| blob | 603a78c8af0666213e3f4924ca7a92278367f6dc |
3 RFC: 793
11 TRANSMISSION CONTROL PROTOCOL
14 DARPA INTERNET PROGRAM
16 PROTOCOL SPECIFICATION
20 September 1981
34 prepared for
36 Defense Advanced Research Projects Agency
37 Information Processing Techniques Office
38 1400 Wilson Boulevard
39 Arlington, Virginia 22209
47 by
49 Information Sciences Institute
50 University of Southern California
51 4676 Admiralty Way
52 Marina del Rey, California 90291
53 \f
54 \f
56 September 1981
57 Transmission Control Protocol
61 TABLE OF CONTENTS
63 PREFACE ........................................................ iii
65 1. INTRODUCTION ..................................................... 1
67 1.1 Motivation .................................................... 1
68 1.2 Scope ......................................................... 2
69 1.3 About This Document ........................................... 2
70 1.4 Interfaces .................................................... 3
71 1.5 Operation ..................................................... 3
73 2. PHILOSOPHY ....................................................... 7
75 2.1 Elements of the Internetwork System ........................... 7
76 2.2 Model of Operation ............................................ 7
77 2.3 The Host Environment .......................................... 8
78 2.4 Interfaces .................................................... 9
79 2.5 Relation to Other Protocols ................................... 9
80 2.6 Reliable Communication ........................................ 9
81 2.7 Connection Establishment and Clearing ........................ 10
82 2.8 Data Communication ........................................... 12
83 2.9 Precedence and Security ...................................... 13
84 2.10 Robustness Principle ......................................... 13
86 3. FUNCTIONAL SPECIFICATION ........................................ 15
88 3.1 Header Format ................................................ 15
89 3.2 Terminology .................................................. 19
90 3.3 Sequence Numbers ............................................. 24
91 3.4 Establishing a connection .................................... 30
92 3.5 Closing a Connection ......................................... 37
93 3.6 Precedence and Security ...................................... 40
94 3.7 Data Communication ........................................... 40
95 3.8 Interfaces ................................................... 44
96 3.9 Event Processing ............................................. 52
98 GLOSSARY ............................................................ 79
100 REFERENCES .......................................................... 85
112 [Page i]
113 \f
115 September 1981
116 Transmission Control Protocol
171 [Page ii]
172 \f
174 September 1981
175 Transmission Control Protocol
179 PREFACE
183 This document describes the DoD Standard Transmission Control Protocol
184 (TCP). There have been nine earlier editions of the ARPA TCP
185 specification on which this standard is based, and the present text
186 draws heavily from them. There have been many contributors to this work
187 both in terms of concepts and in terms of text. This edition clarifies
188 several details and removes the end-of-letter buffer-size adjustments,
189 and redescribes the letter mechanism as a push function.
191 Jon Postel
193 Editor
230 [Page iii]
231 \f
232 \f
235 RFC: 793
236 Replaces: RFC 761
237 IENs: 129, 124, 112, 81,
238 55, 44, 40, 27, 21, 5
240 TRANSMISSION CONTROL PROTOCOL
242 DARPA INTERNET PROGRAM
243 PROTOCOL SPECIFICATION
247 1. INTRODUCTION
249 The Transmission Control Protocol (TCP) is intended for use as a highly
250 reliable host-to-host protocol between hosts in packet-switched computer
251 communication networks, and in interconnected systems of such networks.
253 This document describes the functions to be performed by the
254 Transmission Control Protocol, the program that implements it, and its
255 interface to programs or users that require its services.
257 1.1. Motivation
259 Computer communication systems are playing an increasingly important
260 role in military, government, and civilian environments. This
261 document focuses its attention primarily on military computer
262 communication requirements, especially robustness in the presence of
263 communication unreliability and availability in the presence of
264 congestion, but many of these problems are found in the civilian and
265 government sector as well.
267 As strategic and tactical computer communication networks are
268 developed and deployed, it is essential to provide means of
269 interconnecting them and to provide standard interprocess
270 communication protocols which can support a broad range of
271 applications. In anticipation of the need for such standards, the
272 Deputy Undersecretary of Defense for Research and Engineering has
273 declared the Transmission Control Protocol (TCP) described herein to
274 be a basis for DoD-wide inter-process communication protocol
275 standardization.
