Lines Matching refs:sk
51 * Anonymous : inet_create tidied up (sk->reuse setting)
52 * Alan Cox : inet sockets don't set sk->type!
149 * @sk: Socket to use a capability on or through
157 bool sk_ns_capable(const struct sock *sk,
160 return file_ns_capable(sk->sk_socket->file, user_ns, cap) &&
167 * @sk: Socket to use a capability on or through
174 bool sk_capable(const struct sock *sk, int cap)
176 return sk_ns_capable(sk, &init_user_ns, cap);
182 * @sk: Socket to use a capability on or through
189 bool sk_net_capable(const struct sock *sk, int cap)
191 return sk_ns_capable(sk, sock_net(sk)->user_ns, cap);
259 * sk_callback_lock and sk queues locking rules are per-address-family,
287 * @sk: socket to set it on
293 void sk_set_memalloc(struct sock *sk)
295 sock_set_flag(sk, SOCK_MEMALLOC);
296 sk->sk_allocation |= __GFP_MEMALLOC;
301 void sk_clear_memalloc(struct sock *sk)
303 sock_reset_flag(sk, SOCK_MEMALLOC);
304 sk->sk_allocation &= ~__GFP_MEMALLOC;
314 sk_mem_reclaim(sk);
318 int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
324 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
327 ret = sk->sk_backlog_rcv(sk, skb);
416 static bool sock_needs_netstamp(const struct sock *sk)
418 switch (sk->sk_family) {
427 static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
429 if (sk->sk_flags & flags) {
430 sk->sk_flags &= ~flags;
431 if (sock_needs_netstamp(sk) &&
432 !(sk->sk_flags & SK_FLAGS_TIMESTAMP))
438 int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
441 struct sk_buff_head *list = &sk->sk_receive_queue;
443 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
444 atomic_inc(&sk->sk_drops);
445 trace_sock_rcvqueue_full(sk, skb);
449 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
450 atomic_inc(&sk->sk_drops);
455 skb_set_owner_r(skb, sk);
463 sock_skb_set_dropcount(sk, skb);
467 if (!sock_flag(sk, SOCK_DEAD))
468 sk->sk_data_ready(sk);
473 int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
477 err = sk_filter(sk, skb);
481 return __sock_queue_rcv_skb(sk, skb);
485 int __sk_receive_skb(struct sock *sk, struct sk_buff *skb,
490 if (sk_filter_trim_cap(sk, skb, trim_cap))
495 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
496 atomic_inc(&sk->sk_drops);
500 bh_lock_sock_nested(sk);
502 bh_lock_sock(sk);
503 if (!sock_owned_by_user(sk)) {
507 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
509 rc = sk_backlog_rcv(sk, skb);
511 mutex_release(&sk->sk_lock.dep_map, _RET_IP_);
512 } else if (sk_add_backlog(sk, skb, READ_ONCE(sk->sk_rcvbuf))) {
513 bh_unlock_sock(sk);
514 atomic_inc(&sk->sk_drops);
518 bh_unlock_sock(sk);
521 sock_put(sk);
529 struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
531 struct dst_entry *dst = __sk_dst_get(sk);
534 sk_tx_queue_clear(sk);
535 WRITE_ONCE(sk->sk_dst_pending_confirm, 0);
536 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
545 struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
547 struct dst_entry *dst = sk_dst_get(sk);
550 sk_dst_reset(sk);
559 static int sock_bindtoindex_locked(struct sock *sk, int ifindex)
563 struct net *net = sock_net(sk);
567 if (sk->sk_bound_dev_if && !ns_capable(net->user_ns, CAP_NET_RAW))
574 sk->sk_bound_dev_if = ifindex;
575 if (sk->sk_prot->rehash)
576 sk->sk_prot->rehash(sk);
577 sk_dst_reset(sk);
587 int sock_bindtoindex(struct sock *sk, int ifindex, bool lock_sk)
592 lock_sock(sk);
593 ret = sock_bindtoindex_locked(sk, ifindex);
595 release_sock(sk);
601 static int sock_setbindtodevice(struct sock *sk, sockptr_t optval, int optlen)
605 struct net *net = sock_net(sk);
640 return sock_bindtoindex(sk, index, true);
647 static int sock_getbindtodevice(struct sock *sk, char __user *optval,
652 struct net *net = sock_net(sk);
655 if (sk->sk_bound_dev_if == 0) {
664 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
687 bool sk_mc_loop(struct sock *sk)
691 if (!sk)
693 /* IPV6_ADDRFORM can change sk->sk_family under us. */
694 switch (READ_ONCE(sk->sk_family)) {
696 return inet_sk(sk)->mc_loop;
699 return inet6_sk(sk)->mc_loop;
707 void sock_set_reuseaddr(struct sock *sk)
709 lock_sock(sk);
