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> 1031da177e4SLinus Torvalds #include <net/tcp.h> 1041da177e4SLinus Torvalds #include <net/udp.h> 105ba4e58ecSGerrit Renker #include <net/udplite.h> 106c319b4d7SVasiliy Kulikov #include <net/ping.h> 1071da177e4SLinus Torvalds #include <linux/skbuff.h> 1081da177e4SLinus Torvalds #include <net/sock.h> 1091da177e4SLinus Torvalds #include <net/raw.h> 1101da177e4SLinus Torvalds #include <net/icmp.h> 1111da177e4SLinus Torvalds #include <net/inet_common.h> 112045a0fa0SThomas Graf #include <net/ip_tunnels.h> 1131da177e4SLinus Torvalds #include <net/xfrm.h> 1149b4661bdSPavel Emelyanov #include <net/net_namespace.h> 115aebda156SEric Dumazet #include <net/secure_seq.h> 1161da177e4SLinus Torvalds #ifdef CONFIG_IP_MROUTE 1171da177e4SLinus Torvalds #include <linux/mroute.h> 1181da177e4SLinus Torvalds #endif 1193236b004SDavid Ahern #include <net/l3mdev.h> 120dc13c876SChristoph Hellwig #include <net/compat.h> 1211da177e4SLinus Torvalds 122563e0bb0SYafang Shao #include <trace/events/sock.h> 1231da177e4SLinus Torvalds 1241da177e4SLinus Torvalds /* The inetsw table contains everything that inet_create needs to 1251da177e4SLinus Torvalds * build a new socket. 1261da177e4SLinus Torvalds */ 1271da177e4SLinus Torvalds static struct list_head inetsw[SOCK_MAX]; 1281da177e4SLinus Torvalds static DEFINE_SPINLOCK(inetsw_lock); 1291da177e4SLinus Torvalds 1301da177e4SLinus Torvalds /* New destruction routine */ 1311da177e4SLinus Torvalds 1321da177e4SLinus Torvalds void inet_sock_destruct(struct sock *sk) 1331da177e4SLinus Torvalds { 1341da177e4SLinus Torvalds struct inet_sock *inet = inet_sk(sk); 1351da177e4SLinus Torvalds 1361da177e4SLinus Torvalds __skb_queue_purge(&sk->sk_receive_queue); 1371da177e4SLinus Torvalds __skb_queue_purge(&sk->sk_error_queue); 1381da177e4SLinus Torvalds 1392bb2f5fbSWei Wang sk_mem_reclaim_final(sk); 14095766fffSHideo Aoki 1411da177e4SLinus Torvalds if (sk->sk_type == SOCK_STREAM && sk->sk_state != TCP_CLOSE) { 1423d1427f8SEric Dumazet pr_err("Attempt to release TCP socket in state %d %p\n", 1431da177e4SLinus Torvalds sk->sk_state, sk); 1441da177e4SLinus Torvalds return; 1451da177e4SLinus Torvalds } 1461da177e4SLinus Torvalds if (!sock_flag(sk, SOCK_DEAD)) { 1473d1427f8SEric Dumazet pr_err("Attempt to release alive inet socket %p\n", sk); 1481da177e4SLinus Torvalds return; 1491da177e4SLinus Torvalds } 1501da177e4SLinus Torvalds 151547b792cSIlpo Järvinen WARN_ON(atomic_read(&sk->sk_rmem_alloc)); 15214afee4bSReshetova, Elena WARN_ON(refcount_read(&sk->sk_wmem_alloc)); 153547b792cSIlpo Järvinen WARN_ON(sk->sk_wmem_queued); 154292e6077SPaolo Abeni WARN_ON(sk_forward_alloc_get(sk)); 1551da177e4SLinus Torvalds 156f6d8bd05SEric Dumazet kfree(rcu_dereference_protected(inet->inet_opt, 1)); 1575869b8faSXin Long dst_release(rcu_dereference_protected(sk->sk_dst_cache, 1)); 1588f905c0eSEric Dumazet dst_release(rcu_dereference_protected(sk->sk_rx_dst, 1)); 159e6848976SArnaldo Carvalho de Melo sk_refcnt_debug_dec(sk); 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 */ 220e1cfcbe8SHaishuang Yan tcp_fastopen = 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 338*0968d2a4SKuniyuki 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; 3501da177e4SLinus Torvalds 3511da177e4SLinus Torvalds inet->uc_ttl = -1; 3521da177e4SLinus Torvalds inet->mc_loop = 1; 3531da177e4SLinus Torvalds inet->mc_ttl = 1; 354f771bef9SNivedita Singhvi inet->mc_all = 1; 3551da177e4SLinus Torvalds inet->mc_index = 0; 3561da177e4SLinus Torvalds inet->mc_list = NULL; 3574c507d28SJiri Benc inet->rcv_tos = 0; 3581da177e4SLinus Torvalds 359e6848976SArnaldo Carvalho de Melo sk_refcnt_debug_inc(sk); 3601da177e4SLinus Torvalds 361c720c7e8SEric Dumazet if (inet->inet_num) { 3621da177e4SLinus Torvalds /* It assumes that any protocol which allows 3631da177e4SLinus Torvalds * the user to assign a number at socket 3641da177e4SLinus Torvalds * creation time automatically 3651da177e4SLinus Torvalds * shares. 3661da177e4SLinus Torvalds */ 367c720c7e8SEric Dumazet inet->inet_sport = htons(inet->inet_num); 3681da177e4SLinus Torvalds /* Add to protocol hash chains. */ 369086c653fSCraig Gallek err = sk->sk_prot->hash(sk); 370086c653fSCraig Gallek if (err) { 371086c653fSCraig Gallek sk_common_release(sk); 372086c653fSCraig Gallek goto out; 373086c653fSCraig Gallek } 3741da177e4SLinus Torvalds } 3751da177e4SLinus Torvalds 3761da177e4SLinus Torvalds if (sk->sk_prot->init) { 3771da177e4SLinus Torvalds err = sk->sk_prot->init(sk); 37861023658SDavid Ahern if (err) { 3791da177e4SLinus Torvalds sk_common_release(sk); 38061023658SDavid Ahern goto out; 38161023658SDavid Ahern } 38261023658SDavid Ahern } 38361023658SDavid Ahern 38461023658SDavid Ahern if (!kern) { 38561023658SDavid Ahern err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk); 38661023658SDavid Ahern if (err) { 38761023658SDavid Ahern sk_common_release(sk); 38861023658SDavid Ahern goto out; 38961023658SDavid Ahern } 3901da177e4SLinus Torvalds } 3911da177e4SLinus Torvalds out: 3921da177e4SLinus Torvalds return err; 3931da177e4SLinus Torvalds out_rcu_unlock: 3941da177e4SLinus Torvalds rcu_read_unlock(); 3951da177e4SLinus Torvalds goto out; 3961da177e4SLinus Torvalds } 3971da177e4SLinus Torvalds 3981da177e4SLinus Torvalds 3991da177e4SLinus Torvalds /* 4001da177e4SLinus Torvalds * The peer socket should always be NULL (or else). When we call this 4011da177e4SLinus Torvalds * function we are destroying the object and from then on nobody 4021da177e4SLinus Torvalds * should refer to it. 4031da177e4SLinus Torvalds */ 4041da177e4SLinus Torvalds int inet_release(struct socket *sock) 4051da177e4SLinus Torvalds { 4061da177e4SLinus Torvalds struct sock *sk = sock->sk; 4071da177e4SLinus Torvalds 4081da177e4SLinus Torvalds if (sk) { 4091da177e4SLinus Torvalds long timeout; 4101da177e4SLinus Torvalds 411f5836749SStanislav Fomichev if (!sk->sk_kern_sock) 412f5836749SStanislav Fomichev BPF_CGROUP_RUN_PROG_INET_SOCK_RELEASE(sk); 413f5836749SStanislav Fomichev 4141da177e4SLinus Torvalds /* Applications forget to leave groups before exiting */ 4151da177e4SLinus Torvalds ip_mc_drop_socket(sk); 4161da177e4SLinus Torvalds 4171da177e4SLinus Torvalds /* If linger is set, we don't return until the close 4181da177e4SLinus Torvalds * is complete. Otherwise we return immediately. The 4191da177e4SLinus Torvalds * actually closing is done the same either way. 4201da177e4SLinus Torvalds * 4211da177e4SLinus Torvalds * If the close is due to the process exiting, we never 4221da177e4SLinus Torvalds * linger.. 4231da177e4SLinus Torvalds */ 4241da177e4SLinus Torvalds timeout = 0; 4251da177e4SLinus Torvalds if (sock_flag(sk, SOCK_LINGER) && 4261da177e4SLinus Torvalds !(current->flags & PF_EXITING)) 4271da177e4SLinus Torvalds timeout = sk->sk_lingertime; 4281da177e4SLinus Torvalds sk->sk_prot->close(sk, timeout); 4292b81f816SJakub Kicinski sock->sk = NULL; 4301da177e4SLinus Torvalds } 4311da177e4SLinus Torvalds return 0; 4321da177e4SLinus Torvalds } 4333d1427f8SEric Dumazet EXPORT_SYMBOL(inet_release); 4341da177e4SLinus Torvalds 4351da177e4SLinus Torvalds int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) 4361da177e4SLinus Torvalds { 4371da177e4SLinus Torvalds struct sock *sk = sock->sk; 43877241217SStanislav Fomichev u32 flags = BIND_WITH_LOCK; 4391da177e4SLinus Torvalds int err; 4401da177e4SLinus Torvalds 4411da177e4SLinus Torvalds /* If the socket has its own bind function then use it. (RAW) */ 4421da177e4SLinus Torvalds if (sk->sk_prot->bind) { 4433679d585SAndrey Ignatov return sk->sk_prot->bind(sk, uaddr, addr_len); 4441da177e4SLinus Torvalds } 4451da177e4SLinus Torvalds if (addr_len < sizeof(struct sockaddr_in)) 4463679d585SAndrey Ignatov return -EINVAL; 4471da177e4SLinus Torvalds 4484fbac77dSAndrey Ignatov /* BPF prog is run before any checks are done so that if the prog 4494fbac77dSAndrey Ignatov * changes context in a wrong way it will be caught. 4504fbac77dSAndrey Ignatov */ 45177241217SStanislav Fomichev err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, 4526fc88c35SDave Marchevsky CGROUP_INET4_BIND, &flags); 4534fbac77dSAndrey Ignatov if (err) 4543679d585SAndrey Ignatov return err; 4553679d585SAndrey Ignatov 45677241217SStanislav Fomichev return __inet_bind(sk, uaddr, addr_len, flags); 4573679d585SAndrey Ignatov } 4583679d585SAndrey Ignatov EXPORT_SYMBOL(inet_bind); 4593679d585SAndrey Ignatov 4603679d585SAndrey Ignatov int __inet_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len, 461cb0721c7SStanislav Fomichev u32 flags) 4623679d585SAndrey Ignatov { 4633679d585SAndrey Ignatov struct sockaddr_in *addr = (struct sockaddr_in *)uaddr; 4641da177e4SLinus Torvalds struct inet_sock *inet = inet_sk(sk); 4651da177e4SLinus Torvalds struct net *net = sock_net(sk); 4661da177e4SLinus Torvalds unsigned short snum; 4671da177e4SLinus Torvalds int chk_addr_ret; 4681da177e4SLinus Torvalds u32 tb_id = RT_TABLE_LOCAL; 4691da177e4SLinus Torvalds int err; 4701da177e4SLinus Torvalds 471c349a528SMarcus Meissner if (addr->sin_family != AF_INET) { 47229c486dfSEric Dumazet /* Compatibility games : accept AF_UNSPEC (mapped to AF_INET) 47329c486dfSEric Dumazet * only if s_addr is INADDR_ANY. 47429c486dfSEric Dumazet */ 475c349a528SMarcus Meissner err = -EAFNOSUPPORT; 47629c486dfSEric Dumazet if (addr->sin_family != AF_UNSPEC || 47729c486dfSEric Dumazet addr->sin_addr.s_addr != htonl(INADDR_ANY)) 478d0733d2eSMarcus Meissner goto out; 479c349a528SMarcus Meissner } 480d0733d2eSMarcus Meissner 4813236b004SDavid Ahern tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id; 48230bbaa19SDavid Ahern chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id); 4831da177e4SLinus Torvalds 4841da177e4SLinus Torvalds /* Not specified by any standard per-se, however it breaks too 4851da177e4SLinus Torvalds * many applications when removed. It is unfortunate since 4861da177e4SLinus Torvalds * allowing applications to make a non-local bind solves 4871da177e4SLinus Torvalds * several problems with systems using dynamic addressing. 4881da177e4SLinus Torvalds * (ie. your servers still start up even if your ISDN link 4891da177e4SLinus Torvalds * is temporarily down) 4901da177e4SLinus Torvalds */ 4911da177e4SLinus Torvalds err = -EADDRNOTAVAIL; 4928ff978b8SRiccardo Paolo Bestetti if (!inet_addr_valid_or_nonlocal(net, inet, addr->sin_addr.s_addr, 4938ff978b8SRiccardo Paolo Bestetti chk_addr_ret)) 4941da177e4SLinus Torvalds goto out; 4951da177e4SLinus Torvalds 4961da177e4SLinus Torvalds snum = ntohs(addr->sin_port); 4971da177e4SLinus Torvalds err = -EACCES; 49877241217SStanislav Fomichev if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) && 49977241217SStanislav Fomichev snum && inet_port_requires_bind_service(net, snum) && 5003594698aSEric W. Biederman !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) 5011da177e4SLinus Torvalds goto out; 5021da177e4SLinus Torvalds 5031da177e4SLinus Torvalds /* We keep a pair of addresses. rcv_saddr is the one 5041da177e4SLinus Torvalds * used by hash lookups, and saddr is used for transmit. 