Lines Matching +full:non +full:- +full:urgent
1 // SPDX-License-Identifier: GPL-2.0-or-later
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Florian La Roche, <flla@stud.uni-sb.de>
25 * sk->inuse=1 and was trying to connect
51 * Alan Cox : sk->keepopen now seems to work
53 * Alan Cox : Fixed assorted sk->rqueue->next errors
63 * Alan Cox : FIN with no memory -> CRASH
76 * window non shrink trick.
113 * Alan Cox : Changed the semantics of sk->socket to
164 * Matt Day : poll()->select() match BSD precisely on error
166 * Marc Tamsky : Various sk->prot->retransmits and
167 * sk->retransmits misupdating fixed.
172 * Alan Cox : Look up device on a retransmit - routes may
214 * waiting for final ack in three-way handshake.
322 * Technical note: it is used by multiple contexts non atomically.
338 val--; in tcp_enter_memory_pressure()
353 jiffies_to_msecs(jiffies - val)); in tcp_leave_memory_pressure()
384 while (--retrans) { in retrans_to_secs()
396 u32 rate = READ_ONCE(tp->rate_delivered); in tcp_compute_delivery_rate()
397 u32 intv = READ_ONCE(tp->rate_interval_us); in tcp_compute_delivery_rate()
401 rate64 = (u64)rate * tp->mss_cache * USEC_PER_SEC; in tcp_compute_delivery_rate()
407 /* Address-family independent initialization for a tcp_sock.
417 tp->out_of_order_queue = RB_ROOT; in tcp_init_sock()
418 sk->tcp_rtx_queue = RB_ROOT; in tcp_init_sock()
420 INIT_LIST_HEAD(&tp->tsq_node); in tcp_init_sock()
421 INIT_LIST_HEAD(&tp->tsorted_sent_queue); in tcp_init_sock()
423 icsk->icsk_rto = TCP_TIMEOUT_INIT; in tcp_init_sock()
424 icsk->icsk_rto_min = TCP_RTO_MIN; in tcp_init_sock()
425 icsk->icsk_delack_max = TCP_DELACK_MAX; in tcp_init_sock()
426 tp->mdev_us = jiffies_to_usecs(TCP_TIMEOUT_INIT); in tcp_init_sock()
427 minmax_reset(&tp->rtt_min, tcp_jiffies32, ~0U); in tcp_init_sock()
430 * initial SYN frame in their delayed-ACK and congestion control in tcp_init_sock()
432 * efficiently to them. -DaveM in tcp_init_sock()
437 tp->app_limited = ~0U; in tcp_init_sock()
438 tp->rate_app_limited = 1; in tcp_init_sock()
440 /* See draft-stevens-tcpca-spec-01 for discussion of the in tcp_init_sock()
443 tp->snd_ssthresh = TCP_INFINITE_SSTHRESH; in tcp_init_sock()
444 tp->snd_cwnd_clamp = ~0; in tcp_init_sock()
445 tp->mss_cache = TCP_MSS_DEFAULT; in tcp_init_sock()
447 tp->reordering = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_reordering); in tcp_init_sock()
450 tp->tsoffset = 0; in tcp_init_sock()
451 tp->rack.reo_wnd_steps = 1; in tcp_init_sock()
453 sk->sk_write_space = sk_stream_write_space; in tcp_init_sock()
456 icsk->icsk_sync_mss = tcp_sync_mss; in tcp_init_sock()
458 WRITE_ONCE(sk->sk_sndbuf, READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[1])); in tcp_init_sock()
459 WRITE_ONCE(sk->sk_rcvbuf, READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_rmem[1])); in tcp_init_sock()
462 set_bit(SOCK_SUPPORT_ZC, &sk->sk_socket->flags); in tcp_init_sock()
475 sock_tx_timestamp(sk, tsflags, &shinfo->tx_flags); in tcp_tx_timestamp()
477 tcb->txstamp_ack = 1; in tcp_tx_timestamp()
479 shinfo->tskey = TCP_SKB_CB(skb)->seq + skb->len - 1; in tcp_tx_timestamp()
500 struct sock *sk = sock->sk; in tcp_poll()
529 * Check-me. in tcp_poll()
534 * in state CLOSE_WAIT. One solution is evident --- to set EPOLLHUP in tcp_poll()
540 * why EPOLLHUP is incompatible with EPOLLOUT. --ANK in tcp_poll()
543 * blocking on fresh not-connected or disconnected socket. --ANK in tcp_poll()
545 shutdown = READ_ONCE(sk->sk_shutdown); in tcp_poll()
553 (state != TCP_SYN_RECV || rcu_access_pointer(tp->fastopen_rsk))) { in tcp_poll()
555 u16 urg_data = READ_ONCE(tp->urg_data); in tcp_poll()
558 READ_ONCE(tp->urg_seq) == READ_ONCE(tp->copied_seq) && in tcp_poll()
570 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags); in tcp_poll()
596 if (READ_ONCE(sk->sk_err) || in tcp_poll()
597 !skb_queue_empty_lockless(&sk->sk_error_queue)) in tcp_poll()
612 if (sk->sk_state == TCP_LISTEN) in tcp_ioctl()
613 return -EINVAL; in tcp_ioctl()
620 answ = READ_ONCE(tp->urg_data) && in tcp_ioctl()
621 READ_ONCE(tp->urg_seq) == READ_ONCE(tp->copied_seq); in tcp_ioctl()
624 if (sk->sk_state == TCP_LISTEN) in tcp_ioctl()
625 return -EINVAL; in tcp_ioctl()
627 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) in tcp_ioctl()
630 answ = READ_ONCE(tp->write_seq) - tp->snd_una; in tcp_ioctl()
633 if (sk->sk_state == TCP_LISTEN) in tcp_ioctl()
634 return -EINVAL; in tcp_ioctl()
636 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) in tcp_ioctl()
639 answ = READ_ONCE(tp->write_seq) - in tcp_ioctl()
640 READ_ONCE(tp->snd_nxt); in tcp_ioctl()
643 return -ENOIOCTLCMD; in tcp_ioctl()
653 TCP_SKB_CB(skb)->tcp_flags |= TCPHDR_PSH; in tcp_mark_push()
654 tp->pushed_seq = tp->write_seq; in tcp_mark_push()
659 return after(tp->write_seq, tp->pushed_seq + (tp->max_window >> 1)); in forced_push()
667 tcb->seq = tcb->end_seq = tp->write_seq; in tcp_skb_entail()
668 tcb->tcp_flags = TCPHDR_ACK; in tcp_skb_entail()
671 sk_wmem_queued_add(sk, skb->truesize); in tcp_skb_entail()
672 sk_mem_charge(sk, skb->truesize); in tcp_skb_entail()
673 if (tp->nonagle & TCP_NAGLE_PUSH) in tcp_skb_entail()
674 tp->nonagle &= ~TCP_NAGLE_PUSH; in tcp_skb_entail()
682 tp->snd_up = tp->write_seq; in tcp_mark_urg()
698 return skb->len < size_goal && in tcp_should_autocork()
699 READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_autocorking) && in tcp_should_autocork()
701 refcount_read(&sk->sk_wmem_alloc) > skb->truesize && in tcp_should_autocork()
722 if (!test_bit(TSQ_THROTTLED, &sk->sk_tsq_flags)) { in tcp_push()
724 set_bit(TSQ_THROTTLED, &sk->sk_tsq_flags); in tcp_push()
730 if (refcount_read(&sk->sk_wmem_alloc) > skb->truesize) in tcp_push()
743 struct tcp_splice_state *tss = rd_desc->arg.data; in tcp_splice_data_recv()
746 ret = skb_splice_bits(skb, skb->sk, offset, tss->pipe, in tcp_splice_data_recv()
747 min(rd_desc->count, len), tss->flags); in tcp_splice_data_recv()
749 rd_desc->count -= ret; in tcp_splice_data_recv()
758 .count = tss->len, in __tcp_splice_read()
765 * tcp_splice_read - splice data from TCP socket to a pipe
780 struct sock *sk = sock->sk; in tcp_splice_read()
795 return -ESPIPE; in tcp_splice_read()
801 timeo = sock_rcvtimeo(sk, sock->file->f_flags & O_NONBLOCK); in tcp_splice_read()
811 if (sk->sk_err) { in tcp_splice_read()
815 if (sk->sk_shutdown & RCV_SHUTDOWN) in tcp_splice_read()
817 if (sk->sk_state == TCP_CLOSE) { in tcp_splice_read()
822 ret = -ENOTCONN; in tcp_splice_read()
826 ret = -EAGAIN; in tcp_splice_read()
833 if (!skb_queue_empty(&sk->sk_receive_queue)) in tcp_splice_read()
844 tss.len -= ret; in tcp_splice_read()
852 if (sk->sk_err || sk->sk_state == TCP_CLOSE || in tcp_splice_read()
853 (sk->sk_shutdown & RCV_SHUTDOWN) || in tcp_splice_read()
876 skb->truesize = SKB_TRUESIZE(skb_end_offset(skb)); in tcp_stream_alloc_skb()
879 sk_forced_mem_schedule(sk, skb->truesize); in tcp_stream_alloc_skb()
881 mem_scheduled = sk_wmem_schedule(sk, skb->truesize); in tcp_stream_alloc_skb()
885 skb->ip_summed = CHECKSUM_PARTIAL; in tcp_stream_alloc_skb()
886 INIT_LIST_HEAD(&skb->tcp_tsorted_anchor); in tcp_stream_alloc_skb()
891 sk->sk_prot->enter_memory_pressure(sk); in tcp_stream_alloc_skb()
907 new_size_goal = tcp_bound_to_half_wnd(tp, sk->sk_gso_max_size); in tcp_xmit_size_goal()
910 size_goal = tp->gso_segs * mss_now; in tcp_xmit_size_goal()
913 tp->gso_segs = min_t(u16, new_size_goal / mss_now, in tcp_xmit_size_goal()
914 sk->sk_gso_max_segs); in tcp_xmit_size_goal()
915 size_goal = tp->gso_segs * mss_now; in tcp_xmit_size_goal()
941 if (skb && TCP_SKB_CB(skb)->seq == TCP_SKB_CB(skb)->end_seq) { in tcp_remove_empty_skb()
953 u32 extra = skb->truesize - in tcp_downgrade_zcopy_pure()
957 return -ENOMEM; in tcp_downgrade_zcopy_pure()
960 skb_shinfo(skb)->flags &= ~SKBFL_PURE_ZEROCOPY; in tcp_downgrade_zcopy_pure()
974 * Use whatever is left in sk->sk_forward_alloc and tcp_wmem[0] in tcp_wmem_schedule()
977 left = sock_net(sk)->ipv4.sysctl_tcp_wmem[0] - sk->sk_wmem_queued; in tcp_wmem_schedule()
980 return min(copy, sk->sk_forward_alloc); in tcp_wmem_schedule()
985 if (tp->fastopen_req) { in tcp_free_fastopen_req()
986 kfree(tp->fastopen_req); in tcp_free_fastopen_req()
987 tp->fastopen_req = NULL; in tcp_free_fastopen_req()
996 struct sockaddr *uaddr = msg->msg_name; in tcp_sendmsg_fastopen()
