/*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2014 Apple Computer, Inc. All rights reserved.
*
- * @APPLE_LICENSE_HEADER_START@
+ * @APPLE_OSREFERENCE_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 file contains Original Code and/or Modifications of Original Code
+ * as defined in and that are subject to the Apple Public Source License
+ * Version 2.0 (the 'License'). You may not use this file except in
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * unlawful or unlicensed copies of an Apple operating system, or to
+ * circumvent, violate, or enable the circumvention or violation of, any
+ * terms of an Apple operating system software license agreement.
*
- * This Original Code and all software distributed under the License are
- * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
+ * Please obtain a copy of the License at
+ * http://www.opensource.apple.com/apsl/ and read it before using this file.
+ *
+ * The 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.
+ * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
+ * Please see the License for the specific language governing rights and
+ * limitations under the License.
*
- * @APPLE_LICENSE_HEADER_END@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
*/
/*
* Copyright (c) 1982, 1986, 1993
#define _NETINET_TCP_TIMER_H_
#include <sys/appleapiopts.h>
-/*
- * Definitions of the TCP timers. These timers are counted
- * down PR_SLOWHZ times a second.
- */
-#define TCPT_NTIMERS 4
+#ifdef BSD_KERNEL_PRIVATE
+#include <kern/thread_call.h>
+#endif /* BSD_KERNEL_PRIVATE */
-#define TCPT_REXMT 0 /* retransmit */
-#define TCPT_PERSIST 1 /* retransmit persistence */
-#define TCPT_KEEP 2 /* keep alive */
-#define TCPT_2MSL 3 /* 2*msl quiet time timer */
+/* Keep the external definition the same for binary compatibility */
+#define TCPT_NTIMERS_EXT 4
/*
+ * Definitions of the TCP timers.
+ *
+ * The TCPT_PTO timer is used for probing for a tail loss in a send window.
+ * If this probe gets acknowledged using SACK, it will allow the connection
+ * to enter fast-recovery instead of hitting a retransmit timeout. A probe
+ * timeout will send the last unacknowledged segment to generate more acks
+ * with SACK information which can be used for fast-retransmiting the lost
+ * packets. This will fire in the order of 10ms.
+ *
* The TCPT_REXMT timer is used to force retransmissions.
* The TCP has the TCPT_REXMT timer set whenever segments
* have been sent for which ACKs are expected but not yet
* we retransmit one unacknowledged segment, and do a backoff
* on the retransmit timer.
*
+ * The TCPT_DELACK timer is used for transmitting delayed acknowledgements
+ * if an acknowledgement was delayed in anticipation of a new segment.
+ *
* The TCPT_PERSIST timer is used to keep window size information
* flowing even if the window goes shut. If all previous transmissions
- * have been acknowledged (so that there are no retransmissions in progress),
+ * have been acknowledged(so that there are no retransmissions in progress),
* and the window is too small to bother sending anything, then we start
* the TCPT_PERSIST timer. When it expires, if the window is nonzero,
* we go to transmit state. Otherwise, at intervals send a single byte
* a window update from the peer.
*
* The TCPT_KEEP timer is used to keep connections alive. If an
- * connection is idle (no segments received) for TCPTV_KEEP_INIT amount of time,
- * but not yet established, then we drop the connection. Once the connection
- * is established, if the connection is idle for TCPTV_KEEP_IDLE time
- * (and keepalives have been enabled on the socket), we begin to probe
- * the connection. We force the peer to send us a segment by sending:
+ * connection is idle (no segments received) for TCPTV_KEEP_INIT amount
+ * of time, but not yet established, then we drop the connection.
+ * Once the connection is established, if the connection is idle for
+ * TCPTV_KEEP_IDLE time (and keepalives have been enabled on the socket),
+ * we begin to probe the connection. We force the peer to send us a
+ * segment by sending:
* <SEQ=SND.UNA-1><ACK=RCV.NXT><CTL=ACK>
* This segment is (deliberately) outside the window, and should elicit
* an ack segment in response from the peer. If, despite the TCPT_KEEP
- * initiated segments we cannot elicit a response from a peer in TCPT_MAXIDLE
- * amount of time probing, then we drop the connection.
+ * initiated segments we cannot elicit a response from a peer in
+ * TCPT_MAXIDLE amount of time probing, then we drop the connection.
+ *
+ * The TCPT_2MSL timer is used for keeping the conenction in Time-wait state
+ * before fully closing it so that the connection 4-tuple can be reused.
