[apple/xnu.git] / bsd / netinet / tcp_timer.c

/*
 * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
 *
 * @APPLE_LICENSE_HEADER_START@
 * 
 * The contents of this file constitute Original Code as defined in and
 * are subject to the Apple Public Source License Version 1.1 (the
 * "License").  You may not use this file except in compliance with the
 * License.  Please obtain a copy of the License at
 * http://www.apple.com/publicsource and read it before using this file.
 * 
 * This Original Code and all software distributed under the License are
 * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
 * License for the specific language governing rights and limitations
 * under the License.
 * 
 * @APPLE_LICENSE_HEADER_END@
 */
/*
 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
 *	The Regents of the University of California.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)tcp_timer.c	8.2 (Berkeley) 5/24/95
 * $FreeBSD: src/sys/netinet/tcp_timer.c,v 1.34.2.11 2001/08/22 00:59:12 silby Exp $
 */


#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/mbuf.h>
#include <sys/sysctl.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/protosw.h>

#include <kern/cpu_number.h>	/* before tcp_seq.h, for tcp_random18() */

#include <net/route.h>

#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/in_pcb.h>
#if INET6
#include <netinet6/in6_pcb.h>
#endif
#include <netinet/ip_var.h>
#include <netinet/tcp.h>
#include <netinet/tcp_fsm.h>
#include <netinet/tcp_seq.h>
#include <netinet/tcp_timer.h>
#include <netinet/tcp_var.h>
#include <netinet/tcpip.h>
#if TCPDEBUG
#include <netinet/tcp_debug.h>
#endif
#include <sys/kdebug.h>

#define DBG_FNC_TCP_FAST	NETDBG_CODE(DBG_NETTCP, (5 << 8))
#define DBG_FNC_TCP_SLOW	NETDBG_CODE(DBG_NETTCP, (5 << 8) | 1)

/*
 * NOTE - WARNING
 *
 *
 * 
 *
 */
static int
sysctl_msec_to_ticks SYSCTL_HANDLER_ARGS
{
	int error, s, tt;

	tt = *(int *)oidp->oid_arg1;
	s = tt * 1000 / hz;

	error = sysctl_handle_int(oidp, &s, 0, req);
	if (error || !req->newptr)
		return (error);

	tt = s * hz / 1000;
	if (tt < 1)
		return (EINVAL);

	*(int *)oidp->oid_arg1 = tt;
        return (0);
}

int	tcp_keepinit;
SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPINIT, keepinit, CTLTYPE_INT|CTLFLAG_RW,
    &tcp_keepinit, 0, sysctl_msec_to_ticks, "I", "");

int	tcp_keepidle;
SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPIDLE, keepidle, CTLTYPE_INT|CTLFLAG_RW,
    &tcp_keepidle, 0, sysctl_msec_to_ticks, "I", "");

int	tcp_keepintvl;
SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPINTVL, keepintvl, CTLTYPE_INT|CTLFLAG_RW,
    &tcp_keepintvl, 0, sysctl_msec_to_ticks, "I", "");

int	tcp_delacktime;
SYSCTL_PROC(_net_inet_tcp, TCPCTL_DELACKTIME, delacktime,
    CTLTYPE_INT|CTLFLAG_RW, &tcp_delacktime, 0, sysctl_msec_to_ticks, "I",
    "Time before a delayed ACK is sent");
 
int	tcp_msl;
SYSCTL_PROC(_net_inet_tcp, OID_AUTO, msl, CTLTYPE_INT|CTLFLAG_RW,
    &tcp_msl, 0, sysctl_msec_to_ticks, "I", "Maximum segment lifetime");

static int	always_keepalive = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, always_keepalive, CTLFLAG_RW, 
    &always_keepalive , 0, "Assume SO_KEEPALIVE on all TCP connections");

static int	tcp_keepcnt = TCPTV_KEEPCNT;
	/* max idle probes */
int	tcp_maxpersistidle;
	/* max idle time in persist */
int	tcp_maxidle;

struct	inpcbhead	time_wait_slots[N_TIME_WAIT_SLOTS];
int		cur_tw_slot = 0;

u_long		*delack_bitmask;


void	add_to_time_wait(tp) 
	struct tcpcb	*tp;
{
	int		tw_slot;

	LIST_REMOVE(tp->t_inpcb, inp_list);

	if (tp->t_timer[TCPT_2MSL] == 0) 
	    tp->t_timer[TCPT_2MSL] = 1;

	tp->t_rcvtime += tp->t_timer[TCPT_2MSL] & (N_TIME_WAIT_SLOTS - 1);
	tw_slot = (tp->t_timer[TCPT_2MSL] & (N_TIME_WAIT_SLOTS - 1)) + cur_tw_slot; 
	if (tw_slot >= N_TIME_WAIT_SLOTS)
	    tw_slot -= N_TIME_WAIT_SLOTS;

	LIST_INSERT_HEAD(&time_wait_slots[tw_slot], tp->t_inpcb, inp_list);
}


/*
 * Fast timeout routine for processing delayed acks
 */
void
tcp_fasttimo()
{
    register struct inpcb *inp;
    register struct tcpcb *tp;


    register u_long			i,j;
    register u_long			temp_mask;
    register u_long			elem_base = 0;
    struct inpcbhead			*head;
    int s = splnet();

    static
    int delack_checked = 0;