999 if (!(READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_fastopen) & in tcp_sendmsg_fastopen()
1001 (uaddr && msg->msg_namelen >= sizeof(uaddr->sa_family) && in tcp_sendmsg_fastopen()
1002 uaddr->sa_family == AF_UNSPEC)) in tcp_sendmsg_fastopen()
1003 return -EOPNOTSUPP; in tcp_sendmsg_fastopen()
1004 if (tp->fastopen_req) in tcp_sendmsg_fastopen()
1005 return -EALREADY; /* Another Fast Open is in progress */ in tcp_sendmsg_fastopen()
1007 tp->fastopen_req = kzalloc(sizeof(struct tcp_fastopen_request), in tcp_sendmsg_fastopen()
1008 sk->sk_allocation); in tcp_sendmsg_fastopen()
1009 if (unlikely(!tp->fastopen_req)) in tcp_sendmsg_fastopen()
1010 return -ENOBUFS; in tcp_sendmsg_fastopen()
1011 tp->fastopen_req->data = msg; in tcp_sendmsg_fastopen()
1012 tp->fastopen_req->size = size; in tcp_sendmsg_fastopen()
1013 tp->fastopen_req->uarg = uarg; in tcp_sendmsg_fastopen()
1020 inet->inet_dport = 0; in tcp_sendmsg_fastopen()
1021 sk->sk_route_caps = 0; in tcp_sendmsg_fastopen()
1024 flags = (msg->msg_flags & MSG_DONTWAIT) ? O_NONBLOCK : 0; in tcp_sendmsg_fastopen()
1025 err = __inet_stream_connect(sk->sk_socket, uaddr, in tcp_sendmsg_fastopen()
1026 msg->msg_namelen, flags, 1); in tcp_sendmsg_fastopen()
1030 if (tp->fastopen_req) { in tcp_sendmsg_fastopen()
1031 *copied = tp->fastopen_req->copied; in tcp_sendmsg_fastopen()
1050 flags = msg->msg_flags; in tcp_sendmsg_locked()
1053 if (msg->msg_ubuf) { in tcp_sendmsg_locked()
1054 uarg = msg->msg_ubuf; in tcp_sendmsg_locked()
1055 if (sk->sk_route_caps & NETIF_F_SG) in tcp_sendmsg_locked()
1061 err = -ENOBUFS; in tcp_sendmsg_locked()
1064 if (sk->sk_route_caps & NETIF_F_SG) in tcp_sendmsg_locked()
1067 uarg_to_msgzc(uarg)->zerocopy = 0; in tcp_sendmsg_locked()
1069 } else if (unlikely(msg->msg_flags & MSG_SPLICE_PAGES) && size) { in tcp_sendmsg_locked()
1070 if (sk->sk_route_caps & NETIF_F_SG) in tcp_sendmsg_locked()
1076 !tp->repair) { in tcp_sendmsg_locked()
1078 if (err == -EINPROGRESS && copied_syn > 0) in tcp_sendmsg_locked()
1086 tcp_rate_check_app_limited(sk); /* is sending application-limited? */ in tcp_sendmsg_locked()
1092 if (((1 << sk->sk_state) & ~(TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)) && in tcp_sendmsg_locked()
1099 if (unlikely(tp->repair)) { in tcp_sendmsg_locked()
1100 if (tp->repair_queue == TCP_RECV_QUEUE) { in tcp_sendmsg_locked()
1105 err = -EINVAL; in tcp_sendmsg_locked()
1106 if (tp->repair_queue == TCP_NO_QUEUE) in tcp_sendmsg_locked()
1113 if (msg->msg_controllen) { in tcp_sendmsg_locked()
1116 err = -EINVAL; in tcp_sendmsg_locked()
1130 err = -EPIPE; in tcp_sendmsg_locked()
1131 if (sk->sk_err || (sk->sk_shutdown & SEND_SHUTDOWN)) in tcp_sendmsg_locked()
1139 copy = size_goal - skb->len; in tcp_sendmsg_locked()
1154 skb = tcp_stream_alloc_skb(sk, sk->sk_allocation, in tcp_sendmsg_locked()
1168 if (tp->repair) in tcp_sendmsg_locked()
1169 TCP_SKB_CB(skb)->sacked |= TCPCB_REPAIRED; in tcp_sendmsg_locked()
1178 int i = skb_shinfo(skb)->nr_frags; in tcp_sendmsg_locked()
1184 if (!skb_can_coalesce(skb, i, pfrag->page, in tcp_sendmsg_locked()
1185 pfrag->offset)) { in tcp_sendmsg_locked()
1193 copy = min_t(int, copy, pfrag->size - pfrag->offset); in tcp_sendmsg_locked()
1205 err = skb_copy_to_page_nocache(sk, &msg->msg_iter, skb, in tcp_sendmsg_locked()
1206 pfrag->page, in tcp_sendmsg_locked()
1207 pfrag->offset, in tcp_sendmsg_locked()
1214 skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy); in tcp_sendmsg_locked()
1216 skb_fill_page_desc(skb, i, pfrag->page, in tcp_sendmsg_locked()
1217 pfrag->offset, copy); in tcp_sendmsg_locked()
1218 page_ref_inc(pfrag->page); in tcp_sendmsg_locked()
1220 pfrag->offset += copy; in tcp_sendmsg_locked()
1225 if (!skb->len) in tcp_sendmsg_locked()
1226 skb_shinfo(skb)->flags |= SKBFL_PURE_ZEROCOPY; in tcp_sendmsg_locked()
1235 if (err == -EMSGSIZE || err == -EEXIST) { in tcp_sendmsg_locked()
1250 err = skb_splice_from_iter(skb, &msg->msg_iter, copy, in tcp_sendmsg_locked()
1251 sk->sk_allocation); in tcp_sendmsg_locked()
1253 if (err == -EMSGSIZE) { in tcp_sendmsg_locked()
1262 skb_shinfo(skb)->flags |= SKBFL_SHARED_FRAG; in tcp_sendmsg_locked()
1269 TCP_SKB_CB(skb)->tcp_flags &= ~TCPHDR_PSH; in tcp_sendmsg_locked()
1271 WRITE_ONCE(tp->write_seq, tp->write_seq + copy); in tcp_sendmsg_locked()
1272 TCP_SKB_CB(skb)->end_seq += copy; in tcp_sendmsg_locked()
1278 TCP_SKB_CB(skb)->eor = 1; in tcp_sendmsg_locked()
1282 if (skb->len < size_goal || (flags & MSG_OOB) || unlikely(tp->repair)) in tcp_sendmsg_locked()
1293 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags); in tcp_sendmsg_locked()
1309 tcp_push(sk, flags, mss_now, tp->nonagle, size_goal); in tcp_sendmsg_locked()
1312 /* msg->msg_ubuf is pinned by the caller so we don't take extra refs */ in tcp_sendmsg_locked()
1313 if (uarg && !msg->msg_ubuf) in tcp_sendmsg_locked()
1323 /* msg->msg_ubuf is pinned by the caller so we don't take extra refs */ in tcp_sendmsg_locked()
1324 if (uarg && !msg->msg_ubuf) in tcp_sendmsg_locked()
1328 if (unlikely(tcp_rtx_and_write_queues_empty(sk) && err == -EAGAIN)) { in tcp_sendmsg_locked()
1329 sk->sk_write_space(sk); in tcp_sendmsg_locked()
1350 struct sock *sk = sock->sk; in tcp_splice_eof()
1359 tcp_push(sk, 0, mss_now, tp->nonagle, size_goal); in tcp_splice_eof()
1365 * Handle reading urgent data. BSD has very simple semantics for
1374 if (sock_flag(sk, SOCK_URGINLINE) || !tp->urg_data || in tcp_recv_urg()
1375 tp->urg_data == TCP_URG_READ) in tcp_recv_urg()
1376 return -EINVAL; /* Yes this is right ! */ in tcp_recv_urg()
1378 if (sk->sk_state == TCP_CLOSE && !sock_flag(sk, SOCK_DONE)) in tcp_recv_urg()
1379 return -ENOTCONN; in tcp_recv_urg()
1381 if (tp->urg_data & TCP_URG_VALID) { in tcp_recv_urg()
1383 char c = tp->urg_data; in tcp_recv_urg()
1386 WRITE_ONCE(tp->urg_data, TCP_URG_READ); in tcp_recv_urg()
1388 /* Read urgent data. */ in tcp_recv_urg()
1389 msg->msg_flags |= MSG_OOB; in tcp_recv_urg()
1396 msg->msg_flags |= MSG_TRUNC; in tcp_recv_urg()
1398 return err ? -EFAULT : len; in tcp_recv_urg()
1401 if (sk->sk_state == TCP_CLOSE || (sk->sk_shutdown & RCV_SHUTDOWN)) in tcp_recv_urg()
1408 * Mike <pall@rz.uni-karlsruhe.de> in tcp_recv_urg()
1410 return -EAGAIN; in tcp_recv_urg()
1418 /* XXX -- need to support SO_PEEK_OFF */ in tcp_peek_sndq()
1420 skb_rbtree_walk(skb, &sk->tcp_rtx_queue) { in tcp_peek_sndq()
1421 err = skb_copy_datagram_msg(skb, 0, msg, skb->len); in tcp_peek_sndq()
1424 copied += skb->len; in tcp_peek_sndq()
1427 skb_queue_walk(&sk->sk_write_queue, skb) { in tcp_peek_sndq()
1428 err = skb_copy_datagram_msg(skb, 0, msg, skb->len); in tcp_peek_sndq()
1432 copied += skb->len; in tcp_peek_sndq()
1452 if (/* Once-per-two-segments ACK was not sent by tcp_input.c */ in __tcp_cleanup_rbuf()
1453 tp->rcv_nxt - tp->rcv_wup > icsk->icsk_ack.rcv_mss || in __tcp_cleanup_rbuf()
1461 ((icsk->icsk_ack.pending & ICSK_ACK_PUSHED2) || in __tcp_cleanup_rbuf()
1462 ((icsk->icsk_ack.pending & ICSK_ACK_PUSHED) && in __tcp_cleanup_rbuf()
1464 !atomic_read(&sk->sk_rmem_alloc))) in __tcp_cleanup_rbuf()
1468 /* We send an ACK if we can now advertise a non-zero window in __tcp_cleanup_rbuf()
1474 if (copied > 0 && !time_to_ack && !(sk->sk_shutdown & RCV_SHUTDOWN)) { in __tcp_cleanup_rbuf()
1478 if (2*rcv_window_now <= tp->window_clamp) { in __tcp_cleanup_rbuf()
1496 struct sk_buff *skb = skb_peek(&sk->sk_receive_queue); in tcp_cleanup_rbuf()
1499 WARN(skb && !before(tp->copied_seq, TCP_SKB_CB(skb)->end_seq), in tcp_cleanup_rbuf()
1501 tp->copied_seq, TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt); in tcp_cleanup_rbuf()
1507 __skb_unlink(skb, &sk->sk_receive_queue); in tcp_eat_recv_skb()
1508 if (likely(skb->destructor == sock_rfree)) { in tcp_eat_recv_skb()
1510 skb->destructor = NULL; in tcp_eat_recv_skb()
1511 skb->sk = NULL; in tcp_eat_recv_skb()
1522 while ((skb = skb_peek(&sk->sk_receive_queue)) != NULL) { in tcp_recv_skb()
1523 offset = seq - TCP_SKB_CB(skb)->seq; in tcp_recv_skb()
1524 if (unlikely(TCP_SKB_CB(skb)->tcp_flags & TCPHDR_SYN)) { in tcp_recv_skb()
1526 offset--; in tcp_recv_skb()
1528 if (offset < skb->len || (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN)) { in tcp_recv_skb()
1547 * - It is assumed that the socket was locked by the caller.