710 sk->sk_reuse = SK_CAN_REUSE;
711 release_sock(sk);
715 void sock_set_reuseport(struct sock *sk)
717 lock_sock(sk);
718 sk->sk_reuseport = true;
719 release_sock(sk);
723 void sock_no_linger(struct sock *sk)
725 lock_sock(sk);
726 sk->sk_lingertime = 0;
727 sock_set_flag(sk, SOCK_LINGER);
728 release_sock(sk);
732 void sock_set_priority(struct sock *sk, u32 priority)
734 lock_sock(sk);
735 sk->sk_priority = priority;
736 release_sock(sk);
740 void sock_set_sndtimeo(struct sock *sk, s64 secs)
742 lock_sock(sk);
744 sk->sk_sndtimeo = secs * HZ;
746 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
747 release_sock(sk);
751 static void __sock_set_timestamps(struct sock *sk, bool val, bool new, bool ns)
754 sock_valbool_flag(sk, SOCK_TSTAMP_NEW, new);
755 sock_valbool_flag(sk, SOCK_RCVTSTAMPNS, ns);
756 sock_set_flag(sk, SOCK_RCVTSTAMP);
757 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
759 sock_reset_flag(sk, SOCK_RCVTSTAMP);
760 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
764 void sock_enable_timestamps(struct sock *sk)
766 lock_sock(sk);
767 __sock_set_timestamps(sk, true, false, true);
768 release_sock(sk);
772 void sock_set_keepalive(struct sock *sk)
774 lock_sock(sk);
775 if (sk->sk_prot->keepalive)
776 sk->sk_prot->keepalive(sk, true);
777 sock_valbool_flag(sk, SOCK_KEEPOPEN, true);
778 release_sock(sk);
782 static void __sock_set_rcvbuf(struct sock *sk, int val)
788 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
800 WRITE_ONCE(sk->sk_rcvbuf, max_t(int, val * 2, SOCK_MIN_RCVBUF));
803 void sock_set_rcvbuf(struct sock *sk, int val)
805 lock_sock(sk);
806 __sock_set_rcvbuf(sk, val);
807 release_sock(sk);
811 static void __sock_set_mark(struct sock *sk, u32 val)
813 if (val != sk->sk_mark) {
814 sk->sk_mark = val;
815 sk_dst_reset(sk);
819 void sock_set_mark(struct sock *sk, u32 val)
821 lock_sock(sk);
822 __sock_set_mark(sk, val);
823 release_sock(sk);
836 struct sock *sk = sock->sk;
847 return sock_setbindtodevice(sk, optval, optlen);
857 lock_sock(sk);
864 sock_valbool_flag(sk, SOCK_DBG, valbool);
867 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
870 sk->sk_reuseport = valbool;
879 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
880 sk_dst_reset(sk);
883 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
897 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
898 WRITE_ONCE(sk->sk_sndbuf,
901 sk->sk_write_space(sk);
923 __sock_set_rcvbuf(sk, min_t(u32, val, READ_ONCE(sysctl_rmem_max)));
935 __sock_set_rcvbuf(sk, max(val, 0));
939 if (sk->sk_prot->keepalive)
940 sk->sk_prot->keepalive(sk, valbool);
941 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
945 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
949 sk->sk_no_check_tx = valbool;
954 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
955 sk->sk_priority = val;
970 sock_reset_flag(sk, SOCK_LINGER);
974 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
977 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
978 sock_set_flag(sk, SOCK_LINGER);
993 __sock_set_timestamps(sk, valbool, false, false);
996 __sock_set_timestamps(sk, valbool, true, false);
999 __sock_set_timestamps(sk, valbool, false, true);
1002 __sock_set_timestamps(sk, valbool, true, true);
1012 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
1013 if (sk->sk_protocol == IPPROTO_TCP &&
1014 sk->sk_type == SOCK_STREAM) {
1015 if ((1 << sk->sk_state) &
1020 sk->sk_tskey = tcp_sk(sk)->snd_una;
1022 sk->sk_tskey = 0;
1032 sk->sk_tsflags = val;
1033 sock_valbool_flag(sk, SOCK_TSTAMP_NEW, optname == SO_TIMESTAMPING_NEW);
1036 sock_enable_timestamp(sk,
1039 sock_disable_timestamp(sk,
1047 ret = sock->ops->set_rcvlowat(sk, val);
1049 WRITE_ONCE(sk->sk_rcvlowat, val ? : 1);
1054 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval,
1060 ret = sock_set_timeout(&sk->sk_sndtimeo, optval,
1069 ret = sk_attach_filter(&fprog, sk);
1081 ret = sk_attach_bpf(ufd, sk);
1090 ret = sk_reuseport_attach_filter(&fprog, sk);
1102 ret = sk_reuseport_attach_bpf(ufd, sk);
1107 ret = reuseport_detach_prog(sk);
1111 ret = sk_detach_filter(sk);