    KERNEL_DEBUG(DBG_FNC_TCP_FAST | DBG_FUNC_START, 0,0,0,0,0);

    if (!tcp_delack_enabled) 
	return;

    for (i=0; i < (tcbinfo.hashsize / 32); i++) {
	    if (delack_bitmask[i]) {
		temp_mask = 1;
		for (j=0; j < 32; j++) {
		    if (temp_mask & delack_bitmask[i]) {
			head = &tcbinfo.hashbase[elem_base + j];
			for (inp=head->lh_first; inp != 0; inp = inp->inp_hash.le_next) {
			    delack_checked++;
			    if ((tp = (struct tcpcb *)inp->inp_ppcb) && (tp->t_flags & TF_DELACK)) {
				tp->t_flags &= ~TF_DELACK;
				tp->t_flags |= TF_ACKNOW;
				tcpstat.tcps_delack++;
				(void) tcp_output(tp);
			    }
			}
		    }
		    temp_mask <<=  1;
		}
		delack_bitmask[i] = 0;
	    }
	    elem_base += 32;
    }
    KERNEL_DEBUG(DBG_FNC_TCP_FAST | DBG_FUNC_END, delack_checked,tcpstat.tcps_delack,0,0,0);
    splx(s);

}

/*
 * Tcp protocol timeout routine called every 500 ms.
 * Updates the timers in all active tcb's and
 * causes finite state machine actions if timers expire.
 */
void
tcp_slowtimo()
{
	register struct inpcb *ip, *ipnxt;
	register struct tcpcb *tp;
	register int i;
	int s;
#if TCPDEBUG
	int ostate;
#endif
#if KDEBUG
	static int tws_checked;
#endif

	KERNEL_DEBUG(DBG_FNC_TCP_SLOW | DBG_FUNC_START, 0,0,0,0,0);
	s = splnet();

	tcp_maxidle = tcp_keepcnt * tcp_keepintvl;

	ip = tcb.lh_first;
	if (ip == NULL) {
		splx(s);
		return;
	}
	/*
	 * Search through tcb's and update active timers.
	 */
	for (; ip != NULL; ip = ipnxt) {
		ipnxt = ip->inp_list.le_next;
		tp = intotcpcb(ip);
		if (tp == 0 || tp->t_state == TCPS_LISTEN)
			continue;
		/*
 		 * Bogus state when port owned by SharedIP with loopback as the
		 * only configured interface: BlueBox does not filters loopback
		 */
		if (tp->t_state == TCP_NSTATES)
			continue;      

		for (i = 0; i < TCPT_NTIMERS; i++) {
			if (tp->t_timer[i] && --tp->t_timer[i] == 0) {
#if TCPDEBUG
				ostate = tp->t_state;
#endif
				tp = tcp_timers(tp, i);
				if (tp == NULL)
					goto tpgone;
#if TCPDEBUG
				if (tp->t_inpcb->inp_socket->so_options
				    & SO_DEBUG)
					tcp_trace(TA_USER, ostate, tp,
						  (void *)0,
						  (struct tcphdr *)0,
						  PRU_SLOWTIMO);
#endif
			}
		}
		tp->t_rcvtime++;
		tp->t_starttime++;
		if (tp->t_rtttime)
			tp->t_rtttime++;
tpgone:
		;
	}

#if KDEBUG
	tws_checked = 0;
#endif
	KERNEL_DEBUG(DBG_FNC_TCP_SLOW | DBG_FUNC_NONE, tws_checked,0,0,0,0);

	/*
	 * Process the items in the current time-wait slot
	 */

	for (ip = time_wait_slots[cur_tw_slot].lh_first; ip; ip = ipnxt)
	{
#if KDEBUG
	        tws_checked++;
#endif
		ipnxt = ip->inp_list.le_next;
		tp = intotcpcb(ip);
		if (tp == NULL) { /* tp already closed, remove from list */
			LIST_REMOVE(ip, inp_list);
			continue; 
		}
		if (tp->t_timer[TCPT_2MSL] >= N_TIME_WAIT_SLOTS) {
		    tp->t_timer[TCPT_2MSL] -= N_TIME_WAIT_SLOTS;
		    tp->t_rcvtime += N_TIME_WAIT_SLOTS;
		}
		else
		    tp->t_timer[TCPT_2MSL] = 0;

		if (tp->t_timer[TCPT_2MSL] == 0) 
		    tp = tcp_timers(tp, TCPT_2MSL);
	}

	if (++cur_tw_slot >= N_TIME_WAIT_SLOTS)
		cur_tw_slot = 0;
	tcp_now++;					/* for timestamps */
	splx(s);
	KERNEL_DEBUG(DBG_FNC_TCP_SLOW | DBG_FUNC_END, tws_checked, cur_tw_slot,0,0,0);
}

/*
 * Cancel all timers for TCP tp.
 */
void
tcp_canceltimers(tp)
	struct tcpcb *tp;
{
	register int i;

	for (i = 0; i < TCPT_NTIMERS; i++)
		tp->t_timer[i] = 0;
}

int	tcp_syn_backoff[TCP_MAXRXTSHIFT + 1] =
    { 1, 1, 1, 1, 1, 2, 4, 8, 16, 32, 64, 64, 64 };

int	tcp_backoff[TCP_MAXRXTSHIFT + 1] =
    { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };

static int tcp_totbackoff = 511;	/* sum of tcp_backoff[] */

/*
 * TCP timer processing.
 */
struct tcpcb *
tcp_timers(tp, timer)
	register struct tcpcb *tp;
	int timer;
{
	register int rexmt;
	struct socket *so_tmp;
	struct tcptemp *t_template;

