GFP_KERNEL); 1802acb32ba3SEric Dumazet if (!net->mib.icmpmsg_statistics) 1803923c6586SPavel Emelyanov goto err_icmpmsg_mib; 180457ef42d5SPavel Emelyanov 180557ef42d5SPavel Emelyanov tcp_mib_init(net); 18069b4661bdSPavel Emelyanov return 0; 180757ef42d5SPavel Emelyanov 1808923c6586SPavel Emelyanov err_icmpmsg_mib: 1809698365faSWANG Cong free_percpu(net->mib.icmp_statistics); 1810b60538a0SPavel Emelyanov err_icmp_mib: 1811698365faSWANG Cong free_percpu(net->mib.udplite_statistics); 1812386019d3SPavel Emelyanov err_udplite_mib: 1813698365faSWANG Cong free_percpu(net->mib.udp_statistics); 18142f275f91SPavel Emelyanov err_udp_mib: 1815698365faSWANG Cong free_percpu(net->mib.net_statistics); 181661a7e260SPavel Emelyanov err_net_mib: 1817698365faSWANG Cong free_percpu(net->mib.ip_statistics); 1818a20f5799SPavel Emelyanov err_ip_mib: 1819698365faSWANG Cong free_percpu(net->mib.tcp_statistics); 182057ef42d5SPavel Emelyanov err_tcp_mib: 182157ef42d5SPavel Emelyanov return -ENOMEM; 18229b4661bdSPavel Emelyanov } 18239b4661bdSPavel Emelyanov 18249b4661bdSPavel Emelyanov static __net_exit void ipv4_mib_exit_net(struct net *net) 18259b4661bdSPavel Emelyanov { 1826acb32ba3SEric Dumazet kfree(net->mib.icmpmsg_statistics); 1827698365faSWANG Cong free_percpu(net->mib.icmp_statistics); 1828698365faSWANG Cong free_percpu(net->mib.udplite_statistics); 1829698365faSWANG Cong free_percpu(net->mib.udp_statistics); 1830698365faSWANG Cong free_percpu(net->mib.net_statistics); 1831698365faSWANG Cong free_percpu(net->mib.ip_statistics); 1832698365faSWANG Cong free_percpu(net->mib.tcp_statistics); 1833fc518953SFlorian Westphal #ifdef CONFIG_MPTCP 1834fc518953SFlorian Westphal /* allocated on demand, see mptcp_init_sock() */ 1835fc518953SFlorian Westphal free_percpu(net->mib.mptcp_statistics); 1836fc518953SFlorian Westphal #endif 18379b4661bdSPavel Emelyanov } 18389b4661bdSPavel Emelyanov 18399b4661bdSPavel Emelyanov static __net_initdata struct pernet_operations ipv4_mib_ops = { 18409b4661bdSPavel Emelyanov .init = ipv4_mib_init_net, 18419b4661bdSPavel Emelyanov .exit = ipv4_mib_exit_net, 18429b4661bdSPavel Emelyanov }; 18439b4661bdSPavel Emelyanov 18441da177e4SLinus Torvalds static int __init init_ipv4_mibs(void) 18451da177e4SLinus Torvalds { 1846d89cbbb1SPavel Emelyanov return register_pernet_subsys(&ipv4_mib_ops); 18471da177e4SLinus Torvalds } 18481da177e4SLinus Torvalds 1849c9d8f1a6SCong Wang static __net_init int inet_init_net(struct net *net) 1850c9d8f1a6SCong Wang { 1851c9d8f1a6SCong Wang /* 1852c9d8f1a6SCong Wang * Set defaults for local port range 1853c9d8f1a6SCong Wang */ 1854c9d8f1a6SCong Wang seqlock_init(&net->ipv4.ip_local_ports.lock); 1855c9d8f1a6SCong Wang net->ipv4.ip_local_ports.range[0] = 32768; 185607f4c900SEric Dumazet net->ipv4.ip_local_ports.range[1] = 60999; 1857ba6b918aSCong Wang 1858ba6b918aSCong Wang seqlock_init(&net->ipv4.ping_group_range.lock); 1859ba6b918aSCong Wang /* 1860ba6b918aSCong Wang * Sane defaults - nobody may create ping sockets. 