1548 * - The routine does not block.
1549 * - At present, there is no support for reading OOB data
1558 u32 seq = tp->copied_seq; in tcp_read_sock()
1562 if (sk->sk_state == TCP_LISTEN) in tcp_read_sock()
1563 return -ENOTCONN; in tcp_read_sock()
1565 if (offset < skb->len) { in tcp_read_sock()
1569 len = skb->len - offset; in tcp_read_sock()
1570 /* Stop reading if we hit a patch of urgent data */ in tcp_read_sock()
1571 if (unlikely(tp->urg_data)) { in tcp_read_sock()
1572 u32 urg_offset = tp->urg_seq - seq; in tcp_read_sock()
1595 skb = tcp_recv_skb(sk, seq - 1, &offset); in tcp_read_sock()
1601 if (offset + 1 != skb->len) in tcp_read_sock()
1604 if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN) { in tcp_read_sock()
1610 if (!desc->count) in tcp_read_sock()
1612 WRITE_ONCE(tp->copied_seq, seq); in tcp_read_sock()
1614 WRITE_ONCE(tp->copied_seq, seq); in tcp_read_sock()
1632 if (sk->sk_state == TCP_LISTEN) in tcp_read_skb()
1633 return -ENOTCONN; in tcp_read_skb()
1635 while ((skb = skb_peek(&sk->sk_receive_queue)) != NULL) { in tcp_read_skb()
1639 __skb_unlink(skb, &sk->sk_receive_queue); in tcp_read_skb()
1641 tcp_flags = TCP_SKB_CB(skb)->tcp_flags; in tcp_read_skb()
1660 u32 seq = tp->copied_seq; in tcp_read_done()
1665 if (sk->sk_state == TCP_LISTEN) in tcp_read_done()
1672 used = min_t(size_t, skb->len - offset, left); in tcp_read_done()
1674 left -= used; in tcp_read_done()
1676 if (skb->len > offset + used) in tcp_read_done()
1679 if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN) { in tcp_read_done()
1686 WRITE_ONCE(tp->copied_seq, seq); in tcp_read_done()
1692 tcp_cleanup_rbuf(sk, len - left); in tcp_read_done()
1698 return tcp_inq(sock->sk); in tcp_peek_len()
1707 if (sk->sk_userlocks & SOCK_RCVBUF_LOCK) in tcp_set_rcvlowat()
1708 cap = sk->sk_rcvbuf >> 1; in tcp_set_rcvlowat()
1710 cap = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_rmem[2]) >> 1; in tcp_set_rcvlowat()
1712 WRITE_ONCE(sk->sk_rcvlowat, val ? : 1); in tcp_set_rcvlowat()
1717 if (sk->sk_userlocks & SOCK_RCVBUF_LOCK) in tcp_set_rcvlowat()
1721 if (space > sk->sk_rcvbuf) { in tcp_set_rcvlowat()
1722 WRITE_ONCE(sk->sk_rcvbuf, space); in tcp_set_rcvlowat()
1723 tcp_sk(sk)->window_clamp = val; in tcp_set_rcvlowat()
1732 if (skb->tstamp) in tcp_update_recv_tstamps()
1733 tss->ts[0] = ktime_to_timespec64(skb->tstamp); in tcp_update_recv_tstamps()
1735 tss->ts[0] = (struct timespec64) {0}; in tcp_update_recv_tstamps()
1737 if (skb_hwtstamps(skb)->hwtstamp) in tcp_update_recv_tstamps()
1738 tss->ts[2] = ktime_to_timespec64(skb_hwtstamps(skb)->hwtstamp); in tcp_update_recv_tstamps()
1740 tss->ts[2] = (struct timespec64) {0}; in tcp_update_recv_tstamps()
1750 if (vma->vm_flags & (VM_WRITE | VM_EXEC)) in tcp_mmap()
1751 return -EPERM; in tcp_mmap()
1757 vma->vm_ops = &tcp_vm_ops; in tcp_mmap()
1767 if (unlikely(offset_skb >= skb->len)) in skb_advance_to_frag()
1770 offset_skb -= skb_headlen(skb); in skb_advance_to_frag()
1774 frag = skb_shinfo(skb)->frags; in skb_advance_to_frag()
1780 offset_skb -= skb_frag_size(frag); in skb_advance_to_frag()
1796 if (PageCompound(page) || page->mapping) in can_map_frag()
1826 zc->recv_skip_hint = skb->len - offset; in tcp_zerocopy_set_hint_for_skb()
1837 if (frag == &info->frags[info->nr_frags - 1]) in tcp_zerocopy_set_hint_for_skb()
1841 partial_frag_remainder = skb_frag_size(frag) - frag_offset; in tcp_zerocopy_set_hint_for_skb()
1842 zc->recv_skip_hint -= partial_frag_remainder; in tcp_zerocopy_set_hint_for_skb()
1850 mappable_offset = find_next_mappable_frag(frag, zc->recv_skip_hint); in tcp_zerocopy_set_hint_for_skb()
1851 zc->recv_skip_hint = mappable_offset + partial_frag_remainder; in tcp_zerocopy_set_hint_for_skb()
1861 unsigned long copy_address = (unsigned long)zc->copybuf_address; in receive_fallback_to_copy()
1865 zc->length = 0; in receive_fallback_to_copy()
1866 zc->recv_skip_hint = 0; in receive_fallback_to_copy()
1868 if (copy_address != zc->copybuf_address) in receive_fallback_to_copy()
1869 return -EINVAL; in receive_fallback_to_copy()
1877 tss, &zc->msg_flags); in receive_fallback_to_copy()
1881 zc->copybuf_len = err; in receive_fallback_to_copy()
1882 if (likely(zc->copybuf_len)) { in receive_fallback_to_copy()
1886 skb = tcp_recv_skb(sk, tcp_sk(sk)->copied_seq, &offset); in receive_fallback_to_copy()
1897 unsigned long copy_address = (unsigned long)zc->copybuf_address; in tcp_copy_straggler_data()
1901 if (copy_address != zc->copybuf_address) in tcp_copy_straggler_data()
1902 return -EINVAL; in tcp_copy_straggler_data()
1911 zc->recv_skip_hint -= copylen; in tcp_copy_straggler_data()
1924 u32 offset, copylen = min_t(u32, copybuf_len, zc->recv_skip_hint); in tcp_zc_handle_leftover()
1930 offset = *seq - TCP_SKB_CB(skb)->seq; in tcp_zc_handle_leftover()
1933 if (TCP_SKB_CB(skb)->has_rxtstamp) { in tcp_zc_handle_leftover()
1935 zc->msg_flags |= TCP_CMSG_TS; in tcp_zc_handle_leftover()
1939 zc->copybuf_len = tcp_copy_straggler_data(zc, skb, copylen, &offset, in tcp_zc_handle_leftover()
1941 return zc->copybuf_len < 0 ? 0 : copylen; in tcp_zc_handle_leftover()
1955 if (err == -EBUSY && in tcp_zerocopy_vm_insert_batch_error()
1956 zc->flags & TCP_RECEIVE_ZEROCOPY_FLAG_TLB_CLEAN_HINT) { in tcp_zerocopy_vm_insert_batch_error()
1959 maybe_zap_len = total_bytes_to_map - /* All bytes to map */ in tcp_zerocopy_vm_insert_batch_error()
1974 bytes_mapped = PAGE_SIZE * (leftover_pages - pages_remaining); in tcp_zerocopy_vm_insert_batch_error()
1986 *length -= bytes_not_mapped; in tcp_zerocopy_vm_insert_batch_error()
1987 zc->recv_skip_hint += bytes_not_mapped; in tcp_zerocopy_vm_insert_batch_error()
2007 pages_mapped = pages_to_map - (unsigned int)pages_remaining; in tcp_zerocopy_vm_insert_batch()
2032 msg_control_addr = (unsigned long)zc->msg_control; in tcp_zc_finalize_rx_tstamp()
2035 (__kernel_size_t)zc->msg_controllen; in tcp_zc_finalize_rx_tstamp()
2039 zc->msg_flags = 0; in tcp_zc_finalize_rx_tstamp()
2040 if (zc->msg_control == msg_control_addr && in tcp_zc_finalize_rx_tstamp()
2041 zc->msg_controllen == cmsg_dummy.msg_controllen) { in tcp_zc_finalize_rx_tstamp()
2043 zc->msg_control = (__u64) in tcp_zc_finalize_rx_tstamp()
2045 zc->msg_controllen = in tcp_zc_finalize_rx_tstamp()
2047 zc->msg_flags = (__u32)cmsg_dummy.msg_flags; in tcp_zc_finalize_rx_tstamp()
2058 if (vma->vm_ops != &tcp_vm_ops) { in find_tcp_vma()
2068 if (!vma || vma->vm_ops != &tcp_vm_ops) { in find_tcp_vma()
2082 unsigned long address = (unsigned long)zc->address; in tcp_zerocopy_receive()
2084 s32 copybuf_len = zc->copybuf_len; in tcp_zerocopy_receive()
2090 u32 seq = tp->copied_seq; in tcp_zerocopy_receive()
2096 zc->copybuf_len = 0; in tcp_zerocopy_receive()
2097 zc->msg_flags = 0; in tcp_zerocopy_receive()
2099 if (address & (PAGE_SIZE - 1) || address != zc->address) in tcp_zerocopy_receive()
2100 return -EINVAL; in tcp_zerocopy_receive()
2102 if (sk->sk_state == TCP_LISTEN) in tcp_zerocopy_receive()
2103 return -ENOTCONN; in tcp_zerocopy_receive()
2111 zc->length = 0; in tcp_zerocopy_receive()
2112 zc->recv_skip_hint = inq; in tcp_zerocopy_receive()
2114 return -EIO; in tcp_zerocopy_receive()
2118 vma = find_tcp_vma(current->mm, address, &mmap_locked); in tcp_zerocopy_receive()
2120 return -EINVAL; in tcp_zerocopy_receive()
2122 vma_len = min_t(unsigned long, zc->length, vma->vm_end - address); in tcp_zerocopy_receive()
2124 total_bytes_to_map = avail_len & ~(PAGE_SIZE - 1); in tcp_zerocopy_receive()