#if TCPDEBUG
	int ostate;
#endif

#if INET6
	int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV4) == 0;
#endif /* INET6 */


	switch (timer) {

	/*
	 * 2 MSL timeout in shutdown went off.  If we're closed but
	 * still waiting for peer to close and connection has been idle
	 * too long, or if 2MSL time is up from TIME_WAIT, delete connection
	 * control block.  Otherwise, check again in a bit.
	 */
	case TCPT_2MSL:
		if (tp->t_state != TCPS_TIME_WAIT &&
		    tp->t_rcvtime <= tcp_maxidle) {
			tp->t_timer[TCPT_2MSL] = tcp_keepintvl;
			add_to_time_wait(tp);
		}
		else
			tp = tcp_close(tp);
		break;

	/*
	 * Retransmission timer went off.  Message has not
	 * been acked within retransmit interval.  Back off
	 * to a longer retransmit interval and retransmit one segment.
	 */
	case TCPT_REXMT:
		if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
			tp->t_rxtshift = TCP_MAXRXTSHIFT;
			tcpstat.tcps_timeoutdrop++;
			so_tmp = tp->t_inpcb->inp_socket;
			tp = tcp_drop(tp, tp->t_softerror ?
			    tp->t_softerror : ETIMEDOUT);
			postevent(so_tmp, 0, EV_TIMEOUT);			
			break;
		}

		if (tp->t_rxtshift == 1) {
			/*
			 * first retransmit; record ssthresh and cwnd so they can
			 * be recovered if this turns out to be a "bad" retransmit.
			 * A retransmit is considered "bad" if an ACK for this 
			 * segment is received within RTT/2 interval; the assumption
			 * here is that the ACK was already in flight.  See 
			 * "On Estimating End-to-End Network Path Properties" by
			 * Allman and Paxson for more details.
			 */
			tp->snd_cwnd_prev = tp->snd_cwnd;
			tp->snd_ssthresh_prev = tp->snd_ssthresh;
			tp->t_badrxtwin = tcp_now + (tp->t_srtt >> (TCP_RTT_SHIFT + 1));
		}
		tcpstat.tcps_rexmttimeo++;
		if (tp->t_state == TCPS_SYN_SENT)
			rexmt = TCP_REXMTVAL(tp) * tcp_syn_backoff[tp->t_rxtshift];
		else
			rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
		TCPT_RANGESET(tp->t_rxtcur, rexmt,
			tp->t_rttmin, TCPTV_REXMTMAX);
		tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;

		/*
		 * Disable rfc1323 and rfc1644 if we havn't got any response to
		 * our third SYN to work-around some broken terminal servers 
		 * (most of which have hopefully been retired) that have bad VJ 
		 * header compression code which trashes TCP segments containing 
		 * unknown-to-them TCP options.
		 */
		if ((tp->t_state == TCPS_SYN_SENT) && (tp->t_rxtshift == 3))
				tp->t_flags &= ~(TF_REQ_SCALE|TF_REQ_TSTMP|TF_REQ_CC);
		/*
		 * If losing, let the lower level know and try for
		 * a better route.  Also, if we backed off this far,
		 * our srtt estimate is probably bogus.  Clobber it
		 * so we'll take the next rtt measurement as our srtt;
		 * move the current srtt into rttvar to keep the current
		 * retransmit times until then.
		 */
		if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
#if INET6
			if (isipv6)
				in6_losing(tp->t_inpcb);
			else
#endif /* INET6 */
			in_losing(tp->t_inpcb);
			tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
			tp->t_srtt = 0;
		}
		tp->snd_nxt = tp->snd_una;
		/*
		 * Note:  We overload snd_recover to function also as the
		 * snd_last variable described in RFC 2582
		 */
		tp->snd_recover = tp->snd_max;
		/*
		 * Force a segment to be sent.
		 */
		tp->t_flags |= TF_ACKNOW;
		/*
		 * If timing a segment in this window, stop the timer.
		 */
		tp->t_rtttime = 0;
		/*
		 * Close the congestion window down to one segment
		 * (we'll open it by one segment for each ack we get).
		 * Since we probably have a window's worth of unacked
		 * data accumulated, this "slow start" keeps us from
		 * dumping all that data as back-to-back packets (which
		 * might overwhelm an intermediate gateway).
		 *
		 * There are two phases to the opening: Initially we
		 * open by one mss on each ack.  This makes the window
		 * size increase exponentially with time.  If the
		 * window is larger than the path can handle, this
		 * exponential growth results in dropped packet(s)
		 * almost immediately.  To get more time between
		 * drops but still "push" the network to take advantage
		 * of improving conditions, we switch from exponential
		 * to linear window opening at some threshhold size.
		 * For a threshhold, we use half the current window
		 * size, truncated to a multiple of the mss.
		 *
		 * (the minimum cwnd that will give us exponential
		 * growth is 2 mss.  We don't allow the threshhold
		 * to go below this.)
		 */
		{
		u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
		if (win < 2)
			win = 2;
		tp->snd_cwnd = tp->t_maxseg;
		tp->snd_ssthresh = win * tp->t_maxseg;
		tp->t_dupacks = 0;
		}
		(void) tcp_output(tp);
		break;