1115 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
1118 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
1128 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1133 __sock_set_mark(sk, val);
1137 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
1141 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
1146 ret = sock->ops->set_peek_off(sk, val);
1152 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
1156 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
1162 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
1168 WRITE_ONCE(sk->sk_ll_usec, val);
1184 cmpxchg(&sk->sk_pacing_status,
1188 WRITE_ONCE(sk->sk_max_pacing_rate, ulval);
1189 sk->sk_pacing_rate = min(sk->sk_pacing_rate, ulval);
1193 WRITE_ONCE(sk->sk_incoming_cpu, val);
1198 dst_negative_advice(sk);
1202 if (sk->sk_family == PF_INET || sk->sk_family == PF_INET6) {
1203 if (!((sk->sk_type == SOCK_STREAM &&
1204 sk->sk_protocol == IPPROTO_TCP) ||
1205 (sk->sk_type == SOCK_DGRAM &&
1206 sk->sk_protocol == IPPROTO_UDP)))
1208 } else if (sk->sk_family != PF_RDS) {
1215 sock_valbool_flag(sk, SOCK_ZEROCOPY, valbool);
1235 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1239 sock_valbool_flag(sk, SOCK_TXTIME, true);
1240 sk->sk_clockid = sk_txtime.clockid;
1241 sk->sk_txtime_deadline_mode =
1243 sk->sk_txtime_report_errors =
1248 ret = sock_bindtoindex_locked(sk, val);
1255 release_sock(sk);
1260 static const struct cred *sk_get_peer_cred(struct sock *sk)
1264 spin_lock(&sk->sk_peer_lock);
1265 cred = get_cred(sk->sk_peer_cred);
1266 spin_unlock(&sk->sk_peer_lock);
1299 struct sock *sk = sock->sk;
1324 v.val = sock_flag(sk, SOCK_DBG);
1328 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1332 v.val = sock_flag(sk, SOCK_BROADCAST);
1336 v.val = READ_ONCE(sk->sk_sndbuf);
1340 v.val = READ_ONCE(sk->sk_rcvbuf);
1344 v.val = sk->sk_reuse;
1348 v.val = sk->sk_reuseport;
1352 v.val = sock_flag(sk, SOCK_KEEPOPEN);
1356 v.val = sk->sk_type;
1360 v.val = sk->sk_protocol;
1364 v.val = sk->sk_family;
1368 v.val = -sock_error(sk);
1370 v.val = xchg(&sk->sk_err_soft, 0);
1374 v.val = sock_flag(sk, SOCK_URGINLINE);
1378 v.val = sk->sk_no_check_tx;
1382 v.val = sk->sk_priority;
1387 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
1388 v.ling.l_linger = sk->sk_lingertime / HZ;
1395 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1396 !sock_flag(sk, SOCK_TSTAMP_NEW) &&
1397 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1401 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS) && !sock_flag(sk, SOCK_TSTAMP_NEW);
1405 v.val = sock_flag(sk, SOCK_RCVTSTAMP) && sock_flag(sk, SOCK_TSTAMP_NEW);
1409 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS) && sock_flag(sk, SOCK_TSTAMP_NEW);
1413 v.val = sk->sk_tsflags;
1418 lv = sock_get_timeout(sk->sk_rcvtimeo, &v, SO_RCVTIMEO_OLD == optname);
1423 lv = sock_get_timeout(sk->sk_sndtimeo, &v, SO_SNDTIMEO_OLD == optname);
1427 v.val = READ_ONCE(sk->sk_rcvlowat);
1444 spin_lock(&sk->sk_peer_lock);
1445 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1446 spin_unlock(&sk->sk_peer_lock);
1458 cred = sk_get_peer_cred(sk);
1495 v.val = sk->sk_state == TCP_LISTEN;
1506 v.val = sk->sk_mark;
1510 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
1514 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
1521 v.val = READ_ONCE(sk->sk_peek_off);
1524 v.val = sock_flag(sk, SOCK_NOFCS);
1528 return sock_getbindtodevice(sk, optval, optlen, len);
1531 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1538 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1546 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1551 v.val = READ_ONCE(sk->sk_ll_usec);
1559 v.ulval = READ_ONCE(sk->sk_max_pacing_rate);
1563 READ_ONCE(sk->sk_max_pacing_rate));
1568 v.val = READ_ONCE(sk->sk_incoming_cpu);
1575 sk_get_meminfo(sk, meminfo);
1586 v.val = READ_ONCE(sk->sk_napi_id);
1599 v.val64 = sock_gen_cookie(sk);
1603 v.val = sock_flag(sk, SOCK_ZEROCOPY);
1608 v.txtime.clockid = sk->sk_clockid;
1609 v.txtime.flags |= sk->sk_txtime_deadline_mode ?
1611 v.txtime.flags |= sk->sk_txtime_report_errors ?