2126 if (!(zc->flags & TCP_RECEIVE_ZEROCOPY_FLAG_TLB_CLEAN_HINT)) in tcp_zerocopy_receive()
2129 zc->length = total_bytes_to_map; in tcp_zerocopy_receive()
2130 zc->recv_skip_hint = 0; in tcp_zerocopy_receive()
2132 zc->length = avail_len; in tcp_zerocopy_receive()
2133 zc->recv_skip_hint = avail_len; in tcp_zerocopy_receive()
2136 while (length + PAGE_SIZE <= zc->length) { in tcp_zerocopy_receive()
2140 if (zc->recv_skip_hint < PAGE_SIZE) { in tcp_zerocopy_receive()
2144 if (zc->recv_skip_hint > 0) in tcp_zerocopy_receive()
2146 skb = skb->next; in tcp_zerocopy_receive()
2147 offset = seq - TCP_SKB_CB(skb)->seq; in tcp_zerocopy_receive()
2152 if (TCP_SKB_CB(skb)->has_rxtstamp) { in tcp_zerocopy_receive()
2154 zc->msg_flags |= TCP_CMSG_TS; in tcp_zerocopy_receive()
2156 zc->recv_skip_hint = skb->len - offset; in tcp_zerocopy_receive()
2163 zc->recv_skip_hint); in tcp_zerocopy_receive()
2165 zc->recv_skip_hint = mappable_offset; in tcp_zerocopy_receive()
2172 zc->recv_skip_hint -= PAGE_SIZE; in tcp_zerocopy_receive()
2175 zc->recv_skip_hint < PAGE_SIZE) { in tcp_zerocopy_receive()
2196 mmap_read_unlock(current->mm); in tcp_zerocopy_receive()
2204 WRITE_ONCE(tp->copied_seq, seq); in tcp_zerocopy_receive()
2211 if (length == zc->length) in tcp_zerocopy_receive()
2212 zc->recv_skip_hint = 0; in tcp_zerocopy_receive()
2214 if (!zc->recv_skip_hint && sock_flag(sk, SOCK_DONE)) in tcp_zerocopy_receive()
2215 ret = -EIO; in tcp_zerocopy_receive()
2217 zc->length = length; in tcp_zerocopy_receive()
2229 if (tss->ts[0].tv_sec || tss->ts[0].tv_nsec) { in tcp_recv_timestamp()
2234 .tv_sec = tss->ts[0].tv_sec, in tcp_recv_timestamp()
2235 .tv_nsec = tss->ts[0].tv_nsec, in tcp_recv_timestamp()
2241 .tv_sec = tss->ts[0].tv_sec, in tcp_recv_timestamp()
2242 .tv_nsec = tss->ts[0].tv_nsec, in tcp_recv_timestamp()
2250 .tv_sec = tss->ts[0].tv_sec, in tcp_recv_timestamp()
2251 .tv_usec = tss->ts[0].tv_nsec / 1000, in tcp_recv_timestamp()
2257 .tv_sec = tss->ts[0].tv_sec, in tcp_recv_timestamp()
2258 .tv_usec = tss->ts[0].tv_nsec / 1000, in tcp_recv_timestamp()
2266 if (READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_SOFTWARE) in tcp_recv_timestamp()
2269 tss->ts[0] = (struct timespec64) {0}; in tcp_recv_timestamp()
2272 if (tss->ts[2].tv_sec || tss->ts[2].tv_nsec) { in tcp_recv_timestamp()
2273 if (READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_RAW_HARDWARE) in tcp_recv_timestamp()
2276 tss->ts[2] = (struct timespec64) {0}; in tcp_recv_timestamp()
2280 tss->ts[1] = (struct timespec64) {0}; in tcp_recv_timestamp()
2291 u32 copied_seq = READ_ONCE(tp->copied_seq); in tcp_inq_hint()
2292 u32 rcv_nxt = READ_ONCE(tp->rcv_nxt); in tcp_inq_hint()
2295 inq = rcv_nxt - copied_seq; in tcp_inq_hint()
2296 if (unlikely(inq < 0 || copied_seq != READ_ONCE(tp->copied_seq))) { in tcp_inq_hint()
2298 inq = tp->rcv_nxt - tp->copied_seq; in tcp_inq_hint()
2301 /* After receiving a FIN, tell the user-space to continue reading in tcp_inq_hint()
2302 * by returning a non-zero inq. in tcp_inq_hint()
2313 * tricks with *seq access order and skb->users are not required.
2332 err = -ENOTCONN; in tcp_recvmsg_locked()
2333 if (sk->sk_state == TCP_LISTEN) in tcp_recvmsg_locked()
2336 if (tp->recvmsg_inq) { in tcp_recvmsg_locked()
2338 msg->msg_get_inq = 1; in tcp_recvmsg_locked()
2342 /* Urgent data needs to be handled specially. */ in tcp_recvmsg_locked()
2346 if (unlikely(tp->repair)) { in tcp_recvmsg_locked()
2347 err = -EPERM; in tcp_recvmsg_locked()
2351 if (tp->repair_queue == TCP_SEND_QUEUE) in tcp_recvmsg_locked()
2354 err = -EINVAL; in tcp_recvmsg_locked()
2355 if (tp->repair_queue == TCP_NO_QUEUE) in tcp_recvmsg_locked()
2358 /* 'common' recv queue MSG_PEEK-ing */ in tcp_recvmsg_locked()
2361 seq = &tp->copied_seq; in tcp_recvmsg_locked()
2363 peek_seq = tp->copied_seq; in tcp_recvmsg_locked()
2372 /* Are we at urgent data? Stop if we have read anything or have SIGURG pending. */ in tcp_recvmsg_locked()
2373 if (unlikely(tp->urg_data) && tp->urg_seq == *seq) { in tcp_recvmsg_locked()
2377 copied = timeo ? sock_intr_errno(timeo) : -EAGAIN; in tcp_recvmsg_locked()
2384 last = skb_peek_tail(&sk->sk_receive_queue); in tcp_recvmsg_locked()
2385 skb_queue_walk(&sk->sk_receive_queue, skb) { in tcp_recvmsg_locked()
2390 if (WARN(before(*seq, TCP_SKB_CB(skb)->seq), in tcp_recvmsg_locked()
2392 *seq, TCP_SKB_CB(skb)->seq, tp->rcv_nxt, in tcp_recvmsg_locked()
2396 offset = *seq - TCP_SKB_CB(skb)->seq; in tcp_recvmsg_locked()
2397 if (unlikely(TCP_SKB_CB(skb)->tcp_flags & TCPHDR_SYN)) { in tcp_recvmsg_locked()
2399 offset--; in tcp_recvmsg_locked()
2401 if (offset < skb->len) in tcp_recvmsg_locked()
2403 if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN) in tcp_recvmsg_locked()
2407 *seq, TCP_SKB_CB(skb)->seq, tp->rcv_nxt, flags); in tcp_recvmsg_locked()
2412 if (copied >= target && !READ_ONCE(sk->sk_backlog.tail)) in tcp_recvmsg_locked()
2417 sk->sk_err || in tcp_recvmsg_locked()
2418 sk->sk_state == TCP_CLOSE || in tcp_recvmsg_locked()
2419 (sk->sk_shutdown & RCV_SHUTDOWN) || in tcp_recvmsg_locked()
2426 if (sk->sk_err) { in tcp_recvmsg_locked()
2431 if (sk->sk_shutdown & RCV_SHUTDOWN) in tcp_recvmsg_locked()
2434 if (sk->sk_state == TCP_CLOSE) { in tcp_recvmsg_locked()
2438 copied = -ENOTCONN; in tcp_recvmsg_locked()
2443 copied = -EAGAIN; in tcp_recvmsg_locked()
2466 (peek_seq - copied - urg_hole != tp->copied_seq)) { in tcp_recvmsg_locked()
2468 current->comm, in tcp_recvmsg_locked()
2470 peek_seq = tp->copied_seq; in tcp_recvmsg_locked()
2476 used = skb->len - offset; in tcp_recvmsg_locked()
2480 /* Do we have urgent data here? */ in tcp_recvmsg_locked()
2481 if (unlikely(tp->urg_data)) { in tcp_recvmsg_locked()
2482 u32 urg_offset = tp->urg_seq - *seq; in tcp_recvmsg_locked()
2489 used--; in tcp_recvmsg_locked()
2503 copied = -EFAULT; in tcp_recvmsg_locked()
2510 len -= used; in tcp_recvmsg_locked()
2515 if (unlikely(tp->urg_data) && after(tp->copied_seq, tp->urg_seq)) { in tcp_recvmsg_locked()
2516 WRITE_ONCE(tp->urg_data, 0); in tcp_recvmsg_locked()
2520 if (TCP_SKB_CB(skb)->has_rxtstamp) { in tcp_recvmsg_locked()
2525 if (used + offset < skb->len) in tcp_recvmsg_locked()
2528 if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN) in tcp_recvmsg_locked()
2543 * on connected socket. I was just happy when found this 8) --ANK in tcp_recvmsg_locked()
2572 skb_queue_empty_lockless(&sk->sk_receive_queue) && in tcp_recvmsg()
2573 sk->sk_state == TCP_ESTABLISHED) in tcp_recvmsg()
2580 if ((cmsg_flags || msg->msg_get_inq) && ret >= 0) { in tcp_recvmsg()
2583 if (msg->msg_get_inq) { in tcp_recvmsg()
2584 msg->msg_inq = tcp_inq_hint(sk); in tcp_recvmsg()
2587 sizeof(msg->msg_inq), &msg->msg_inq); in tcp_recvmsg()
2596 int oldstate = sk->sk_state; in tcp_set_state()
2626 * above-mentioned anonymous enum in the vmlinux DWARF and hence BTF in tcp_set_state()
2644 sk->sk_prot->unhash(sk); in tcp_set_state()
2645 if (inet_csk(sk)->icsk_bind_hash && in tcp_set_state()
2646 !(sk->sk_userlocks & SOCK_BINDPORT_LOCK)) in tcp_set_state()
2687 int next = (int)new_state[sk->sk_state]; in tcp_close_state()
2710 if ((1 << sk->sk_state) & in tcp_shutdown()
2756 net_info_ratelimited("out of memory -- consider tuning tcp_mem\n"); in tcp_check_oom()
2766 WRITE_ONCE(sk->sk_shutdown, SHUTDOWN_MASK); in __tcp_close()
2768 if (sk->sk_state == TCP_LISTEN) { in __tcp_close()
2778 * descriptor close, not protocol-sourced closes, because the in __tcp_close()
2781 while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) { in __tcp_close()