	/*
	 * Persistance timer into zero window.
	 * Force a byte to be output, if possible.
	 */
	case TCPT_PERSIST:
		tcpstat.tcps_persisttimeo++;
		/*
		 * Hack: if the peer is dead/unreachable, we do not
		 * time out if the window is closed.  After a full
		 * backoff, drop the connection if the idle time
		 * (no responses to probes) reaches the maximum
		 * backoff that we would use if retransmitting.
		 */
		if (tp->t_rxtshift == TCP_MAXRXTSHIFT &&
		    (tp->t_rcvtime >= tcp_maxpersistidle ||
		    tp->t_rcvtime >= TCP_REXMTVAL(tp) * tcp_totbackoff)) {
			tcpstat.tcps_persistdrop++;
			so_tmp = tp->t_inpcb->inp_socket;
			tp = tcp_drop(tp, ETIMEDOUT);
			postevent(so_tmp, 0, EV_TIMEOUT);
			break;
		}
		tcp_setpersist(tp);
		tp->t_force = 1;
		(void) tcp_output(tp);
		tp->t_force = 0;
		break;

	/*
	 * Keep-alive timer went off; send something
	 * or drop connection if idle for too long.
	 */
	case TCPT_KEEP:
		tcpstat.tcps_keeptimeo++;
		if (tp->t_state < TCPS_ESTABLISHED)
			goto dropit;
		if ((always_keepalive ||
		    tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE) &&
		    tp->t_state <= TCPS_CLOSING) {
		    	if (tp->t_rcvtime >= TCP_KEEPIDLE(tp) + tcp_maxidle)
				goto dropit;
			/*
			 * Send a packet designed to force a response
			 * if the peer is up and reachable:
			 * either an ACK if the connection is still alive,
			 * or an RST if the peer has closed the connection
			 * due to timeout or reboot.
			 * Using sequence number tp->snd_una-1
			 * causes the transmitted zero-length segment
			 * to lie outside the receive window;
			 * by the protocol spec, this requires the
			 * correspondent TCP to respond.
			 */
			tcpstat.tcps_keepprobe++;
			t_template = tcp_maketemplate(tp);
			if (t_template) {
				tcp_respond(tp, t_template->tt_ipgen,
				    &t_template->tt_t, (struct mbuf *)NULL,
				    tp->rcv_nxt, tp->snd_una - 1, 0);
				(void) m_free(dtom(t_template));
			}
			tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
		} else
			tp->t_timer[TCPT_KEEP] = TCP_KEEPIDLE(tp);
		break;