1616 v.val = sk->sk_bound_dev_if;
1641 static inline void sock_lock_init(struct sock *sk)
1643 if (sk->sk_kern_sock)
1645 sk,
1646 af_family_kern_slock_key_strings[sk->sk_family],
1647 af_family_kern_slock_keys + sk->sk_family,
1648 af_family_kern_key_strings[sk->sk_family],
1649 af_family_kern_keys + sk->sk_family);
1652 sk,
1653 af_family_slock_key_strings[sk->sk_family],
1654 af_family_slock_keys + sk->sk_family,
1655 af_family_key_strings[sk->sk_family],
1656 af_family_keys + sk->sk_family);
1684 struct sock *sk;
1689 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1690 if (!sk)
1691 return sk;
1693 sk_prot_clear_nulls(sk, prot->obj_size);
1695 sk = kmalloc(prot->obj_size, priority);
1697 if (sk != NULL) {
1698 if (security_sk_alloc(sk, family, priority))
1703 sk_tx_queue_clear(sk);
1706 return sk;
1709 security_sk_free(sk);
1712 kmem_cache_free(slab, sk);
1714 kfree(sk);
1718 static void sk_prot_free(struct proto *prot, struct sock *sk)
1726 cgroup_sk_free(&sk->sk_cgrp_data);
1727 mem_cgroup_sk_free(sk);
1728 security_sk_free(sk);
1730 kmem_cache_free(slab, sk);
1732 kfree(sk);
1747 struct sock *sk;
1749 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1750 if (sk) {
1751 sk->sk_family = family;
1756 sk->sk_prot = sk->sk_prot_creator = prot;
1757 sk->sk_kern_sock = kern;
1758 sock_lock_init(sk);
1759 sk->sk_net_refcnt = kern ? 0 : 1;
1760 if (likely(sk->sk_net_refcnt)) {
1765 sock_net_set(sk, net);
1766 refcount_set(&sk->sk_wmem_alloc, 1);
1768 mem_cgroup_sk_alloc(sk);
1769 cgroup_sk_alloc(&sk->sk_cgrp_data);
1770 sock_update_classid(&sk->sk_cgrp_data);
1771 sock_update_netprioidx(&sk->sk_cgrp_data);
1772 sk_tx_queue_clear(sk);
1775 return sk;
1784 struct sock *sk = container_of(head, struct sock, sk_rcu);
1787 if (sk->sk_destruct)
1788 sk->sk_destruct(sk);
1790 filter = rcu_dereference_check(sk->sk_filter,
1791 refcount_read(&sk->sk_wmem_alloc) == 0);
1793 sk_filter_uncharge(sk, filter);
1794 RCU_INIT_POINTER(sk->sk_filter, NULL);
1797 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1800 bpf_sk_storage_free(sk);
1803 if (atomic_read(&sk->sk_omem_alloc))
1805 __func__, atomic_read(&sk->sk_omem_alloc));
1807 if (sk->sk_frag.page) {
1808 put_page(sk->sk_frag.page);
1809 sk->sk_frag.page = NULL;
1812 /* We do not need to acquire sk->sk_peer_lock, we are the last user. */
1813 put_cred(sk->sk_peer_cred);
1814 put_pid(sk->sk_peer_pid);
1816 if (likely(sk->sk_net_refcnt))
1817 put_net(sock_net(sk));
1818 sk_prot_free(sk->sk_prot_creator, sk);
1821 void sk_destruct(struct sock *sk)
1823 bool use_call_rcu = sock_flag(sk, SOCK_RCU_FREE);
1825 if (rcu_access_pointer(sk->sk_reuseport_cb)) {
1826 reuseport_detach_sock(sk);
1831 call_rcu(&sk->sk_rcu, __sk_destruct);
1833 __sk_destruct(&sk->sk_rcu);
1836 static void __sk_free(struct sock *sk)
1838 if (likely(sk->sk_net_refcnt))
1839 sock_inuse_add(sock_net(sk), -1);
1841 if (unlikely(sk->sk_net_refcnt && sock_diag_has_destroy_listeners(sk)))
1842 sock_diag_broadcast_destroy(sk);
1844 sk_destruct(sk);
1847 void sk_free(struct sock *sk)
1852 * If not null, sock_wfree() will call __sk_free(sk) later
1854 if (refcount_dec_and_test(&sk->sk_wmem_alloc))
1855 __sk_free(sk);
1859 static void sk_init_common(struct sock *sk)
1861 skb_queue_head_init(&sk->sk_receive_queue);
1862 skb_queue_head_init(&sk->sk_write_queue);
1863 skb_queue_head_init(&sk->sk_error_queue);
1865 rwlock_init(&sk->sk_callback_lock);
1866 lockdep_set_class_and_name(&sk->sk_receive_queue.lock,
1867 af_rlock_keys + sk->sk_family,
1868 af_family_rlock_key_strings[sk->sk_family]);
1869 lockdep_set_class_and_name(&sk->sk_write_queue.lock,
1870 af_wlock_keys + sk->sk_family,
1871 af_family_wlock_key_strings[sk->sk_family]);
1872 lockdep_set_class_and_name(&sk->sk_error_queue.lock,
1873 af_elock_keys + sk->sk_family,
1874 af_family_elock_key_strings[sk->sk_family]);
1875 lockdep_set_class_and_name(&sk->sk_callback_lock,
1876 af_callback_keys + sk->sk_family,
1877 af_family_clock_key_strings[sk->sk_family]);
1882 * @sk: the socket to clone
1887 struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
1889 struct proto *prot = READ_ONCE(sk->sk_prot);
1894 newsk = sk_prot_alloc(prot, priority, sk->sk_family);
1898 sock_copy(newsk, sk);
1927 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1932 /* sk->sk_memcg will be populated at accept() time */