2782 u32 len = TCP_SKB_CB(skb)->end_seq - TCP_SKB_CB(skb)->seq; in __tcp_close()
2784 if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN) in __tcp_close()
2785 len--; in __tcp_close()
2790 /* If socket has been already reset (e.g. in tcp_reset()) - kill it. */ in __tcp_close()
2791 if (sk->sk_state == TCP_CLOSE) in __tcp_close()
2798 * advertise a zero window, then kill -9 the FTP client, wheee... in __tcp_close()
2801 if (unlikely(tcp_sk(sk)->repair)) { in __tcp_close()
2802 sk->sk_prot->disconnect(sk, 0); in __tcp_close()
2807 tcp_send_active_reset(sk, sk->sk_allocation); in __tcp_close()
2808 } else if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) { in __tcp_close()
2810 sk->sk_prot->disconnect(sk, 0); in __tcp_close()
2817 /* RED-PEN. Formally speaking, we have broken TCP state in __tcp_close()
2820 * TCP_ESTABLISHED -> TCP_FIN_WAIT1 in __tcp_close()
2821 * TCP_SYN_RECV -> TCP_FIN_WAIT1 (forget it, it's impossible) in __tcp_close()
2822 * TCP_CLOSE_WAIT -> TCP_LAST_ACK in __tcp_close()
2828 * if Linux state is FIN-WAIT-1, but FIN is still not sent. in __tcp_close()
2831 * we enter time-wait state, when it is not required really in __tcp_close()
2836 * --ANK in __tcp_close()
2837 * XXX (TFO) - To start off we don't support SYN+ACK+FIN in __tcp_close()
2839 * probably need API support or TCP_CORK SYN-ACK until in __tcp_close()
2848 state = sk->sk_state; in __tcp_close()
2860 if (state != TCP_CLOSE && sk->sk_state == TCP_CLOSE) in __tcp_close()
2867 * our end. If they send after that then tough - BUT: long enough in __tcp_close()
2868 * that we won't make the old 4*rto = almost no time - whoops in __tcp_close()
2874 * linger2 option. --ANK in __tcp_close()
2877 if (sk->sk_state == TCP_FIN_WAIT2) { in __tcp_close()
2879 if (READ_ONCE(tp->linger2) < 0) { in __tcp_close()
2889 tmo - TCP_TIMEWAIT_LEN); in __tcp_close()
2896 if (sk->sk_state != TCP_CLOSE) { in __tcp_close()
2908 if (sk->sk_state == TCP_CLOSE) { in __tcp_close()
2911 req = rcu_dereference_protected(tcp_sk(sk)->fastopen_rsk, in __tcp_close()
2913 /* We could get here with a non-NULL req if the socket is in __tcp_close()
2948 struct rb_node *p = rb_first(&sk->tcp_rtx_queue); in tcp_rtx_queue_purge()
2950 tcp_sk(sk)->highest_sack = NULL; in tcp_rtx_queue_purge()
2956 * list_del(&skb->tcp_tsorted_anchor) in tcp_rtx_queue_purge()
2968 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) { in tcp_write_queue_purge()
2973 INIT_LIST_HEAD(&tcp_sk(sk)->tsorted_sent_queue); in tcp_write_queue_purge()
2975 tcp_sk(sk)->packets_out = 0; in tcp_write_queue_purge()
2976 inet_csk(sk)->icsk_backoff = 0; in tcp_write_queue_purge()
2984 int old_state = sk->sk_state; in tcp_disconnect()
2993 } else if (unlikely(tp->repair)) { in tcp_disconnect()
2994 WRITE_ONCE(sk->sk_err, ECONNABORTED); in tcp_disconnect()
2996 (tp->snd_nxt != tp->write_seq && in tcp_disconnect()
3002 WRITE_ONCE(sk->sk_err, ECONNRESET); in tcp_disconnect()
3004 WRITE_ONCE(sk->sk_err, ECONNRESET); in tcp_disconnect()
3007 __skb_queue_purge(&sk->sk_receive_queue); in tcp_disconnect()
3008 WRITE_ONCE(tp->copied_seq, tp->rcv_nxt); in tcp_disconnect()
3009 WRITE_ONCE(tp->urg_data, 0); in tcp_disconnect()
3012 skb_rbtree_purge(&tp->out_of_order_queue); in tcp_disconnect()
3014 inet->inet_dport = 0; in tcp_disconnect()
3018 WRITE_ONCE(sk->sk_shutdown, 0); in tcp_disconnect()
3020 tp->srtt_us = 0; in tcp_disconnect()
3021 tp->mdev_us = jiffies_to_usecs(TCP_TIMEOUT_INIT); in tcp_disconnect()
3022 tp->rcv_rtt_last_tsecr = 0; in tcp_disconnect()
3024 seq = tp->write_seq + tp->max_window + 2; in tcp_disconnect()
3027 WRITE_ONCE(tp->write_seq, seq); in tcp_disconnect()
3029 icsk->icsk_backoff = 0; in tcp_disconnect()
3030 icsk->icsk_probes_out = 0; in tcp_disconnect()
3031 icsk->icsk_probes_tstamp = 0; in tcp_disconnect()
3032 icsk->icsk_rto = TCP_TIMEOUT_INIT; in tcp_disconnect()
3033 icsk->icsk_rto_min = TCP_RTO_MIN; in tcp_disconnect()
3034 icsk->icsk_delack_max = TCP_DELACK_MAX; in tcp_disconnect()
3035 tp->snd_ssthresh = TCP_INFINITE_SSTHRESH; in tcp_disconnect()
3037 tp->snd_cwnd_cnt = 0; in tcp_disconnect()
3038 tp->is_cwnd_limited = 0; in tcp_disconnect()
3039 tp->max_packets_out = 0; in tcp_disconnect()
3040 tp->window_clamp = 0; in tcp_disconnect()
3041 tp->delivered = 0; in tcp_disconnect()
3042 tp->delivered_ce = 0; in tcp_disconnect()
3043 if (icsk->icsk_ca_ops->release) in tcp_disconnect()
3044 icsk->icsk_ca_ops->release(sk); in tcp_disconnect()
3045 memset(icsk->icsk_ca_priv, 0, sizeof(icsk->icsk_ca_priv)); in tcp_disconnect()
3046 icsk->icsk_ca_initialized = 0; in tcp_disconnect()
3048 tp->is_sack_reneg = 0; in tcp_disconnect()
3050 tp->total_retrans = 0; in tcp_disconnect()
3055 icsk->icsk_ack.rcv_mss = TCP_MIN_MSS; in tcp_disconnect()
3056 memset(&tp->rx_opt, 0, sizeof(tp->rx_opt)); in tcp_disconnect()
3058 dst_release(xchg((__force struct dst_entry **)&sk->sk_rx_dst, NULL)); in tcp_disconnect()
3060 tp->compressed_ack = 0; in tcp_disconnect()
3061 tp->segs_in = 0; in tcp_disconnect()
3062 tp->segs_out = 0; in tcp_disconnect()
3063 tp->bytes_sent = 0; in tcp_disconnect()
3064 tp->bytes_acked = 0; in tcp_disconnect()
3065 tp->bytes_received = 0; in tcp_disconnect()
3066 tp->bytes_retrans = 0; in tcp_disconnect()
3067 tp->data_segs_in = 0; in tcp_disconnect()
3068 tp->data_segs_out = 0; in tcp_disconnect()
3069 tp->duplicate_sack[0].start_seq = 0; in tcp_disconnect()
3070 tp->duplicate_sack[0].end_seq = 0; in tcp_disconnect()
3071 tp->dsack_dups = 0; in tcp_disconnect()
3072 tp->reord_seen = 0; in tcp_disconnect()
3073 tp->retrans_out = 0; in tcp_disconnect()
3074 tp->sacked_out = 0; in tcp_disconnect()
3075 tp->tlp_high_seq = 0; in tcp_disconnect()
3076 tp->last_oow_ack_time = 0; in tcp_disconnect()
3077 tp->plb_rehash = 0; in tcp_disconnect()
3079 tp->app_limited = ~0U; in tcp_disconnect()
3080 tp->rate_app_limited = 1; in tcp_disconnect()
3081 tp->rack.mstamp = 0; in tcp_disconnect()
3082 tp->rack.advanced = 0; in tcp_disconnect()
3083 tp->rack.reo_wnd_steps = 1; in tcp_disconnect()
3084 tp->rack.last_delivered = 0; in tcp_disconnect()
3085 tp->rack.reo_wnd_persist = 0; in tcp_disconnect()
3086 tp->rack.dsack_seen = 0; in tcp_disconnect()
3087 tp->syn_data_acked = 0; in tcp_disconnect()
3088 tp->rx_opt.saw_tstamp = 0; in tcp_disconnect()
3089 tp->rx_opt.dsack = 0; in tcp_disconnect()
3090 tp->rx_opt.num_sacks = 0; in tcp_disconnect()
3091 tp->rcv_ooopack = 0; in tcp_disconnect()
3097 tp->fastopen_client_fail = 0; in tcp_disconnect()
3099 WARN_ON(inet->inet_num && !icsk->icsk_bind_hash); in tcp_disconnect()
3101 if (sk->sk_frag.page) { in tcp_disconnect()
3102 put_page(sk->sk_frag.page); in tcp_disconnect()
3103 sk->sk_frag.page = NULL; in tcp_disconnect()
3104 sk->sk_frag.offset = 0; in tcp_disconnect()
3113 return sockopt_ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN) && in tcp_can_repair_sock()
3114 (sk->sk_state != TCP_LISTEN); in tcp_can_repair_sock()
3121 if (!tp->repair) in tcp_repair_set_window()
3122 return -EPERM; in tcp_repair_set_window()
3125 return -EINVAL; in tcp_repair_set_window()
3128 return -EFAULT; in tcp_repair_set_window()
3131 return -EINVAL; in tcp_repair_set_window()
3133 if (after(opt.snd_wl1, tp->rcv_nxt + opt.rcv_wnd)) in tcp_repair_set_window()
3134 return -EINVAL; in tcp_repair_set_window()
3136 if (after(opt.rcv_wup, tp->rcv_nxt)) in tcp_repair_set_window()
3137 return -EINVAL; in tcp_repair_set_window()
3139 tp->snd_wl1 = opt.snd_wl1; in tcp_repair_set_window()