#if TCPDEBUG
	if (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)
		tcp_trace(TA_USER, ostate, tp, (void *)0, (struct tcphdr *)0,
			  PRU_SLOWTIMO);
#endif
	dropit:
		tcpstat.tcps_keepdrops++;
		so_tmp = tp->t_inpcb->inp_socket;
		tp = tcp_drop(tp, ETIMEDOUT);
		postevent(so_tmp, 0, EV_TIMEOUT);
		break;
	}
	return (tp);
}
Commit	Line	Data
1c79356b A	1	/*
	2	* Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
	3	*
	4	* @APPLE_LICENSE_HEADER_START@
	5	*
e5568f75 A	6	* The contents of this file constitute Original Code as defined in and
	7	* are subject to the Apple Public Source License Version 1.1 (the
	8	* "License"). You may not use this file except in compliance with the
	9	* License. Please obtain a copy of the License at
	10	* http://www.apple.com/publicsource and read it before using this file.
1c79356b	11	*
e5568f75 A	12	* This Original Code and all software distributed under the License are
e5568f75 A	13	* distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
1c79356b A	14	* EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
1c79356b A	15	* INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
e5568f75 A	16	* FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the
	17	* License for the specific language governing rights and limitations
	18	* under the License.
1c79356b A	19	*
	20	* @APPLE_LICENSE_HEADER_END@
	21	*/
	22	/*
	23	* Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
	24	* The Regents of the University of California. All rights reserved.
	25	*
	26	* Redistribution and use in source and binary forms, with or without
	27	* modification, are permitted provided that the following conditions
	28	* are met:
	29	* 1. Redistributions of source code must retain the above copyright
	30	* notice, this list of conditions and the following disclaimer.
	31	* 2. Redistributions in binary form must reproduce the above copyright
	32	* notice, this list of conditions and the following disclaimer in the
	33	* documentation and/or other materials provided with the distribution.
	34	* 3. All advertising materials mentioning features or use of this software
	35	* must display the following acknowledgement:
	36	* This product includes software developed by the University of
	37	* California, Berkeley and its contributors.
	38	* 4. Neither the name of the University nor the names of its contributors
	39	* may be used to endorse or promote products derived from this software
	40	* without specific prior written permission.
	41	*
	42	* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
	43	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	44	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	45	* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
	46	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	47	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	48	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	49	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	50	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	51	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	52	* SUCH DAMAGE.
	53	*
	54	* @(#)tcp_timer.c 8.2 (Berkeley) 5/24/95
9bccf70c	55	* $FreeBSD: src/sys/netinet/tcp_timer.c,v 1.34.2.11 2001/08/22 00:59:12 silby Exp $
1c79356b A	56	*/
1c79356b A	57
1c79356b A	58
	59	#include <sys/param.h>
	60	#include <sys/systm.h>
	61	#include <sys/kernel.h>
9bccf70c	62	#include <sys/mbuf.h>
1c79356b A	63	#include <sys/sysctl.h>
	64	#include <sys/socket.h>
	65	#include <sys/socketvar.h>
	66	#include <sys/protosw.h>
	67
	68	#include <kern/cpu_number.h> /* before tcp_seq.h, for tcp_random18() */
	69
	70	#include <net/route.h>
	71
	72	#include <netinet/in.h>
	73	#include <netinet/in_systm.h>
1c79356b	74	#include <netinet/in_pcb.h>
9bccf70c A	75	#if INET6
	76	#include <netinet6/in6_pcb.h>
	77	#endif
1c79356b A	78	#include <netinet/ip_var.h>
	79	#include <netinet/tcp.h>
	80	#include <netinet/tcp_fsm.h>
	81	#include <netinet/tcp_seq.h>
	82	#include <netinet/tcp_timer.h>
	83	#include <netinet/tcp_var.h>
	84	#include <netinet/tcpip.h>
	85	#if TCPDEBUG
	86	#include <netinet/tcp_debug.h>
	87	#endif
	88	#include <sys/kdebug.h>
	89
	90	#define DBG_FNC_TCP_FAST NETDBG_CODE(DBG_NETTCP, (5 << 8))
	91	#define DBG_FNC_TCP_SLOW NETDBG_CODE(DBG_NETTCP, (5 << 8) \| 1)
	92
55e303ae A	93	/*
	94	* NOTE - WARNING
	95	*
	96	*
	97	*
	98	*
	99	*/
9bccf70c A	100	static int
	101	sysctl_msec_to_ticks SYSCTL_HANDLER_ARGS
	102	{
	103	int error, s, tt;
1c79356b	104
9bccf70c A	105	tt = (int )oidp->oid_arg1;
9bccf70c A	106	s = tt * 1000 / hz;
1c79356b	107
9bccf70c A	108	error = sysctl_handle_int(oidp, &s, 0, req);
	109	if (error \|\| !req->newptr)
	110	return (error);
	111
	112	tt = s * hz / 1000;
	113	if (tt < 1)
	114	return (EINVAL);
	115
	116	(int )oidp->oid_arg1 = tt;
	117	return (0);
	118	}
1c79356b	119
9bccf70c A	120	int tcp_keepinit;
	121	SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPINIT, keepinit, CTLTYPE_INT\|CTLFLAG_RW,
