diff options
Diffstat (limited to 'net/ipv4/tcp.c')
| -rw-r--r-- | net/ipv4/tcp.c | 63 |
1 files changed, 41 insertions, 22 deletions
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index aa890aef3..9f7ad441e 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -5,7 +5,7 @@ * * Implementation of the Transmission Control Protocol(TCP). * - * Version: $Id: tcp.c,v 1.161 2000/01/31 01:21:16 davem Exp $ + * Version: $Id: tcp.c,v 1.163 2000/02/08 21:27:13 davem Exp $ * * Authors: Ross Biro, <bir7@leland.Stanford.Edu> * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> @@ -1106,8 +1106,8 @@ static void cleanup_rbuf(struct sock *sk, int copied) { struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp); struct sk_buff *skb; - int time_to_ack; - + int time_to_ack = 0; + /* NOTE! The socket must be locked, so that we don't get * a messed-up receive queue. */ @@ -1117,13 +1117,39 @@ static void cleanup_rbuf(struct sock *sk, int copied) tcp_eat_skb(sk, skb); } - /* Delayed ACKs frequently hit locked sockets during bulk receive. */ - time_to_ack = tp->ack.blocked && tp->ack.pending; -#ifdef CONFIG_TCP_MORE_COARSE_ACKS - if (tp->ack.pending && - (tp->rcv_nxt - tp->rcv_wup) > tp->ack.rcv_mss) - time_to_ack = 1; + if (tp->ack.pending) { + /* Delayed ACKs frequently hit locked sockets during bulk receive. */ + if (tp->ack.blocked +#ifdef TCP_MORE_COARSE_ACKS + /* Once-per-two-segments ACK was not sent by tcp_input.c */ + || tp->rcv_nxt - tp->rcv_wup > tp->ack.rcv_mss #endif + /* + * If this read emptied read buffer, we send ACK when: + * + * -- ATO estimator diverged. In this case it is useless + * to delay ACK, it will miss in any case. + * + * -- The second condition is triggered when we did not + * ACK 8 segments not depending of their size. + * Linux senders allocate full-sized frame even for one byte + * packets, so that default queue for MTU=8K can hold + * only 8 packets. Note, that no other workarounds + * but counting packets are possible. If sender selected + * a small sndbuf or have larger mtu lockup will still + * occur. Well, not lockup, but 10-20msec gap. + * It is essentially dead lockup for 1Gib ethernet + * and loopback :-). The value 8 covers all reasonable + * cases and we may receive packet of any size + * with maximal possible rate now. + */ + || (copied > 0 && + (tp->ack.ato >= TCP_DELACK_MAX || tp->ack.rcv_segs > 7) && + !tp->ack.pingpong && + atomic_read(&sk->rmem_alloc) == 0)) { + time_to_ack = 1; + } + } /* We send an ACK if we can now advertise a non-zero window * which has been raised "significantly". @@ -1135,14 +1161,12 @@ static void cleanup_rbuf(struct sock *sk, int copied) __u32 rcv_window_now = tcp_receive_window(tp); __u32 new_window = __tcp_select_window(sk); - /* We won't be raising the window any further than - * the window-clamp allows. Our window selection - * also keeps things a nice multiple of MSS. These - * checks are necessary to prevent spurious ACKs - * which don't advertize a larger window. + /* Send ACK now, if this read freed lots of space + * in our buffer. Certainly, new_window is new window. + * We can advertise it now, if it is not less than current one. + * "Lots" means "at least twice" here. */ - if((new_window && (new_window >= rcv_window_now * 2)) && - ((rcv_window_now + tp->ack.rcv_mss) <= tp->window_clamp)) + if(new_window && new_window >= 2*rcv_window_now) time_to_ack = 1; } if (time_to_ack) @@ -1408,11 +1432,6 @@ do_prequeue: copied += chunk; } } -#ifdef CONFIG_TCP_MORE_COARSE_ACKS - if (tp->ack.pending && - (tp->rcv_nxt - tp->rcv_wup) > tp->ack.rcv_mss) - tcp_send_ack(sk); -#endif } continue; @@ -1472,7 +1491,7 @@ do_prequeue: skb->used = 1; tcp_eat_skb(sk, skb); -#ifdef CONFIG_TCP_LESS_COARSE_ACKS +#ifdef TCP_LESS_COARSE_ACKS /* Possible improvement. When sender is faster than receiver, * traffic looks like: fill window ... wait for window open ... * fill window. We lose at least one rtt, because call |