1938 filter = rcu_dereference(sk->sk_filter);
1948 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk, sk))) {
1961 if (bpf_sk_storage_clone(sk, newsk)) {
1986 * is the same as sk->sk_prot->socks, as this field was copied
2002 if (sock_needs_netstamp(sk) && newsk->sk_flags & SK_FLAGS_TIMESTAMP)
2009 void sk_free_unlock_clone(struct sock *sk)
2013 sk->sk_destruct = NULL;
2014 bh_unlock_sock(sk);
2015 sk_free(sk);
2019 void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
2023 sk->sk_route_caps = dst->dev->features | sk->sk_route_forced_caps;
2024 if (sk->sk_route_caps & NETIF_F_GSO)
2025 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
2026 sk->sk_route_caps &= ~sk->sk_route_nocaps;
2027 if (sk_can_gso(sk)) {
2029 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
2031 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
2032 sk->sk_gso_max_size = dst->dev->gso_max_size;
2036 sk->sk_gso_max_segs = max_segs;
2037 sk_dst_set(sk, dst);
2051 struct sock *sk = skb->sk;
2054 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
2059 WARN_ON(refcount_sub_and_test(len - 1, &sk->sk_wmem_alloc));
2060 sk->sk_write_space(sk);
2067 if (refcount_sub_and_test(len, &sk->sk_wmem_alloc))
2068 __sk_free(sk);
2077 struct sock *sk = skb->sk;
2079 if (refcount_sub_and_test(skb->truesize, &sk->sk_wmem_alloc))
2080 __sk_free(sk);
2083 void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
2086 skb->sk = sk;
2088 if (unlikely(!sk_fullsock(sk))) {
2090 sock_hold(sk);
2095 skb_set_hash_from_sk(skb, sk);
2097 * We used to take a refcount on sk, but following operation
2101 refcount_add(skb->truesize, &sk->sk_wmem_alloc);
2121 * But we also want to keep skb->sk set because some packet schedulers
2129 if (can_skb_orphan_partial(skb) && skb_set_owner_sk_safe(skb, skb->sk))
2141 struct sock *sk = skb->sk;
2144 atomic_sub(len, &sk->sk_rmem_alloc);
2145 sk_mem_uncharge(sk, len);
2155 sock_put(skb->sk);
2165 if (sk_is_refcounted(skb->sk))
2166 sock_gen_put(skb->sk);
2171 kuid_t sock_i_uid(struct sock *sk)
2175 read_lock_bh(&sk->sk_callback_lock);
2176 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
2177 read_unlock_bh(&sk->sk_callback_lock);
2182 unsigned long __sock_i_ino(struct sock *sk)
2186 read_lock(&sk->sk_callback_lock);
2187 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
2188 read_unlock(&sk->sk_callback_lock);
2193 unsigned long sock_i_ino(struct sock *sk)
2198 ino = __sock_i_ino(sk);
2207 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
2211 refcount_read(&sk->sk_wmem_alloc) < READ_ONCE(sk->sk_sndbuf)) {
2215 skb_set_owner_w(skb, sk);
2225 struct sock *sk = skb->sk;
2227 atomic_sub(skb->truesize, &sk->sk_omem_alloc);
2230 struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size,
2236 if (atomic_read(&sk->sk_omem_alloc) + SKB_TRUESIZE(size) >
2244 atomic_add(skb->truesize, &sk->sk_omem_alloc);
2245 skb->sk = sk;
2253 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
2258 atomic_read(&sk->sk_omem_alloc) + size < optmem_max) {
2263 atomic_add(size, &sk->sk_omem_alloc);
2267 atomic_sub(size, &sk->sk_omem_alloc);
2277 static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
2286 atomic_sub(size, &sk->sk_omem_alloc);
2289 void sock_kfree_s(struct sock *sk, void *mem, int size)
2291 __sock_kfree_s(sk, mem, size, false);
2295 void sock_kzfree_s(struct sock *sk, void *mem, int size)
2297 __sock_kfree_s(sk, mem, size, true);
2304 static long sock_wait_for_wmem(struct sock *sk, long timeo)
2308 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2314 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
2315 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
2316 if (refcount_read(&sk->sk_wmem_alloc) < READ_ONCE(sk->sk_sndbuf))
2318 if (READ_ONCE(sk->sk_shutdown) & SEND_SHUTDOWN)
2320 if (READ_ONCE(sk->sk_err))
2324 finish_wait(sk_sleep(sk), &wait);
2333 struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
2341 timeo = sock_sndtimeo(sk, noblock);
2343 err = sock_error(sk);
2348 if (READ_ONCE(sk->sk_shutdown) & SEND_SHUTDOWN)
2351 if (sk_wmem_alloc_get(sk) < READ_ONCE(sk->sk_sndbuf))
2354 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2355 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
2361 timeo = sock_wait_for_wmem(sk, timeo);
2364 errcode, sk->sk_allocation);
