Rework TCP window scaling (RFC1323) to properly scale the send window
right from the beginning and partly clean up the differences in handling between SYN_SENT and SYN_RCVD (syncache). Further changes to this code to come. This is a first incremental step to a general overhaul and streamlining of the TCP code. PR: kern/15095 PR: kern/92690 (partly) Reviewed by: qingli (and tested with ANVL) Sponsored by: TCP/IP Optimization Fundraise 2005
This commit is contained in:
Andre Oppermann
parent
215e7c161a
commit
464fcfbc5c
Notes:
svn2git
2020-12-20 02:59:44 +00:00
svn path=/head/; revision=156125
@ -778,12 +778,6 @@ tcp_input(m, off0)
|
||||
if (tp->t_state == TCPS_CLOSED)
|
||||
goto drop;
|
||||
|
||||
/* Unscale the window into a 32-bit value. */
|
||||
if ((thflags & TH_SYN) == 0)
|
||||
tiwin = th->th_win << tp->snd_scale;
|
||||
else
|
||||
tiwin = th->th_win;
|
||||
|
||||
#ifdef MAC
|
||||
INP_LOCK_ASSERT(inp);
|
||||
if (mac_check_inpcb_deliver(inp, m))
|
||||
@ -842,7 +836,7 @@ tcp_input(m, off0)
|
||||
/*
|
||||
* Could not complete 3-way handshake,
|
||||
* connection is being closed down, and
|
||||
* syncache will free mbuf.
|
||||
* syncache has free'd mbuf.
|
||||
*/
|
||||
INP_UNLOCK(inp);
|
||||
INP_INFO_WUNLOCK(&tcbinfo);
|
||||
@ -863,11 +857,6 @@ tcp_input(m, off0)
|
||||
tp->snd_up = tp->snd_una;
|
||||
tp->snd_max = tp->snd_nxt = tp->iss + 1;
|
||||
tp->last_ack_sent = tp->rcv_nxt;
|
||||
/*
|
||||
* RFC1323: The window in SYN & SYN/ACK
|
||||
* segments is never scaled.
|
||||
*/
|
||||
tp->snd_wnd = tiwin; /* unscaled */
|
||||
goto after_listen;
|
||||
}
|
||||
if (thflags & TH_RST) {
|
||||
@ -988,12 +977,12 @@ tcp_input(m, off0)
|
||||
}
|
||||
/*
|
||||
* Segment passed TAO tests.
|
||||
* XXX: Can't happen at the moment.
|
||||
*/
|
||||
INP_UNLOCK(inp);
|
||||
inp = sotoinpcb(so);
|
||||
INP_LOCK(inp);
|
||||
tp = intotcpcb(inp);
|
||||
tp->snd_wnd = tiwin;
|
||||
tp->t_starttime = ticks;
|
||||
tp->t_state = TCPS_ESTABLISHED;
|
||||
|
||||
@ -1008,6 +997,7 @@ tcp_input(m, off0)
|
||||
tp->t_flags |= (TF_DELACK | TF_NEEDSYN);
|
||||
else
|
||||
tp->t_flags |= (TF_ACKNOW | TF_NEEDSYN);
|
||||
tiwin = th->th_win << tp->snd_scale;
|
||||
tcpstat.tcps_connects++;
|
||||
soisconnected(so);
|
||||
goto trimthenstep6;
|
||||
@ -1087,6 +1077,13 @@ tcp_input(m, off0)
|
||||
if (TCPS_HAVEESTABLISHED(tp->t_state))
|
||||
callout_reset(tp->tt_keep, tcp_keepidle, tcp_timer_keep, tp);
|
||||
|
||||
/*
|
||||
* Unscale the window into a 32-bit value.
|
||||
* This value is bogus for the TCPS_SYN_SENT state
|
||||
* and is overwritten later.
|
||||
*/
|
||||
tiwin = th->th_win << tp->snd_scale;
|
||||
|
||||
/*
|
||||
* Process options only when we get SYN/ACK back. The SYN case
|
||||
* for incoming connections is handled in tcp_syncache.