5051da177e4SLinus Torvalds * 5061da177e4SLinus Torvalds * In the BSD API these are the same except where it 5071da177e4SLinus Torvalds * would be illegal to use them (multicast/broadcast) in 5081da177e4SLinus Torvalds * which case the sending device address is used. 5091da177e4SLinus Torvalds */ 510cb0721c7SStanislav Fomichev if (flags & BIND_WITH_LOCK) 5111da177e4SLinus Torvalds lock_sock(sk); 5121da177e4SLinus Torvalds 5131da177e4SLinus Torvalds /* Check these errors (active socket, double bind). */ 5141da177e4SLinus Torvalds err = -EINVAL; 515c720c7e8SEric Dumazet if (sk->sk_state != TCP_CLOSE || inet->inet_num) 5161da177e4SLinus Torvalds goto out_release_sock; 5171da177e4SLinus Torvalds 518c720c7e8SEric Dumazet inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr; 5191da177e4SLinus Torvalds if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) 520c720c7e8SEric Dumazet inet->inet_saddr = 0; /* Use device */ 5211da177e4SLinus Torvalds 5221da177e4SLinus Torvalds /* Make sure we are allowed to bind here. */ 523aac3fc32SAndrey Ignatov if (snum || !(inet->bind_address_no_port || 524cb0721c7SStanislav Fomichev (flags & BIND_FORCE_ADDRESS_NO_PORT))) { 525aac3fc32SAndrey Ignatov if (sk->sk_prot->get_port(sk, snum)) { 526c720c7e8SEric Dumazet inet->inet_saddr = inet->inet_rcv_saddr = 0; 5271da177e4SLinus Torvalds err = -EADDRINUSE; 5281da177e4SLinus Torvalds goto out_release_sock; 5291da177e4SLinus Torvalds } 5308086fbafSStanislav Fomichev if (!(flags & BIND_FROM_BPF)) { 531aac3fc32SAndrey Ignatov err = BPF_CGROUP_RUN_PROG_INET4_POST_BIND(sk); 532aac3fc32SAndrey Ignatov if (err) { 533aac3fc32SAndrey Ignatov inet->inet_saddr = inet->inet_rcv_saddr = 0; 53491a760b2SMenglong Dong if (sk->sk_prot->put_port) 53591a760b2SMenglong Dong sk->sk_prot->put_port(sk); 536aac3fc32SAndrey Ignatov goto out_release_sock; 537aac3fc32SAndrey Ignatov } 538aac3fc32SAndrey Ignatov } 5398086fbafSStanislav Fomichev } 5401da177e4SLinus Torvalds 541c720c7e8SEric Dumazet if (inet->inet_rcv_saddr) 5421da177e4SLinus Torvalds sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 5431da177e4SLinus Torvalds if (snum) 5441da177e4SLinus Torvalds sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 545c720c7e8SEric Dumazet inet->inet_sport = htons(inet->inet_num); 546c720c7e8SEric Dumazet inet->inet_daddr = 0; 547c720c7e8SEric Dumazet inet->inet_dport = 0; 5481da177e4SLinus Torvalds sk_dst_reset(sk); 5491da177e4SLinus Torvalds err = 0; 5501da177e4SLinus Torvalds out_release_sock: 551cb0721c7SStanislav Fomichev if (flags & BIND_WITH_LOCK) 5521da177e4SLinus Torvalds release_sock(sk); 5531da177e4SLinus Torvalds out: 5541da177e4SLinus Torvalds return err; 5551da177e4SLinus Torvalds } 5561da177e4SLinus Torvalds 5571da177e4SLinus Torvalds int inet_dgram_connect(struct socket *sock, struct sockaddr *uaddr, 5581da177e4SLinus Torvalds int addr_len, int flags) 5591da177e4SLinus Torvalds { 5601da177e4SLinus Torvalds struct sock *sk = sock->sk; 561d74bad4eSAndrey Ignatov int err; 5621da177e4SLinus Torvalds 5636503d961SChangli Gao if (addr_len < sizeof(uaddr->sa_family)) 5646503d961SChangli Gao return -EINVAL; 5651da177e4SLinus Torvalds if (uaddr->sa_family == AF_UNSPEC) 5661da177e4SLinus Torvalds return sk->sk_prot->disconnect(sk, flags); 5671da177e4SLinus Torvalds 568d74bad4eSAndrey Ignatov if (BPF_CGROUP_PRE_CONNECT_ENABLED(sk)) { 569d74bad4eSAndrey Ignatov err = sk->sk_prot->pre_connect(sk, uaddr, addr_len); 570d74bad4eSAndrey Ignatov if (err) 571d74bad4eSAndrey Ignatov return err; 572d74bad4eSAndrey Ignatov } 573d74bad4eSAndrey Ignatov 574dcd01eeaSEric Dumazet if (data_race(!inet_sk(sk)->inet_num) && inet_autobind(sk)) 5751da177e4SLinus Torvalds return -EAGAIN; 576e3192690SJoe Perches return sk->sk_prot->connect(sk, uaddr, addr_len); 5771da177e4SLinus Torvalds } 5783d1427f8SEric Dumazet EXPORT_SYMBOL(inet_dgram_connect); 5791da177e4SLinus Torvalds 580783237e8SYuchung Cheng static long inet_wait_for_connect(struct sock *sk, long timeo, int writebias) 5811da177e4SLinus Torvalds { 58214135f30SWANG Cong DEFINE_WAIT_FUNC(wait, woken_wake_function); 5831da177e4SLinus Torvalds 58414135f30SWANG Cong add_wait_queue(sk_sleep(sk), &wait); 585783237e8SYuchung Cheng sk->sk_write_pending += writebias; 5861da177e4SLinus Torvalds 5871da177e4SLinus Torvalds /* Basic assumption: if someone sets sk->sk_err, he _must_ 5881da177e4SLinus Torvalds * change state of the socket from TCP_SYN_*. 5891da177e4SLinus Torvalds * Connect() does not allow to get error notifications 5901da177e4SLinus Torvalds * without closing the socket. 5911da177e4SLinus Torvalds */ 5921da177e4SLinus Torvalds while ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { 5931da177e4SLinus Torvalds release_sock(sk); 59414135f30SWANG Cong timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); 5951da177e4SLinus Torvalds lock_sock(sk); 5961da177e4SLinus Torvalds if (signal_pending(current) || !timeo) 5971da177e4SLinus Torvalds break; 5981da177e4SLinus Torvalds } 59914135f30SWANG Cong remove_wait_queue(sk_sleep(sk), &wait); 600783237e8SYuchung Cheng sk->sk_write_pending -= writebias; 6011da177e4SLinus Torvalds return timeo; 6021da177e4SLinus Torvalds } 6031da177e4SLinus Torvalds 6041da177e4SLinus Torvalds /* 6051da177e4SLinus Torvalds * Connect to a remote host. There is regrettably still a little 6061da177e4SLinus Torvalds * TCP 'magic' in here. 6071da177e4SLinus Torvalds */ 608cf60af03SYuchung Cheng int __inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, 6093979ad7eSWilly Tarreau int addr_len, int flags, int is_sendmsg) 6101da177e4SLinus Torvalds { 6111da177e4SLinus Torvalds struct sock *sk = sock->sk; 6121da177e4SLinus Torvalds int err; 6131da177e4SLinus Torvalds long timeo; 6141da177e4SLinus Torvalds 61519f6d3f3SWei Wang /* 61619f6d3f3SWei Wang * uaddr can be NULL and addr_len can be 0 if: 61719f6d3f3SWei Wang * sk is a TCP fastopen active socket and 61819f6d3f3SWei Wang * TCP_FASTOPEN_CONNECT sockopt is set and 61919f6d3f3SWei Wang * we already have a valid cookie for this socket. 62019f6d3f3SWei Wang * In this case, user can call write() after connect(). 62119f6d3f3SWei Wang * write() will invoke tcp_sendmsg_fastopen() which calls 62219f6d3f3SWei Wang * __inet_stream_connect(). 62319f6d3f3SWei Wang */ 62419f6d3f3SWei Wang if (uaddr) { 6256503d961SChangli Gao if (addr_len < sizeof(uaddr->sa_family)) 6266503d961SChangli Gao return -EINVAL; 6276503d961SChangli Gao 6281da177e4SLinus Torvalds if (uaddr->sa_family == AF_UNSPEC) { 6291da177e4SLinus Torvalds err = sk->sk_prot->disconnect(sk, flags); 6301da177e4SLinus Torvalds sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; 6311da177e4SLinus Torvalds goto out; 6321da177e4SLinus Torvalds } 63319f6d3f3SWei Wang } 6341da177e4SLinus Torvalds 6351da177e4SLinus Torvalds switch (sock->state) { 6361da177e4SLinus Torvalds default: 6371da177e4SLinus Torvalds err = -EINVAL; 6381da177e4SLinus Torvalds goto out; 6391da177e4SLinus Torvalds case SS_CONNECTED: 6401da177e4SLinus Torvalds err = -EISCONN; 6411da177e4SLinus Torvalds goto out; 6421da177e4SLinus Torvalds case SS_CONNECTING: 64319f6d3f3SWei Wang if (inet_sk(sk)->defer_connect) 6443979ad7eSWilly Tarreau err = is_sendmsg ? -EINPROGRESS : -EISCONN; 64519f6d3f3SWei Wang else 6461da177e4SLinus Torvalds err = -EALREADY; 6471da177e4SLinus Torvalds /* Fall out of switch with err, set for this state */ 6481da177e4SLinus Torvalds break; 6491da177e4SLinus Torvalds case SS_UNCONNECTED: 6501da177e4SLinus Torvalds err = -EISCONN; 6511da177e4SLinus Torvalds if (sk->sk_state != TCP_CLOSE) 6521da177e4SLinus Torvalds goto out; 6531da177e4SLinus Torvalds 654d74bad4eSAndrey Ignatov if (BPF_CGROUP_PRE_CONNECT_ENABLED(sk)) { 655d74bad4eSAndrey Ignatov err = sk->sk_prot->pre_connect(sk, uaddr, addr_len); 656d74bad4eSAndrey Ignatov if (err) 657d74bad4eSAndrey Ignatov goto out; 658d74bad4eSAndrey Ignatov } 659d74bad4eSAndrey Ignatov 6601da177e4SLinus Torvalds err = sk->sk_prot->connect(sk, uaddr, addr_len); 6611da177e4SLinus Torvalds if (err < 0) 6621da177e4SLinus Torvalds goto out; 6631da177e4SLinus Torvalds 6641da177e4SLinus Torvalds sock->state = SS_CONNECTING; 6651da177e4SLinus Torvalds 66619f6d3f3SWei Wang if (!err && inet_sk(sk)->defer_connect) 66719f6d3f3SWei Wang goto out; 66819f6d3f3SWei Wang 6691da177e4SLinus Torvalds /* Just entered SS_CONNECTING state; the only 6701da177e4SLinus Torvalds * difference is that return value in non-blocking 6711da177e4SLinus Torvalds * case is EINPROGRESS, rather than EALREADY. 6721da177e4SLinus Torvalds */ 6731da177e4SLinus Torvalds err = -EINPROGRESS; 6741da177e4SLinus Torvalds break; 6751da177e4SLinus Torvalds } 6761da177e4SLinus Torvalds 6771da177e4SLinus Torvalds timeo = sock_sndtimeo(sk, flags & O_NONBLOCK); 6781da177e4SLinus Torvalds 6791da177e4SLinus Torvalds if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) { 680783237e8SYuchung Cheng int writebias = (sk->sk_protocol == IPPROTO_TCP) && 681783237e8SYuchung Cheng tcp_sk(sk)->fastopen_req && 682783237e8SYuchung Cheng tcp_sk(sk)->fastopen_req->data ? 1 : 0; 683783237e8SYuchung Cheng 6841da177e4SLinus Torvalds /* Error code is set above */ 685783237e8SYuchung Cheng if (!timeo || !inet_wait_for_connect(sk, timeo, writebias)) 6861da177e4SLinus Torvalds goto out; 6871da177e4SLinus Torvalds 6881da177e4SLinus Torvalds err = sock_intr_errno(timeo); 6891da177e4SLinus Torvalds if (signal_pending(current)) 6901da177e4SLinus Torvalds goto out; 6911da177e4SLinus Torvalds } 6921da177e4SLinus Torvalds 6931da177e4SLinus Torvalds /* Connection was closed by RST, timeout, ICMP error 6941da177e4SLinus Torvalds * or another process disconnected us. 6951da177e4SLinus Torvalds */ 6961da177e4SLinus Torvalds if (sk->sk_state == TCP_CLOSE) 6971da177e4SLinus Torvalds goto sock_error; 6981da177e4SLinus Torvalds 6991da177e4SLinus Torvalds /* sk->sk_err may be not zero now, if RECVERR was ordered by user 7001da177e4SLinus Torvalds * and error was received after socket entered established state. 7011da177e4SLinus Torvalds * Hence, it is handled normally after connect() return successfully. 7021da177e4SLinus Torvalds */ 7031da177e4SLinus Torvalds 7041da177e4SLinus Torvalds sock->state = SS_CONNECTED; 7051da177e4SLinus Torvalds err = 0; 7061da177e4SLinus Torvalds out: 7071da177e4SLinus Torvalds return err; 7081da177e4SLinus Torvalds 7091da177e4SLinus Torvalds sock_error: 7101da177e4SLinus Torvalds err = sock_error(sk) ? : -ECONNABORTED; 7111da177e4SLinus Torvalds sock->state = SS_UNCONNECTED; 7121da177e4SLinus Torvalds if (sk->sk_prot->disconnect(sk, flags)) 7131da177e4SLinus Torvalds sock->state = SS_DISCONNECTING; 7141da177e4SLinus Torvalds goto out; 7151da177e4SLinus Torvalds } 716cf60af03SYuchung Cheng EXPORT_SYMBOL(__inet_stream_connect); 717cf60af03SYuchung Cheng 718cf60af03SYuchung Cheng int inet_stream_connect(struct socket *sock, struct sockaddr *uaddr, 719cf60af03SYuchung Cheng int addr_len, int flags) 720cf60af03SYuchung Cheng { 721cf60af03SYuchung Cheng int err; 722cf60af03SYuchung Cheng 723cf60af03SYuchung Cheng lock_sock(sock->sk); 7243979ad7eSWilly Tarreau err = __inet_stream_connect(sock, uaddr, addr_len, flags, 0); 725cf60af03SYuchung Cheng release_sock(sock->sk); 726cf60af03SYuchung Cheng return err; 727cf60af03SYuchung Cheng } 7283d1427f8SEric Dumazet EXPORT_SYMBOL(inet_stream_connect); 7291da177e4SLinus Torvalds 7301da177e4SLinus Torvalds /* 7311da177e4SLinus Torvalds * Accept a pending connection. The TCP layer now gives BSD semantics. 