2366 skb_set_owner_w(skb, sk);
2377 struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
2380 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
2384 int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
2391 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2410 if (!sock_flag(sk, SOCK_TXTIME))
2427 int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
2438 ret = __sock_cmsg_send(sk, msg, cmsg, sockc);
2446 static void sk_enter_memory_pressure(struct sock *sk)
2448 if (!sk->sk_prot->enter_memory_pressure)
2451 sk->sk_prot->enter_memory_pressure(sk);
2454 static void sk_leave_memory_pressure(struct sock *sk)
2456 if (sk->sk_prot->leave_memory_pressure) {
2457 sk->sk_prot->leave_memory_pressure(sk);
2459 unsigned long *memory_pressure = sk->sk_prot->memory_pressure;
2512 bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
2514 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
2517 sk_enter_memory_pressure(sk);
2518 sk_stream_moderate_sndbuf(sk);
2523 static void __lock_sock(struct sock *sk)
2524 __releases(&sk->sk_lock.slock)
2525 __acquires(&sk->sk_lock.slock)
2530 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
2532 spin_unlock_bh(&sk->sk_lock.slock);
2534 spin_lock_bh(&sk->sk_lock.slock);
2535 if (!sock_owned_by_user(sk))
2538 finish_wait(&sk->sk_lock.wq, &wait);
2541 void __release_sock(struct sock *sk)
2542 __releases(&sk->sk_lock.slock)
2543 __acquires(&sk->sk_lock.slock)
2547 while ((skb = sk->sk_backlog.head) != NULL) {
2548 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
2550 spin_unlock_bh(&sk->sk_lock.slock);
2557 sk_backlog_rcv(sk, skb);
2564 spin_lock_bh(&sk->sk_lock.slock);
2571 sk->sk_backlog.len = 0;
2574 void __sk_flush_backlog(struct sock *sk)
2576 spin_lock_bh(&sk->sk_lock.slock);
2577 __release_sock(sk);
2578 spin_unlock_bh(&sk->sk_lock.slock);
2583 * @sk: sock to wait on
2587 * Now socket state including sk->sk_err is changed only under lock,
2592 int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
2597 add_wait_queue(sk_sleep(sk), &wait);
2598 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
2599 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb, &wait);
2600 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
2601 remove_wait_queue(sk_sleep(sk), &wait);
2608 * @sk: socket
2615 int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind)
2617 struct proto *prot = sk->sk_prot;
2618 long allocated = sk_memory_allocated_add(sk, amt);
2621 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
2622 !(charged = mem_cgroup_charge_skmem(sk->sk_memcg, amt)))
2626 if (allocated <= sk_prot_mem_limits(sk, 0)) {
2627 sk_leave_memory_pressure(sk);
2632 if (allocated > sk_prot_mem_limits(sk, 1))
2633 sk_enter_memory_pressure(sk);
2636 if (allocated > sk_prot_mem_limits(sk, 2))
2641 if (atomic_read(&sk->sk_rmem_alloc) < sk_get_rmem0(sk, prot))
2645 int wmem0 = sk_get_wmem0(sk, prot);
2647 if (sk->sk_type == SOCK_STREAM) {
2648 if (sk->sk_wmem_queued < wmem0)
2650 } else if (refcount_read(&sk->sk_wmem_alloc) < wmem0) {
2655 if (sk_has_memory_pressure(sk)) {
2658 if (!sk_under_memory_pressure(sk))
2660 alloc = sk_sockets_allocated_read_positive(sk);
2661 if (sk_prot_mem_limits(sk, 2) > alloc *
2662 sk_mem_pages(sk->sk_wmem_queued +
2663 atomic_read(&sk->sk_rmem_alloc) +
2664 sk->sk_forward_alloc))
2670 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2671 sk_stream_moderate_sndbuf(sk);
2676 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2681 trace_sock_exceed_buf_limit(sk, prot, allocated, kind);
2683 sk_memory_allocated_sub(sk, amt);
2685 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2686 mem_cgroup_uncharge_skmem(sk->sk_memcg, amt);
2694 * @sk: socket
2702 int __sk_mem_schedule(struct sock *sk, int size, int kind)
2706 sk->sk_forward_alloc += amt << SK_MEM_QUANTUM_SHIFT;
2707 ret = __sk_mem_raise_allocated(sk, size, amt, kind);
2709 sk->sk_forward_alloc -= amt << SK_MEM_QUANTUM_SHIFT;
2716 * @sk: socket
2721 void __sk_mem_reduce_allocated(struct sock *sk, int amount)
2723 sk_memory_allocated_sub(sk, amount);
2725 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2726 mem_cgroup_uncharge_skmem(sk->sk_memcg, amount);
2728 if (sk_under_global_memory_pressure(sk) &&
2729 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2730 sk_leave_memory_pressure(sk);