277 TCP is a connection-oriented, end-to-end reliable protocol designed to
278 fit into a layered hierarchy of protocols which support multi-network
279 applications. The TCP provides for reliable inter-process
280 communication between pairs of processes in host computers attached to
281 distinct but interconnected computer communication networks. Very few
282 assumptions are made as to the reliability of the communication
283 protocols below the TCP layer. TCP assumes it can obtain a simple,
284 potentially unreliable datagram service from the lower level
285 protocols. In principle, the TCP should be able to operate above a
286 wide spectrum of communication systems ranging from hard-wired
287 connections to packet-switched or circuit-switched networks.
290 [Page 1]
291 \f
293 September 1981
294 Transmission Control Protocol
295 Introduction
299 TCP is based on concepts first described by Cerf and Kahn in [1]. The
300 TCP fits into a layered protocol architecture just above a basic
301 Internet Protocol [2] which provides a way for the TCP to send and
302 receive variable-length segments of information enclosed in internet
303 datagram "envelopes". The internet datagram provides a means for
304 addressing source and destination TCPs in different networks. The
305 internet protocol also deals with any fragmentation or reassembly of
306 the TCP segments required to achieve transport and delivery through
307 multiple networks and interconnecting gateways. The internet protocol
308 also carries information on the precedence, security classification
309 and compartmentation of the TCP segments, so this information can be
310 communicated end-to-end across multiple networks.
312 Protocol Layering
314 +---------------------+
315 | higher-level |
316 +---------------------+
317 | TCP |
318 +---------------------+
319 | internet protocol |
320 +---------------------+
321 |communication network|
322 +---------------------+
324 Figure 1
326 Much of this document is written in the context of TCP implementations
327 which are co-resident with higher level protocols in the host
328 computer. Some computer systems will be connected to networks via
329 front-end computers which house the TCP and internet protocol layers,
330 as well as network specific software. The TCP specification describes
331 an interface to the higher level protocols which appears to be
332 implementable even for the front-end case, as long as a suitable
333 host-to-front end protocol is implemented.
335 1.2. Scope
337 The TCP is intended to provide a reliable process-to-process
338 communication service in a multinetwork environment. The TCP is
339 intended to be a host-to-host protocol in common use in multiple
340 networks.
342 1.3. About this Document
344 This document represents a specification of the behavior required of
345 any TCP implementation, both in its interactions with higher level
346 protocols and in its interactions with other TCPs. The rest of this
349 [Page 2]
350 \f
352 September 1981
353 Transmission Control Protocol
354 Introduction
358 section offers a very brief view of the protocol interfaces and
359 operation. Section 2 summarizes the philosophical basis for the TCP
360 design. Section 3 offers both a detailed description of the actions
361 required of TCP when various events occur (arrival of new segments,
362 user calls, errors, etc.) and the details of the formats of TCP
363 segments.
365 1.4. Interfaces
367 The TCP interfaces on one side to user or application processes and on
368 the other side to a lower level protocol such as Internet Protocol.
370 The interface between an application process and the TCP is
371 illustrated in reasonable detail. This interface consists of a set of
372 calls much like the calls an operating system provides to an
373 application process for manipulating files. For example, there are
374 calls to open and close connections and to send and receive data on
375 established connections. It is also expected that the TCP can
376 asynchronously communicate with application programs. Although
377 considerable freedom is permitted to TCP implementors to design
378 interfaces which are appropriate to a particular operating system
379 environment, a minimum functionality is required at the TCP/user
380 interface for any valid implementation.
382 The interface between TCP and lower level protocol is essentially
383 unspecified except that it is assumed there is a mechanism whereby the
384 two levels can asynchronously pass information to each other.
385 Typically, one expects the lower level protocol to specify this
386 interface. TCP is designed to work in a very general environment of
387 interconnected networks. The lower level protocol which is assumed
388 throughout this document is the Internet Protocol [2].
390 1.5. Operation
392 As noted above, the primary purpose of the TCP is to provide reliable,
393 securable logical circuit or connection service between pairs of
394 processes. To provide this service on top of a less reliable internet
395 communication system requires facilities in the following areas:
397 Basic Data Transfer
398 Reliability
399 Flow Control
400 Multiplexing
401 Connections
402 Precedence and Security
404 The basic operation of the TCP in each of these areas is described in
405 the following paragraphs.