7321da177e4SLinus Torvalds */ 7331da177e4SLinus Torvalds 734cdfbabfbSDavid Howells int inet_accept(struct socket *sock, struct socket *newsock, int flags, 735cdfbabfbSDavid Howells bool kern) 7361da177e4SLinus Torvalds { 7371da177e4SLinus Torvalds struct sock *sk1 = sock->sk; 7381da177e4SLinus Torvalds int err = -EINVAL; 739cdfbabfbSDavid Howells struct sock *sk2 = sk1->sk_prot->accept(sk1, flags, &err, kern); 7401da177e4SLinus Torvalds 7411da177e4SLinus Torvalds if (!sk2) 7421da177e4SLinus Torvalds goto do_err; 7431da177e4SLinus Torvalds 7441da177e4SLinus Torvalds lock_sock(sk2); 7451da177e4SLinus Torvalds 7461eddceadSEric Dumazet sock_rps_record_flow(sk2); 747547b792cSIlpo Järvinen WARN_ON(!((1 << sk2->sk_state) & 7488336886fSJerry Chu (TCPF_ESTABLISHED | TCPF_SYN_RECV | 7498336886fSJerry Chu TCPF_CLOSE_WAIT | TCPF_CLOSE))); 7501da177e4SLinus Torvalds 7511da177e4SLinus Torvalds sock_graft(sk2, newsock); 7521da177e4SLinus Torvalds 7531da177e4SLinus Torvalds newsock->state = SS_CONNECTED; 7541da177e4SLinus Torvalds err = 0; 7551da177e4SLinus Torvalds release_sock(sk2); 7561da177e4SLinus Torvalds do_err: 7571da177e4SLinus Torvalds return err; 7581da177e4SLinus Torvalds } 7593d1427f8SEric Dumazet EXPORT_SYMBOL(inet_accept); 7601da177e4SLinus Torvalds 7611da177e4SLinus Torvalds /* 7621da177e4SLinus Torvalds * This does both peername and sockname. 7631da177e4SLinus Torvalds */ 7641da177e4SLinus Torvalds int inet_getname(struct socket *sock, struct sockaddr *uaddr, 7659b2c45d4SDenys Vlasenko int peer) 7661da177e4SLinus Torvalds { 7671da177e4SLinus Torvalds struct sock *sk = sock->sk; 7681da177e4SLinus Torvalds struct inet_sock *inet = inet_sk(sk); 76938bfd8f5SCyrill Gorcunov DECLARE_SOCKADDR(struct sockaddr_in *, sin, uaddr); 7701da177e4SLinus Torvalds 7711da177e4SLinus Torvalds sin->sin_family = AF_INET; 7729dfc685eSEric Dumazet lock_sock(sk); 7731da177e4SLinus Torvalds if (peer) { 774c720c7e8SEric Dumazet if (!inet->inet_dport || 7751da177e4SLinus Torvalds (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && 7769dfc685eSEric Dumazet peer == 1)) { 7779dfc685eSEric Dumazet release_sock(sk); 7781da177e4SLinus Torvalds return -ENOTCONN; 7799dfc685eSEric Dumazet } 780c720c7e8SEric Dumazet sin->sin_port = inet->inet_dport; 781c720c7e8SEric Dumazet sin->sin_addr.s_addr = inet->inet_daddr; 7829dfc685eSEric Dumazet BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 7839dfc685eSEric Dumazet CGROUP_INET4_GETPEERNAME); 7841da177e4SLinus Torvalds } else { 785c720c7e8SEric Dumazet __be32 addr = inet->inet_rcv_saddr; 7861da177e4SLinus Torvalds if (!addr) 787c720c7e8SEric Dumazet addr = inet->inet_saddr; 788c720c7e8SEric Dumazet sin->sin_port = inet->inet_sport; 7891da177e4SLinus Torvalds sin->sin_addr.s_addr = addr; 7909dfc685eSEric Dumazet BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 7919dfc685eSEric Dumazet CGROUP_INET4_GETSOCKNAME); 792a9ed15daSStanislav Fomichev } 7939dfc685eSEric Dumazet release_sock(sk); 7941da177e4SLinus Torvalds memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 7959b2c45d4SDenys Vlasenko return sizeof(*sin); 7961da177e4SLinus Torvalds } 7973d1427f8SEric Dumazet EXPORT_SYMBOL(inet_getname); 7981da177e4SLinus Torvalds 799e4730936SPaolo Abeni int inet_send_prepare(struct sock *sk) 8001da177e4SLinus Torvalds { 801c58dc01bSDavid S. Miller sock_rps_record_flow(sk); 802fec5e652STom Herbert 8031da177e4SLinus Torvalds /* We may need to bind the socket. */ 804dcd01eeaSEric Dumazet if (data_race(!inet_sk(sk)->inet_num) && !sk->sk_prot->no_autobind && 8057ba42910SChangli Gao inet_autobind(sk)) 8061da177e4SLinus Torvalds return -EAGAIN; 8071da177e4SLinus Torvalds 808e4730936SPaolo Abeni return 0; 809e4730936SPaolo Abeni } 810e4730936SPaolo Abeni EXPORT_SYMBOL_GPL(inet_send_prepare); 811e4730936SPaolo Abeni 812e4730936SPaolo Abeni int inet_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 813e4730936SPaolo Abeni { 814e4730936SPaolo Abeni struct sock *sk = sock->sk; 815e4730936SPaolo Abeni 816e4730936SPaolo Abeni if (unlikely(inet_send_prepare(sk))) 817e4730936SPaolo Abeni return -EAGAIN; 818e4730936SPaolo Abeni 8196f24080eSPaolo Abeni return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udp_sendmsg, 8206f24080eSPaolo Abeni sk, msg, size); 8211da177e4SLinus Torvalds } 8223d1427f8SEric Dumazet EXPORT_SYMBOL(inet_sendmsg); 8231da177e4SLinus Torvalds 8247ba42910SChangli Gao ssize_t inet_sendpage(struct socket *sock, struct page *page, int offset, 8253d1427f8SEric Dumazet size_t size, int flags) 8261da177e4SLinus Torvalds { 8271da177e4SLinus Torvalds struct sock *sk = sock->sk; 8281da177e4SLinus Torvalds 829e4730936SPaolo Abeni if (unlikely(inet_send_prepare(sk))) 8301da177e4SLinus Torvalds return -EAGAIN; 8311da177e4SLinus Torvalds 8321da177e4SLinus Torvalds if (sk->sk_prot->sendpage) 8331da177e4SLinus Torvalds return sk->sk_prot->sendpage(sk, page, offset, size, flags); 8341da177e4SLinus Torvalds return sock_no_sendpage(sock, page, offset, size, flags); 8351da177e4SLinus Torvalds } 8367ba42910SChangli Gao EXPORT_SYMBOL(inet_sendpage); 8371da177e4SLinus Torvalds 8386f24080eSPaolo Abeni INDIRECT_CALLABLE_DECLARE(int udp_recvmsg(struct sock *, struct msghdr *, 839ec095263SOliver Hartkopp size_t, int, int *)); 8401b784140SYing Xue int inet_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, 8411b784140SYing Xue int flags) 842fec5e652STom Herbert { 843fec5e652STom Herbert struct sock *sk = sock->sk; 844fec5e652STom Herbert int addr_len = 0; 845fec5e652STom Herbert int err; 846fec5e652STom Herbert 847e3f2c4a3SSoheil Hassas Yeganeh if (likely(!(flags & MSG_ERRQUEUE))) 848c58dc01bSDavid S. Miller sock_rps_record_flow(sk); 849fec5e652STom Herbert 8506f24080eSPaolo Abeni err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udp_recvmsg, 851ec095263SOliver Hartkopp sk, msg, size, flags, &addr_len); 852fec5e652STom Herbert if (err >= 0) 853fec5e652STom Herbert msg->msg_namelen = addr_len; 854fec5e652STom Herbert return err; 855fec5e652STom Herbert } 856fec5e652STom Herbert EXPORT_SYMBOL(inet_recvmsg); 8571da177e4SLinus Torvalds 8581da177e4SLinus Torvalds int inet_shutdown(struct socket *sock, int how) 8591da177e4SLinus Torvalds { 8601da177e4SLinus Torvalds struct sock *sk = sock->sk; 8611da177e4SLinus Torvalds int err = 0; 8621da177e4SLinus Torvalds 8631da177e4SLinus Torvalds /* This should really check to make sure 8641da177e4SLinus Torvalds * the socket is a TCP socket. (WHY AC...) 8651da177e4SLinus Torvalds */ 8661da177e4SLinus Torvalds how++; /* maps 0->1 has the advantage of making bit 1 rcvs and 8671da177e4SLinus Torvalds 1->2 bit 2 snds. 8681da177e4SLinus Torvalds 2->3 */ 8691da177e4SLinus Torvalds if ((how & ~SHUTDOWN_MASK) || !how) /* MAXINT->0 */ 8701da177e4SLinus Torvalds return -EINVAL; 8711da177e4SLinus Torvalds 8721da177e4SLinus Torvalds lock_sock(sk); 8731da177e4SLinus Torvalds if (sock->state == SS_CONNECTING) { 8741da177e4SLinus Torvalds if ((1 << sk->sk_state) & 8751da177e4SLinus Torvalds (TCPF_SYN_SENT | TCPF_SYN_RECV | TCPF_CLOSE)) 8761da177e4SLinus Torvalds sock->state = SS_DISCONNECTING; 8771da177e4SLinus Torvalds else 8781da177e4SLinus Torvalds sock->state = SS_CONNECTED; 8791da177e4SLinus Torvalds } 8801da177e4SLinus Torvalds 8811da177e4SLinus Torvalds switch (sk->sk_state) { 8821da177e4SLinus Torvalds case TCP_CLOSE: 8831da177e4SLinus Torvalds err = -ENOTCONN; 8841da177e4SLinus Torvalds /* Hack to wake up other listeners, who can poll for 885a9a08845SLinus Torvalds EPOLLHUP, even on eg. unconnected UDP sockets -- RR */ 886a8eceea8SJoe Perches fallthrough; 8871da177e4SLinus Torvalds default: 8881da177e4SLinus Torvalds sk->sk_shutdown |= how; 8891da177e4SLinus Torvalds if (sk->sk_prot->shutdown) 8901da177e4SLinus Torvalds sk->sk_prot->shutdown(sk, how); 8911da177e4SLinus Torvalds break; 8921da177e4SLinus Torvalds 8931da177e4SLinus Torvalds /* Remaining two branches are temporary solution for missing 8941da177e4SLinus Torvalds * close() in multithreaded environment. It is _not_ a good idea, 8951da177e4SLinus Torvalds * but we have no choice until close() is repaired at VFS level. 8961da177e4SLinus Torvalds */ 8971da177e4SLinus Torvalds case TCP_LISTEN: 8981da177e4SLinus Torvalds if (!(how & RCV_SHUTDOWN)) 8991da177e4SLinus Torvalds break; 900a8eceea8SJoe Perches fallthrough; 9011da177e4SLinus Torvalds case TCP_SYN_SENT: 9021da177e4SLinus Torvalds err = sk->sk_prot->disconnect(sk, O_NONBLOCK); 9031da177e4SLinus Torvalds sock->state = err ? SS_DISCONNECTING : SS_UNCONNECTED; 9041da177e4SLinus Torvalds break; 9051da177e4SLinus Torvalds } 9061da177e4SLinus Torvalds 9071da177e4SLinus Torvalds /* Wake up anyone sleeping in poll. */ 9081da177e4SLinus Torvalds sk->sk_state_change(sk); 9091da177e4SLinus Torvalds release_sock(sk); 9101da177e4SLinus Torvalds return err; 9111da177e4SLinus Torvalds } 9123d1427f8SEric Dumazet EXPORT_SYMBOL(inet_shutdown); 9131da177e4SLinus Torvalds 9141da177e4SLinus Torvalds /* 9151da177e4SLinus Torvalds * ioctl() calls you can issue on an INET socket. Most of these are 9161da177e4SLinus Torvalds * device configuration and stuff and very rarely used. Some ioctls 9171da177e4SLinus Torvalds * pass on to the socket itself. 9181da177e4SLinus Torvalds * 9191da177e4SLinus Torvalds * NOTE: I like the idea of a module for the config stuff. ie ifconfig 9201da177e4SLinus Torvalds * loads the devconfigure module does its configuring and unloads it. 9211da177e4SLinus Torvalds * There's a good 20K of config code hanging around the kernel. 9221da177e4SLinus Torvalds */ 9231da177e4SLinus Torvalds 9241da177e4SLinus Torvalds int inet_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 9251da177e4SLinus Torvalds { 9261da177e4SLinus Torvalds struct sock *sk = sock->sk; 9271da177e4SLinus Torvalds int err = 0; 9283b1e0a65SYOSHIFUJI Hideaki struct net *net = sock_net(sk); 92903aef17bSAl Viro void __user *p = (void __user *)arg; 93003aef17bSAl Viro struct ifreq ifr; 931ca25c300SAl Viro struct rtentry rt; 9321da177e4SLinus Torvalds 9331da177e4SLinus Torvalds switch (cmd) { 9341da177e4SLinus Torvalds case SIOCADDRT: 9351da177e4SLinus Torvalds case SIOCDELRT: 936ca25c300SAl Viro if (copy_from_user(&rt, p, sizeof(struct rtentry))) 937ca25c300SAl Viro return -EFAULT; 938ca25c300SAl Viro err = ip_rt_ioctl(net, cmd, &rt); 939ca25c300SAl Viro break; 9401da177e4SLinus Torvalds case SIOCRTMSG: 941ca25c300SAl Viro err = -EINVAL; 9421da177e4SLinus Torvalds break; 9431da177e4SLinus Torvalds case SIOCDARP: 9441da177e4SLinus Torvalds case SIOCGARP: 9451da177e4SLinus Torvalds case SIOCSARP: 946e5b13cb1SDenis V. Lunev err = arp_ioctl(net, cmd, (void __user *)arg); 9471da177e4SLinus Torvalds break; 9481da177e4SLinus Torvalds case SIOCGIFADDR: 9491da177e4SLinus Torvalds case SIOCGIFBRDADDR: 9501da177e4SLinus Torvalds case SIOCGIFNETMASK: 9511da177e4SLinus Torvalds case SIOCGIFDSTADDR: 95203aef17bSAl Viro case SIOCGIFPFLAGS: 95329c49648SArnd Bergmann if (get_user_ifreq(&ifr, NULL, p)) 95403aef17bSAl Viro return -EFAULT; 95503aef17bSAl Viro err = devinet_ioctl(net, cmd, &ifr); 95629c49648SArnd Bergmann if (!err && put_user_ifreq(&ifr, p)) 95703aef17bSAl Viro err = -EFAULT; 95803aef17bSAl Viro break; 95903aef17bSAl Viro 96003aef17bSAl Viro case SIOCSIFADDR: 96103aef17bSAl Viro case SIOCSIFBRDADDR: 96203aef17bSAl Viro case SIOCSIFNETMASK: 9631da177e4SLinus Torvalds case SIOCSIFDSTADDR: 9641da177e4SLinus Torvalds case SIOCSIFPFLAGS: 9651da177e4SLinus Torvalds case SIOCSIFFLAGS: 96629c49648SArnd Bergmann if (get_user_ifreq(&ifr, NULL, p)) 96703aef17bSAl Viro