	122	&tcp_keepinit, 0, sysctl_msec_to_ticks, "I", "");
	123
	124	int tcp_keepidle;
	125	SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPIDLE, keepidle, CTLTYPE_INT\|CTLFLAG_RW,
	126	&tcp_keepidle, 0, sysctl_msec_to_ticks, "I", "");
	127
	128	int tcp_keepintvl;
	129	SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPINTVL, keepintvl, CTLTYPE_INT\|CTLFLAG_RW,
	130	&tcp_keepintvl, 0, sysctl_msec_to_ticks, "I", "");
	131
	132	int tcp_delacktime;
	133	SYSCTL_PROC(_net_inet_tcp, TCPCTL_DELACKTIME, delacktime,
	134	CTLTYPE_INT\|CTLFLAG_RW, &tcp_delacktime, 0, sysctl_msec_to_ticks, "I",
	135	"Time before a delayed ACK is sent");
	136
	137	int tcp_msl;
	138	SYSCTL_PROC(_net_inet_tcp, OID_AUTO, msl, CTLTYPE_INT\|CTLFLAG_RW,
	139	&tcp_msl, 0, sysctl_msec_to_ticks, "I", "Maximum segment lifetime");
1c79356b A	140
1c79356b A	141	static int always_keepalive = 0;
9bccf70c A	142	SYSCTL_INT(_net_inet_tcp, OID_AUTO, always_keepalive, CTLFLAG_RW,
9bccf70c A	143	&always_keepalive , 0, "Assume SO_KEEPALIVE on all TCP connections");
1c79356b A	144
	145	static int tcp_keepcnt = TCPTV_KEEPCNT;
	146	/* max idle probes */
9bccf70c	147	int tcp_maxpersistidle;
1c79356b A	148	/* max idle time in persist */
	149	int tcp_maxidle;
	150
1c79356b A	151	struct inpcbhead time_wait_slots[N_TIME_WAIT_SLOTS];
	152	int cur_tw_slot = 0;
	153
	154	u_long *delack_bitmask;
1c79356b A	155
	156
	157	void add_to_time_wait(tp)
	158	struct tcpcb *tp;
	159	{
	160	int tw_slot;
	161
	162	LIST_REMOVE(tp->t_inpcb, inp_list);
	163
	164	if (tp->t_timer[TCPT_2MSL] == 0)
	165	tp->t_timer[TCPT_2MSL] = 1;
	166
9bccf70c	167	tp->t_rcvtime += tp->t_timer[TCPT_2MSL] & (N_TIME_WAIT_SLOTS - 1);
1c79356b A	168	tw_slot = (tp->t_timer[TCPT_2MSL] & (N_TIME_WAIT_SLOTS - 1)) + cur_tw_slot;
	169	if (tw_slot >= N_TIME_WAIT_SLOTS)
	170	tw_slot -= N_TIME_WAIT_SLOTS;
	171
	172	LIST_INSERT_HEAD(&time_wait_slots[tw_slot], tp->t_inpcb, inp_list);
	173	}
	174
	175
	176
	177
	178
	179	/*
	180	* Fast timeout routine for processing delayed acks
	181	*/
	182	void
	183	tcp_fasttimo()
	184	{
	185	register struct inpcb *inp;
	186	register struct tcpcb *tp;
	187
	188
	189	register u_long i,j;
	190	register u_long temp_mask;
	191	register u_long elem_base = 0;
	192	struct inpcbhead *head;
	193	int s = splnet();
	194
	195	static
	196	int delack_checked = 0;
	197
	198	KERNEL_DEBUG(DBG_FNC_TCP_FAST \| DBG_FUNC_START, 0,0,0,0,0);
	199
	200	if (!tcp_delack_enabled)
	201	return;
	202
9bccf70c	203	for (i=0; i < (tcbinfo.hashsize / 32); i++) {
1c79356b A	204	if (delack_bitmask[i]) {
	205	temp_mask = 1;
	206	for (j=0; j < 32; j++) {
	207	if (temp_mask & delack_bitmask[i]) {
	208	head = &tcbinfo.hashbase[elem_base + j];
	209	for (inp=head->lh_first; inp != 0; inp = inp->inp_hash.le_next) {
	210	delack_checked++;
	211	if ((tp = (struct tcpcb *)inp->inp_ppcb) && (tp->t_flags & TF_DELACK)) {
	212	tp->t_flags &= ~TF_DELACK;
	213	tp->t_flags \|= TF_ACKNOW;
	214	tcpstat.tcps_delack++;
	215	(void) tcp_output(tp);
	216	}
	217	}
	218	}
	219	temp_mask <<= 1;
	220	}
	221	delack_bitmask[i] = 0;
	222	}
	223	elem_base += 32;
1c79356b	224	}
1c79356b A	225	KERNEL_DEBUG(DBG_FNC_TCP_FAST \| DBG_FUNC_END, delack_checked,tcpstat.tcps_delack,0,0,0);
	226	splx(s);
	227
	228	}
	229
	230	/*
	231	* Tcp protocol timeout routine called every 500 ms.
	232	* Updates the timers in all active tcb's and
	233	* causes finite state machine actions if timers expire.
	234	*/
	235	void
	236	tcp_slowtimo()
	237	{
	238	register struct inpcb ip, ipnxt;
	239	register struct tcpcb *tp;
	240	register int i;
	241	int s;
	242	#if TCPDEBUG
	243	int ostate;
	244	#endif
	245	#if KDEBUG
	246	static int tws_checked;
	247	#endif
	248
	249	KERNEL_DEBUG(DBG_FNC_TCP_SLOW \| DBG_FUNC_START, 0,0,0,0,0);
	250	s = splnet();
	251
	252	tcp_maxidle = tcp_keepcnt * tcp_keepintvl;
	253
	254	ip = tcb.lh_first;
	255	if (ip == NULL) {
	256	splx(s);
	257	return;
	258	}
	259	/*
	260	* Search through tcb's and update active timers.
	261	*/
	262	for (; ip != NULL; ip = ipnxt) {
	263	ipnxt = ip->inp_list.le_next;
	264	tp = intotcpcb(ip);
	265	if (tp == 0 \|\| tp->t_state == TCPS_LISTEN)
	266	continue;
9bccf70c A	267	/*
	268	* Bogus state when port owned by SharedIP with loopback as the
	269	* only configured interface: BlueBox does not filters loopback
	270	*/
	271	if (tp->t_state == TCP_NSTATES)
	272	continue;
	273
1c79356b A	274	for (i = 0; i < TCPT_NTIMERS; i++) {
	275	if (tp->t_timer[i] && --tp->t_timer[i] == 0) {
	276	#if TCPDEBUG
	277	ostate = tp->t_state;
	278	#endif
	279	tp = tcp_timers(tp, i);
	280	if (tp == NULL)
	281	goto tpgone;
	282	#if TCPDEBUG
	283	if (tp->t_inpcb->inp_socket->so_options
	284	& SO_DEBUG)
	285	tcp_trace(TA_USER, ostate, tp,
	286	(void *)0,
	287	(struct tcphdr *)0,
	288	PRU_SLOWTIMO);
	289	#endif
	290	}
	291	}
9bccf70c A	292	tp->t_rcvtime++;
	293	tp->t_starttime++;