*/
+#ifdef BSD_KERNEL_PRIVATE
-#ifdef PRIVATE
+#define TCPT_PTO 0 /* Probe timeout */
+#define TCPT_DELAYFR 1 /* Delay recovery if there is reordering */
+#define TCPT_REXMT 2 /* retransmit */
+#define TCPT_DELACK 3 /* delayed ack */
+#define TCPT_PERSIST 4 /* retransmit persistence */
+#define TCPT_KEEP 5 /* keep alive */
+#define TCPT_2MSL 6 /* 2*msl quiet time timer */
+#if MPTCP
+#define TCPT_JACK_RXMT 7 /* retransmit timer for join ack */
+#define TCPT_MAX 7
+#else /* MPTCP */
+#define TCPT_MAX 6
+#endif /* !MPTCP */
+#define TCPT_NONE (TCPT_MAX + 1)
+#define TCPT_NTIMERS (TCPT_MAX + 1)
+
+/* External definitions */
+#define TCPT_REXMT_EXT 0
+#define TCPT_PERSIST_EXT 1
+#define TCPT_KEEP_EXT 2
+#define TCPT_2MSL_EXT 3
+#define TCPT_DELACK_EXT 4
+
+#else /* !BSD_KERNEL_PRIVATE */
+#define TCPT_REXMT 0 /* retransmit */
+#define TCPT_PERSIST 1 /* retransmit persistence */
+#define TCPT_KEEP 2 /* keep alive */
+#define TCPT_2MSL 3 /* 2*msl quiet time timer */
+#define TCPT_DELACK 4 /* delayed ack timer */
+#if MPTCP
+#define TCPT_JACK_RXMT 5 /* retransmit timer for join ack */
+#define TCPT_MAX 5
+#else /* MPTCP */
+#define TCPT_MAX 4
+#endif /* !MPTCP */
+#define TCPT_NONE (TCPT_MAX + 1)
+#define TCPT_NTIMERS (TCPT_MAX + 1)
+
+#endif /* BSD_KERNEL_PRIVATE */
+
+#ifdef BSD_KERNEL_PRIVATE
/*
* Time constants.
*/
-#define TCPTV_MSL ( 30*PR_SLOWHZ) /* max seg lifetime (hah!) */
-#define TCPTV_SRTTBASE 0 /* base roundtrip time;
- if 0, no idea yet */
-#define TCPTV_RTOBASE ( 3*PR_SLOWHZ) /* assumed RTO if no info */
-#define TCPTV_SRTTDFLT ( 3*PR_SLOWHZ) /* assumed RTT if no info */
-
-#define TCPTV_PERSMIN ( 5*PR_SLOWHZ) /* retransmit persistence */
-#define TCPTV_PERSMAX ( 60*PR_SLOWHZ) /* maximum persist interval */
-
-#define TCPTV_KEEP_INIT ( 75*PR_SLOWHZ) /* initial connect keep alive */
-#define TCPTV_KEEP_IDLE (120*60*PR_SLOWHZ) /* dflt time before probing */
-#define TCPTV_KEEPINTVL ( 75*PR_SLOWHZ) /* default probe interval */
+#define TCPTV_MSL ( 15*TCP_RETRANSHZ) /* max seg lifetime */
+#define TCPTV_SRTTBASE 0 /* base roundtrip time; if 0, no idea yet */
+#define TCPTV_RTOBASE ( 1*TCP_RETRANSHZ) /* assumed RTO if no info */
+#define TCPTV_SRTTDFLT ( 1*TCP_RETRANSHZ) /* assumed RTT if no info */
+#define TCPTV_PERSMIN ( 5*TCP_RETRANSHZ) /* retransmit persistence */
+#define TCPTV_PERSMAX ( 60*TCP_RETRANSHZ) /* maximum persist interval */
+
+#define TCPTV_KEEP_INIT ( 75*TCP_RETRANSHZ) /* connect keep alive */
+#define TCPTV_KEEP_IDLE (120*60*TCP_RETRANSHZ) /* time before probing */
+#define TCPTV_KEEPINTVL ( 75*TCP_RETRANSHZ) /* default probe interval */
#define TCPTV_KEEPCNT 8 /* max probes before drop */
-#define TCPTV_MIN ( 1*PR_SLOWHZ) /* minimum allowable value */
-#define TCPTV_REXMTMAX ( 64*PR_SLOWHZ) /* max allowable REXMT value */
+#define TCPTV_REXMTMAX ( 64*TCP_RETRANSHZ ) /* max REXMT value */
+#define TCPTV_REXMTMIN ( TCP_RETRANSHZ/33 ) /* min REXMT for non-local connections */
+
+/*
+ * Window for counting received bytes to see if ack-stretching
+ * can start (default 100 ms)
+ */
+#define TCPTV_UNACKWIN ( TCP_RETRANSHZ/10 )
+
+/* Receiver idle time, avoid ack-stretching after this idle time */
+#define TCPTV_MAXRCVIDLE (TCP_RETRANSHZ/5 )
-#define TCPTV_TWTRUNC 8 /* RTO factor to truncate TW */
+/*
+ * No ack stretching during slow-start, until we see some packets.