return -EFAULT; 96803aef17bSAl Viro err = devinet_ioctl(net, cmd, &ifr); 9691da177e4SLinus Torvalds break; 9701da177e4SLinus Torvalds default: 971b5e5fa5eSChristoph Hellwig if (sk->sk_prot->ioctl) 972b5e5fa5eSChristoph Hellwig err = sk->sk_prot->ioctl(sk, cmd, arg); 973b5e5fa5eSChristoph Hellwig else 974b5e5fa5eSChristoph Hellwig err = -ENOIOCTLCMD; 9751da177e4SLinus Torvalds break; 9761da177e4SLinus Torvalds } 9771da177e4SLinus Torvalds return err; 9781da177e4SLinus Torvalds } 9793d1427f8SEric Dumazet EXPORT_SYMBOL(inet_ioctl); 9801da177e4SLinus Torvalds 981709b46e8SEric W. Biederman #ifdef CONFIG_COMPAT 982dc13c876SChristoph Hellwig static int inet_compat_routing_ioctl(struct sock *sk, unsigned int cmd, 983dc13c876SChristoph Hellwig struct compat_rtentry __user *ur) 984dc13c876SChristoph Hellwig { 985dc13c876SChristoph Hellwig compat_uptr_t rtdev; 986dc13c876SChristoph Hellwig struct rtentry rt; 987dc13c876SChristoph Hellwig 988dc13c876SChristoph Hellwig if (copy_from_user(&rt.rt_dst, &ur->rt_dst, 989dc13c876SChristoph Hellwig 3 * sizeof(struct sockaddr)) || 990dc13c876SChristoph Hellwig get_user(rt.rt_flags, &ur->rt_flags) || 991dc13c876SChristoph Hellwig get_user(rt.rt_metric, &ur->rt_metric) || 992dc13c876SChristoph Hellwig get_user(rt.rt_mtu, &ur->rt_mtu) || 993dc13c876SChristoph Hellwig get_user(rt.rt_window, &ur->rt_window) || 994dc13c876SChristoph Hellwig get_user(rt.rt_irtt, &ur->rt_irtt) || 995dc13c876SChristoph Hellwig get_user(rtdev, &ur->rt_dev)) 996dc13c876SChristoph Hellwig return -EFAULT; 997dc13c876SChristoph Hellwig 998dc13c876SChristoph Hellwig rt.rt_dev = compat_ptr(rtdev); 999dc13c876SChristoph Hellwig return ip_rt_ioctl(sock_net(sk), cmd, &rt); 1000dc13c876SChristoph Hellwig } 1001dc13c876SChristoph Hellwig 1002686dc6b6SGerrit Renker static int inet_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 1003709b46e8SEric W. Biederman { 1004dc13c876SChristoph Hellwig void __user *argp = compat_ptr(arg); 1005709b46e8SEric W. Biederman struct sock *sk = sock->sk; 1006709b46e8SEric W. Biederman 1007dc13c876SChristoph Hellwig switch (cmd) { 1008dc13c876SChristoph Hellwig case SIOCADDRT: 1009dc13c876SChristoph Hellwig case SIOCDELRT: 1010dc13c876SChristoph Hellwig return inet_compat_routing_ioctl(sk, cmd, argp); 1011dc13c876SChristoph Hellwig default: 1012dc13c876SChristoph Hellwig if (!sk->sk_prot->compat_ioctl) 1013dc13c876SChristoph Hellwig return -ENOIOCTLCMD; 1014dc13c876SChristoph Hellwig return sk->sk_prot->compat_ioctl(sk, cmd, arg); 1015709b46e8SEric W. Biederman } 1016dc13c876SChristoph Hellwig } 1017dc13c876SChristoph Hellwig #endif /* CONFIG_COMPAT */ 1018709b46e8SEric W. Biederman 101990ddc4f0SEric Dumazet const struct proto_ops inet_stream_ops = { 10201da177e4SLinus Torvalds .family = PF_INET, 10211da177e4SLinus Torvalds .owner = THIS_MODULE, 10221da177e4SLinus Torvalds .release = inet_release, 10231da177e4SLinus Torvalds .bind = inet_bind, 10241da177e4SLinus Torvalds .connect = inet_stream_connect, 10251da177e4SLinus Torvalds .socketpair = sock_no_socketpair, 10261da177e4SLinus Torvalds .accept = inet_accept, 10271da177e4SLinus Torvalds .getname = inet_getname, 1028a11e1d43SLinus Torvalds .poll = tcp_poll, 10291da177e4SLinus Torvalds .ioctl = inet_ioctl, 1030c7cbdbf2SArnd Bergmann .gettstamp = sock_gettstamp, 10311da177e4SLinus Torvalds .listen = inet_listen, 10321da177e4SLinus Torvalds .shutdown = inet_shutdown, 10331da177e4SLinus Torvalds .setsockopt = sock_common_setsockopt, 10341da177e4SLinus Torvalds .getsockopt = sock_common_getsockopt, 10357ba42910SChangli Gao .sendmsg = inet_sendmsg, 1036fec5e652STom Herbert .recvmsg = inet_recvmsg, 103705255b82SEric Dumazet #ifdef CONFIG_MMU 103893ab6cc6SEric Dumazet .mmap = tcp_mmap, 103905255b82SEric Dumazet #endif 10407ba42910SChangli Gao .sendpage = inet_sendpage, 10419c55e01cSJens Axboe .splice_read = tcp_splice_read, 104232035585STom Herbert .read_sock = tcp_read_sock, 1043306b13ebSTom Herbert .sendmsg_locked = tcp_sendmsg_locked, 1044306b13ebSTom Herbert .sendpage_locked = tcp_sendpage_locked, 104532035585STom Herbert .peek_len = tcp_peek_len, 10463fdadf7dSDmitry Mishin #ifdef CONFIG_COMPAT 1047709b46e8SEric W. Biederman .compat_ioctl = inet_compat_ioctl, 10483fdadf7dSDmitry Mishin #endif 1049d1361840SEric Dumazet .set_rcvlowat = tcp_set_rcvlowat, 10501da177e4SLinus Torvalds }; 10513d1427f8SEric Dumazet EXPORT_SYMBOL(inet_stream_ops); 10521da177e4SLinus Torvalds 105390ddc4f0SEric Dumazet const struct proto_ops inet_dgram_ops = { 10541da177e4SLinus Torvalds .family = PF_INET, 10551da177e4SLinus Torvalds .owner = THIS_MODULE, 10561da177e4SLinus Torvalds .release = inet_release, 10571da177e4SLinus Torvalds .bind = inet_bind, 10581da177e4SLinus Torvalds .connect = inet_dgram_connect, 10591da177e4SLinus Torvalds .socketpair = sock_no_socketpair, 10601da177e4SLinus Torvalds .accept = sock_no_accept, 10611da177e4SLinus Torvalds .getname = inet_getname, 1062a11e1d43SLinus Torvalds .poll = udp_poll, 10631da177e4SLinus Torvalds .ioctl = inet_ioctl, 1064c7cbdbf2SArnd Bergmann .gettstamp = sock_gettstamp, 10651da177e4SLinus Torvalds .listen = sock_no_listen, 10661da177e4SLinus Torvalds .shutdown = inet_shutdown, 10671da177e4SLinus Torvalds .setsockopt = sock_common_setsockopt, 10681da177e4SLinus Torvalds .getsockopt = sock_common_getsockopt, 10691da177e4SLinus Torvalds .sendmsg = inet_sendmsg, 1070d7f57118SCong Wang .read_sock = udp_read_sock, 1071fec5e652STom Herbert .recvmsg = inet_recvmsg, 10721da177e4SLinus Torvalds .mmap = sock_no_mmap, 10731da177e4SLinus Torvalds .sendpage = inet_sendpage, 1074627d2d6bSsamanthakumar .set_peek_off = sk_set_peek_off, 1075543d9cfeSArnaldo Carvalho de Melo #ifdef CONFIG_COMPAT 1076709b46e8SEric W. Biederman .compat_ioctl = inet_compat_ioctl, 1077543d9cfeSArnaldo Carvalho de Melo #endif 10781da177e4SLinus Torvalds }; 10793d1427f8SEric Dumazet EXPORT_SYMBOL(inet_dgram_ops); 10801da177e4SLinus Torvalds 10811da177e4SLinus Torvalds /* 10821da177e4SLinus Torvalds * For SOCK_RAW sockets; should be the same as inet_dgram_ops but without 1083a11e1d43SLinus Torvalds * udp_poll 10841da177e4SLinus Torvalds */ 108590ddc4f0SEric Dumazet static const struct proto_ops inet_sockraw_ops = { 10861da177e4SLinus Torvalds .family = PF_INET, 10871da177e4SLinus Torvalds .owner = THIS_MODULE, 10881da177e4SLinus Torvalds .release = inet_release, 10891da177e4SLinus Torvalds .bind = inet_bind, 10901da177e4SLinus Torvalds .connect = inet_dgram_connect, 10911da177e4SLinus Torvalds .socketpair = sock_no_socketpair, 10921da177e4SLinus Torvalds .accept = sock_no_accept, 10931da177e4SLinus Torvalds .getname = inet_getname, 1094a11e1d43SLinus Torvalds .poll = datagram_poll, 10951da177e4SLinus Torvalds .ioctl = inet_ioctl, 1096c7cbdbf2SArnd Bergmann .gettstamp = sock_gettstamp, 10971da177e4SLinus Torvalds .listen = sock_no_listen, 10981da177e4SLinus Torvalds .shutdown = inet_shutdown, 10991da177e4SLinus Torvalds .setsockopt = sock_common_setsockopt, 11001da177e4SLinus Torvalds .getsockopt = sock_common_getsockopt, 11011da177e4SLinus Torvalds .sendmsg = inet_sendmsg, 1102fec5e652STom Herbert .recvmsg = inet_recvmsg, 11031da177e4SLinus Torvalds .mmap = sock_no_mmap, 11041da177e4SLinus Torvalds .sendpage = inet_sendpage, 1105543d9cfeSArnaldo Carvalho de Melo #ifdef CONFIG_COMPAT 1106709b46e8SEric W. Biederman .compat_ioctl = inet_compat_ioctl, 1107543d9cfeSArnaldo Carvalho de Melo #endif 11081da177e4SLinus Torvalds }; 11091da177e4SLinus Torvalds 1110ec1b4cf7SStephen Hemminger static const struct net_proto_family inet_family_ops = { 11111da177e4SLinus Torvalds .family = PF_INET, 11121da177e4SLinus Torvalds .create = inet_create, 11131da177e4SLinus Torvalds .owner = THIS_MODULE, 11141da177e4SLinus Torvalds }; 11151da177e4SLinus Torvalds 11161da177e4SLinus Torvalds /* Upon startup we insert all the elements in inetsw_array[] into 11171da177e4SLinus Torvalds * the linked list inetsw. 11181da177e4SLinus Torvalds */ 11191da177e4SLinus Torvalds static struct inet_protosw inetsw_array[] = 11201da177e4SLinus Torvalds { 11211da177e4SLinus Torvalds { 11221da177e4SLinus Torvalds .type = SOCK_STREAM, 11231da177e4SLinus Torvalds .protocol = IPPROTO_TCP, 11241da177e4SLinus Torvalds .prot = &tcp_prot, 11251da177e4SLinus Torvalds .ops = &inet_stream_ops, 1126d83d8461SArnaldo Carvalho de Melo .flags = INET_PROTOSW_PERMANENT | 1127d83d8461SArnaldo Carvalho de Melo INET_PROTOSW_ICSK, 11281da177e4SLinus Torvalds }, 11291da177e4SLinus Torvalds 11301da177e4SLinus Torvalds { 11311da177e4SLinus Torvalds .type = SOCK_DGRAM, 11321da177e4SLinus Torvalds .protocol = IPPROTO_UDP, 11331da177e4SLinus Torvalds .prot = &udp_prot, 11341da177e4SLinus Torvalds .ops = &inet_dgram_ops, 11351da177e4SLinus Torvalds .flags = INET_PROTOSW_PERMANENT, 11361da177e4SLinus Torvalds }, 11371da177e4SLinus Torvalds 1138c319b4d7SVasiliy Kulikov { 1139c319b4d7SVasiliy Kulikov .type = SOCK_DGRAM, 1140c319b4d7SVasiliy Kulikov .protocol = IPPROTO_ICMP, 1141c319b4d7SVasiliy Kulikov .prot = &ping_prot, 114277d4b1d3SEric Dumazet .ops = &inet_sockraw_ops, 1143c319b4d7SVasiliy Kulikov .flags = INET_PROTOSW_REUSE, 1144c319b4d7SVasiliy Kulikov }, 11451da177e4SLinus Torvalds 11461da177e4SLinus Torvalds { 11471da177e4SLinus Torvalds .type = SOCK_RAW, 11481da177e4SLinus Torvalds .protocol = IPPROTO_IP, /* wild card */ 11491da177e4SLinus Torvalds .prot = &raw_prot, 11501da177e4SLinus Torvalds .ops = &inet_sockraw_ops, 11511da177e4SLinus Torvalds .flags = INET_PROTOSW_REUSE, 11521da177e4SLinus Torvalds } 11531da177e4SLinus Torvalds }; 11541da177e4SLinus Torvalds 1155c40f6fffSDenis Cheng #define INETSW_ARRAY_LEN ARRAY_SIZE(inetsw_array) 11561da177e4SLinus Torvalds 11571da177e4SLinus Torvalds void inet_register_protosw(struct inet_protosw *p) 11581da177e4SLinus Torvalds { 11591da177e4SLinus Torvalds struct list_head *lh; 11601da177e4SLinus Torvalds struct inet_protosw *answer; 11611da177e4SLinus Torvalds int protocol = p->protocol; 11621da177e4SLinus Torvalds struct list_head *last_perm; 11631da177e4SLinus Torvalds 11641da177e4SLinus Torvalds spin_lock_bh(&inetsw_lock); 11651da177e4SLinus Torvalds 11661da177e4SLinus Torvalds if (p->type >= SOCK_MAX) 11671da177e4SLinus Torvalds goto out_illegal; 11681da177e4SLinus Torvalds 11691da177e4SLinus Torvalds /* If we are trying to override a permanent protocol, bail. */ 11701da177e4SLinus Torvalds last_perm = &inetsw[p->type]; 11711da177e4SLinus Torvalds list_for_each(lh, &inetsw[p->type]) { 11721da177e4SLinus Torvalds answer = list_entry(lh, struct inet_protosw, list); 11731da177e4SLinus Torvalds /* Check only the non-wild match. */ 1174f6c53334SJunwei Zhang if ((INET_PROTOSW_PERMANENT & answer->flags) == 0) 11751da177e4SLinus Torvalds break; 1176f6c53334SJunwei Zhang if (protocol == answer->protocol) 1177f6c53334SJunwei Zhang goto out_permanent; 11781da177e4SLinus Torvalds last_perm = lh; 11791da177e4SLinus Torvalds } 11801da177e4SLinus Torvalds 11811da177e4SLinus Torvalds /* Add the new entry after the last permanent entry if any, so that 11821da177e4SLinus Torvalds * the new entry does not override a permanent entry when matched with 11831da177e4SLinus Torvalds * a wild-card protocol. But it is allowed to override any existing 11841da177e4SLinus Torvalds * non-permanent entry. This means that when we remove this entry, the 11851da177e4SLinus Torvalds * system automatically returns to the old behavior. 