Add code to remember where in the send queue of mbufs the last packet was
sent from. This change avoid a linear search through all mbufs when using large TCP windows, and therefore permit high-speed connections on long distances. Tested on a 1 Gigabit connection between Luleå and San Francisco, a distance of about 15000km. With TCP windows of just over 20 Mbytes it could keep up with 950Mbit/s. After discussions with Matt Thomas and Jason Thorpe.
This commit is contained in:
ragge
parent
d189d63466
commit
679db94879
|
|
@ -1,4 +1,4 @@
|
|||
/* $NetBSD: tcp_input.c,v 1.169 2003/06/15 02:49:33 matt Exp $ */
|
||||
/* $NetBSD: tcp_input.c,v 1.170 2003/06/29 18:58:26 ragge Exp $ */
|
||||
|
||||
/*
|
||||
* Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
|
||||
|
|
@ -152,7 +152,7 @@
|
|||
*/
|
||||
|
||||
#include <sys/cdefs.h>
|
||||
__KERNEL_RCSID(0, "$NetBSD: tcp_input.c,v 1.169 2003/06/15 02:49:33 matt Exp $");
|
||||
__KERNEL_RCSID(0, "$NetBSD: tcp_input.c,v 1.170 2003/06/29 18:58:26 ragge Exp $");
|
||||
|
||||
#include "opt_inet.h"
|
||||
#include "opt_ipsec.h"
|
||||
|
|
@ -1457,6 +1457,9 @@ after_listen:
|
|||
tcpstat.tcps_rcvackbyte += acked;
|
||||
ND6_HINT(tp);
|
||||
sbdrop(&so->so_snd, acked);
|
||||
if (tp->t_lastm != NULL)
|
||||
tp->t_lastoff -= acked;
|
||||
|
||||
/*
|
||||
* We want snd_recover to track snd_una to
|
||||
* avoid sequence wraparound problems for
|
||||
|
|
@ -2069,11 +2072,15 @@ after_listen:
|
|||
ND6_HINT(tp);
|
||||
if (acked > so->so_snd.sb_cc) {
|
||||
tp->snd_wnd -= so->so_snd.sb_cc;
|
||||
tp->t_lastm = NULL;
|
||||
sbdrop(&so->so_snd, (int)so->so_snd.sb_cc);
|
||||
ourfinisacked = 1;
|
||||
} else {
|
||||
sbdrop(&so->so_snd, acked);
|
||||
tp->snd_wnd -= acked;
|
||||
tp->t_lastoff -= acked;
|
||||
if (tp->t_lastoff <= 0)
|
||||
tp->t_lastm = NULL;
|
||||
ourfinisacked = 0;
|
||||
}
|
||||
sowwakeup(so);
|
||||
|
|
|
|||
|
|
@ -1,4 +1,4 @@
|
|||
/* $NetBSD: tcp_output.c,v 1.91 2003/05/17 17:16:20 itojun Exp $ */
|
||||
/* $NetBSD: tcp_output.c,v 1.92 2003/06/29 18:58:27 ragge Exp $ */
|
||||
|
||||
/*
|
||||
* Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
|
||||
|
|
@ -142,7 +142,7 @@
|
|||
*/
|
||||
|
||||
#include <sys/cdefs.h>
|
||||
__KERNEL_RCSID(0, "$NetBSD: tcp_output.c,v 1.91 2003/05/17 17:16:20 itojun Exp $");
|
||||
__KERNEL_RCSID(0, "$NetBSD: tcp_output.c,v 1.92 2003/06/29 18:58:27 ragge Exp $");
|
||||
|
||||
#include "opt_inet.h"
|
||||
#include "opt_ipsec.h"
|
||||
|
|
@ -446,7 +446,44 @@ tcp_build_datapkt(struct tcpcb *tp, struct socket *so, int off,
|
|||
m->m_len += len;
|
||||
TCP_OUTPUT_COUNTER_INCR(&tcp_output_copysmall);
|
||||
} else {
|
||||
m->m_next = m_copy(so->so_snd.sb_mb, off, (int) len);
|
||||
struct mbuf *m0;
|
||||
|
||||
/*
|
||||
* To avoid traversing the whole sb_mb chain for correct
|
||||
* data to send, remember last sent mbuf, its offset and
|
||||
* the sent size. When called the next time, see if the
|
||||
* data to send is the directly following the previous
|
||||
* transfer. This is important for large TCP windows.