3140 tp->snd_wnd = opt.snd_wnd; in tcp_repair_set_window()
3141 tp->max_window = opt.max_window; in tcp_repair_set_window()
3143 tp->rcv_wnd = opt.rcv_wnd; in tcp_repair_set_window()
3144 tp->rcv_wup = opt.rcv_wup; in tcp_repair_set_window()
3158 return -EFAULT; in tcp_repair_options_est()
3161 len -= sizeof(opt); in tcp_repair_options_est()
3165 tp->rx_opt.mss_clamp = opt.opt_val; in tcp_repair_options_est()
3174 return -EFBIG; in tcp_repair_options_est()
3176 tp->rx_opt.snd_wscale = snd_wscale; in tcp_repair_options_est()
3177 tp->rx_opt.rcv_wscale = rcv_wscale; in tcp_repair_options_est()
3178 tp->rx_opt.wscale_ok = 1; in tcp_repair_options_est()
3183 return -EINVAL; in tcp_repair_options_est()
3185 tp->rx_opt.sack_ok |= TCP_SACK_SEEN; in tcp_repair_options_est()
3189 return -EINVAL; in tcp_repair_options_est()
3191 tp->rx_opt.tstamp_ok = 1; in tcp_repair_options_est()
3214 /* When set indicates to always queue non-full frames. Later the user clears
3228 tp->nonagle |= TCP_NAGLE_CORK; in __tcp_sock_set_cork()
3230 tp->nonagle &= ~TCP_NAGLE_CORK; in __tcp_sock_set_cork()
3231 if (tp->nonagle & TCP_NAGLE_OFF) in __tcp_sock_set_cork()
3232 tp->nonagle |= TCP_NAGLE_PUSH; in __tcp_sock_set_cork()
3254 tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH; in __tcp_sock_set_nodelay()
3257 tcp_sk(sk)->nonagle &= ~TCP_NAGLE_OFF; in __tcp_sock_set_nodelay()
3277 if ((1 << sk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT) && in __tcp_sock_set_quickack()
3279 inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_PUSHED; in __tcp_sock_set_quickack()
3297 return -EINVAL; in tcp_sock_set_syncnt()
3299 WRITE_ONCE(inet_csk(sk)->icsk_syn_retries, val); in tcp_sock_set_syncnt()
3310 return -EINVAL; in tcp_sock_set_user_timeout()
3312 WRITE_ONCE(inet_csk(sk)->icsk_user_timeout, val); in tcp_sock_set_user_timeout()
3322 return -EINVAL; in tcp_sock_set_keepidle_locked()
3325 WRITE_ONCE(tp->keepalive_time, val * HZ); in tcp_sock_set_keepidle_locked()
3327 !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) { in tcp_sock_set_keepidle_locked()
3330 if (tp->keepalive_time > elapsed) in tcp_sock_set_keepidle_locked()
3331 elapsed = tp->keepalive_time - elapsed; in tcp_sock_set_keepidle_locked()
3354 return -EINVAL; in tcp_sock_set_keepintvl()
3356 WRITE_ONCE(tcp_sk(sk)->keepalive_intvl, val * HZ); in tcp_sock_set_keepintvl()
3364 return -EINVAL; in tcp_sock_set_keepcnt()
3367 WRITE_ONCE(tcp_sk(sk)->keepalive_probes, val); in tcp_sock_set_keepcnt()
3377 if (sk->sk_state != TCP_CLOSE) in tcp_set_window_clamp()
3378 return -EINVAL; in tcp_set_window_clamp()
3379 tp->window_clamp = 0; in tcp_set_window_clamp()
3381 u32 new_rcv_ssthresh, old_window_clamp = tp->window_clamp; in tcp_set_window_clamp()
3388 tp->window_clamp = new_window_clamp; in tcp_set_window_clamp()
3393 __tcp_adjust_rcv_ssthresh(sk, tp->window_clamp); in tcp_set_window_clamp()
3396 new_rcv_ssthresh = min(tp->rcv_wnd, tp->window_clamp); in tcp_set_window_clamp()
3397 tp->rcv_ssthresh = max(new_rcv_ssthresh, in tcp_set_window_clamp()
3398 tp->rcv_ssthresh); in tcp_set_window_clamp()
3422 return -EINVAL; in do_tcp_setsockopt()
3425 min_t(long, TCP_CA_NAME_MAX-1, optlen)); in do_tcp_setsockopt()
3427 return -EFAULT; in do_tcp_setsockopt()
3432 sockopt_ns_capable(sock_net(sk)->user_ns, in do_tcp_setsockopt()
3441 return -EINVAL; in do_tcp_setsockopt()
3444 min_t(long, TCP_ULP_NAME_MAX - 1, in do_tcp_setsockopt()
3447 return -EFAULT; in do_tcp_setsockopt()
3464 return -EINVAL; in do_tcp_setsockopt()
3467 return -EFAULT; in do_tcp_setsockopt()
3480 return -EINVAL; in do_tcp_setsockopt()
3483 return -EFAULT; in do_tcp_setsockopt()
3497 WRITE_ONCE(tp->linger2, -1); in do_tcp_setsockopt()
3499 WRITE_ONCE(tp->linger2, TCP_FIN_TIMEOUT_MAX); in do_tcp_setsockopt()
3501 WRITE_ONCE(tp->linger2, val * HZ); in do_tcp_setsockopt()
3505 WRITE_ONCE(icsk->icsk_accept_queue.rskq_defer_accept, in do_tcp_setsockopt()
3520 err = -EINVAL; in do_tcp_setsockopt()
3523 tp->rx_opt.user_mss = val; in do_tcp_setsockopt()
3532 err = -EINVAL; in do_tcp_setsockopt()
3534 tp->thin_lto = val; in do_tcp_setsockopt()
3539 err = -EINVAL; in do_tcp_setsockopt()
3544 err = -EPERM; in do_tcp_setsockopt()
3546 tp->repair = 1; in do_tcp_setsockopt()
3547 sk->sk_reuse = SK_FORCE_REUSE; in do_tcp_setsockopt()
3548 tp->repair_queue = TCP_NO_QUEUE; in do_tcp_setsockopt()
3550 tp->repair = 0; in do_tcp_setsockopt()
3551 sk->sk_reuse = SK_NO_REUSE; in do_tcp_setsockopt()
3554 tp->repair = 0; in do_tcp_setsockopt()
3555 sk->sk_reuse = SK_NO_REUSE; in do_tcp_setsockopt()
3557 err = -EINVAL; in do_tcp_setsockopt()
3562 if (!tp->repair) in do_tcp_setsockopt()
3563 err = -EPERM; in do_tcp_setsockopt()
3565 tp->repair_queue = val; in do_tcp_setsockopt()
3567 err = -EINVAL; in do_tcp_setsockopt()
3571 if (sk->sk_state != TCP_CLOSE) { in do_tcp_setsockopt()
3572 err = -EPERM; in do_tcp_setsockopt()
3573 } else if (tp->repair_queue == TCP_SEND_QUEUE) { in do_tcp_setsockopt()
3575 err = -EPERM; in do_tcp_setsockopt()
3577 WRITE_ONCE(tp->write_seq, val); in do_tcp_setsockopt()
3578 } else if (tp->repair_queue == TCP_RECV_QUEUE) { in do_tcp_setsockopt()
3579 if (tp->rcv_nxt != tp->copied_seq) { in do_tcp_setsockopt()
3580 err = -EPERM; in do_tcp_setsockopt()
3582 WRITE_ONCE(tp->rcv_nxt, val); in do_tcp_setsockopt()
3583 WRITE_ONCE(tp->copied_seq, val); in do_tcp_setsockopt()
3586 err = -EINVAL; in do_tcp_setsockopt()
3591 if (!tp->repair) in do_tcp_setsockopt()
3592 err = -EINVAL; in do_tcp_setsockopt()
3593 else if (sk->sk_state == TCP_ESTABLISHED && !tp->bytes_sent) in do_tcp_setsockopt()
3596 err = -EPERM; in do_tcp_setsockopt()
3609 err = -EINVAL; in do_tcp_setsockopt()
3611 tp->save_syn = val; in do_tcp_setsockopt()
3624 err = -EPERM; in do_tcp_setsockopt()
3633 /* If this is the first TCP-AO setsockopt() on the socket, in do_tcp_setsockopt()
3637 if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) in do_tcp_setsockopt()
3639 if (rcu_dereference_protected(tcp_sk(sk)->ao_info, in do_tcp_setsockopt()
3642 if (tp->repair) in do_tcp_setsockopt()
3644 err = -EISCONN; in do_tcp_setsockopt()
3647 err = tp->af_specific->ao_parse(sk, optname, optval, optlen); in do_tcp_setsockopt()
3654 err = tp->af_specific->md5_parse(sk, optname, optval, optlen); in do_tcp_setsockopt()
3658 if (val >= 0 && ((1 << sk->sk_state) & (TCPF_CLOSE | in do_tcp_setsockopt()
3664 err = -EINVAL; in do_tcp_setsockopt()
3669 err = -EINVAL; in do_tcp_setsockopt()
3670 } else if (READ_ONCE(net->ipv4.sysctl_tcp_fastopen) & in do_tcp_setsockopt()
3672 if (sk->sk_state == TCP_CLOSE) in do_tcp_setsockopt()
3673 tp->fastopen_connect = val; in do_tcp_setsockopt()
3675 err = -EINVAL; in do_tcp_setsockopt()
3677 err = -EOPNOTSUPP; in do_tcp_setsockopt()
3682 err = -EINVAL; in do_tcp_setsockopt()
3683 else if (!((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) in do_tcp_setsockopt()
3684 err = -EINVAL; in do_tcp_setsockopt()
3686 tp->fastopen_no_cookie = val; in do_tcp_setsockopt()
3689 if (!tp->repair) { in do_tcp_setsockopt()
3690 err = -EPERM; in do_tcp_setsockopt()
3697 tp->tcp_usec_ts = val & 1; in do_tcp_setsockopt()
3698 WRITE_ONCE(tp->tsoffset, val - tcp_clock_ts(tp->tcp_usec_ts)); in do_tcp_setsockopt()
3704 WRITE_ONCE(tp->notsent_lowat, val); in do_tcp_setsockopt()
3705 sk->sk_write_space(sk); in do_tcp_setsockopt()
3709 err = -EINVAL; in do_tcp_setsockopt()
3711 tp->recvmsg_inq = val; in do_tcp_setsockopt()
3716 WRITE_ONCE(tp->tcp_tx_delay, val); in do_tcp_setsockopt()