1861ba6b918aSCong Wang * Boot scripts should set this to distro-specific group. 1862ba6b918aSCong Wang */ 1863ba6b918aSCong Wang net->ipv4.ping_group_range.range[0] = make_kgid(&init_user_ns, 1); 1864ba6b918aSCong Wang net->ipv4.ping_group_range.range[1] = make_kgid(&init_user_ns, 0); 1865049bbf58SEzequiel Garcia 1866049bbf58SEzequiel Garcia /* Default values for sysctl-controlled parameters. 1867049bbf58SEzequiel Garcia * We set them here, in case sysctl is not compiled. 1868049bbf58SEzequiel Garcia */ 1869049bbf58SEzequiel Garcia net->ipv4.sysctl_ip_default_ttl = IPDEFTTL; 1870432e05d3SPetr Machata net->ipv4.sysctl_ip_fwd_update_priority = 1; 1871049bbf58SEzequiel Garcia net->ipv4.sysctl_ip_dynaddr = 0; 1872049bbf58SEzequiel Garcia net->ipv4.sysctl_ip_early_demux = 1; 1873dddb64bcSsubashab@codeaurora.org net->ipv4.sysctl_udp_early_demux = 1; 1874dddb64bcSsubashab@codeaurora.org net->ipv4.sysctl_tcp_early_demux = 1; 18754f80116dSRoopa Prabhu net->ipv4.sysctl_nexthop_compat_mode = 1; 18764548b683SKrister Johansen #ifdef CONFIG_SYSCTL 18774548b683SKrister Johansen net->ipv4.sysctl_ip_prot_sock = PROT_SOCK; 18784548b683SKrister Johansen #endif 1879049bbf58SEzequiel Garcia 18801714020eSNikolay Borisov /* Some igmp sysctl, whose values are always used */ 18811714020eSNikolay Borisov net->ipv4.sysctl_igmp_max_memberships = 20; 18821714020eSNikolay Borisov net->ipv4.sysctl_igmp_max_msf = 10; 18831714020eSNikolay Borisov /* IGMP reports for link-local multicast groups are enabled by default */ 18841714020eSNikolay Borisov net->ipv4.sysctl_igmp_llm_reports = 1; 18851714020eSNikolay Borisov net->ipv4.sysctl_igmp_qrv = 2; 18861714020eSNikolay Borisov 1887680aea08SAmit Cohen net->ipv4.sysctl_fib_notify_on_flag_change = 0; 1888680aea08SAmit Cohen 1889c9d8f1a6SCong Wang return 0; 1890c9d8f1a6SCong Wang } 1891c9d8f1a6SCong Wang 1892c9d8f1a6SCong Wang static __net_initdata struct pernet_operations af_inet_ops = { 1893c9d8f1a6SCong Wang .init = inet_init_net, 1894c9d8f1a6SCong Wang }; 1895c9d8f1a6SCong Wang 1896c9d8f1a6SCong Wang static int __init init_inet_pernet_ops(void) 1897c9d8f1a6SCong Wang { 1898c9d8f1a6SCong Wang return register_pernet_subsys(&af_inet_ops); 1899c9d8f1a6SCong Wang } 1900c9d8f1a6SCong Wang 19011da177e4SLinus Torvalds static int ipv4_proc_init(void); 19021da177e4SLinus Torvalds 190330e224d7SHerbert Xu /* 190430e224d7SHerbert Xu * IP protocol layer initialiser 190530e224d7SHerbert Xu */ 190630e224d7SHerbert Xu 190722061d80SVlad Yasevich static struct packet_offload ip_packet_offload __read_mostly = { 190822061d80SVlad Yasevich .type = cpu_to_be16(ETH_P_IP), 1909f191a1d1SVlad Yasevich .callbacks = { 1910f4c50d99SHerbert Xu .gso_segment = inet_gso_segment, 191173cc19f1SHerbert Xu .gro_receive = inet_gro_receive, 191273cc19f1SHerbert Xu .gro_complete = inet_gro_complete, 1913f191a1d1SVlad Yasevich }, 191430e224d7SHerbert Xu }; 191530e224d7SHerbert Xu 1916cb32f511SEric Dumazet static const struct net_offload ipip_offload = { 1917cb32f511SEric Dumazet .callbacks = { 1918418e897eSWillem de Bruijn .gso_segment = ipip_gso_segment, 1919fac8e0f5SJesse Gross .gro_receive = ipip_gro_receive, 1920b8cba75bSJesse Gross .gro_complete = ipip_gro_complete, 1921cb32f511SEric Dumazet }, 1922cb32f511SEric Dumazet }; 1923cb32f511SEric Dumazet 192493b1b31fSTonghao Zhang static int __init ipip_offload_init(void) 192593b1b31fSTonghao Zhang { 192693b1b31fSTonghao Zhang return inet_add_offload(&ipip_offload, IPPROTO_IPIP); 192793b1b31fSTonghao Zhang } 192893b1b31fSTonghao Zhang 1929808a8f88SVlad Yasevich static int __init ipv4_offload_init(void) 1930808a8f88SVlad Yasevich { 1931808a8f88SVlad Yasevich /* 1932808a8f88SVlad Yasevich * Add offloads 1933808a8f88SVlad Yasevich */ 1934da5bab07SDaniel Borkmann if (udpv4_offload_init() < 0) 1935808a8f88SVlad Yasevich pr_crit("%s: Cannot add UDP protocol offload\n", __func__); 193628850dc7SDaniel Borkmann if (tcpv4_offload_init() < 0) 193728850dc7SDaniel Borkmann pr_crit("%s: Cannot add TCP protocol offload\n", __func__); 193893b1b31fSTonghao Zhang if (ipip_offload_init() < 0) 193993b1b31fSTonghao Zhang pr_crit("%s: Cannot add IPIP protocol offload\n", __func__); 1940808a8f88SVlad Yasevich 1941808a8f88SVlad Yasevich dev_add_offload(&ip_packet_offload); 1942808a8f88SVlad Yasevich return 0; 1943808a8f88SVlad Yasevich } 1944808a8f88SVlad Yasevich 1945808a8f88SVlad Yasevich fs_initcall(ipv4_offload_init); 1946808a8f88SVlad Yasevich 1947808a8f88SVlad Yasevich static struct packet_type ip_packet_type __read_mostly = { 1948808a8f88SVlad Yasevich .type = cpu_to_be16(ETH_P_IP), 1949808a8f88SVlad Yasevich .func = ip_rcv, 195017266ee9SEdward Cree .list_func = ip_list_rcv, 1951808a8f88SVlad Yasevich }; 1952808a8f88SVlad Yasevich 19531da177e4SLinus Torvalds static int __init inet_init(void) 19541da177e4SLinus Torvalds { 19551da177e4SLinus Torvalds struct inet_protosw *q; 19561da177e4SLinus Torvalds struct list_head *r; 1957b6246f4dSColin Ian King int rc; 19581da177e4SLinus Torvalds 1959b4772ef8SEyal Birger sock_skb_cb_check_size(sizeof(struct inet_skb_parm)); 19601da177e4SLinus Torvalds 19610daf07e5SEric Dumazet raw_hashinfo_init(&raw_v4_hashinfo); 19620daf07e5SEric Dumazet 19631da177e4SLinus Torvalds rc = proto_register(&tcp_prot, 1); 19641da177e4SLinus Torvalds if (rc) 1965122ff243SWANG Cong goto out; 19661da177e4SLinus Torvalds 19671da177e4SLinus Torvalds rc = proto_register(&udp_prot, 1); 19681da177e4SLinus Torvalds if (rc) 19691da177e4SLinus Torvalds goto out_unregister_tcp_proto; 19701da177e4SLinus Torvalds 19711da177e4SLinus Torvalds rc = proto_register(&raw_prot, 1); 19721da177e4SLinus Torvalds if (rc) 19731da177e4SLinus Torvalds goto out_unregister_udp_proto; 19741da177e4SLinus Torvalds 1975c319b4d7SVasiliy Kulikov rc = proto_register(&ping_prot, 1); 1976c319b4d7SVasiliy Kulikov if (rc) 1977c319b4d7SVasiliy Kulikov goto out_unregister_raw_proto; 1978c319b4d7SVasiliy Kulikov 19791da177e4SLinus Torvalds /* 19801da177e4SLinus Torvalds * Tell SOCKET that we are alive... 