408 [Page 3]
409 \f
411 September 1981
412 Transmission Control Protocol
413 Introduction
417 Basic Data Transfer:
419 The TCP is able to transfer a continuous stream of octets in each
420 direction between its users by packaging some number of octets into
421 segments for transmission through the internet system. In general,
422 the TCPs decide when to block and forward data at their own
423 convenience.
425 Sometimes users need to be sure that all the data they have
426 submitted to the TCP has been transmitted. For this purpose a push
427 function is defined. To assure that data submitted to a TCP is
428 actually transmitted the sending user indicates that it should be
429 pushed through to the receiving user. A push causes the TCPs to
430 promptly forward and deliver data up to that point to the receiver.
431 The exact push point might not be visible to the receiving user and
432 the push function does not supply a record boundary marker.
434 Reliability:
436 The TCP must recover from data that is damaged, lost, duplicated, or
437 delivered out of order by the internet communication system. This
438 is achieved by assigning a sequence number to each octet
439 transmitted, and requiring a positive acknowledgment (ACK) from the
440 receiving TCP. If the ACK is not received within a timeout
441 interval, the data is retransmitted. At the receiver, the sequence
442 numbers are used to correctly order segments that may be received
443 out of order and to eliminate duplicates. Damage is handled by
444 adding a checksum to each segment transmitted, checking it at the
445 receiver, and discarding damaged segments.
447 As long as the TCPs continue to function properly and the internet
448 system does not become completely partitioned, no transmission
449 errors will affect the correct delivery of data. TCP recovers from
450 internet communication system errors.
452 Flow Control:
454 TCP provides a means for the receiver to govern the amount of data
455 sent by the sender. This is achieved by returning a "window" with
456 every ACK indicating a range of acceptable sequence numbers beyond
457 the last segment successfully received. The window indicates an
458 allowed number of octets that the sender may transmit before
459 receiving further permission.
467 [Page 4]
468 \f
470 September 1981
471 Transmission Control Protocol
472 Introduction
476 Multiplexing:
478 To allow for many processes within a single Host to use TCP
479 communication facilities simultaneously, the TCP provides a set of
480 addresses or ports within each host. Concatenated with the network
481 and host addresses from the internet communication layer, this forms
482 a socket. A pair of sockets uniquely identifies each connection.
483 That is, a socket may be simultaneously used in multiple
484 connections.
486 The binding of ports to processes is handled independently by each
487 Host. However, it proves useful to attach frequently used processes
488 (e.g., a "logger" or timesharing service) to fixed sockets which are
489 made known to the public. These services can then be accessed
490 through the known addresses. Establishing and learning the port
491 addresses of other processes may involve more dynamic mechanisms.
493 Connections:
495 The reliability and flow control mechanisms described above require
496 that TCPs initialize and maintain certain status information for
497 each data stream. The combination of this information, including
498 sockets, sequence numbers, and window sizes, is called a connection.
499 Each connection is uniquely specified by a pair of sockets
500 identifying its two sides.
502 When two processes wish to communicate, their TCP's must first
503 establish a connection (initialize the status information on each
504 side). When their communication is complete, the connection is
505 terminated or closed to free the resources for other uses.
507 Since connections must be established between unreliable hosts and
508 over the unreliable internet communication system, a handshake
509 mechanism with clock-based sequence numbers is used to avoid
510 erroneous initialization of connections.
512 Precedence and Security:
514 The users of TCP may indicate the security and precedence of their
515 communication. Provision is made for default values to be used when
516 these features are not needed.
526 [Page 5]
527 \f
529 September 1981
530 Transmission Control Protocol
585 [Page 6]
586 \f
588 September 1981
589 Transmission Control Protocol
593 2. PHILOSOPHY
595 2.1. Elements of the Internetwork System
597 The internetwork environment consists of hosts connected to networks
598 which are in turn interconnected via gateways. It is assumed here
599 that the networks may be either local networks (e.g., the ETHERNET) or
600 large networks (e.g., the ARPANET), but in any case are based on
601 packet switching technology. The active agents that produce and
602 consume messages are processes. Various levels of protocols in the
603 networks, the gateways, and the hosts support an interprocess
604 communication system that provides two-way data flow on logical
605 connections between process ports.