	294	if (tp->t_rtttime)
	295	tp->t_rtttime++;
1c79356b A	296	tpgone:
	297	;
	298	}
	299
	300	#if KDEBUG
	301	tws_checked = 0;
	302	#endif
	303	KERNEL_DEBUG(DBG_FNC_TCP_SLOW \| DBG_FUNC_NONE, tws_checked,0,0,0,0);
	304
	305	/*
	306	* Process the items in the current time-wait slot
	307	*/
	308
	309	for (ip = time_wait_slots[cur_tw_slot].lh_first; ip; ip = ipnxt)
	310	{
	311	#if KDEBUG
	312	tws_checked++;
	313	#endif
	314	ipnxt = ip->inp_list.le_next;
	315	tp = intotcpcb(ip);
ab86ba33 A	316	if (tp == NULL) { /* tp already closed, remove from list */
	317	LIST_REMOVE(ip, inp_list);
	318	continue;
	319	}
1c79356b A	320	if (tp->t_timer[TCPT_2MSL] >= N_TIME_WAIT_SLOTS) {
1c79356b A	321	tp->t_timer[TCPT_2MSL] -= N_TIME_WAIT_SLOTS;
9bccf70c	322	tp->t_rcvtime += N_TIME_WAIT_SLOTS;
1c79356b A	323	}
	324	else
	325	tp->t_timer[TCPT_2MSL] = 0;
	326
	327	if (tp->t_timer[TCPT_2MSL] == 0)
	328	tp = tcp_timers(tp, TCPT_2MSL);
	329	}
	330
	331	if (++cur_tw_slot >= N_TIME_WAIT_SLOTS)
	332	cur_tw_slot = 0;
1c79356b A	333	tcp_now++; /* for timestamps */
	334	splx(s);
	335	KERNEL_DEBUG(DBG_FNC_TCP_SLOW \| DBG_FUNC_END, tws_checked, cur_tw_slot,0,0,0);
	336	}
	337
	338	/*
	339	* Cancel all timers for TCP tp.
	340	*/
	341	void
	342	tcp_canceltimers(tp)
	343	struct tcpcb *tp;
	344	{
	345	register int i;
	346
	347	for (i = 0; i < TCPT_NTIMERS; i++)
	348	tp->t_timer[i] = 0;
	349	}
	350
9bccf70c A	351	int tcp_syn_backoff[TCP_MAXRXTSHIFT + 1] =
	352	{ 1, 1, 1, 1, 1, 2, 4, 8, 16, 32, 64, 64, 64 };
	353
1c79356b A	354	int tcp_backoff[TCP_MAXRXTSHIFT + 1] =
	355	{ 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
	356
	357	static int tcp_totbackoff = 511; /* sum of tcp_backoff[] */
	358
	359	/*
	360	* TCP timer processing.
	361	*/
	362	struct tcpcb *
	363	tcp_timers(tp, timer)
	364	register struct tcpcb *tp;
	365	int timer;
	366	{
	367	register int rexmt;
	368	struct socket *so_tmp;
9bccf70c A	369	struct tcptemp *t_template;
9bccf70c A	370
55e303ae A	371	#if TCPDEBUG
	372	int ostate;
	373	#endif
	374
1c79356b A	375	#if INET6
	376	int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV4) == 0;
	377	#endif /* INET6 */
	378
9bccf70c	379
1c79356b A	380	switch (timer) {
	381
	382	/*
	383	* 2 MSL timeout in shutdown went off. If we're closed but
	384	* still waiting for peer to close and connection has been idle
	385	* too long, or if 2MSL time is up from TIME_WAIT, delete connection
	386	* control block. Otherwise, check again in a bit.
	387	*/
	388	case TCPT_2MSL:
	389	if (tp->t_state != TCPS_TIME_WAIT &&
9bccf70c	390	tp->t_rcvtime <= tcp_maxidle) {
1c79356b A	391	tp->t_timer[TCPT_2MSL] = tcp_keepintvl;
	392	add_to_time_wait(tp);
	393	}
	394	else
	395	tp = tcp_close(tp);
	396	break;
	397
	398	/*
	399	* Retransmission timer went off. Message has not
	400	* been acked within retransmit interval. Back off
	401	* to a longer retransmit interval and retransmit one segment.
	402	*/
	403	case TCPT_REXMT:
	404	if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
	405	tp->t_rxtshift = TCP_MAXRXTSHIFT;
	406	tcpstat.tcps_timeoutdrop++;
	407	so_tmp = tp->t_inpcb->inp_socket;
	408	tp = tcp_drop(tp, tp->t_softerror ?
	409	tp->t_softerror : ETIMEDOUT);
	410	postevent(so_tmp, 0, EV_TIMEOUT);
	411	break;
	412	}
9bccf70c A	413
	414	if (tp->t_rxtshift == 1) {
	415	/*
	416	* first retransmit; record ssthresh and cwnd so they can
	417	* be recovered if this turns out to be a "bad" retransmit.
	418	* A retransmit is considered "bad" if an ACK for this
	419	* segment is received within RTT/2 interval; the assumption
	420	* here is that the ACK was already in flight. See
	421	* "On Estimating End-to-End Network Path Properties" by
	422	* Allman and Paxson for more details.
	423	*/
	424	tp->snd_cwnd_prev = tp->snd_cwnd;
	425	tp->snd_ssthresh_prev = tp->snd_ssthresh;
	426	tp->t_badrxtwin = tcp_now + (tp->t_srtt >> (TCP_RTT_SHIFT + 1));
	427	}
1c79356b	428	tcpstat.tcps_rexmttimeo++;
9bccf70c A	429	if (tp->t_state == TCPS_SYN_SENT)
	430	rexmt = TCP_REXMTVAL(tp) * tcp_syn_backoff[tp->t_rxtshift];
	431	else
	432	rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
1c79356b	433	TCPT_RANGESET(tp->t_rxtcur, rexmt,
9bccf70c	434	tp->t_rttmin, TCPTV_REXMTMAX);
1c79356b	435	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
9bccf70c A	436
	437	/*
	438	* Disable rfc1323 and rfc1644 if we havn't got any response to
	439	* our third SYN to work-around some broken terminal servers
	440	* (most of which have hopefully been retired) that have bad VJ
	441	* header compression code which trashes TCP segments containing
	442	* unknown-to-them TCP options.
	443	*/
	444	if ((tp->t_state == TCPS_SYN_SENT) && (tp->t_rxtshift == 3))
	445	tp->t_flags &= ~(TF_REQ_SCALE\|TF_REQ_TSTMP\|TF_REQ_CC);
1c79356b A	446	/*
	447	* If losing, let the lower level know and try for
	448	* a better route. Also, if we backed off this far,
	449	* our srtt estimate is probably bogus. Clobber it