19811da177e4SLinus Torvalds */ 19821da177e4SLinus Torvalds 19831da177e4SLinus Torvalds (void)sock_register(&inet_family_ops); 19841da177e4SLinus Torvalds 1985bd7b1533SAl Viro #ifdef CONFIG_SYSCTL 1986bd7b1533SAl Viro ip_static_sysctl_init(); 1987bd7b1533SAl Viro #endif 1988bd7b1533SAl Viro 19891da177e4SLinus Torvalds /* 19901da177e4SLinus Torvalds * Add all the base protocols. 19911da177e4SLinus Torvalds */ 19921da177e4SLinus Torvalds 19931da177e4SLinus Torvalds if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0) 1994058bd4d2SJoe Perches pr_crit("%s: Cannot add ICMP protocol\n", __func__); 19951da177e4SLinus Torvalds if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0) 1996058bd4d2SJoe Perches pr_crit("%s: Cannot add UDP protocol\n", __func__); 19971da177e4SLinus Torvalds if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0) 1998058bd4d2SJoe Perches pr_crit("%s: Cannot add TCP protocol\n", __func__); 19991da177e4SLinus Torvalds #ifdef CONFIG_IP_MULTICAST 20001da177e4SLinus Torvalds if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0) 2001058bd4d2SJoe Perches pr_crit("%s: Cannot add IGMP protocol\n", __func__); 20021da177e4SLinus Torvalds #endif 20031da177e4SLinus Torvalds 20041da177e4SLinus Torvalds /* Register the socket-side information for inet_create. */ 20051da177e4SLinus Torvalds for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r) 20061da177e4SLinus Torvalds INIT_LIST_HEAD(r); 20071da177e4SLinus Torvalds 20081da177e4SLinus Torvalds for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q) 20091da177e4SLinus Torvalds inet_register_protosw(q); 20101da177e4SLinus Torvalds 20111da177e4SLinus Torvalds /* 20121da177e4SLinus Torvalds * Set the ARP module up 20131da177e4SLinus Torvalds */ 20141da177e4SLinus Torvalds 20151da177e4SLinus Torvalds arp_init(); 20161da177e4SLinus Torvalds 20171da177e4SLinus Torvalds /* 20181da177e4SLinus Torvalds * Set the IP module up 20191da177e4SLinus Torvalds */ 20201da177e4SLinus Torvalds 20211da177e4SLinus Torvalds ip_init(); 20221da177e4SLinus Torvalds 2023e22e45fcSMuchun Song /* Initialise per-cpu ipv4 mibs */ 2024e22e45fcSMuchun Song if (init_ipv4_mibs()) 2025e22e45fcSMuchun Song panic("%s: Cannot init ipv4 mibs\n", __func__); 2026e22e45fcSMuchun Song 20271da177e4SLinus Torvalds /* Setup TCP slab cache for open requests. */ 20281da177e4SLinus Torvalds tcp_init(); 20291da177e4SLinus Torvalds 203095766fffSHideo Aoki /* Setup UDP memory threshold */ 203195766fffSHideo Aoki udp_init(); 203295766fffSHideo Aoki 2033ba4e58ecSGerrit Renker /* Add UDP-Lite (RFC 3828) */ 2034ba4e58ecSGerrit Renker udplite4_register(); 20351da177e4SLinus Torvalds 20366897445fSMike Manning raw_init(); 20376897445fSMike Manning 2038c319b4d7SVasiliy Kulikov ping_init(); 2039c319b4d7SVasiliy Kulikov 20401da177e4SLinus Torvalds /* 20411da177e4SLinus Torvalds * Set the ICMP layer up 20421da177e4SLinus Torvalds */ 20431da177e4SLinus Torvalds 2044a5710d65SDenis V. Lunev if (icmp_init() < 0) 2045a5710d65SDenis V. Lunev panic("Failed to create the ICMP control socket.