607 The term packet is used generically here to mean the data of one
608 transaction between a host and its network. The format of data blocks
609 exchanged within the a network will generally not be of concern to us.
611 Hosts are computers attached to a network, and from the communication
612 network's point of view, are the sources and destinations of packets.
613 Processes are viewed as the active elements in host computers (in
614 accordance with the fairly common definition of a process as a program
615 in execution). Even terminals and files or other I/O devices are
616 viewed as communicating with each other through the use of processes.
617 Thus, all communication is viewed as inter-process communication.
619 Since a process may need to distinguish among several communication
620 streams between itself and another process (or processes), we imagine
621 that each process may have a number of ports through which it
622 communicates with the ports of other processes.
624 2.2. Model of Operation
626 Processes transmit data by calling on the TCP and passing buffers of
627 data as arguments. The TCP packages the data from these buffers into
628 segments and calls on the internet module to transmit each segment to
629 the destination TCP. The receiving TCP places the data from a segment
630 into the receiving user's buffer and notifies the receiving user. The
631 TCPs include control information in the segments which they use to
632 ensure reliable ordered data transmission.
634 The model of internet communication is that there is an internet
635 protocol module associated with each TCP which provides an interface
636 to the local network. This internet module packages TCP segments
637 inside internet datagrams and routes these datagrams to a destination
638 internet module or intermediate gateway. To transmit the datagram
639 through the local network, it is embedded in a local network packet.
641 The packet switches may perform further packaging, fragmentation, or
644 [Page 7]
645 \f
647 September 1981
648 Transmission Control Protocol
649 Philosophy
653 other operations to achieve the delivery of the local packet to the
654 destination internet module.
656 At a gateway between networks, the internet datagram is "unwrapped"
657 from its local packet and examined to determine through which network
658 the internet datagram should travel next. The internet datagram is
659 then "wrapped" in a local packet suitable to the next network and
660 routed to the next gateway, or to the final destination.
662 A gateway is permitted to break up an internet datagram into smaller
663 internet datagram fragments if this is necessary for transmission
664 through the next network. To do this, the gateway produces a set of
665 internet datagrams; each carrying a fragment. Fragments may be
666 further broken into smaller fragments at subsequent gateways. The
667 internet datagram fragment format is designed so that the destination
668 internet module can reassemble fragments into internet datagrams.
670 A destination internet module unwraps the segment from the datagram
671 (after reassembling the datagram, if necessary) and passes it to the
672 destination TCP.
674 This simple model of the operation glosses over many details. One
675 important feature is the type of service. This provides information
676 to the gateway (or internet module) to guide it in selecting the
677 service parameters to be used in traversing the next network.
678 Included in the type of service information is the precedence of the
679 datagram. Datagrams may also carry security information to permit
680 host and gateways that operate in multilevel secure environments to
681 properly segregate datagrams for security considerations.
683 2.3. The Host Environment
685 The TCP is assumed to be a module in an operating system. The users
686 access the TCP much like they would access the file system. The TCP
687 may call on other operating system functions, for example, to manage
688 data structures. The actual interface to the network is assumed to be
689 controlled by a device driver module. The TCP does not call on the
690 network device driver directly, but rather calls on the internet
691 datagram protocol module which may in turn call on the device driver.
693 The mechanisms of TCP do not preclude implementation of the TCP in a
694 front-end processor. However, in such an implementation, a
695 host-to-front-end protocol must provide the functionality to support
696 the type of TCP-user interface described in this document.
703 [Page 8]
704 \f
706 September 1981
707 Transmission Control Protocol
708 Philosophy
712 2.4. Interfaces
714 The TCP/user interface provides for calls made by the user on the TCP
715 to OPEN or CLOSE a connection, to SEND or RECEIVE data, or to obtain
716 STATUS about a connection. These calls are like other calls from user
717 programs on the operating system, for example, the calls to open, read
718 from, and close a file.
720 The TCP/internet interface provides calls to send and receive
721 datagrams addressed to TCP modules in hosts anywhere in the internet
722 system. These calls have parameters for passing the address, type of
723 service, precedence, security, and other control information.