2736 * @sk: socket
2739 void __sk_mem_reclaim(struct sock *sk, int amount)
2742 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
2743 __sk_mem_reduce_allocated(sk, amount);
2747 int sk_set_peek_off(struct sock *sk, int val)
2749 WRITE_ONCE(sk->sk_peek_off, val);
2818 int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *m, size_t len)
2854 sock_update_netprioidx(&sock->sk->sk_cgrp_data);
2855 sock_update_classid(&sock->sk->sk_cgrp_data);
2873 ssize_t sock_no_sendpage_locked(struct sock *sk, struct page *page,
2883 res = kernel_sendmsg_locked(sk, &msg, &iov, 1, size);
2893 static void sock_def_wakeup(struct sock *sk)
2898 wq = rcu_dereference(sk->sk_wq);
2904 static void sock_def_error_report(struct sock *sk)
2909 wq = rcu_dereference(sk->sk_wq);
2912 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
2916 void sock_def_readable(struct sock *sk)
2921 wq = rcu_dereference(sk->sk_wq);
2925 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
2929 static void sock_def_write_space(struct sock *sk)
2938 if ((refcount_read(&sk->sk_wmem_alloc) << 1) <= READ_ONCE(sk->sk_sndbuf)) {
2939 wq = rcu_dereference(sk->sk_wq);
2945 if (sock_writeable(sk))
2946 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
2952 static void sock_def_destruct(struct sock *sk)
2956 void sk_send_sigurg(struct sock *sk)
2958 if (sk->sk_socket && sk->sk_socket->file)
2959 if (send_sigurg(&sk->sk_socket->file->f_owner))
2960 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
2964 void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2968 sock_hold(sk);
2972 void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2975 __sock_put(sk);
2979 void sk_stop_timer_sync(struct sock *sk, struct timer_list *timer)
2982 __sock_put(sk);
2986 void sock_init_data_uid(struct socket *sock, struct sock *sk, kuid_t uid)
2988 sk_init_common(sk);
2989 sk->sk_send_head = NULL;
2991 timer_setup(&sk->sk_timer, NULL, 0);
2993 sk->sk_allocation = GFP_KERNEL;
2994 sk->sk_rcvbuf = READ_ONCE(sysctl_rmem_default);
2995 sk->sk_sndbuf = READ_ONCE(sysctl_wmem_default);
2996 sk->sk_state = TCP_CLOSE;
2997 sk_set_socket(sk, sock);
2999 sock_set_flag(sk, SOCK_ZAPPED);
3002 sk->sk_type = sock->type;
3003 RCU_INIT_POINTER(sk->sk_wq, &sock->wq);
3004 sock->sk = sk;
3006 RCU_INIT_POINTER(sk->sk_wq, NULL);
3008 sk->sk_uid = uid;
3010 rwlock_init(&sk->sk_callback_lock);
3011 if (sk->sk_kern_sock)
3013 &sk->sk_callback_lock,
3014 af_kern_callback_keys + sk->sk_family,
3015 af_family_kern_clock_key_strings[sk->sk_family]);
3018 &sk->sk_callback_lock,
3019 af_callback_keys + sk->sk_family,
3020 af_family_clock_key_strings[sk->sk_family]);
3022 sk->sk_state_change = sock_def_wakeup;
3023 sk->sk_data_ready = sock_def_readable;
3024 sk->sk_write_space = sock_def_write_space;
3025 sk->sk_error_report = sock_def_error_report;
3026 sk->sk_destruct = sock_def_destruct;
3028 sk->sk_frag.page = NULL;
3029 sk->sk_frag.offset = 0;
3030 sk->sk_peek_off = -1;
3032 sk->sk_peer_pid = NULL;
3033 sk->sk_peer_cred = NULL;
3034 spin_lock_init(&sk->sk_peer_lock);
3036 sk->sk_write_pending = 0;
3037 sk->sk_rcvlowat = 1;
3038 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
3039 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
3041 sk->sk_stamp = SK_DEFAULT_STAMP;
3043 seqlock_init(&sk->sk_stamp_seq);
3045 atomic_set(&sk->sk_zckey, 0);
3048 sk->sk_napi_id = 0;
3049 sk->sk_ll_usec = READ_ONCE(sysctl_net_busy_read);
3052 sk->sk_max_pacing_rate = ~0UL;
3053 sk->sk_pacing_rate = ~0UL;
3054 WRITE_ONCE(sk->sk_pacing_shift, 10);
3055 sk->sk_incoming_cpu = -1;
3057 sk_rx_queue_clear(sk);
3063 refcount_set(&sk->sk_refcnt, 1);
3064 atomic_set(&sk->sk_drops, 0);
3068 void sock_init_data(struct socket *sock, struct sock *sk)
3072 make_kuid(sock_net(sk)->user_ns, 0);
3074 sock_init_data_uid(sock, sk, uid);
3078 void lock_sock_nested(struct sock *sk, int subclass)
3081 spin_lock_bh(&sk->sk_lock.slock);
3082 if (sk->sk_lock.owned)
3083 __lock_sock(sk);
3084 sk->sk_lock.owned = 1;
3085 spin_unlock(&sk->sk_lock.slock);
3089 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
3094 void release_sock(struct sock *sk)
3096 spin_lock_bh(&sk->sk_lock.slock);
3097 if (sk->sk_backlog.tail)
3098 __release_sock(sk);
3100 /* Warning : release_cb() might need to release sk ownership,
3101 * ie call sock_release_ownership(sk) before us.