\n"); 20461da177e4SLinus Torvalds 20471da177e4SLinus Torvalds /* 20481da177e4SLinus Torvalds * Initialise the multicast router 20491da177e4SLinus Torvalds */ 20501da177e4SLinus Torvalds #if defined(CONFIG_IP_MROUTE) 205103d2f897SWang Chen if (ip_mr_init()) 2052058bd4d2SJoe Perches pr_crit("%s: Cannot init ipv4 mroute\n", __func__); 20531da177e4SLinus Torvalds #endif 2054c9d8f1a6SCong Wang 2055c9d8f1a6SCong Wang if (init_inet_pernet_ops()) 2056c9d8f1a6SCong Wang pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__); 20571da177e4SLinus Torvalds 20581da177e4SLinus Torvalds ipv4_proc_init(); 20591da177e4SLinus Torvalds 20601da177e4SLinus Torvalds ipfrag_init(); 20611da177e4SLinus Torvalds 206230e224d7SHerbert Xu dev_add_pack(&ip_packet_type); 206330e224d7SHerbert Xu 2064045a0fa0SThomas Graf ip_tunnel_core_init(); 2065045a0fa0SThomas Graf 20661da177e4SLinus Torvalds rc = 0; 20671da177e4SLinus Torvalds out: 20681da177e4SLinus Torvalds return rc; 2069c319b4d7SVasiliy Kulikov out_unregister_raw_proto: 2070c319b4d7SVasiliy Kulikov proto_unregister(&raw_prot); 20711da177e4SLinus Torvalds out_unregister_udp_proto: 20721da177e4SLinus Torvalds proto_unregister(&udp_prot); 2073321efff7SOlaf Kirch out_unregister_tcp_proto: 2074321efff7SOlaf Kirch proto_unregister(&tcp_prot); 20751da177e4SLinus Torvalds goto out; 20761da177e4SLinus Torvalds } 20771da177e4SLinus Torvalds 2078a536e077SHeiko Carstens fs_initcall(inet_init); 20791da177e4SLinus Torvalds 20801da177e4SLinus Torvalds /* ------------------------------------------------------------------------ */ 20811da177e4SLinus Torvalds 20821da177e4SLinus Torvalds #ifdef CONFIG_PROC_FS 20831da177e4SLinus Torvalds static int __init ipv4_proc_init(void) 20841da177e4SLinus Torvalds { 20851da177e4SLinus Torvalds int rc = 0; 20861da177e4SLinus Torvalds 20871da177e4SLinus Torvalds if (raw_proc_init()) 20881da177e4SLinus Torvalds goto out_raw; 20891da177e4SLinus Torvalds if (tcp4_proc_init()) 20901da177e4SLinus Torvalds goto out_tcp; 20911da177e4SLinus Torvalds if (udp4_proc_init()) 20921da177e4SLinus Torvalds goto out_udp; 2093c319b4d7SVasiliy Kulikov if (ping_proc_init()) 2094c319b4d7SVasiliy Kulikov goto out_ping; 20951da177e4SLinus Torvalds if (ip_misc_proc_init()) 20961da177e4SLinus Torvalds goto out_misc; 20971da177e4SLinus Torvalds out: 20981da177e4SLinus Torvalds return rc; 20991da177e4SLinus Torvalds out_misc: 2100c319b4d7SVasiliy Kulikov ping_proc_exit(); 2101c319b4d7SVasiliy Kulikov out_ping: 21021da177e4SLinus Torvalds udp4_proc_exit(); 21031da177e4SLinus Torvalds out_udp: 21041da177e4SLinus Torvalds tcp4_proc_exit(); 21051da177e4SLinus Torvalds out_tcp: 21061da177e4SLinus Torvalds raw_proc_exit(); 21071da177e4SLinus Torvalds out_raw: 21081da177e4SLinus Torvalds rc = -ENOMEM; 21091da177e4SLinus Torvalds goto out; 21101da177e4SLinus Torvalds } 21111da177e4SLinus Torvalds 21121da177e4SLinus Torvalds #else /* CONFIG_PROC_FS */ 21131da177e4SLinus Torvalds static int __init ipv4_proc_init(void) 21141da177e4SLinus Torvalds { 21151da177e4SLinus Torvalds return 0; 21161da177e4SLinus Torvalds } 21171da177e4SLinus Torvalds #endif /* CONFIG_PROC_FS */ 2118