725 2.5. Relation to Other Protocols
727 The following diagram illustrates the place of the TCP in the protocol
728 hierarchy:
731 +------+ +-----+ +-----+ +-----+
732 |Telnet| | FTP | |Voice| ... | | Application Level
733 +------+ +-----+ +-----+ +-----+
734 | | | |
735 +-----+ +-----+ +-----+
736 | TCP | | RTP | ... | | Host Level
737 +-----+ +-----+ +-----+
738 | | |
739 +-------------------------------+
740 | Internet Protocol & ICMP | Gateway Level
741 +-------------------------------+
742 |
743 +---------------------------+
744 | Local Network Protocol | Network Level
745 +---------------------------+
747 Protocol Relationships
749 Figure 2.
751 It is expected that the TCP will be able to support higher level
752 protocols efficiently. It should be easy to interface higher level
753 protocols like the ARPANET Telnet or AUTODIN II THP to the TCP.
755 2.6. Reliable Communication
757 A stream of data sent on a TCP connection is delivered reliably and in
758 order at the destination.
762 [Page 9]
763 \f
765 September 1981
766 Transmission Control Protocol
767 Philosophy
771 Transmission is made reliable via the use of sequence numbers and
772 acknowledgments. Conceptually, each octet of data is assigned a
773 sequence number. The sequence number of the first octet of data in a
774 segment is transmitted with that segment and is called the segment
775 sequence number. Segments also carry an acknowledgment number which
776 is the sequence number of the next expected data octet of
777 transmissions in the reverse direction. When the TCP transmits a
778 segment containing data, it puts a copy on a retransmission queue and
779 starts a timer; when the acknowledgment for that data is received, the
780 segment is deleted from the queue. If the acknowledgment is not
781 received before the timer runs out, the segment is retransmitted.
783 An acknowledgment by TCP does not guarantee that the data has been
784 delivered to the end user, but only that the receiving TCP has taken
785 the responsibility to do so.
787 To govern the flow of data between TCPs, a flow control mechanism is
788 employed. The receiving TCP reports a "window" to the sending TCP.
789 This window specifies the number of octets, starting with the
790 acknowledgment number, that the receiving TCP is currently prepared to
791 receive.
793 2.7. Connection Establishment and Clearing
795 To identify the separate data streams that a TCP may handle, the TCP
796 provides a port identifier. Since port identifiers are selected
797 independently by each TCP they might not be unique. To provide for
798 unique addresses within each TCP, we concatenate an internet address
799 identifying the TCP with a port identifier to create a socket which
800 will be unique throughout all networks connected together.
802 A connection is fully specified by the pair of sockets at the ends. A
803 local socket may participate in many connections to different foreign
804 sockets. A connection can be used to carry data in both directions,
805 that is, it is "full duplex".
807 TCPs are free to associate ports with processes however they choose.
808 However, several basic concepts are necessary in any implementation.
809 There must be well-known sockets which the TCP associates only with
810 the "appropriate" processes by some means. We envision that processes
811 may "own" ports, and that processes can initiate connections only on
812 the ports they own. (Means for implementing ownership is a local
813 issue, but we envision a Request Port user command, or a method of
814 uniquely allocating a group of ports to a given process, e.g., by
815 associating the high order bits of a port name with a given process.)
817 A connection is specified in the OPEN call by the local port and
818 foreign socket arguments. In return, the TCP supplies a (short) local
821 [Page 10]
822 \f
824 September 1981
825 Transmission Control Protocol
826 Philosophy
830 connection name by which the user refers to the connection in
831 subsequent calls. There are several things that must be remembered
832 about a connection. To store this information we imagine that there
833 is a data structure called a Transmission Control Block (TCB). One
834 implementation strategy would have the local connection name be a
835 pointer to the TCB for this connection. The OPEN call also specifies
836 whether the connection establishment is to be actively pursued, or to
837 be passively waited for.
839 A passive OPEN request means that the process wants to accept incoming
840 connection requests rather than attempting to initiate a connection.
841 Often the process requesting a passive OPEN will accept a connection
842 request from any caller. In this case a foreign socket of all zeros
843 is used to denote an unspecified socket. Unspecified foreign sockets
844 are allowed only on passive OPENs.
846 A service process that wished to provide services for unknown other
847 processes would issue a passive OPEN request with an unspecified
848 foreign socket. Then a connection could be made with any process that
849 requested a connection to this local socket. It would help if this
850 local socket were known to be associated with this service.