|
||||
*/
|
||||
if (off > 8*1024) { /* Only for long chains */
|
||||
if (tp->t_lastm == NULL ||
|
||||
(tp->t_lastoff + tp->t_lastlen) != off) {
|
||||
/* Prediction failed */
|
||||
tp->t_lastm = so->so_snd.sb_mb;
|
||||
tp->t_inoff = off;
|
||||
} else {
|
||||
tp->t_inoff += tp->t_lastlen;
|
||||
tp->t_lastoff = off - tp->t_lastoff;
|
||||
}
|
||||
|
||||
/* Traverse forward to next packet */
|
||||
while (tp->t_inoff > 0) {
|
||||
if (tp->t_lastm == NULL)
|
||||
panic("tp->t_lastm == NULL");
|
||||
if (tp->t_inoff < tp->t_lastm->m_len)
|
||||
break;
|
||||
tp->t_inoff -= tp->t_lastm->m_len;
|
||||
tp->t_lastm = tp->t_lastm->m_next;
|
||||
}
|
||||
|
||||
tp->t_lastoff = off;
|
||||
tp->t_lastlen = len;
|
||||
m0 = tp->t_lastm;
|
||||
off = tp->t_inoff;
|
||||
} else
|
||||
m0 = so->so_snd.sb_mb;
|
||||
|
||||
m->m_next = m_copy(m0, off, (int) len);
|
||||
if (m->m_next == NULL) {
|
||||
m_freem(m);
|
||||
return (ENOBUFS);
|
||||
|
|
|
|||
|
|
@ -1,4 +1,4 @@
|
|||
/* $NetBSD: tcp_subr.c,v 1.140 2003/06/23 11:02:15 martin Exp $ */
|
||||
/* $NetBSD: tcp_subr.c,v 1.141 2003/06/29 18:58:27 ragge Exp $ */
|
||||
|
||||
/*
|
||||
* Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
|
||||
|
|
@ -102,7 +102,7 @@
|
|||
*/
|
||||
|
||||
#include <sys/cdefs.h>
|
||||
__KERNEL_RCSID(0, "$NetBSD: tcp_subr.c,v 1.140 2003/06/23 11:02:15 martin Exp $");
|
||||
__KERNEL_RCSID(0, "$NetBSD: tcp_subr.c,v 1.141 2003/06/29 18:58:27 ragge Exp $");
|
||||
|
||||
#include "opt_inet.h"
|
||||
#include "opt_ipsec.h"
|
||||
|
|
@ -916,6 +916,9 @@ tcp_newtcpcb(family, aux)
|
|||
tp->t_segsz = tcp_mssdflt;
|
||||
LIST_INIT(&tp->t_sc);
|
||||
|
||||
tp->t_lastm = NULL;
|
||||
tp->t_lastoff = 0;
|
||||
|
||||
callout_init(&tp->t_delack_ch);
|
||||
for (i = 0; i < TCPT_NTIMERS; i++)
|
||||
TCP_TIMER_INIT(tp, i);
|
||||
|
|
|
|||
|
|
@ -1,4 +1,4 @@
|
|||
/* $NetBSD: tcp_var.h,v 1.100 2003/06/28 14:22:09 darrenr Exp $ */
|
||||
/* $NetBSD: tcp_var.h,v 1.101 2003/06/29 18:58:28 ragge Exp $ */
|
||||
|
||||
/*
|
||||
* Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
|
||||
|
|
@ -261,6 +261,12 @@ struct tcpcb {
|
|||
|
||||
/* pointer for syn cache entries*/
|
||||
LIST_HEAD(, syn_cache) t_sc; /* list of entries by this tcb */
|
||||
|
||||
/* prediction of next mbuf when using large window sizes */
|
||||
struct mbuf *t_lastm; /* last mbuf that data was sent from */
|
||||
int t_inoff; /* data offset in previous mbuf */
|
||||
int t_lastoff; /* last data address in mbuf chain */
|
||||
int t_lastlen; /* last length read from mbuf chain */
|
||||
};
|
||||
|
||||
#ifdef _KERNEL
|
||||
|
|
|
|||
Loading…
Reference in New Issue