+ * By the time the receiver gets 512 packets, the senders cwnd
+ * should open by a few hundred packets consdering the
+ * slow-start progression.
+ */
+#define TCP_RCV_SS_PKTCOUNT 512
-#define TCP_LINGERTIME 120 /* linger at most 2 minutes */
+/* Receiver idle time, for rcv socket buffer resizing */
+#define TCPTV_RCVBUFIDLE (TCP_RETRANSHZ/2)
+#define TCPTV_TWTRUNC 8 /* RTO factor to truncate TW */
-#define TCP_MAXRXTSHIFT 12 /* maximum retransmits */
+#define TCP_LINGERTIME 120 /* linger at most 2 minutes */
-#define TCPTV_DELACK PR_FASTHZ /* 125ms timeout */
+#define TCP_MAXRXTSHIFT 12 /* maximum retransmits */
#ifdef TCPTIMERS
static char *tcptimers[] =
- { "REXMT", "PERSIST", "KEEP", "2MSL" };
-#endif
+ { "REXMT", "PERSIST", "KEEP", "2MSL" , "DELACK"};
+#endif /* TCPTIMERS */
-#ifdef KERNEL
/*
- * Force a time value to be in a certain range.
+ * Persist, keep, 2msl and MPTCP's join-ack timer as slow timers which can
+ * be coalesced at a higher granularity (500 ms).
+ *
+ * Rexmt and delayed ack timers are considered as fast timers which run
+ * in the order of 100ms.
+ *
+ * Probe timeout is a quick timer which will run in the order of 10ms.
*/
-#define TCPT_RANGESET(tv, value, tvmin, tvmax) do { \
- (tv) = (value); \
- if ((u_long)(tv) < (u_long)(tvmin)) \
+#define IS_TIMER_HZ_500MS(i) ((i) >= TCPT_PERSIST)
+#define IS_TIMER_HZ_100MS(i) ((i) >= TCPT_REXMT && (i) < TCPT_PERSIST)
+#define IS_TIMER_HZ_10MS(i) ((i) < TCPT_REXMT)
+
+struct tcptimerlist;
+
+struct tcptimerentry {
+ LIST_ENTRY(tcptimerentry) le; /* links for timer list */
+ uint32_t timer_start; /* tcp clock when the timer was started */
+ uint16_t index; /* index of lowest timer that needs to run first */
+ uint16_t mode; /* Bit-wise OR of timers that are active */
+ uint32_t runtime; /* deadline at which the first timer has to fire */
+};
+
+LIST_HEAD(timerlisthead, tcptimerentry);
+
+struct tcptimerlist {
+ struct timerlisthead lhead; /* head of the list */
+ lck_mtx_t *mtx; /* lock to protect the list */
+ lck_attr_t *mtx_attr; /* mutex attributes */
+ lck_grp_t *mtx_grp; /* mutex group definition */
+ lck_grp_attr_t *mtx_grp_attr; /* mutex group attributes */
+ thread_call_t call; /* call entry */
+ uint32_t runtime; /* time at which this list is going to run */
+ uint32_t entries; /* Number of entries on the list */
+ uint32_t maxentries; /* Max number of entries at any time */
+
+ /* Set desired mode when timer list running */
+ boolean_t running; /* Set when timer list is being processed */
+ boolean_t scheduled; /* set when the timer is scheduled */
+#define TCP_TIMERLIST_10MS_MODE 0x1
+#define TCP_TIMERLIST_100MS_MODE 0x2
+#define TCP_TIMERLIST_500MS_MODE 0x4
+ uint32_t mode; /* Current mode of the timer */
+ uint32_t pref_mode; /* Preferred mode set by a connection */
+ uint32_t pref_offset; /* Preferred offset set by a connection */
+ uint32_t idleruns; /* Number of times the list has been idle in fast mode */
+ struct tcptimerentry *next_te; /* next timer entry pointer to process */
+ u_int16_t probe_if_index; /* Interface index that needs to send probes */
+
+};
+
+/* number of idle runs allowed for TCP timer list in fast or quick modes */
+#define TCP_FASTMODE_IDLERUN_MAX 10
+
+/*
+ * Minimum retransmit timeout is set to 30ms. We add a slop of
+ * 200 ms to the retransmit value to account for processing
+ * variance and delayed ack. This extra 200ms will help to avoid
+ * spurious retransmits by taking into consideration the receivers
+ * that wait for delayed ack timer instead of generating an ack
+ * for every two packets.