|
||||
@ -1094,9 +1091,12 @@ tcp_input(m, off0)
|
||||
*/
|
||||
tcp_dooptions(&to, optp, optlen, thflags & TH_SYN);
|
||||
if (tp->t_state == TCPS_SYN_SENT && (thflags & TH_SYN)) {
|
||||
if (to.to_flags & TOF_SCALE) {
|
||||
if ((to.to_flags & TOF_SCALE) &&
|
||||
(tp->t_flags & TF_REQ_SCALE)) {
|
||||
tp->t_flags |= TF_RCVD_SCALE;
|
||||
tp->requested_s_scale = to.to_requested_s_scale;
|
||||
tp->snd_scale = to.to_requested_s_scale;
|
||||
tp->snd_wnd = th->th_win << tp->snd_scale;
|
||||
tiwin = tp->snd_wnd;
|
||||
}
|
||||
if (to.to_flags & TOF_TS) {
|
||||
tp->t_flags |= TF_RCVD_TSTMP;
|
||||
@ -1365,7 +1365,9 @@ tcp_input(m, off0)
|
||||
}
|
||||
if ((thflags & TH_SYN) == 0)
|
||||
goto drop;
|
||||
tp->snd_wnd = th->th_win; /* initial send window */
|
||||
|
||||
/* Initial send window, already scaled. */
|
||||
tp->snd_wnd = th->th_win;
|
||||
|
||||
tp->irs = th->th_seq;
|
||||
tcp_rcvseqinit(tp);
|
||||
@ -1380,7 +1382,6 @@ tcp_input(m, off0)
|
||||
/* Do window scaling on this connection? */
|
||||
if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
|
||||
(TF_RCVD_SCALE|TF_REQ_SCALE)) {
|
||||
tp->snd_scale = tp->requested_s_scale;
|
||||
tp->rcv_scale = tp->request_r_scale;
|
||||
}
|
||||
tp->rcv_adv += tp->rcv_wnd;
|
||||
@ -1793,8 +1794,8 @@ tcp_input(m, off0)
|
||||
/* Do window scaling? */
|
||||
if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
|
||||
(TF_RCVD_SCALE|TF_REQ_SCALE)) {
|
||||
tp->snd_scale = tp->requested_s_scale;
|
||||
tp->rcv_scale = tp->request_r_scale;
|
||||
tp->snd_wnd = tiwin;
|
||||
}
|
||||
/*
|
||||
* Make transitions:
|
||||
@ -2035,8 +2036,8 @@ tcp_input(m, off0)
|
||||
/* Do window scaling? */
|
||||
if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
|
||||
(TF_RCVD_SCALE|TF_REQ_SCALE)) {
|
||||
tp->snd_scale = tp->requested_s_scale;
|
||||
tp->rcv_scale = tp->request_r_scale;
|
||||
/* Send window already scaled. */
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@ -778,12 +778,6 @@ tcp_input(m, off0)
|
||||
if (tp->t_state == TCPS_CLOSED)
|
||||
goto drop;
|
||||
|
||||
/* Unscale the window into a 32-bit value. */
|
||||
if ((thflags & TH_SYN) == 0)
|
||||
tiwin = th->th_win << tp->snd_scale;
|
||||
else
|
||||
tiwin = th->th_win;
|
||||
|
||||
#ifdef MAC
|
||||
INP_LOCK_ASSERT(inp);
|
||||
if (mac_check_inpcb_deliver(inp, m))
|
||||
@ -842,7 +836,7 @@ tcp_input(m, off0)
|
||||
/*
|
||||
* Could not complete 3-way handshake,
|
||||
* connection is being closed down, and
|
||||
* syncache will free mbuf.
|
||||
* syncache has free'd mbuf.
|
||||
*/
|
||||
INP_UNLOCK(inp);
|
||||
INP_INFO_WUNLOCK(&tcbinfo);
|
||||
@ -863,11 +857,6 @@ tcp_input(m, off0)
|
||||
tp->snd_up = tp->snd_una;
|
||||
tp->snd_max = tp->snd_nxt = tp->iss + 1;
|
||||
tp->last_ack_sent = tp->rcv_nxt;
|
||||
/*
|
||||
* RFC1323: The window in SYN & SYN/ACK
|
||||
* segments is never scaled.
|
||||
*/
|
||||
tp->snd_wnd = tiwin; /* unscaled */
|
||||
goto after_listen;
|
||||
}
|
||||
if (thflags & TH_RST) {
|
||||
@ -988,12 +977,12 @@ tcp_input(m, off0)
|
||||
}
|
||||
/*
|
||||
* Segment passed TAO tests.
|
||||
* XXX: Can't happen at the moment.
|
||||
*/
|
||||
INP_UNLOCK(inp);
|
||||
inp = sotoinpcb(so);
|
||||
INP_LOCK(inp);
|
||||
tp = intotcpcb(inp);
|
||||
tp->snd_wnd = tiwin;
|
||||
tp->t_starttime = ticks;
|
||||
tp->t_state = TCPS_ESTABLISHED;
|
||||
|
||||
@ -1008,6 +997,7 @@ tcp_input(m, off0)
|
||||
tp->t_flags |= (TF_DELACK | TF_NEEDSYN);
|
||||
else
|
||||
tp->t_flags |= (TF_ACKNOW | TF_NEEDSYN);
|
||||
tiwin = th->th_win << tp->snd_scale;
|
||||
tcpstat.tcps_connects++;
|
||||
soisconnected(so);
|
||||
goto trimthenstep6;
|
||||
@ -1087,6 +1077,13 @@ tcp_input(m, off0)
|
||||
if (TCPS_HAVEESTABLISHED(tp->t_state))
|
||||
callout_reset(tp->tt_keep, tcp_keepidle, tcp_timer_keep, tp);
|
||||
|
||||
/*
|
||||
* Unscale the window into a 32-bit value.