852 Well-known sockets are a convenient mechanism for a priori associating
853 a socket address with a standard service. For instance, the
854 "Telnet-Server" process is permanently assigned to a particular
855 socket, and other sockets are reserved for File Transfer, Remote Job
856 Entry, Text Generator, Echoer, and Sink processes (the last three
857 being for test purposes). A socket address might be reserved for
858 access to a "Look-Up" service which would return the specific socket
859 at which a newly created service would be provided. The concept of a
860 well-known socket is part of the TCP specification, but the assignment
861 of sockets to services is outside this specification. (See [4].)
863 Processes can issue passive OPENs and wait for matching active OPENs
864 from other processes and be informed by the TCP when connections have
865 been established. Two processes which issue active OPENs to each
866 other at the same time will be correctly connected. This flexibility
867 is critical for the support of distributed computing in which
868 components act asynchronously with respect to each other.
870 There are two principal cases for matching the sockets in the local
871 passive OPENs and an foreign active OPENs. In the first case, the
872 local passive OPENs has fully specified the foreign socket. In this
873 case, the match must be exact. In the second case, the local passive
874 OPENs has left the foreign socket unspecified. In this case, any
875 foreign socket is acceptable as long as the local sockets match.
876 Other possibilities include partially restricted matches.
880 [Page 11]
881 \f
883 September 1981
884 Transmission Control Protocol
885 Philosophy
889 If there are several pending passive OPENs (recorded in TCBs) with the
890 same local socket, an foreign active OPEN will be matched to a TCB
891 with the specific foreign socket in the foreign active OPEN, if such a
892 TCB exists, before selecting a TCB with an unspecified foreign socket.
894 The procedures to establish connections utilize the synchronize (SYN)
895 control flag and involves an exchange of three messages. This
896 exchange has been termed a three-way hand shake [3].
898 A connection is initiated by the rendezvous of an arriving segment
899 containing a SYN and a waiting TCB entry each created by a user OPEN
900 command. The matching of local and foreign sockets determines when a
901 connection has been initiated. The connection becomes "established"
902 when sequence numbers have been synchronized in both directions.
904 The clearing of a connection also involves the exchange of segments,
905 in this case carrying the FIN control flag.
907 2.8. Data Communication
909 The data that flows on a connection may be thought of as a stream of
910 octets. The sending user indicates in each SEND call whether the data
911 in that call (and any preceeding calls) should be immediately pushed
912 through to the receiving user by the setting of the PUSH flag.
914 A sending TCP is allowed to collect data from the sending user and to
915 send that data in segments at its own convenience, until the push
916 function is signaled, then it must send all unsent data. When a
917 receiving TCP sees the PUSH flag, it must not wait for more data from
918 the sending TCP before passing the data to the receiving process.
920 There is no necessary relationship between push functions and segment
921 boundaries. The data in any particular segment may be the result of a
922 single SEND call, in whole or part, or of multiple SEND calls.
924 The purpose of push function and the PUSH flag is to push data through
925 from the sending user to the receiving user. It does not provide a
926 record service.
928 There is a coupling between the push function and the use of buffers
929 of data that cross the TCP/user interface. Each time a PUSH flag is
930 associated with data placed into the receiving user's buffer, the
931 buffer is returned to the user for processing even if the buffer is
932 not filled. If data arrives that fills the user's buffer before a
933 PUSH is seen, the data is passed to the user in buffer size units.
935 TCP also provides a means to communicate to the receiver of data that
936 at some point further along in the data stream than the receiver is
939 [Page 12]
940 \f
942 September 1981
943 Transmission Control Protocol
944 Philosophy
948 currently reading there is urgent data. TCP does not attempt to
949 define what the user specifically does upon being notified of pending
950 urgent data, but the general notion is that the receiving process will
951 take action to process the urgent data quickly.
953 2.9. Precedence and Security
955 The TCP makes use of the internet protocol type of service field and
956 security option to provide precedence and security on a per connection
957 basis to TCP users. Not all TCP modules will necessarily function in
958 a multilevel secure environment; some may be limited to unclassified
959 use only, and others may operate at only one security level and
960 compartment. Consequently, some TCP implementations and services to
961 users may be limited to a subset of the multilevel secure case.