3103 if (sk->sk_prot->release_cb)
3104 sk->sk_prot->release_cb(sk);
3106 sock_release_ownership(sk);
3107 if (waitqueue_active(&sk->sk_lock.wq))
3108 wake_up(&sk->sk_lock.wq);
3109 spin_unlock_bh(&sk->sk_lock.slock);
3115 * @sk: socket
3126 bool lock_sock_fast(struct sock *sk)
3129 spin_lock_bh(&sk->sk_lock.slock);
3131 if (!sk->sk_lock.owned)
3137 __lock_sock(sk);
3138 sk->sk_lock.owned = 1;
3139 spin_unlock(&sk->sk_lock.slock);
3143 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
3152 struct sock *sk = sock->sk;
3155 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
3156 ts = ktime_to_timespec64(sock_read_timestamp(sk));
3161 sock_write_timestamp(sk, kt);
3188 void sock_enable_timestamp(struct sock *sk, enum sock_flags flag)
3190 if (!sock_flag(sk, flag)) {
3191 unsigned long previous_flags = sk->sk_flags;
3193 sock_set_flag(sk, flag);
3199 if (sock_needs_netstamp(sk) &&
3205 int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
3213 skb = sock_dequeue_err_skb(sk);
3226 sock_recv_timestamp(msg, sk, skb);
3251 struct sock *sk = sock->sk;
3253 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
3254 return READ_ONCE(sk->sk_prot)->getsockopt(sk, level, optname, optval, optlen);
3261 struct sock *sk = sock->sk;
3265 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
3279 struct sock *sk = sock->sk;
3281 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
3282 return READ_ONCE(sk->sk_prot)->setsockopt(sk, level, optname, optval, optlen);
3286 void sk_common_release(struct sock *sk)
3288 if (sk->sk_prot->destroy)
3289 sk->sk_prot->destroy(sk);
3299 sk->sk_prot->unhash(sk);
3313 sock_orphan(sk);
3315 xfrm_sk_free_policy(sk);
3317 sk_refcnt_debug_release(sk);
3319 sock_put(sk);
3323 void sk_get_meminfo(const struct sock *sk, u32 *mem)
3327 mem[SK_MEMINFO_RMEM_ALLOC] = sk_rmem_alloc_get(sk);
3328 mem[SK_MEMINFO_RCVBUF] = READ_ONCE(sk->sk_rcvbuf);
3329 mem[SK_MEMINFO_WMEM_ALLOC] = sk_wmem_alloc_get(sk);
3330 mem[SK_MEMINFO_SNDBUF] = READ_ONCE(sk->sk_sndbuf);
3331 mem[SK_MEMINFO_FWD_ALLOC] = sk->sk_forward_alloc;
3332 mem[SK_MEMINFO_WMEM_QUEUED] = READ_ONCE(sk->sk_wmem_queued);
3333 mem[SK_MEMINFO_OPTMEM] = atomic_read(&sk->sk_omem_alloc);
3334 mem[SK_MEMINFO_BACKLOG] = READ_ONCE(sk->sk_backlog.len);
3335 mem[SK_MEMINFO_DROPS] = atomic_read(&sk->sk_drops);
3724 struct sock *sk = p;
3726 return !skb_queue_empty_lockless(&sk->sk_receive_queue) ||
3727 sk_busy_loop_timeout(sk, start_time);
3732 int sock_bind_add(struct sock *sk, struct sockaddr *addr, int addr_len)
3734 if (!sk->sk_prot->bind_add)
3736 return sk->sk_prot->bind_add(sk, addr, addr_len);