3719 err = -ENOPROTOOPT; in do_tcp_setsockopt()
3734 return READ_ONCE(icsk->icsk_af_ops)->setsockopt(sk, level, optname, in tcp_setsockopt()
3747 stats[i] = tp->chrono_stat[i - 1]; in tcp_get_info_chrono_stats()
3748 if (i == tp->chrono_type) in tcp_get_info_chrono_stats()
3749 stats[i] += tcp_jiffies32 - tp->chrono_start; in tcp_get_info_chrono_stats()
3754 info->tcpi_busy_time = total; in tcp_get_info_chrono_stats()
3755 info->tcpi_rwnd_limited = stats[TCP_CHRONO_RWND_LIMITED]; in tcp_get_info_chrono_stats()
3756 info->tcpi_sndbuf_limited = stats[TCP_CHRONO_SNDBUF_LIMITED]; in tcp_get_info_chrono_stats()
3770 if (sk->sk_type != SOCK_STREAM) in tcp_get_info()
3773 info->tcpi_state = inet_sk_state_load(sk); in tcp_get_info()
3776 rate = READ_ONCE(sk->sk_pacing_rate); in tcp_get_info()
3778 info->tcpi_pacing_rate = rate64; in tcp_get_info()
3780 rate = READ_ONCE(sk->sk_max_pacing_rate); in tcp_get_info()
3782 info->tcpi_max_pacing_rate = rate64; in tcp_get_info()
3784 info->tcpi_reordering = tp->reordering; in tcp_get_info()
3785 info->tcpi_snd_cwnd = tcp_snd_cwnd(tp); in tcp_get_info()
3787 if (info->tcpi_state == TCP_LISTEN) { in tcp_get_info()
3789 * tcpi_unacked -> Number of children ready for accept() in tcp_get_info()
3790 * tcpi_sacked -> max backlog in tcp_get_info()
3792 info->tcpi_unacked = READ_ONCE(sk->sk_ack_backlog); in tcp_get_info()
3793 info->tcpi_sacked = READ_ONCE(sk->sk_max_ack_backlog); in tcp_get_info()
3799 info->tcpi_ca_state = icsk->icsk_ca_state; in tcp_get_info()
3800 info->tcpi_retransmits = icsk->icsk_retransmits; in tcp_get_info()
3801 info->tcpi_probes = icsk->icsk_probes_out; in tcp_get_info()
3802 info->tcpi_backoff = icsk->icsk_backoff; in tcp_get_info()
3804 if (tp->rx_opt.tstamp_ok) in tcp_get_info()
3805 info->tcpi_options |= TCPI_OPT_TIMESTAMPS; in tcp_get_info()
3807 info->tcpi_options |= TCPI_OPT_SACK; in tcp_get_info()
3808 if (tp->rx_opt.wscale_ok) { in tcp_get_info()
3809 info->tcpi_options |= TCPI_OPT_WSCALE; in tcp_get_info()
3810 info->tcpi_snd_wscale = tp->rx_opt.snd_wscale; in tcp_get_info()
3811 info->tcpi_rcv_wscale = tp->rx_opt.rcv_wscale; in tcp_get_info()
3814 if (tp->ecn_flags & TCP_ECN_OK) in tcp_get_info()
3815 info->tcpi_options |= TCPI_OPT_ECN; in tcp_get_info()
3816 if (tp->ecn_flags & TCP_ECN_SEEN) in tcp_get_info()
3817 info->tcpi_options |= TCPI_OPT_ECN_SEEN; in tcp_get_info()
3818 if (tp->syn_data_acked) in tcp_get_info()
3819 info->tcpi_options |= TCPI_OPT_SYN_DATA; in tcp_get_info()
3820 if (tp->tcp_usec_ts) in tcp_get_info()
3821 info->tcpi_options |= TCPI_OPT_USEC_TS; in tcp_get_info()
3823 info->tcpi_rto = jiffies_to_usecs(icsk->icsk_rto); in tcp_get_info()
3824 info->tcpi_ato = jiffies_to_usecs(min_t(u32, icsk->icsk_ack.ato, in tcp_get_info()
3826 info->tcpi_snd_mss = tp->mss_cache; in tcp_get_info()
3827 info->tcpi_rcv_mss = icsk->icsk_ack.rcv_mss; in tcp_get_info()
3829 info->tcpi_unacked = tp->packets_out; in tcp_get_info()
3830 info->tcpi_sacked = tp->sacked_out; in tcp_get_info()
3832 info->tcpi_lost = tp->lost_out; in tcp_get_info()
3833 info->tcpi_retrans = tp->retrans_out; in tcp_get_info()
3836 info->tcpi_last_data_sent = jiffies_to_msecs(now - tp->lsndtime); in tcp_get_info()
3837 info->tcpi_last_data_recv = jiffies_to_msecs(now - icsk->icsk_ack.lrcvtime); in tcp_get_info()
3838 info->tcpi_last_ack_recv = jiffies_to_msecs(now - tp->rcv_tstamp); in tcp_get_info()
3840 info->tcpi_pmtu = icsk->icsk_pmtu_cookie; in tcp_get_info()
3841 info->tcpi_rcv_ssthresh = tp->rcv_ssthresh; in tcp_get_info()
3842 info->tcpi_rtt = tp->srtt_us >> 3; in tcp_get_info()
3843 info->tcpi_rttvar = tp->mdev_us >> 2; in tcp_get_info()
3844 info->tcpi_snd_ssthresh = tp->snd_ssthresh; in tcp_get_info()
3845 info->tcpi_advmss = tp->advmss; in tcp_get_info()
3847 info->tcpi_rcv_rtt = tp->rcv_rtt_est.rtt_us >> 3; in tcp_get_info()
3848 info->tcpi_rcv_space = tp->rcvq_space.space; in tcp_get_info()
3850 info->tcpi_total_retrans = tp->total_retrans; in tcp_get_info()
3852 info->tcpi_bytes_acked = tp->bytes_acked; in tcp_get_info()
3853 info->tcpi_bytes_received = tp->bytes_received; in tcp_get_info()
3854 info->tcpi_notsent_bytes = max_t(int, 0, tp->write_seq - tp->snd_nxt); in tcp_get_info()
3857 info->tcpi_segs_out = tp->segs_out; in tcp_get_info()
3860 info->tcpi_segs_in = READ_ONCE(tp->segs_in); in tcp_get_info()
3861 info->tcpi_data_segs_in = READ_ONCE(tp->data_segs_in); in tcp_get_info()
3863 info->tcpi_min_rtt = tcp_min_rtt(tp); in tcp_get_info()
3864 info->tcpi_data_segs_out = tp->data_segs_out; in tcp_get_info()
3866 info->tcpi_delivery_rate_app_limited = tp->rate_app_limited ? 1 : 0; in tcp_get_info()
3869 info->tcpi_delivery_rate = rate64; in tcp_get_info()
3870 info->tcpi_delivered = tp->delivered; in tcp_get_info()
3871 info->tcpi_delivered_ce = tp->delivered_ce; in tcp_get_info()
3872 info->tcpi_bytes_sent = tp->bytes_sent; in tcp_get_info()
3873 info->tcpi_bytes_retrans = tp->bytes_retrans; in tcp_get_info()
3874 info->tcpi_dsack_dups = tp->dsack_dups; in tcp_get_info()
3875 info->tcpi_reord_seen = tp->reord_seen; in tcp_get_info()
3876 info->tcpi_rcv_ooopack = tp->rcv_ooopack; in tcp_get_info()
3877 info->tcpi_snd_wnd = tp->snd_wnd; in tcp_get_info()
3878 info->tcpi_rcv_wnd = tp->rcv_wnd; in tcp_get_info()
3879 info->tcpi_rehash = tp->plb_rehash + tp->timeout_rehash; in tcp_get_info()
3880 info->tcpi_fastopen_client_fail = tp->fastopen_client_fail; in tcp_get_info()
3882 info->tcpi_total_rto = tp->total_rto; in tcp_get_info()
3883 info->tcpi_total_rto_recoveries = tp->total_rto_recoveries; in tcp_get_info()
3884 info->tcpi_total_rto_time = tp->total_rto_time; in tcp_get_info()
3885 if (tp->rto_stamp) in tcp_get_info()
3886 info->tcpi_total_rto_time += tcp_clock_ms() - tp->rto_stamp; in tcp_get_info()
3928 if (skb->protocol == htons(ETH_P_IP)) in tcp_skb_ttl_or_hop_limit()
3929 return ip_hdr(skb)->ttl; in tcp_skb_ttl_or_hop_limit()
3930 else if (skb->protocol == htons(ETH_P_IPV6)) in tcp_skb_ttl_or_hop_limit()
3931 return ipv6_hdr(skb)->hop_limit; in tcp_skb_ttl_or_hop_limit()
3958 tp->data_segs_out, TCP_NLA_PAD); in tcp_get_timestamping_opt_stats()
3960 tp->total_retrans, TCP_NLA_PAD); in tcp_get_timestamping_opt_stats()
3962 rate = READ_ONCE(sk->sk_pacing_rate); in tcp_get_timestamping_opt_stats()
3970 nla_put_u32(stats, TCP_NLA_REORDERING, tp->reordering); in tcp_get_timestamping_opt_stats()
3973 nla_put_u8(stats, TCP_NLA_RECUR_RETRANS, inet_csk(sk)->icsk_retransmits); in tcp_get_timestamping_opt_stats()
3974 nla_put_u8(stats, TCP_NLA_DELIVERY_RATE_APP_LMT, !!tp->rate_app_limited); in tcp_get_timestamping_opt_stats()
3975 nla_put_u32(stats, TCP_NLA_SND_SSTHRESH, tp->snd_ssthresh); in tcp_get_timestamping_opt_stats()
3976 nla_put_u32(stats, TCP_NLA_DELIVERED, tp->delivered); in tcp_get_timestamping_opt_stats()
3977 nla_put_u32(stats, TCP_NLA_DELIVERED_CE, tp->delivered_ce); in tcp_get_timestamping_opt_stats()
3979 nla_put_u32(stats, TCP_NLA_SNDQ_SIZE, tp->write_seq - tp->snd_una); in tcp_get_timestamping_opt_stats()
3980 nla_put_u8(stats, TCP_NLA_CA_STATE, inet_csk(sk)->icsk_ca_state); in tcp_get_timestamping_opt_stats()
3982 nla_put_u64_64bit(stats, TCP_NLA_BYTES_SENT, tp->bytes_sent, in tcp_get_timestamping_opt_stats()
3984 nla_put_u64_64bit(stats, TCP_NLA_BYTES_RETRANS, tp->bytes_retrans, in tcp_get_timestamping_opt_stats()
3986 nla_put_u32(stats, TCP_NLA_DSACK_DUPS, tp->dsack_dups); in tcp_get_timestamping_opt_stats()
3987 nla_put_u32(stats, TCP_NLA_REORD_SEEN, tp->reord_seen); in tcp_get_timestamping_opt_stats()