963 TCP modules which operate in a multilevel secure environment must
964 properly mark outgoing segments with the security, compartment, and
965 precedence. Such TCP modules must also provide to their users or
966 higher level protocols such as Telnet or THP an interface to allow
967 them to specify the desired security level, compartment, and
968 precedence of connections.
970 2.10. Robustness Principle
972 TCP implementations will follow a general principle of robustness: be
973 conservative in what you do, be liberal in what you accept from
974 others.
998 [Page 13]
999 \f
1001 September 1981
1002 Transmission Control Protocol
1057 [Page 14]
1058 \f
1060 September 1981
1061 Transmission Control Protocol
1065 3. FUNCTIONAL SPECIFICATION
1067 3.1. Header Format
1069 TCP segments are sent as internet datagrams. The Internet Protocol
1070 header carries several information fields, including the source and
1071 destination host addresses [2]. A TCP header follows the internet
1072 header, supplying information specific to the TCP protocol. This
1073 division allows for the existence of host level protocols other than
1074 TCP.
1076 TCP Header Format
1079 0 1 2 3
1080 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
1081 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1082 | Source Port | Destination Port |
1083 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1084 | Sequence Number |
1085 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1086 | Acknowledgment Number |
1087 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1088 | Data | |U|A|P|R|S|F| |
1089 | Offset| Reserved |R|C|S|S|Y|I| Window |
1090 | | |G|K|H|T|N|N| |
1091 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1092 | Checksum | Urgent Pointer |
1093 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1094 | Options | Padding |
1095 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1096 | data |
1097 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1099 TCP Header Format
1101 Note that one tick mark represents one bit position.
1103 Figure 3.
1105 Source Port: 16 bits
1107 The source port number.
1109 Destination Port: 16 bits
1111 The destination port number.
1116 [Page 15]
1117 \f
1119 September 1981
1120 Transmission Control Protocol
1121 Functional Specification
1125 Sequence Number: 32 bits
1127 The sequence number of the first data octet in this segment (except
1128 when SYN is present). If SYN is present the sequence number is the
1129 initial sequence number (ISN) and the first data octet is ISN+1.
1131 Acknowledgment Number: 32 bits
1133 If the ACK control bit is set this field contains the value of the
1134 next sequence number the sender of the segment is expecting to
1135 receive. Once a connection is established this is always sent.
1137 Data Offset: 4 bits
1139 The number of 32 bit words in the TCP Header. This indicates where
1140 the data begins. The TCP header (even one including options) is an
1141 integral number of 32 bits long.
1143 Reserved: 6 bits
1145 Reserved for future use. Must be zero.
1147 Control Bits: 6 bits (from left to right):
1149 URG: Urgent Pointer field significant
1150 ACK: Acknowledgment field significant
1151 PSH: Push Function
1152 RST: Reset the connection
1153 SYN: Synchronize sequence numbers
1154 FIN: No more data from sender
1156 Window: 16 bits
1158 The number of data octets beginning with the one indicated in the
1159 acknowledgment field which the sender of this segment is willing to
1160 accept.
1162 Checksum: 16 bits
1164 The checksum field is the 16 bit one's complement of the one's
1165 complement sum of all 16 bit words in the header and text. If a
1166 segment contains an odd number of header and text octets to be
1167 checksummed, the last octet is padded on the right with zeros to
1168 form a 16 bit word for checksum purposes. The pad is not
1169 transmitted as part of the segment. While computing the checksum,
1170 the checksum field itself is replaced with zeros.
1172 The checksum also covers a 96 bit pseudo header conceptually
1175 [Page 16]
1176 \f
1178 September 1981
1179 Transmission Control Protocol
1180 Functional Specification
1184 prefixed to the TCP header. This pseudo header contains the Source
1185 Address, the Destination Address, the Protocol, and TCP length.
1186 This gives the TCP protection against misrouted segments. This
1187 information is carried in the Internet Protocol and is transferred
1188 across the TCP/Network interface in the arguments or results of
1189 calls by the TCP on the IP.
1191 +--------+--------+--------+--------+
1192 | Source Address |
1193 +--------+--------+--------+--------+
1194 | Destination Address |
1195 +--------+--------+--------+--------+
1196 | zero | PTCL | TCP Length |
1197 +--------+--------+--------+--------+
1199 The TCP Length is the TCP header length plus the data length in
1200 octets (this is not an explicitly transmitted quantity, but is
1201 computed), and it does not count the 12 octets of the pseudo
1202 header.