+ *
+ * On a local link, the minimum retransmit timeout is 100ms and
+ * variance is set to 0. This will make the sender a little bit more
+ * aggressive on local link. When the connection is not established yet,
+ * there is no need to add an extra 200ms to retransmit timeout because
+ * the initial value is high (1s) and delayed ack is not a problem in
+ * that case.
+ */
+#define TCPTV_REXMTSLOP ( TCP_RETRANSHZ/5 ) /* extra 200 ms slop */
+
+/* macro to decide when retransmit slop (described above) should be added */
+#define TCP_ADD_REXMTSLOP(tp) (tp->t_state >= TCPS_ESTABLISHED)
+
+#define TCPT_RANGESET(tv, value, tvmin, tvmax, addslop) do { \
+ (tv) = ((addslop) ? tcp_rexmt_slop : 0) + (value); \
+ if ((uint32_t)(tv) < (uint32_t)(tvmin)) \
(tv) = (tvmin); \
- else if ((u_long)(tv) > (u_long)(tvmax)) \
+ else if ((uint32_t)(tv) > (uint32_t)(tvmax)) \
(tv) = (tvmax); \
} while(0)
-#define TCP_KEEPIDLE(tp) \
- (tp->t_keepidle && (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE) ? \
- tp->t_keepidle : tcp_keepidle)
+#define TCP_CONN_KEEPIDLE(tp) \
+ ((tp)->t_keepidle && \
+ ((tp)->t_inpcb->inp_socket->so_options & SO_KEEPALIVE) ? \
+ (tp)->t_keepidle : tcp_keepidle)
+#define TCP_CONN_KEEPINIT(tp) \
+ (((tp)->t_keepinit > 0) ? (tp)->t_keepinit : tcp_keepinit)
+#define TCP_CONN_KEEPCNT(tp) \
+ (((tp)->t_keepcnt > 0) ? (tp)->t_keepcnt : tcp_keepcnt)
+#define TCP_CONN_KEEPINTVL(tp) \
+ (((tp)->t_keepintvl > 0) ? (tp)->t_keepintvl : tcp_keepintvl)
+#define TCP_CONN_MAXIDLE(tp) \
+ (TCP_CONN_KEEPCNT(tp) * TCP_CONN_KEEPINTVL(tp))
+
+#define TCP_IDLETIMEOUT(tp) \
+ (((TCP_ADD_REXMTSLOP(tp)) ? 0 : tcp_rexmt_slop) + tp->t_rxtcur)
+
+TAILQ_HEAD(tcptailq, tcpcb);
-extern int tcp_keepinit; /* time to establish connection */
-extern int tcp_keepidle; /* time before keepalive probes begin */
-extern int tcp_keepintvl; /* time between keepalive probes */
-extern int tcp_maxidle; /* time to drop after starting probes */
-extern int tcp_delacktime; /* time before sending a delayed ACK */
+extern int tcp_keepinit; /* time to establish connection */
+extern int tcp_keepidle; /* time before keepalive probes begin */
+extern int tcp_keepintvl; /* time between keepalive probes */
+extern int tcp_keepcnt; /* number of keepalives */
+extern int tcp_delack; /* delayed ack timer */
extern int tcp_maxpersistidle;
extern int tcp_msl;
-extern int tcp_ttl; /* time to live for TCP segs */
+extern int tcp_ttl; /* time to live for TCP segs */
extern int tcp_backoff[];
+extern int tcp_rexmt_slop;
+extern u_int32_t tcp_max_persist_timeout; /* Maximum persistence for Zero Window Probes */
-void tcp_timer_2msl(void *xtp);
-void tcp_timer_keep(void *xtp);
-void tcp_timer_persist(void *xtp);
-void tcp_timer_rexmt(void *xtp);
-void tcp_timer_delack(void *xtp);
+#define OFFSET_FROM_START(tp, off) ((tcp_now + (off)) - (tp)->tentry.timer_start)
-#endif /* KERNEL */
-#endif /* PRIVATE */
+#endif /* BSD_KERNEL_PRIVATE */
#endif /* !_NETINET_TCP_TIMER_H_ */
projects / apple / xnu.git / blobdiff