3988 nla_put_u32(stats, TCP_NLA_SRTT, tp->srtt_us >> 3); in tcp_get_timestamping_opt_stats()
3989 nla_put_u16(stats, TCP_NLA_TIMEOUT_REHASH, tp->timeout_rehash); in tcp_get_timestamping_opt_stats()
3991 max_t(int, 0, tp->write_seq - tp->snd_nxt)); in tcp_get_timestamping_opt_stats()
3992 nla_put_u64_64bit(stats, TCP_NLA_EDT, orig_skb->skb_mstamp_ns, in tcp_get_timestamping_opt_stats()
3998 nla_put_u32(stats, TCP_NLA_REHASH, tp->plb_rehash + tp->timeout_rehash); in tcp_get_timestamping_opt_stats()
4011 return -EFAULT; in do_tcp_getsockopt()
4016 return -EINVAL; in do_tcp_getsockopt()
4020 val = tp->mss_cache; in do_tcp_getsockopt()
4021 if (tp->rx_opt.user_mss && in do_tcp_getsockopt()
4022 ((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) in do_tcp_getsockopt()
4023 val = tp->rx_opt.user_mss; in do_tcp_getsockopt()
4024 if (tp->repair) in do_tcp_getsockopt()
4025 val = tp->rx_opt.mss_clamp; in do_tcp_getsockopt()
4028 val = !!(tp->nonagle&TCP_NAGLE_OFF); in do_tcp_getsockopt()
4031 val = !!(tp->nonagle&TCP_NAGLE_CORK); in do_tcp_getsockopt()
4043 val = READ_ONCE(icsk->icsk_syn_retries) ? : in do_tcp_getsockopt()
4044 READ_ONCE(net->ipv4.sysctl_tcp_syn_retries); in do_tcp_getsockopt()
4047 val = READ_ONCE(tp->linger2); in do_tcp_getsockopt()
4049 val = (val ? : READ_ONCE(net->ipv4.sysctl_tcp_fin_timeout)) / HZ; in do_tcp_getsockopt()
4052 val = READ_ONCE(icsk->icsk_accept_queue.rskq_defer_accept); in do_tcp_getsockopt()
4057 val = tp->window_clamp; in do_tcp_getsockopt()
4063 return -EFAULT; in do_tcp_getsockopt()
4069 return -EFAULT; in do_tcp_getsockopt()
4071 return -EFAULT; in do_tcp_getsockopt()
4081 return -EFAULT; in do_tcp_getsockopt()
4083 ca_ops = icsk->icsk_ca_ops; in do_tcp_getsockopt()
4084 if (ca_ops && ca_ops->get_info) in do_tcp_getsockopt()
4085 sz = ca_ops->get_info(sk, ~0U, &attr, &info); in do_tcp_getsockopt()
4089 return -EFAULT; in do_tcp_getsockopt()
4091 return -EFAULT; in do_tcp_getsockopt()
4100 return -EFAULT; in do_tcp_getsockopt()
4103 return -EFAULT; in do_tcp_getsockopt()
4104 if (copy_to_sockptr(optval, icsk->icsk_ca_ops->name, len)) in do_tcp_getsockopt()
4105 return -EFAULT; in do_tcp_getsockopt()
4110 return -EFAULT; in do_tcp_getsockopt()
4112 if (!icsk->icsk_ulp_ops) { in do_tcp_getsockopt()
4115 return -EFAULT; in do_tcp_getsockopt()
4119 return -EFAULT; in do_tcp_getsockopt()
4120 if (copy_to_sockptr(optval, icsk->icsk_ulp_ops->name, len)) in do_tcp_getsockopt()
4121 return -EFAULT; in do_tcp_getsockopt()
4129 return -EFAULT; in do_tcp_getsockopt()
4135 return -EFAULT; in do_tcp_getsockopt()
4137 return -EFAULT; in do_tcp_getsockopt()
4141 val = tp->thin_lto; in do_tcp_getsockopt()
4149 val = tp->repair; in do_tcp_getsockopt()
4153 if (tp->repair) in do_tcp_getsockopt()
4154 val = tp->repair_queue; in do_tcp_getsockopt()
4156 return -EINVAL; in do_tcp_getsockopt()
4163 return -EFAULT; in do_tcp_getsockopt()
4166 return -EINVAL; in do_tcp_getsockopt()
4168 if (!tp->repair) in do_tcp_getsockopt()
4169 return -EPERM; in do_tcp_getsockopt()
4171 opt.snd_wl1 = tp->snd_wl1; in do_tcp_getsockopt()
4172 opt.snd_wnd = tp->snd_wnd; in do_tcp_getsockopt()
4173 opt.max_window = tp->max_window; in do_tcp_getsockopt()
4174 opt.rcv_wnd = tp->rcv_wnd; in do_tcp_getsockopt()
4175 opt.rcv_wup = tp->rcv_wup; in do_tcp_getsockopt()
4178 return -EFAULT; in do_tcp_getsockopt()
4182 if (tp->repair_queue == TCP_SEND_QUEUE) in do_tcp_getsockopt()
4183 val = tp->write_seq; in do_tcp_getsockopt()
4184 else if (tp->repair_queue == TCP_RECV_QUEUE) in do_tcp_getsockopt()
4185 val = tp->rcv_nxt; in do_tcp_getsockopt()
4187 return -EINVAL; in do_tcp_getsockopt()
4191 val = READ_ONCE(icsk->icsk_user_timeout); in do_tcp_getsockopt()
4195 val = READ_ONCE(icsk->icsk_accept_queue.fastopenq.max_qlen); in do_tcp_getsockopt()
4199 val = tp->fastopen_connect; in do_tcp_getsockopt()
4203 val = tp->fastopen_no_cookie; in do_tcp_getsockopt()
4207 val = READ_ONCE(tp->tcp_tx_delay); in do_tcp_getsockopt()
4211 val = tcp_clock_ts(tp->tcp_usec_ts) + READ_ONCE(tp->tsoffset); in do_tcp_getsockopt()
4212 if (tp->tcp_usec_ts) in do_tcp_getsockopt()
4218 val = READ_ONCE(tp->notsent_lowat); in do_tcp_getsockopt()
4221 val = tp->recvmsg_inq; in do_tcp_getsockopt()
4224 val = tp->save_syn; in do_tcp_getsockopt()
4228 return -EFAULT; in do_tcp_getsockopt()
4231 if (tp->saved_syn) { in do_tcp_getsockopt()
4232 if (len < tcp_saved_syn_len(tp->saved_syn)) { in do_tcp_getsockopt()
4233 len = tcp_saved_syn_len(tp->saved_syn); in do_tcp_getsockopt()
4236 return -EFAULT; in do_tcp_getsockopt()
4239 return -EINVAL; in do_tcp_getsockopt()
4241 len = tcp_saved_syn_len(tp->saved_syn); in do_tcp_getsockopt()
4244 return -EFAULT; in do_tcp_getsockopt()
4246 if (copy_to_sockptr(optval, tp->saved_syn->data, len)) { in do_tcp_getsockopt()
4248 return -EFAULT; in do_tcp_getsockopt()
4256 return -EFAULT; in do_tcp_getsockopt()
4267 return -EFAULT; in do_tcp_getsockopt()
4270 return -EINVAL; in do_tcp_getsockopt()
4273 len - sizeof(zc)); in do_tcp_getsockopt()
4275 return err == 0 ? -EINVAL : err; in do_tcp_getsockopt()
4278 return -EFAULT; in do_tcp_getsockopt()
4281 return -EFAULT; in do_tcp_getsockopt()
4283 return -EINVAL; in do_tcp_getsockopt()
4285 return -EINVAL; in do_tcp_getsockopt()
4321 err = -EFAULT; in do_tcp_getsockopt()
4327 return -EPERM; in do_tcp_getsockopt()
4343 return -ENOPROTOOPT; in do_tcp_getsockopt()
4347 return -EFAULT; in do_tcp_getsockopt()
4349 return -EFAULT; in do_tcp_getsockopt()
4372 return READ_ONCE(icsk->icsk_af_ops)->getsockopt(sk, level, optname, in tcp_getsockopt()
4380 int tcp_md5_sigpool_id = -1;
4392 * id would stay the same. Re-write the id only for the case in tcp_md5_alloc_sigpool()
4416 u8 keylen = READ_ONCE(key->keylen); /* paired with WRITE_ONCE() in tcp_md5_do_add */ in tcp_md5_hash_key()
4419 sg_init_one(&sg, key->key, keylen); in tcp_md5_hash_key()
4420 ahash_request_set_crypt(hp->req, &sg, NULL, keylen); in tcp_md5_hash_key()
4423 * key->key under us in tcp_md5_hash_key()
4425 return data_race(crypto_ahash_update(hp->req)); in tcp_md5_hash_key()
4435 /* This gets called for each TCP segment that has TCP-MD5 option. in tcp_inbound_md5_hash()
4456 * IPv4-mapped case. in tcp_inbound_md5_hash()
4461 genhash = tp->af_specific->calc_md5_hash(newhash, key, in tcp_inbound_md5_hash()
4496 req = rcu_dereference_protected(tcp_sk(sk)->fastopen_rsk, 1); in tcp_done()
4498 if (sk->sk_state == TCP_SYN_SENT || sk->sk_state == TCP_SYN_RECV) in tcp_done()
4506 WRITE_ONCE(sk->sk_shutdown, SHUTDOWN_MASK); in tcp_done()
4509 sk->sk_state_change(sk); in tcp_done()
4523 inet_csk_reqsk_queue_drop(req->rsk_listener, req); in tcp_abort()
4530 refcount_inc(&tw->tw_refcnt); in tcp_abort()
4542 if (sk->sk_state == TCP_LISTEN) { in tcp_abort()
4552 WRITE_ONCE(sk->sk_err, err); in tcp_abort()
4556 if (tcp_need_reset(sk->sk_state)) in tcp_abort()
4600 /* TX read-mostly hotpath cache lines */ in tcp_struct_check()
4610 /* TXRX read-mostly hotpath cache lines */ in tcp_struct_check()
4621 /* RX read-mostly hotpath cache lines */ in tcp_struct_check()
4636 /* TX read-write hotpath cache lines */ in tcp_struct_check()
4656 /* TXRX read-write hotpath cache lines */ in tcp_struct_check()
4672 /* RX read-write hotpath cache lines */ in tcp_struct_check()
4768 /* Set per-socket limits to no more than 1/128 the pressure threshold */ in tcp_init()
4769 limit = nr_free_buffer_pages() << (PAGE_SHIFT - 7); in tcp_init()