11861da177e4SLinus Torvalds */ 11871da177e4SLinus Torvalds list_add_rcu(&p->list, last_perm); 11881da177e4SLinus Torvalds out: 11891da177e4SLinus Torvalds spin_unlock_bh(&inetsw_lock); 11901da177e4SLinus Torvalds 11911da177e4SLinus Torvalds return; 11921da177e4SLinus Torvalds 11931da177e4SLinus Torvalds out_permanent: 1194058bd4d2SJoe Perches pr_err("Attempt to override permanent protocol %d\n", protocol); 11951da177e4SLinus Torvalds goto out; 11961da177e4SLinus Torvalds 11971da177e4SLinus Torvalds out_illegal: 1198058bd4d2SJoe Perches pr_err("Ignoring attempt to register invalid socket type %d\n", 11991da177e4SLinus Torvalds p->type); 12001da177e4SLinus Torvalds goto out; 12011da177e4SLinus Torvalds } 12023d1427f8SEric Dumazet EXPORT_SYMBOL(inet_register_protosw); 12031da177e4SLinus Torvalds 12041da177e4SLinus Torvalds void inet_unregister_protosw(struct inet_protosw *p) 12051da177e4SLinus Torvalds { 12061da177e4SLinus Torvalds if (INET_PROTOSW_PERMANENT & p->flags) { 1207058bd4d2SJoe Perches pr_err("Attempt to unregister permanent protocol %d\n", 12081da177e4SLinus Torvalds p->protocol); 12091da177e4SLinus Torvalds } else { 12101da177e4SLinus Torvalds spin_lock_bh(&inetsw_lock); 12111da177e4SLinus Torvalds list_del_rcu(&p->list); 12121da177e4SLinus Torvalds spin_unlock_bh(&inetsw_lock); 12131da177e4SLinus Torvalds 12141da177e4SLinus Torvalds synchronize_net(); 12151da177e4SLinus Torvalds } 12161da177e4SLinus Torvalds } 12173d1427f8SEric Dumazet EXPORT_SYMBOL(inet_unregister_protosw); 12181da177e4SLinus Torvalds 121932519f11SArnaldo Carvalho de Melo static int inet_sk_reselect_saddr(struct sock *sk) 122032519f11SArnaldo Carvalho de Melo { 122132519f11SArnaldo Carvalho de Melo struct inet_sock *inet = inet_sk(sk); 1222c720c7e8SEric Dumazet __be32 old_saddr = inet->inet_saddr; 1223c720c7e8SEric Dumazet __be32 daddr = inet->inet_daddr; 12246e869138SDavid S. Miller struct flowi4 *fl4; 1225b23dd4feSDavid S. Miller struct rtable *rt; 1226b23dd4feSDavid S. Miller __be32 new_saddr; 1227f6d8bd05SEric Dumazet struct ip_options_rcu *inet_opt; 122832519f11SArnaldo Carvalho de Melo 1229f6d8bd05SEric Dumazet inet_opt = rcu_dereference_protected(inet->inet_opt, 12301e1d04e6SHannes Frederic Sowa lockdep_sock_is_held(sk)); 1231f6d8bd05SEric Dumazet if (inet_opt && inet_opt->opt.srr) 1232f6d8bd05SEric Dumazet daddr = inet_opt->opt.faddr; 123332519f11SArnaldo Carvalho de Melo 123432519f11SArnaldo Carvalho de Melo /* Query new route. */ 12356e869138SDavid S. Miller fl4 = &inet->cork.fl.u.ip4; 123667e1e2f4SGuillaume Nault rt = ip_route_connect(fl4, daddr, 0, sk->sk_bound_dev_if, 123767e1e2f4SGuillaume Nault sk->sk_protocol, inet->inet_sport, 123867e1e2f4SGuillaume Nault inet->inet_dport, sk); 1239b23dd4feSDavid S. Miller if (IS_ERR(rt)) 1240b23dd4feSDavid S. Miller return PTR_ERR(rt); 124132519f11SArnaldo Carvalho de Melo 1242d8d1f30bSChangli Gao sk_setup_caps(sk, &rt->dst); 124332519f11SArnaldo Carvalho de Melo 12446e869138SDavid S. Miller new_saddr = fl4->saddr; 124532519f11SArnaldo Carvalho de Melo 124632519f11SArnaldo Carvalho de Melo if (new_saddr == old_saddr) 124732519f11SArnaldo Carvalho de Melo return 0; 124832519f11SArnaldo Carvalho de Melo 1249e49e4affSKuniyuki Iwashima if (READ_ONCE(sock_net(sk)->ipv4.sysctl_ip_dynaddr) > 1) { 1250058bd4d2SJoe Perches pr_info("%s(): shifting inet->saddr from %pI4 to %pI4\n", 1251673d57e7SHarvey Harrison __func__, &old_saddr, &new_saddr); 125232519f11SArnaldo Carvalho de Melo } 125332519f11SArnaldo Carvalho de Melo 1254c720c7e8SEric Dumazet inet->inet_saddr = inet->inet_rcv_saddr = new_saddr; 125532519f11SArnaldo Carvalho de Melo 125632519f11SArnaldo Carvalho de Melo /* 125732519f11SArnaldo Carvalho de Melo * XXX The only one ugly spot where we need to 125832519f11SArnaldo Carvalho de Melo * XXX really change the sockets identity after 125932519f11SArnaldo Carvalho de Melo * XXX it has entered the hashes. -DaveM 126032519f11SArnaldo Carvalho de Melo * 126132519f11SArnaldo Carvalho de Melo * Besides that, it does not check for connection 126232519f11SArnaldo Carvalho de Melo * uniqueness. Wait for troubles. 126332519f11SArnaldo Carvalho de Melo */ 1264086c653fSCraig Gallek return __sk_prot_rehash(sk); 126532519f11SArnaldo Carvalho de Melo } 126632519f11SArnaldo Carvalho de Melo 126732519f11SArnaldo Carvalho de Melo int inet_sk_rebuild_header(struct sock *sk) 126832519f11SArnaldo Carvalho de Melo { 126932519f11SArnaldo Carvalho de Melo struct inet_sock *inet = inet_sk(sk); 127032519f11SArnaldo Carvalho de Melo struct rtable *rt = (struct rtable *)__sk_dst_check(sk, 0); 12713ca3c68eSAl Viro __be32 daddr; 1272f6d8bd05SEric Dumazet struct ip_options_rcu *inet_opt; 12736e869138SDavid S. Miller struct flowi4 *fl4; 127432519f11SArnaldo Carvalho de Melo int err; 127532519f11SArnaldo Carvalho de Melo 127632519f11SArnaldo Carvalho de Melo /* Route is OK, nothing to do. */ 127732519f11SArnaldo Carvalho de Melo if (rt) 127832519f11SArnaldo Carvalho de Melo return 0; 127932519f11SArnaldo Carvalho de Melo 128032519f11SArnaldo Carvalho de Melo /* Reroute. */ 1281f6d8bd05SEric Dumazet rcu_read_lock(); 1282f6d8bd05SEric Dumazet inet_opt = rcu_dereference(inet->inet_opt); 1283c720c7e8SEric Dumazet daddr = inet->inet_daddr; 1284f6d8bd05SEric Dumazet if (inet_opt && inet_opt->opt.srr) 1285f6d8bd05SEric Dumazet daddr = inet_opt->opt.faddr; 1286f6d8bd05SEric Dumazet rcu_read_unlock(); 12876e869138SDavid S. Miller fl4 = &inet->cork.fl.u.ip4; 12886e869138SDavid S. Miller rt = ip_route_output_ports(sock_net(sk), fl4, sk, daddr, inet->inet_saddr, 128978fbfd8aSDavid S. Miller inet->inet_dport, inet->inet_sport, 129078fbfd8aSDavid S. Miller sk->sk_protocol, RT_CONN_FLAGS(sk), 129178fbfd8aSDavid S. Miller sk->sk_bound_dev_if); 1292b23dd4feSDavid S. Miller if (!IS_ERR(rt)) { 1293b23dd4feSDavid S. Miller err = 0; 1294d8d1f30bSChangli Gao sk_setup_caps(sk, &rt->dst); 1295b23dd4feSDavid S. Miller } else { 1296b23dd4feSDavid S. Miller err = PTR_ERR(rt); 1297b23dd4feSDavid S. Miller 129832519f11SArnaldo Carvalho de Melo /* Routing failed... */ 129932519f11SArnaldo Carvalho de Melo sk->sk_route_caps = 0; 130032519f11SArnaldo Carvalho de Melo /* 130132519f11SArnaldo Carvalho de Melo * Other protocols have to map its equivalent state to TCP_SYN_SENT. 130232519f11SArnaldo Carvalho de Melo * DCCP maps its DCCP_REQUESTING state to TCP_SYN_SENT. -acme 130332519f11SArnaldo Carvalho de Melo */ 1304e49e4affSKuniyuki Iwashima if (!READ_ONCE(sock_net(sk)->ipv4.sysctl_ip_dynaddr) || 130532519f11SArnaldo Carvalho de Melo sk->sk_state != TCP_SYN_SENT || 130632519f11SArnaldo Carvalho de Melo (sk->sk_userlocks & SOCK_BINDADDR_LOCK) || 130732519f11SArnaldo Carvalho de Melo (err = inet_sk_reselect_saddr(sk)) != 0) 130832519f11SArnaldo Carvalho de Melo sk->sk_err_soft = -err; 130932519f11SArnaldo Carvalho de Melo } 131032519f11SArnaldo Carvalho de Melo 131132519f11SArnaldo Carvalho de Melo return err; 131232519f11SArnaldo Carvalho de Melo } 131332519f11SArnaldo Carvalho de Melo EXPORT_SYMBOL(inet_sk_rebuild_header); 131432519f11SArnaldo Carvalho de Melo 1315563e0bb0SYafang Shao void inet_sk_set_state(struct sock *sk, int state) 1316563e0bb0SYafang Shao { 1317563e0bb0SYafang Shao trace_inet_sock_set_state(sk, sk->sk_state, state); 1318563e0bb0SYafang Shao sk->sk_state = state; 1319563e0bb0SYafang Shao } 1320563e0bb0SYafang Shao EXPORT_SYMBOL(inet_sk_set_state); 1321563e0bb0SYafang Shao 1322563e0bb0SYafang Shao void inet_sk_state_store(struct sock *sk, int newstate) 1323563e0bb0SYafang Shao { 1324563e0bb0SYafang Shao trace_inet_sock_set_state(sk, sk->sk_state, newstate); 1325563e0bb0SYafang Shao smp_store_release(&sk->sk_state, newstate); 1326563e0bb0SYafang Shao } 1327563e0bb0SYafang Shao 1328b8921ca8STom Herbert struct sk_buff *inet_gso_segment(struct sk_buff *skb, 1329c8f44affSMichał Mirosław netdev_features_t features) 1330f4c50d99SHerbert Xu { 13310c19f846SWillem de Bruijn bool udpfrag = false, fixedid = false, gso_partial, encap; 1332f4c50d99SHerbert Xu struct sk_buff *segs = ERR_PTR(-EINVAL); 1333bca49f84SVlad Yasevich const struct net_offload *ops; 13340c19f846SWillem de Bruijn unsigned int offset = 0; 1335f9242b6bSDavid S. Miller struct iphdr *iph; 1336802ab55aSAlexander Duyck int proto, tot_len; 13373347c960SEric Dumazet int nhoff; 1338f4c50d99SHerbert Xu int ihl; 1339f4c50d99SHerbert Xu int id; 1340f4c50d99SHerbert Xu 13413347c960SEric Dumazet skb_reset_network_header(skb); 13423347c960SEric Dumazet nhoff = skb_network_header(skb) - skb_mac_header(skb); 1343bbcf467dSHerbert Xu if (unlikely(!pskb_may_pull(skb, sizeof(*iph)))) 1344f4c50d99SHerbert Xu goto out; 1345f4c50d99SHerbert Xu 1346eddc9ec5SArnaldo Carvalho de Melo iph = ip_hdr(skb); 1347f4c50d99SHerbert Xu ihl = iph->ihl * 4; 1348f4c50d99SHerbert Xu if (ihl < sizeof(*iph)) 1349f4c50d99SHerbert Xu goto out; 1350f4c50d99SHerbert Xu 135147d27aadSEric Dumazet id = ntohs(iph->id); 135247d27aadSEric Dumazet proto = iph->protocol; 135347d27aadSEric Dumazet 135447d27aadSEric Dumazet /* Warning: after this point, iph might be no longer valid */ 1355bbcf467dSHerbert Xu if (unlikely(!pskb_may_pull(skb, ihl))) 1356f4c50d99SHerbert Xu goto out; 135747d27aadSEric Dumazet __skb_pull(skb, ihl); 1358f4c50d99SHerbert Xu 13598c3a897bSEric Dumazet encap = SKB_GSO_CB(skb)->encap_level > 0; 13608c3a897bSEric Dumazet if (encap) 13611e16aa3dSFlorian Westphal features &= skb->dev->hw_enc_features; 13623347c960SEric Dumazet SKB_GSO_CB(skb)->encap_level += ihl; 136373136267SPravin B Shelar 1364badff6d0SArnaldo Carvalho de Melo skb_reset_transport_header(skb); 136547d27aadSEric Dumazet 1366f4c50d99SHerbert Xu segs = ERR_PTR(-EPROTONOSUPPORT); 1367f4c50d99SHerbert Xu 1368cbc53e08SAlexander Duyck if (!skb->encapsulation || encap) { 13690c19f846SWillem de Bruijn udpfrag = !!(skb_shinfo(skb)->gso_type & SKB_GSO_UDP); 1370cbc53e08SAlexander Duyck fixedid = !!(skb_shinfo(skb)->gso_type & SKB_GSO_TCP_FIXEDID); 1371cbc53e08SAlexander Duyck 1372cbc53e08SAlexander Duyck /* fixed ID is invalid if DF bit is not set */ 1373a5108878SArnaldo Carvalho de Melo if (fixedid && !(ip_hdr(skb)->frag_off & htons(IP_DF))) 1374cbc53e08SAlexander Duyck goto out; 1375cbc53e08SAlexander Duyck } 1376dcd60771SEric Dumazet 1377bca49f84SVlad Yasevich ops = rcu_dereference(inet_offloads[proto]); 1378cc20ccedSTao Liu if (likely(ops && ops->callbacks.gso_segment)) { 1379f191a1d1SVlad Yasevich segs = ops->callbacks.gso_segment(skb, features); 1380cc20ccedSTao Liu if (!segs) 1381cc20ccedSTao Liu skb->network_header = skb_mac_header(skb) + nhoff - skb->head; 1382cc20ccedSTao Liu } 1383f4c50d99SHerbert Xu 138450c3a487SYOSHIFUJI Hideaki / 吉藤英明 if (IS_ERR_OR_NULL(segs)) 1385f4c50d99SHerbert Xu goto out; 1386f4c50d99SHerbert Xu 138707b26c94SSteffen Klassert gso_partial = !!