	450	* so we'll take the next rtt measurement as our srtt;
	451	* move the current srtt into rttvar to keep the current
	452	* retransmit times until then.
	453	*/
	454	if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
	455	#if INET6
	456	if (isipv6)
	457	in6_losing(tp->t_inpcb);
	458	else
	459	#endif /* INET6 */
	460	in_losing(tp->t_inpcb);
	461	tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
	462	tp->t_srtt = 0;
	463	}
	464	tp->snd_nxt = tp->snd_una;
9bccf70c A	465	/*
	466	* Note: We overload snd_recover to function also as the
	467	* snd_last variable described in RFC 2582
	468	*/
	469	tp->snd_recover = tp->snd_max;
1c79356b A	470	/*
	471	* Force a segment to be sent.
	472	*/
	473	tp->t_flags \|= TF_ACKNOW;
	474	/*
	475	* If timing a segment in this window, stop the timer.
	476	*/
9bccf70c	477	tp->t_rtttime = 0;
1c79356b A	478	/*
	479	* Close the congestion window down to one segment
	480	* (we'll open it by one segment for each ack we get).
	481	* Since we probably have a window's worth of unacked
	482	* data accumulated, this "slow start" keeps us from
	483	* dumping all that data as back-to-back packets (which
	484	* might overwhelm an intermediate gateway).
	485	*
	486	* There are two phases to the opening: Initially we
	487	* open by one mss on each ack. This makes the window
	488	* size increase exponentially with time. If the
	489	* window is larger than the path can handle, this
	490	* exponential growth results in dropped packet(s)
	491	* almost immediately. To get more time between
	492	* drops but still "push" the network to take advantage
	493	* of improving conditions, we switch from exponential
	494	* to linear window opening at some threshhold size.
	495	* For a threshhold, we use half the current window
	496	* size, truncated to a multiple of the mss.
	497	*
	498	* (the minimum cwnd that will give us exponential
	499	* growth is 2 mss. We don't allow the threshhold
	500	* to go below this.)
	501	*/
	502	{
	503	u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
	504	if (win < 2)
	505	win = 2;
	506	tp->snd_cwnd = tp->t_maxseg;
	507	tp->snd_ssthresh = win * tp->t_maxseg;
	508	tp->t_dupacks = 0;
	509	}
	510	(void) tcp_output(tp);
	511	break;
	512
	513	/*
	514	* Persistance timer into zero window.
	515	* Force a byte to be output, if possible.
	516	*/
	517	case TCPT_PERSIST:
	518	tcpstat.tcps_persisttimeo++;
	519	/*
	520	* Hack: if the peer is dead/unreachable, we do not
	521	* time out if the window is closed. After a full
	522	* backoff, drop the connection if the idle time
	523	* (no responses to probes) reaches the maximum
	524	* backoff that we would use if retransmitting.
	525	*/
	526	if (tp->t_rxtshift == TCP_MAXRXTSHIFT &&
9bccf70c A	527	(tp->t_rcvtime >= tcp_maxpersistidle \|\|
9bccf70c A	528	tp->t_rcvtime >= TCP_REXMTVAL(tp) * tcp_totbackoff)) {
1c79356b A	529	tcpstat.tcps_persistdrop++;
	530	so_tmp = tp->t_inpcb->inp_socket;
	531	tp = tcp_drop(tp, ETIMEDOUT);
	532	postevent(so_tmp, 0, EV_TIMEOUT);
	533	break;
	534	}
	535	tcp_setpersist(tp);
	536	tp->t_force = 1;
	537	(void) tcp_output(tp);
	538	tp->t_force = 0;
	539	break;
	540
	541	/*
	542	* Keep-alive timer went off; send something
	543	* or drop connection if idle for too long.
	544	*/
	545	case TCPT_KEEP:
	546	tcpstat.tcps_keeptimeo++;
	547	if (tp->t_state < TCPS_ESTABLISHED)
	548	goto dropit;
	549	if ((always_keepalive \|\|
	550	tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE) &&
	551	tp->t_state <= TCPS_CLOSING) {
55e303ae	552	if (tp->t_rcvtime >= TCP_KEEPIDLE(tp) + tcp_maxidle)
1c79356b A	553	goto dropit;
	554	/*
	555	* Send a packet designed to force a response
	556	* if the peer is up and reachable:
	557	* either an ACK if the connection is still alive,
	558	* or an RST if the peer has closed the connection
	559	* due to timeout or reboot.
	560	* Using sequence number tp->snd_una-1
	561	* causes the transmitted zero-length segment
	562	* to lie outside the receive window;
	563	* by the protocol spec, this requires the
	564	* correspondent TCP to respond.
	565	*/
	566	tcpstat.tcps_keepprobe++;
9bccf70c A	567	t_template = tcp_maketemplate(tp);
	568	if (t_template) {
	569	tcp_respond(tp, t_template->tt_ipgen,
	570	&t_template->tt_t, (struct mbuf *)NULL,
	571	tp->rcv_nxt, tp->snd_una - 1, 0);
	572	(void) m_free(dtom(t_template));
	573	}
1c79356b A	574	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
1c79356b A	575	} else
55e303ae	576	tp->t_timer[TCPT_KEEP] = TCP_KEEPIDLE(tp);
1c79356b	577	break;
9bccf70c A	578
	579	#if TCPDEBUG
	580	if (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)
	581	tcp_trace(TA_USER, ostate, tp, (void )0, (struct tcphdr )0,
	582	PRU_SLOWTIMO);
	583	#endif
1c79356b A	584	dropit:
	585	tcpstat.tcps_keepdrops++;
	586	so_tmp = tp->t_inpcb->inp_socket;
	587	tp = tcp_drop(tp, ETIMEDOUT);
	588	postevent(so_tmp, 0, EV_TIMEOUT);
	589	break;
	590	}
	591	return (tp);
	592	}