|
||||
* This value is bogus for the TCPS_SYN_SENT state
|
||||
* and is overwritten later.
|
||||
*/
|
||||
tiwin = th->th_win << tp->snd_scale;
|
||||
|
||||
/*
|
||||
* Process options only when we get SYN/ACK back. The SYN case
|
||||
* for incoming connections is handled in tcp_syncache.
|
||||
@ -1094,9 +1091,12 @@ tcp_input(m, off0)
|
||||
*/
|
||||
tcp_dooptions(&to, optp, optlen, thflags & TH_SYN);
|
||||
if (tp->t_state == TCPS_SYN_SENT && (thflags & TH_SYN)) {
|
||||
if (to.to_flags & TOF_SCALE) {
|
||||
if ((to.to_flags & TOF_SCALE) &&
|
||||
(tp->t_flags & TF_REQ_SCALE)) {
|
||||
tp->t_flags |= TF_RCVD_SCALE;
|
||||
tp->requested_s_scale = to.to_requested_s_scale;
|
||||
tp->snd_scale = to.to_requested_s_scale;
|
||||
tp->snd_wnd = th->th_win << tp->snd_scale;
|
||||
tiwin = tp->snd_wnd;
|
||||
}
|
||||
if (to.to_flags & TOF_TS) {
|
||||
tp->t_flags |= TF_RCVD_TSTMP;
|
||||
@ -1365,7 +1365,9 @@ tcp_input(m, off0)
|
||||
}
|
||||
if ((thflags & TH_SYN) == 0)
|
||||
goto drop;
|
||||
tp->snd_wnd = th->th_win; /* initial send window */
|
||||
|
||||
/* Initial send window, already scaled. */
|
||||
tp->snd_wnd = th->th_win;
|
||||
|
||||
tp->irs = th->th_seq;
|
||||
tcp_rcvseqinit(tp);
|
||||
@ -1380,7 +1382,6 @@ tcp_input(m, off0)
|
||||
/* Do window scaling on this connection? */
|
||||
if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
|
||||
(TF_RCVD_SCALE|TF_REQ_SCALE)) {
|
||||
tp->snd_scale = tp->requested_s_scale;
|
||||
tp->rcv_scale = tp->request_r_scale;
|
||||
}
|
||||
tp->rcv_adv += tp->rcv_wnd;
|
||||
@ -1793,8 +1794,8 @@ tcp_input(m, off0)
|
||||
/* Do window scaling? */
|
||||
if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
|
||||
(TF_RCVD_SCALE|TF_REQ_SCALE)) {
|
||||
tp->snd_scale = tp->requested_s_scale;
|
||||
tp->rcv_scale = tp->request_r_scale;
|
||||
tp->snd_wnd = tiwin;
|
||||
}
|
||||
/*
|
||||
* Make transitions:
|
||||
@ -2035,8 +2036,8 @@ tcp_input(m, off0)
|
||||
/* Do window scaling? */
|
||||
if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
|
||||
(TF_RCVD_SCALE|TF_REQ_SCALE)) {
|
||||
tp->snd_scale = tp->requested_s_scale;
|
||||
tp->rcv_scale = tp->request_r_scale;
|
||||
/* Send window already scaled. */
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@ -682,7 +682,7 @@ syncache_socket(sc, lso, m)
|
||||
tp->t_flags |= TF_NOOPT;
|
||||
if (sc->sc_flags & SCF_WINSCALE) {
|
||||
tp->t_flags |= TF_REQ_SCALE|TF_RCVD_SCALE;
|
||||
tp->requested_s_scale = sc->sc_requested_s_scale;
|
||||
tp->snd_scale = sc->sc_requested_s_scale;
|
||||
tp->request_r_scale = sc->sc_request_r_scale;
|
||||
}
|
||||
if (sc->sc_flags & SCF_TIMESTAMP) {
|
||||
|
||||
@ -173,7 +173,7 @@ struct tcpcb {
|
||||
u_char snd_scale; /* window scaling for send window */
|
||||
u_char rcv_scale; /* window scaling for recv window */
|
||||
u_char request_r_scale; /* pending window scaling */
|
||||
u_char requested_s_scale;
|
||||
u_char requested_s_scale; /* unused, to be reused later */
|
||||
u_long ts_recent; /* timestamp echo data */
|
||||
|
||||
u_long ts_recent_age; /* when last updated */
|
||||
|
||||
Loading…
Reference in New Issue
Block a user