(skb_shinfo(segs)->gso_type & SKB_GSO_PARTIAL); 138807b26c94SSteffen Klassert 1389f4c50d99SHerbert Xu skb = segs; 1390f4c50d99SHerbert Xu do { 13913347c960SEric Dumazet iph = (struct iphdr *)(skb_mac_header(skb) + nhoff); 13920c19f846SWillem de Bruijn if (udpfrag) { 13930c19f846SWillem de Bruijn iph->frag_off = htons(offset >> 3); 13940c19f846SWillem de Bruijn if (skb->next) 13950c19f846SWillem de Bruijn iph->frag_off |= htons(IP_MF); 13960c19f846SWillem de Bruijn offset += skb->len - nhoff - ihl; 13970c19f846SWillem de Bruijn tot_len = skb->len - nhoff; 13980c19f846SWillem de Bruijn } else if (skb_is_gso(skb)) { 1399802ab55aSAlexander Duyck if (!fixedid) { 1400802ab55aSAlexander Duyck iph->id = htons(id); 1401802ab55aSAlexander Duyck id += skb_shinfo(skb)->gso_segs; 1402490ab081SPravin B Shelar } 140307b26c94SSteffen Klassert 140407b26c94SSteffen Klassert if (gso_partial) 1405802ab55aSAlexander Duyck tot_len = skb_shinfo(skb)->gso_size + 1406802ab55aSAlexander Duyck SKB_GSO_CB(skb)->data_offset + 1407802ab55aSAlexander Duyck skb->head - (unsigned char *)iph; 140807b26c94SSteffen Klassert else 140907b26c94SSteffen Klassert tot_len = skb->len - nhoff; 1410802ab55aSAlexander Duyck } else { 1411802ab55aSAlexander Duyck if (!fixedid) 1412802ab55aSAlexander Duyck iph->id = htons(id++); 1413802ab55aSAlexander Duyck tot_len = skb->len - nhoff; 1414802ab55aSAlexander Duyck } 1415802ab55aSAlexander Duyck iph->tot_len = htons(tot_len); 141647d27aadSEric Dumazet ip_send_check(iph); 14178c3a897bSEric Dumazet if (encap) 14183347c960SEric Dumazet skb_reset_inner_headers(skb); 14193347c960SEric Dumazet skb->network_header = (u8 *)iph - skb->head; 1420c56cae23SToke Høiland-Jørgensen skb_reset_mac_len(skb); 1421f4c50d99SHerbert Xu } while ((skb = skb->next)); 1422f4c50d99SHerbert Xu 1423f4c50d99SHerbert Xu out: 1424f4c50d99SHerbert Xu return segs; 1425f4c50d99SHerbert Xu } 1426f4c50d99SHerbert Xu 1427418e897eSWillem de Bruijn static struct sk_buff *ipip_gso_segment(struct sk_buff *skb, 1428418e897eSWillem de Bruijn netdev_features_t features) 1429418e897eSWillem de Bruijn { 1430418e897eSWillem de Bruijn if (!(skb_shinfo(skb)->gso_type & SKB_GSO_IPXIP4)) 1431418e897eSWillem de Bruijn return ERR_PTR(-EINVAL); 1432418e897eSWillem de Bruijn 1433418e897eSWillem de Bruijn return inet_gso_segment(skb, features); 1434418e897eSWillem de Bruijn } 1435418e897eSWillem de Bruijn 1436d4546c25SDavid Miller struct sk_buff *inet_gro_receive(struct list_head *head, struct sk_buff *skb) 143773cc19f1SHerbert Xu { 1438bca49f84SVlad Yasevich const struct net_offload *ops; 1439d4546c25SDavid Miller struct sk_buff *pp = NULL; 1440b71d1d42SEric Dumazet const struct iphdr *iph; 1441d4546c25SDavid Miller struct sk_buff *p; 1442a5b1cf28SHerbert Xu unsigned int hlen; 1443a5b1cf28SHerbert Xu unsigned int off; 14441075f3f6SHerbert Xu unsigned int id; 144573cc19f1SHerbert Xu int flush = 1; 144673cc19f1SHerbert Xu int proto; 144773cc19f1SHerbert Xu 1448a5b1cf28SHerbert Xu off = skb_gro_offset(skb); 1449a5b1cf28SHerbert Xu hlen = off + sizeof(*iph); 1450a5b1cf28SHerbert Xu iph = skb_gro_header_fast(skb, off); 1451a5b1cf28SHerbert Xu if (skb_gro_header_hard(skb, hlen)) { 1452a5b1cf28SHerbert Xu iph = skb_gro_header_slow(skb, hlen, off); 145386911732SHerbert Xu if (unlikely(!iph)) 145473cc19f1SHerbert Xu goto out; 1455a5b1cf28SHerbert Xu } 145673cc19f1SHerbert Xu 1457f9242b6bSDavid S. Miller proto = iph->protocol; 145873cc19f1SHerbert Xu 1459bca49f84SVlad Yasevich ops = rcu_dereference(inet_offloads[proto]); 1460f191a1d1SVlad Yasevich if (!ops || !ops->callbacks.gro_receive) 1461fc1ca334SEric Dumazet goto out; 146273cc19f1SHerbert Xu 1463a5ad24beSHerbert Xu if (*(u8 *)iph != 0x45) 1464fc1ca334SEric Dumazet goto out; 146573cc19f1SHerbert Xu 14669b83e031SSteffen Klassert if (ip_is_fragment(iph)) 1467fc1ca334SEric Dumazet goto out; 14689b83e031SSteffen Klassert 1469a9e050f4SEric Dumazet if (unlikely(ip_fast_csum((u8 *)iph, 5))) 1470fc1ca334SEric Dumazet goto out; 147173cc19f1SHerbert Xu 14720eae88f3SEric Dumazet id = ntohl(*(__be32 *)&iph->id); 1473db8caf3dSEric Dumazet flush = (u16)((ntohl(*(__be32 *)iph) ^ skb_gro_len(skb)) | (id & ~IP_DF)); 14741075f3f6SHerbert Xu id >>= 16; 147573cc19f1SHerbert Xu 1476d4546c25SDavid Miller list_for_each_entry(p, head, list) { 147773cc19f1SHerbert Xu struct iphdr *iph2; 14781530545eSAlexander Duyck u16 flush_id; 147973cc19f1SHerbert Xu 148073cc19f1SHerbert Xu if (!NAPI_GRO_CB(p)->same_flow) 148173cc19f1SHerbert Xu continue; 148273cc19f1SHerbert Xu 1483299603e8SJerry Chu iph2 = (struct iphdr *)(p->data + off); 1484299603e8SJerry Chu /* The above works because, with the exception of the top 1485299603e8SJerry Chu * (inner most) layer, we only aggregate pkts with the same 1486299603e8SJerry Chu * hdr length so all the hdrs we'll need to verify will start 1487299603e8SJerry Chu * at the same offset. 1488299603e8SJerry Chu */ 1489a5ad24beSHerbert Xu if ((iph->protocol ^ iph2->protocol) | 14900eae88f3SEric Dumazet ((__force u32)iph->saddr ^ (__force u32)iph2->saddr) | 14910eae88f3SEric Dumazet ((__force u32)iph->daddr ^ (__force u32)iph2->daddr)) { 149273cc19f1SHerbert Xu NAPI_GRO_CB(p)->same_flow = 0; 149373cc19f1SHerbert Xu continue; 149473cc19f1SHerbert Xu } 149573cc19f1SHerbert Xu 149673cc19f1SHerbert Xu /* All fields must match except length and checksum. */ 149773cc19f1SHerbert Xu NAPI_GRO_CB(p)->flush |= 1498a5ad24beSHerbert Xu (iph->ttl ^ iph2->ttl) | 1499a9e050f4SEric Dumazet (iph->tos ^ iph2->tos) | 1500bf5a755fSJerry Chu ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF)); 150173cc19f1SHerbert Xu 150273cc19f1SHerbert Xu NAPI_GRO_CB(p)->flush |= flush; 15031530545eSAlexander Duyck 15041530545eSAlexander Duyck /* We need to store of the IP ID check to be included later 15051530545eSAlexander Duyck * when we can verify that this packet does in fact belong 15061530545eSAlexander Duyck * to a given flow. 15071530545eSAlexander Duyck */ 15081530545eSAlexander Duyck flush_id = (u16)(id - ntohs(iph2->id)); 15091530545eSAlexander Duyck 15101530545eSAlexander Duyck /* This bit of code makes it much easier for us to identify 15111530545eSAlexander Duyck * the cases where we are doing atomic vs non-atomic IP ID 15121530545eSAlexander Duyck * checks. Specifically an atomic check can return IP ID 15131530545eSAlexander Duyck * values 0 - 0xFFFF, while a non-atomic check can only 15141530545eSAlexander Duyck * return 0 or 0xFFFF. 15151530545eSAlexander Duyck */ 15161530545eSAlexander Duyck if (!NAPI_GRO_CB(p)->is_atomic || 15171530545eSAlexander Duyck !(iph->frag_off & htons(IP_DF))) { 15181530545eSAlexander Duyck flush_id ^= NAPI_GRO_CB(p)->count; 15191530545eSAlexander Duyck flush_id = flush_id ? 0xFFFF : 0; 152073cc19f1SHerbert Xu } 152173cc19f1SHerbert Xu 15221530545eSAlexander Duyck /* If the previous IP ID value was based on an atomic 15231530545eSAlexander Duyck * datagram we can overwrite the value and ignore it. 15241530545eSAlexander Duyck */ 15251530545eSAlexander Duyck if (NAPI_GRO_CB(skb)->is_atomic) 15261530545eSAlexander Duyck NAPI_GRO_CB(p)->flush_id = flush_id; 15271530545eSAlexander Duyck else 15281530545eSAlexander Duyck NAPI_GRO_CB(p)->flush_id |= flush_id; 15291530545eSAlexander Duyck } 15301530545eSAlexander Duyck 15311530545eSAlexander Duyck NAPI_GRO_CB(skb)->is_atomic = !!(iph->frag_off & htons(IP_DF)); 153273cc19f1SHerbert Xu NAPI_GRO_CB(skb)->flush |= flush; 1533299603e8SJerry Chu skb_set_network_header(skb, off); 1534299603e8SJerry Chu /* The above will be needed by the transport layer if there is one 1535299603e8SJerry Chu * immediately following this IP hdr. 1536299603e8SJerry Chu */ 1537299603e8SJerry Chu 15382c804d0fSEric Dumazet /* Note : No need to call skb_gro_postpull_rcsum() here, 15392c804d0fSEric Dumazet * as we already checked checksum over ipv4 header was 0 15402c804d0fSEric Dumazet */ 154186911732SHerbert Xu skb_gro_pull(skb, sizeof(*iph)); 154286911732SHerbert Xu skb_set_transport_header(skb, skb_gro_offset(skb)); 154373cc19f1SHerbert Xu 1544028e0a47SPaolo Abeni pp = indirect_call_gro_receive(tcp4_gro_receive, udp4_gro_receive, 1545028e0a47SPaolo Abeni ops->callbacks.gro_receive, head, skb); 154673cc19f1SHerbert Xu 154773cc19f1SHerbert Xu out: 15485f114163SSteffen Klassert skb_gro_flush_final(skb, pp, flush); 154973cc19f1SHerbert Xu 155073cc19f1SHerbert Xu return pp; 155173cc19f1SHerbert Xu } 155273cc19f1SHerbert Xu 1553d4546c25SDavid Miller static struct sk_buff *ipip_gro_receive(struct list_head *head, 1554fac8e0f5SJesse Gross struct sk_buff *skb) 1555fac8e0f5SJesse Gross { 1556fac8e0f5SJesse Gross if (NAPI_GRO_CB(skb)->encap_mark) { 1557fac8e0f5SJesse Gross NAPI_GRO_CB(skb)->flush = 1; 1558fac8e0f5SJesse Gross return NULL; 1559fac8e0f5SJesse Gross } 1560fac8e0f5SJesse Gross 1561fac8e0f5SJesse Gross NAPI_GRO_CB(skb)->encap_mark = 1; 1562fac8e0f5SJesse Gross 1563fac8e0f5SJesse Gross return inet_gro_receive(head, skb); 1564fac8e0f5SJesse Gross } 1565fac8e0f5SJesse Gross 1566822c8685SDeepa Dinamani #define SECONDS_PER_DAY 86400 1567822c8685SDeepa Dinamani 1568822c8685SDeepa Dinamani /* inet_current_timestamp - Return IP network timestamp 1569822c8685SDeepa Dinamani * 1570822c8685SDeepa Dinamani * Return milliseconds since midnight in network byte order. 1571822c8685SDeepa Dinamani */ 1572822c8685SDeepa Dinamani __be32 inet_current_timestamp(void) 1573822c8685SDeepa Dinamani { 1574822c8685SDeepa Dinamani u32 secs; 1575822c8685SDeepa Dinamani u32 msecs; 1576822c8685SDeepa Dinamani struct timespec64 ts; 1577822c8685SDeepa Dinamani 1578822c8685SDeepa Dinamani ktime_get_real_ts64(&ts); 1579822c8685SDeepa Dinamani 1580822c8685SDeepa Dinamani /* Get secs since midnight. */ 1581822c8685SDeepa Dinamani (void)div_u64_rem(ts.tv_sec, SECONDS_PER_DAY, &secs); 1582822c8685SDeepa Dinamani /* Convert to msecs. */ 1583822c8685SDeepa Dinamani msecs = secs * MSEC_PER_SEC; 1584822c8685SDeepa Dinamani /* Convert nsec to msec. */ 1585822c8685SDeepa Dinamani msecs += (u32)ts.tv_nsec / NSEC_PER_MSEC; 1586822c8685SDeepa Dinamani 1587822c8685SDeepa Dinamani /* Convert to network byte order. */ 15883ba9d300SDeepa Dinamani return htonl(msecs); 1589822c8685SDeepa Dinamani } 1590822c8685SDeepa Dinamani EXPORT_SYMBOL(inet_current_timestamp); 1591822c8685SDeepa Dinamani 1592f4713a3dSWillem de Bruijn int inet_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len) 1593f4713a3dSWillem de Bruijn { 1594f4713a3dSWillem de Bruijn if (sk->sk_family == AF_INET) 1595f4713a3dSWillem de Bruijn return ip_recv_error(sk, msg, len, addr_len); 1596f4713a3dSWillem de Bruijn #if IS_ENABLED(CONFIG_IPV6) 1597f4713a3dSWillem de Bruijn if (sk->sk_family == AF_INET6) 1598f4713a3dSWillem de Bruijn return pingv6_ops.ipv6_recv_error(sk, msg, len, addr_len); 1599f4713a3dSWillem de Bruijn #endif 1600f4713a3dSWillem de Bruijn return -EINVAL; 1601f4713a3dSWillem de Bruijn } 1602f4713a3dSWillem de Bruijn 1603b8921ca8STom Herbert int inet_gro_complete(struct sk_buff *skb, int nhoff) 160473cc19f1SHerbert Xu { 1605299603e8SJerry Chu __be16 newlen = htons(skb->len - nhoff); 1606299603e8SJerry Chu struct iphdr *iph = (struct iphdr *)(skb->data + nhoff); 1607bca49f84SVlad Yasevich const struct net_offload *ops; 1608f9242b6bSDavid S. Miller int proto = iph->protocol; 1609f9242b6bSDavid S. Miller int err = -ENOSYS; 161073cc19f1SHerbert Xu 1611294acf1cSPaolo Abeni if (skb->encapsulation) { 1612294acf1cSPaolo Abeni skb_set_inner_protocol(skb, cpu_to_be16(ETH_P_IP)); 1613c3caf119SJerry Chu skb_set_inner_network_header(skb, nhoff); 1614294acf1cSPaolo Abeni } 1615c3caf119SJerry Chu 161673cc19f1SHerbert Xu csum_replace2(&iph->check, iph->tot_len, newlen); 161773cc19f1SHerbert Xu iph->tot_len = newlen; 161873cc19f1SHerbert Xu 1619bca49f84SVlad Yasevich ops = rcu_dereference(inet_offloads[proto]); 1620f191a1d1SVlad Yasevich if (WARN_ON(!ops || !ops->callbacks.gro_complete)) 1621627b94f7SEric Dumazet goto out; 162273cc19f1SHerbert Xu 1623299603e8SJerry Chu /* Only need to add sizeof(*iph) to get to the next hdr below 1624299603e8SJerry Chu * because any hdr with option will have been flushed in 1625299603e8SJerry Chu * inet_gro_receive(). 1626299603e8SJerry Chu */ 1627028e0a47SPaolo Abeni err = INDIRECT_CALL_2(ops->callbacks.gro_complete, 1628028e0a47SPaolo Abeni tcp4_gro_complete, udp4_gro_complete, 1629028e0a47SPaolo Abeni skb, nhoff + sizeof(*iph)); 163073cc19f1SHerbert Xu 1631627b94f7SEric Dumazet out: 163273cc19f1SHerbert Xu return err; 163373cc19f1SHerbert Xu } 163473cc19f1SHerbert Xu 1635b8cba75bSJesse Gross static int ipip_gro_complete(struct sk_buff *skb, int nhoff) 1636b8cba75bSJesse Gross { 1637b8cba75bSJesse Gross skb->encapsulation = 1; 16387e13318dSTom Herbert skb_shinfo(skb)->gso_type |= SKB_GSO_IPXIP4; 1639b8cba75bSJesse Gross return inet_gro_complete(skb, nhoff); 1640b8cba75bSJesse Gross } 1641b8cba75bSJesse Gross 1642eee4fe4dSDenis V. Lunev int inet_ctl_sock_create(struct sock **sk, unsigned short family, 16435677242fSDenis V. Lunev unsigned short type, unsigned char protocol, 16445677242fSDenis V. Lunev struct net *net) 16453d58b5faSDenis V. Lunev { 1646eee4fe4dSDenis V. Lunev struct socket *sock; 164726abe143SEric W. Biederman int rc = sock_create_kern(net, family, type, protocol, &sock); 16483d58b5faSDenis V. Lunev 16493d58b5faSDenis V. Lunev if (rc == 0) { 1650eee4fe4dSDenis V. Lunev *sk = sock->sk; 1651eee4fe4dSDenis V. Lunev (*sk)->sk_allocation = GFP_ATOMIC; 16523d58b5faSDenis V. Lunev /* 16533d58b5faSDenis V. Lunev * Unhash it so that IP input processing does not even see it, 16543d58b5faSDenis V. Lunev * we do not wish this socket to see incoming packets. 16553d58b5faSDenis V. Lunev */ 1656eee4fe4dSDenis V. Lunev (*sk)->sk_prot->unhash(*sk); 16573d58b5faSDenis V. Lunev } 16583d58b5faSDenis V. Lunev return rc; 16593d58b5faSDenis V. Lunev } 16603d58b5faSDenis V. Lunev EXPORT_SYMBOL_GPL(inet_ctl_sock_create); 16613d58b5faSDenis V. Lunev 1662698365faSWANG Cong unsigned long snmp_fold_field(void __percpu *mib, int offt) 16635e0f0435SHerbert Xu { 16645e0f0435SHerbert Xu unsigned long res = 0; 1665698365faSWANG Cong int i; 16665e0f0435SHerbert Xu 1667698365faSWANG Cong for_each_possible_cpu(i) 1668c4c6bc31SRaghavendra K T res += snmp_get_cpu_field(mib, i, offt); 16695e0f0435SHerbert Xu return res; 16705e0f0435SHerbert Xu } 16715e0f0435SHerbert Xu EXPORT_SYMBOL_GPL(snmp_fold_field); 16725e0f0435SHerbert Xu 16734ce3c183SEric Dumazet #if BITS_PER_LONG==32 16744ce3c183SEric Dumazet 167580ec1927SDavid S. Miller u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offt, 1676c4c6bc31SRaghavendra K T size_t syncp_offset) 16774ce3c183SEric Dumazet { 16788f0ea0feSEric Dumazet void *bhptr; 16794ce3c183SEric Dumazet struct u64_stats_sync *syncp; 16808f0ea0feSEric Dumazet u64 v; 16814ce3c183SEric Dumazet unsigned int start; 16824ce3c183SEric Dumazet 1683698365faSWANG Cong bhptr = per_cpu_ptr(mib, cpu); 16844ce3c183SEric Dumazet syncp = (struct u64_stats_sync *)(bhptr + syncp_offset); 16854ce3c183SEric Dumazet do { 168657a7744eSEric W. Biederman start = u64_stats_fetch_begin_irq(syncp); 16878f0ea0feSEric Dumazet v = *(((u64 *)bhptr) + offt); 168857a7744eSEric W. Biederman } while (u64_stats_fetch_retry_irq(syncp, start)); 16894ce3c183SEric Dumazet 1690c4c6bc31SRaghavendra K T return v; 1691c4c6bc31SRaghavendra K T } 1692c4c6bc31SRaghavendra K T EXPORT_SYMBOL_GPL(snmp_get_cpu_field64); 1693c4c6bc31SRaghavendra K T 1694c4c6bc31SRaghavendra K T u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_offset) 1695c4c6bc31SRaghavendra K T { 1696c4c6bc31SRaghavendra K T u64 res = 0; 1697c4c6bc31SRaghavendra K T int cpu; 1698c4c6bc31SRaghavendra K T 1699c4c6bc31SRaghavendra K T for_each_possible_cpu(cpu) { 1700d1bfc625SMadalin Bucur res += snmp_get_cpu_field64(mib, cpu, offt, syncp_offset); 17014ce3c183SEric Dumazet } 17024ce3c183SEric Dumazet return res; 17034ce3c183SEric Dumazet } 17044ce3c183SEric Dumazet EXPORT_SYMBOL_GPL(snmp_fold_field64); 17054ce3c183SEric Dumazet #endif 17064ce3c183SEric Dumazet 17071da177e4SLinus Torvalds #ifdef CONFIG_IP_MULTICAST 170832613090SAlexey Dobriyan static const struct net_protocol igmp_protocol = { 17091da177e4SLinus Torvalds .handler = igmp_rcv, 17101da177e4SLinus Torvalds }; 17111da177e4SLinus Torvalds #endif 17121da177e4SLinus Torvalds 1713a8e3bb34SDavid Ahern /* thinking of making this const? Don't. 1714a8e3bb34SDavid Ahern * early_demux can change based on sysctl. 1715a8e3bb34SDavid Ahern */ 171603157937SDavid Ahern static struct net_protocol tcp_protocol = { 171741063e9dSDavid S. Miller .early_demux = tcp_v4_early_demux, 1718dddb64bcSsubashab@codeaurora.org .early_demux_handler = tcp_v4_early_demux, 17191da177e4SLinus Torvalds .handler = tcp_v4_rcv, 17201da177e4SLinus Torvalds .err_handler = tcp_v4_err, 17211da177e4SLinus Torvalds .no_policy = 1, 17228ed1dc44SHannes Frederic Sowa .icmp_strict_tag_validation = 1, 17231da177e4SLinus Torvalds }; 17241da177e4SLinus Torvalds 1725a8e3bb34SDavid Ahern /* thinking of making this const? Don't. 1726a8e3bb34SDavid Ahern * early_demux can change based on sysctl. 1727a8e3bb34SDavid Ahern */ 172803157937SDavid Ahern static struct net_protocol udp_protocol = { 1729421b3885SShawn Bohrer .early_demux = udp_v4_early_demux, 1730dddb64bcSsubashab@codeaurora.org .early_demux_handler = udp_v4_early_demux, 17311da177e4SLinus Torvalds .handler = udp_rcv, 17321da177e4SLinus Torvalds .err_handler = udp_err, 17331da177e4SLinus Torvalds .no_policy = 1, 17341da177e4SLinus Torvalds }; 17351da177e4SLinus Torvalds 173632613090SAlexey Dobriyan static const struct net_protocol icmp_protocol = { 17371da177e4SLinus Torvalds .handler = icmp_rcv, 17385b052042SLi Wei .err_handler = icmp_err, 17398b7817f3SHerbert Xu .no_policy = 1, 17401da177e4SLinus Torvalds }; 17411da177e4SLinus Torvalds 17429b4661bdSPavel Emelyanov static __net_init int ipv4_mib_init_net(struct net *net) 17439b4661bdSPavel Emelyanov { 1744827da44cSJohn Stultz int i; 1745827da44cSJohn Stultz 1746698365faSWANG Cong net->mib.tcp_statistics = alloc_percpu(struct tcp_mib); 1747698365faSWANG Cong if (!net->mib.tcp_statistics) 174857ef42d5SPavel Emelyanov goto err_tcp_mib; 1749698365faSWANG Cong net->mib.ip_statistics = alloc_percpu(struct ipstats_mib); 1750698365faSWANG Cong if (!net->mib.ip_statistics) 1751a20f5799SPavel Emelyanov goto err_ip_mib; 1752827da44cSJohn Stultz 1753827da44cSJohn Stultz for_each_possible_cpu(i) { 1754827da44cSJohn Stultz struct ipstats_mib *af_inet_stats; 1755698365faSWANG Cong af_inet_stats = per_cpu_ptr(net->mib.ip_statistics, i); 1756827da44cSJohn Stultz u64_stats_init(&af_inet_stats->syncp); 1757827da44cSJohn Stultz } 1758827da44cSJohn Stultz 1759698365faSWANG Cong net->mib.net_statistics = alloc_percpu(struct linux_mib); 1760698365faSWANG Cong if (!net->mib.net_statistics) 176161a7e260SPavel Emelyanov goto err_net_mib; 1762698365faSWANG Cong net->mib.udp_statistics = alloc_percpu(struct udp_mib); 1763698365faSWANG Cong if (!net->mib.udp_statistics) 17642f275f91SPavel Emelyanov goto err_udp_mib; 1765698365faSWANG Cong net->mib.udplite_statistics = alloc_percpu(struct udp_mib); 1766698365faSWANG Cong if (!net->mib.udplite_statistics) 1767386019d3SPavel Emelyanov goto err_udplite_mib; 1768698365faSWANG Cong net->mib.icmp_statistics = alloc_percpu(struct icmp_mib); 1769698365faSWANG Cong if (!net->mib.icmp_statistics) 1770b60538a0SPavel Emelyanov goto err_icmp_mib; 1771acb32ba3SEric Dumazet net->mib.icmpmsg_statistics = kzalloc(sizeof(struct icmpmsg_mib), 1772acb32ba3SEric Dumazet GFP_KERNEL); 1773acb32ba3SEric Dumazet if (!net->mib.icmpmsg_statistics) 1774923c6586SPavel Emelyanov goto err_icmpmsg_mib; 177557ef42d5SPavel Emelyanov 177657ef42d5SPavel Emelyanov tcp_mib_init(net); 17779b4661bdSPavel Emelyanov return 0; 177857ef42d5SPavel Emelyanov 1779923c6586SPavel Emelyanov err_icmpmsg_mib: 1780698365faSWANG Cong free_percpu(net->mib.icmp_statistics); 1781b60538a0SPavel Emelyanov err_icmp_mib: 1782698365faSWANG Cong free_percpu(net->mib.udplite_statistics); 1783386019d3SPavel Emelyanov err_udplite_mib: 1784698365faSWANG Cong free_percpu(net->mib.udp_statistics); 17852f275f91SPavel Emelyanov err_udp_mib: 1786698365faSWANG Cong free_percpu(net->mib.net_statistics); 178761a7e260SPavel Emelyanov err_net_mib: 1788698365faSWANG Cong free_percpu(net->mib.ip_statistics); 1789a20f5799SPavel Emelyanov err_ip_mib: 1790698365faSWANG Cong free_percpu(net->mib.tcp_statistics); 179157ef42d5SPavel Emelyanov err_tcp_mib: 179257ef42d5SPavel Emelyanov return -ENOMEM; 17939b4661bdSPavel Emelyanov } 17949b4661bdSPavel Emelyanov 17959b4661bdSPavel Emelyanov static __net_exit void ipv4_mib_exit_net(struct net *net) 17969b4661bdSPavel Emelyanov { 1797acb32ba3SEric Dumazet kfree(net->mib.icmpmsg_statistics); 1798698365faSWANG Cong free_percpu(net->mib.icmp_statistics); 1799698365faSWANG Cong free_percpu(net->mib.udplite_statistics); 1800698365faSWANG Cong free_percpu(net->mib.udp_statistics); 1801698365faSWANG Cong free_percpu(net->mib.net_statistics); 1802698365faSWANG Cong free_percpu(net->mib.ip_statistics); 1803698365faSWANG Cong free_percpu(net->mib.tcp_statistics); 1804fc518953SFlorian Westphal #ifdef CONFIG_MPTCP 1805fc518953SFlorian Westphal /* allocated on demand, see mptcp_init_sock() */ 1806fc518953SFlorian Westphal free_percpu(net->mib.mptcp_statistics); 1807fc518953SFlorian Westphal #endif 18089b4661bdSPavel Emelyanov } 18099b4661bdSPavel Emelyanov 18109b4661bdSPavel Emelyanov static __net_initdata struct pernet_operations ipv4_mib_ops = { 18119b4661bdSPavel Emelyanov .init = ipv4_mib_init_net, 18129b4661bdSPavel Emelyanov .exit = ipv4_mib_exit_net, 18139b4661bdSPavel Emelyanov }; 18149b4661bdSPavel Emelyanov 18151da177e4SLinus Torvalds static int __init init_ipv4_mibs(void) 18161da177e4SLinus Torvalds { 1817d89cbbb1SPavel Emelyanov return register_pernet_subsys(&ipv4_mib_ops); 18181da177e4SLinus Torvalds } 18191da177e4SLinus Torvalds 1820c9d8f1a6SCong Wang static __net_init int inet_init_net(struct net *net) 1821c9d8f1a6SCong Wang { 1822c9d8f1a6SCong Wang /* 1823c9d8f1a6SCong Wang * Set defaults for local port range 1824c9d8f1a6SCong Wang */ 1825c9d8f1a6SCong Wang seqlock_init(&net->ipv4.ip_local_ports.lock); 1826c9d8f1a6SCong Wang net->ipv4.ip_local_ports.range[0] = 32768; 182707f4c900SEric Dumazet net->ipv4.ip_local_ports.range[1] = 60999; 1828ba6b918aSCong Wang 1829ba6b918aSCong Wang seqlock_init(&net->ipv4.ping_group_range.lock); 1830ba6b918aSCong Wang /* 1831ba6b918aSCong Wang * Sane defaults - nobody may create ping sockets. 1832ba6b918aSCong Wang * Boot scripts should set this to distro-specific group. 1833ba6b918aSCong Wang */ 1834ba6b918aSCong Wang net->ipv4.ping_group_range.range[0] = make_kgid(&init_user_ns, 1); 1835ba6b918aSCong Wang net->ipv4.ping_group_range.range[1] = make_kgid(&init_user_ns, 0); 1836049bbf58SEzequiel Garcia 1837049bbf58SEzequiel Garcia /* Default values for sysctl-controlled parameters. 1838049bbf58SEzequiel Garcia * We set them here, in case sysctl is not compiled. 1839049bbf58SEzequiel Garcia */ 1840049bbf58SEzequiel Garcia net->ipv4.sysctl_ip_default_ttl = IPDEFTTL; 1841432e05d3SPetr Machata net->ipv4.sysctl_ip_fwd_update_priority = 1; 1842049bbf58SEzequiel Garcia net->ipv4.sysctl_ip_dynaddr = 0; 1843049bbf58SEzequiel Garcia net->ipv4.sysctl_ip_early_demux = 1; 1844dddb64bcSsubashab@codeaurora.org net->ipv4.sysctl_udp_early_demux = 1; 1845dddb64bcSsubashab@codeaurora.org net->ipv4.sysctl_tcp_early_demux = 1; 18464f80116dSRoopa Prabhu net->ipv4.sysctl_nexthop_compat_mode = 1; 18474548b683SKrister Johansen #ifdef CONFIG_SYSCTL 18484548b683SKrister Johansen net->ipv4.sysctl_ip_prot_sock = PROT_SOCK; 18494548b683SKrister Johansen #endif 1850049bbf58SEzequiel Garcia 18511714020eSNikolay Borisov /* Some igmp sysctl, whose values are always used */ 18521714020eSNikolay Borisov net->ipv4.sysctl_igmp_max_memberships = 20; 18531714020eSNikolay Borisov net->ipv4.sysctl_igmp_max_msf = 10; 18541714020eSNikolay Borisov /* IGMP reports for link-local multicast groups are enabled by default */ 18551714020eSNikolay Borisov net->ipv4.sysctl_igmp_llm_reports = 1; 18561714020eSNikolay Borisov net->ipv4.sysctl_igmp_qrv = 2; 18571714020eSNikolay Borisov 1858680aea08SAmit Cohen net->ipv4.sysctl_fib_notify_on_flag_change = 0; 1859680aea08SAmit Cohen 1860c9d8f1a6SCong Wang return 0; 1861c9d8f1a6SCong Wang } 1862c9d8f1a6SCong Wang 1863c9d8f1a6SCong Wang static __net_initdata struct pernet_operations af_inet_ops = { 1864c9d8f1a6SCong Wang .init = inet_init_net, 1865c9d8f1a6SCong Wang }; 1866c9d8f1a6SCong Wang 1867c9d8f1a6SCong Wang static int __init init_inet_pernet_ops(void) 1868c9d8f1a6SCong Wang { 1869c9d8f1a6SCong Wang return register_pernet_subsys(&af_inet_ops); 1870c9d8f1a6SCong Wang } 1871c9d8f1a6SCong Wang 18721da177e4SLinus Torvalds static int ipv4_proc_init(void); 18731da177e4SLinus Torvalds 187430e224d7SHerbert Xu /* 187530e224d7SHerbert Xu * IP protocol layer initialiser 187630e224d7SHerbert Xu */ 187730e224d7SHerbert Xu 187822061d80SVlad Yasevich static struct packet_offload ip_packet_offload __read_mostly = { 187922061d80SVlad Yasevich .type = cpu_to_be16(ETH_P_IP), 1880f191a1d1SVlad Yasevich .callbacks = { 1881f4c50d99SHerbert Xu .gso_segment = inet_gso_segment, 188273cc19f1SHerbert Xu .gro_receive = inet_gro_receive, 188373cc19f1SHerbert Xu .gro_complete = inet_gro_complete, 1884f191a1d1SVlad Yasevich }, 188530e224d7SHerbert Xu }; 188630e224d7SHerbert Xu 1887cb32f511SEric Dumazet static const struct net_offload ipip_offload = { 1888cb32f511SEric Dumazet .callbacks = { 1889418e897eSWillem de Bruijn .gso_segment = ipip_gso_segment, 1890fac8e0f5SJesse Gross .gro_receive = ipip_gro_receive, 1891b8cba75bSJesse Gross .gro_complete = ipip_gro_complete, 1892cb32f511SEric Dumazet }, 1893cb32f511SEric Dumazet }; 1894cb32f511SEric Dumazet 189593b1b31fSTonghao Zhang static int __init ipip_offload_init(void) 189693b1b31fSTonghao Zhang { 189793b1b31fSTonghao Zhang return inet_add_offload(&ipip_offload, IPPROTO_IPIP); 189893b1b31fSTonghao Zhang } 189993b1b31fSTonghao Zhang 1900808a8f88SVlad Yasevich static int __init ipv4_offload_init(void) 1901808a8f88SVlad Yasevich { 1902808a8f88SVlad Yasevich /* 1903808a8f88SVlad Yasevich * Add offloads 1904808a8f88SVlad Yasevich */ 1905da5bab07SDaniel Borkmann if (udpv4_offload_init() < 0) 1906808a8f88SVlad Yasevich pr_crit("%s: Cannot add UDP protocol offload\n", __func__); 190728850dc7SDaniel Borkmann if (tcpv4_offload_init() < 0) 190828850dc7SDaniel Borkmann pr_crit("%s: Cannot add TCP protocol offload\n", __func__); 190993b1b31fSTonghao Zhang if (ipip_offload_init() < 0) 191093b1b31fSTonghao Zhang pr_crit("%s: Cannot add IPIP protocol offload\n", __func__); 1911808a8f88SVlad Yasevich 1912808a8f88SVlad Yasevich dev_add_offload(&ip_packet_offload); 1913808a8f88SVlad Yasevich return 0; 1914808a8f88SVlad Yasevich } 1915808a8f88SVlad Yasevich 1916808a8f88SVlad Yasevich fs_initcall(ipv4_offload_init); 1917808a8f88SVlad Yasevich 1918808a8f88SVlad Yasevich static struct packet_type ip_packet_type __read_mostly = { 1919808a8f88SVlad Yasevich .type = cpu_to_be16(ETH_P_IP), 1920808a8f88SVlad Yasevich .func = ip_rcv, 192117266ee9SEdward Cree .list_func = ip_list_rcv, 1922808a8f88SVlad Yasevich }; 1923808a8f88SVlad Yasevich 19241da177e4SLinus Torvalds static int __init inet_init(void) 19251da177e4SLinus Torvalds { 19261da177e4SLinus Torvalds struct inet_protosw *q; 19271da177e4SLinus Torvalds struct list_head *r; 1928b6246f4dSColin Ian King int rc; 19291da177e4SLinus Torvalds 1930b4772ef8SEyal Birger sock_skb_cb_check_size(sizeof(struct inet_skb_parm)); 19311da177e4SLinus Torvalds 19321da177e4SLinus Torvalds rc = proto_register(&tcp_prot, 1); 19331da177e4SLinus Torvalds if (rc) 1934122ff243SWANG Cong goto out; 19351da177e4SLinus Torvalds 19361da177e4SLinus Torvalds rc = proto_register(&udp_prot, 1); 19371da177e4SLinus Torvalds if (rc) 19381da177e4SLinus Torvalds goto out_unregister_tcp_proto; 19391da177e4SLinus Torvalds 19401da177e4SLinus Torvalds rc = proto_register(&raw_prot, 1); 19411da177e4SLinus Torvalds if (rc) 19421da177e4SLinus Torvalds goto out_unregister_udp_proto; 19431da177e4SLinus Torvalds 1944c319b4d7SVasiliy Kulikov rc = proto_register(&ping_prot, 1); 1945c319b4d7SVasiliy Kulikov if (rc) 1946c319b4d7SVasiliy Kulikov goto out_unregister_raw_proto; 1947c319b4d7SVasiliy Kulikov 19481da177e4SLinus Torvalds /* 19491da177e4SLinus Torvalds * Tell SOCKET that we are alive... 19501da177e4SLinus Torvalds */ 19511da177e4SLinus Torvalds 19521da177e4SLinus Torvalds (void)sock_register(&inet_family_ops); 19531da177e4SLinus Torvalds 1954bd7b1533SAl Viro #ifdef CONFIG_SYSCTL 1955bd7b1533SAl Viro ip_static_sysctl_init(); 1956bd7b1533SAl Viro #endif 1957bd7b1533SAl Viro 19581da177e4SLinus Torvalds /* 19591da177e4SLinus Torvalds * Add all the base protocols. 19601da177e4SLinus Torvalds */ 19611da177e4SLinus Torvalds 19621da177e4SLinus Torvalds if (inet_add_protocol(&icmp_protocol, IPPROTO_ICMP) < 0) 1963058bd4d2SJoe Perches pr_crit("%s: Cannot add ICMP protocol\n", __func__); 19641da177e4SLinus Torvalds if (inet_add_protocol(&udp_protocol, IPPROTO_UDP) < 0) 1965058bd4d2SJoe Perches pr_crit("%s: Cannot add UDP protocol\n", __func__); 19661da177e4SLinus Torvalds if (inet_add_protocol(&tcp_protocol, IPPROTO_TCP) < 0) 1967058bd4d2SJoe Perches pr_crit("%s: Cannot add TCP protocol\n", __func__); 19681da177e4SLinus Torvalds #ifdef CONFIG_IP_MULTICAST 19691da177e4SLinus Torvalds if (inet_add_protocol(&igmp_protocol, IPPROTO_IGMP) < 0) 1970058bd4d2SJoe Perches pr_crit("%s: Cannot add IGMP protocol\n", __func__); 19711da177e4SLinus Torvalds #endif 19721da177e4SLinus Torvalds 19731da177e4SLinus Torvalds /* Register the socket-side information for inet_create. */ 19741da177e4SLinus Torvalds for (r = &inetsw[0]; r < &inetsw[SOCK_MAX]; ++r) 19751da177e4SLinus Torvalds INIT_LIST_HEAD(r); 19761da177e4SLinus Torvalds 19771da177e4SLinus Torvalds for (q = inetsw_array; q < &inetsw_array[INETSW_ARRAY_LEN]; ++q) 19781da177e4SLinus Torvalds inet_register_protosw(q); 19791da177e4SLinus Torvalds 19801da177e4SLinus Torvalds /* 19811da177e4SLinus Torvalds * Set the ARP module up 19821da177e4SLinus Torvalds */ 19831da177e4SLinus Torvalds 19841da177e4SLinus Torvalds arp_init(); 19851da177e4SLinus Torvalds 19861da177e4SLinus Torvalds /* 19871da177e4SLinus Torvalds * Set the IP module up 19881da177e4SLinus Torvalds */ 19891da177e4SLinus Torvalds 19901da177e4SLinus Torvalds ip_init(); 19911da177e4SLinus Torvalds 1992e22e45fcSMuchun Song /* Initialise per-cpu ipv4 mibs */ 1993e22e45fcSMuchun Song if (init_ipv4_mibs()) 1994e22e45fcSMuchun Song panic("%s: Cannot init ipv4 mibs\n", __func__); 1995e22e45fcSMuchun Song 19961da177e4SLinus Torvalds /* Setup TCP slab cache for open requests. */ 19971da177e4SLinus Torvalds tcp_init(); 19981da177e4SLinus Torvalds 199995766fffSHideo Aoki /* Setup UDP memory threshold */ 200095766fffSHideo Aoki udp_init(); 200195766fffSHideo Aoki 2002ba4e58ecSGerrit Renker /* Add UDP-Lite (RFC 3828) */ 2003ba4e58ecSGerrit Renker udplite4_register(); 20041da177e4SLinus Torvalds 20056897445fSMike Manning raw_init(); 20066897445fSMike Manning 2007c319b4d7SVasiliy Kulikov ping_init(); 2008c319b4d7SVasiliy Kulikov 20091da177e4SLinus Torvalds /* 20101da177e4SLinus Torvalds * Set the ICMP layer up 20111da177e4SLinus Torvalds */ 20121da177e4SLinus Torvalds 2013a5710d65SDenis V. Lunev if (icmp_init() < 0) 2014a5710d65SDenis V. Lunev panic("Failed to create the ICMP control socket.\n"); 20151da177e4SLinus Torvalds 20161da177e4SLinus Torvalds /* 20171da177e4SLinus Torvalds * Initialise the multicast router 20181da177e4SLinus Torvalds */ 20191da177e4SLinus Torvalds #if defined(CONFIG_IP_MROUTE) 202003d2f897SWang Chen if (ip_mr_init()) 2021058bd4d2SJoe Perches pr_crit("%s: Cannot init ipv4 mroute\n", __func__); 20221da177e4SLinus Torvalds #endif 2023c9d8f1a6SCong Wang 2024c9d8f1a6SCong Wang if (init_inet_pernet_ops()) 2025c9d8f1a6SCong Wang pr_crit("%s: Cannot init ipv4 inet pernet ops\n", __func__); 20261da177e4SLinus Torvalds 20271da177e4SLinus Torvalds ipv4_proc_init(); 20281da177e4SLinus Torvalds 20291da177e4SLinus Torvalds ipfrag_init(); 20301da177e4SLinus Torvalds 203130e224d7SHerbert Xu dev_add_pack(&ip_packet_type); 203230e224d7SHerbert Xu 2033045a0fa0SThomas Graf ip_tunnel_core_init(); 2034045a0fa0SThomas Graf 20351da177e4SLinus Torvalds rc = 0; 20361da177e4SLinus Torvalds out: 20371da177e4SLinus Torvalds return rc; 2038c319b4d7SVasiliy Kulikov out_unregister_raw_proto: 2039c319b4d7SVasiliy Kulikov proto_unregister(&raw_prot); 20401da177e4SLinus Torvalds out_unregister_udp_proto: 20411da177e4SLinus Torvalds proto_unregister(&udp_prot); 2042321efff7SOlaf Kirch out_unregister_tcp_proto: 2043321efff7SOlaf Kirch proto_unregister(&tcp_prot); 20441da177e4SLinus Torvalds goto out; 20451da177e4SLinus Torvalds } 20461da177e4SLinus Torvalds 2047a536e077SHeiko Carstens fs_initcall(inet_init); 20481da177e4SLinus Torvalds 20491da177e4SLinus Torvalds /* ------------------------------------------------------------------------ */ 20501da177e4SLinus Torvalds 20511da177e4SLinus Torvalds #ifdef CONFIG_PROC_FS 20521da177e4SLinus Torvalds static int __init ipv4_proc_init(void) 20531da177e4SLinus Torvalds { 20541da177e4SLinus Torvalds int rc = 0; 20551da177e4SLinus Torvalds 20561da177e4SLinus Torvalds if (raw_proc_init()) 20571da177e4SLinus Torvalds goto out_raw; 20581da177e4SLinus Torvalds if (tcp4_proc_init()) 20591da177e4SLinus Torvalds goto out_tcp; 20601da177e4SLinus Torvalds if (udp4_proc_init()) 20611da177e4SLinus Torvalds goto out_udp; 2062c319b4d7SVasiliy Kulikov if (ping_proc_init()) 2063c319b4d7SVasiliy Kulikov goto out_ping; 20641da177e4SLinus Torvalds if (ip_misc_proc_init()) 20651da177e4SLinus Torvalds goto out_misc; 20661da177e4SLinus Torvalds out: 20671da177e4SLinus Torvalds return rc; 20681da177e4SLinus Torvalds out_misc: 2069c319b4d7SVasiliy Kulikov ping_proc_exit(); 2070c319b4d7SVasiliy Kulikov out_ping: 20711da177e4SLinus Torvalds udp4_proc_exit(); 20721da177e4SLinus Torvalds out_udp: 20731da177e4SLinus Torvalds tcp4_proc_exit(); 20741da177e4SLinus Torvalds out_tcp: 20751da177e4SLinus Torvalds raw_proc_exit(); 20761da177e4SLinus Torvalds out_raw: 20771da177e4SLinus Torvalds rc = -ENOMEM; 20781da177e4SLinus Torvalds goto out; 20791da177e4SLinus Torvalds } 20801da177e4SLinus Torvalds 20811da177e4SLinus Torvalds #else /* CONFIG_PROC_FS */ 20821da177e4SLinus Torvalds static int __init ipv4_proc_init(void) 20831da177e4SLinus Torvalds { 20841da177e4SLinus Torvalds return 0; 20851da177e4SLinus Torvalds } 20861da177e4SLinus Torvalds #endif /* CONFIG_PROC_FS */ 2087