1204 Urgent Pointer: 16 bits
1206 This field communicates the current value of the urgent pointer as a
1207 positive offset from the sequence number in this segment. The
1208 urgent pointer points to the sequence number of the octet following
1209 the urgent data. This field is only be interpreted in segments with
1210 the URG control bit set.
1212 Options: variable
1214 Options may occupy space at the end of the TCP header and are a
1215 multiple of 8 bits in length. All options are included in the
1216 checksum. An option may begin on any octet boundary. There are two
1217 cases for the format of an option:
1219 Case 1: A single octet of option-kind.
1221 Case 2: An octet of option-kind, an octet of option-length, and
1222 the actual option-data octets.
1224 The option-length counts the two octets of option-kind and
1225 option-length as well as the option-data octets.
1227 Note that the list of options may be shorter than the data offset
1228 field might imply. The content of the header beyond the
1229 End-of-Option option must be header padding (i.e., zero).
1231 A TCP must implement all options.
1234 [Page 17]
1235 \f
1237 September 1981
1238 Transmission Control Protocol
1239 Functional Specification
1243 Currently defined options include (kind indicated in octal):
1245 Kind Length Meaning
1246 ---- ------ -------
1247 0 - End of option list.
1248 1 - No-Operation.
1249 2 4 Maximum Segment Size.
1252 Specific Option Definitions
1254 End of Option List
1256 +--------+
1257 |00000000|
1258 +--------+
1259 Kind=0
1261 This option code indicates the end of the option list. This
1262 might not coincide with the end of the TCP header according to
1263 the Data Offset field. This is used at the end of all options,
1264 not the end of each option, and need only be used if the end of
1265 the options would not otherwise coincide with the end of the TCP
1266 header.
1268 No-Operation
1270 +--------+
1271 |00000001|
1272 +--------+
1273 Kind=1
1275 This option code may be used between options, for example, to
1276 align the beginning of a subsequent option on a word boundary.
1277 There is no guarantee that senders will use this option, so
1278 receivers must be prepared to process options even if they do
1279 not begin on a word boundary.
1281 Maximum Segment Size
1283 +--------+--------+---------+--------+
1284 |00000010|00000100| max seg size |
1285 +--------+--------+---------+--------+
1286 Kind=2 Length=4
1293 [Page 18]
1294 \f
1296 September 1981
1297 Transmission Control Protocol
1298 Functional Specification
1302 Maximum Segment Size Option Data: 16 bits
1304 If this option is present, then it communicates the maximum
1305 receive segment size at the TCP which sends this segment.
1306 This field must only be sent in the initial connection request
1307 (i.e., in segments with the SYN control bit set). If this
1308 option is not used, any segment size is allowed.
1310 Padding: variable
1312 The TCP header padding is used to ensure that the TCP header ends
1313 and data begins on a 32 bit boundary. The padding is composed of
1314 zeros.
1316 3.2. Terminology
1318 Before we can discuss very much about the operation of the TCP we need
1319 to introduce some detailed terminology. The maintenance of a TCP
1320 connection requires the remembering of several variables. We conceive
1321 of these variables being stored in a connection record called a
1322 Transmission Control Block or TCB. Among the variables stored in the
1323 TCB are the local and remote socket numbers, the security and
1324 precedence of the connection, pointers to the user's send and receive
1325 buffers, pointers to the retransmit queue and to the current segment.
1326 In addition several variables relating to the send and receive
1327 sequence numbers are stored in the TCB.
1329 Send Sequence Variables
1331 SND.UNA - send unacknowledged
1332 SND.NXT - send next
1333 SND.WND - send window
1334 SND.UP - send urgent pointer
1335 SND.WL1 - segment sequence number used for last window update
1336 SND.WL2 - segment acknowledgment number used for last window
1337 update
1338 ISS - initial send sequence number
1340 Receive Sequence Variables
1342 RCV.NXT - receive next
1343 RCV.WND - receive window
1344 RCV.UP - receive urgent pointer
1345 IRS - initial receive sequence number
1352 [Page 19]
1353 \f
1355 September 1981
1356 Transmission Control Protocol
1357 Functional Specification
1361 The following diagrams may help to relate some of these variables to
1362 the sequence space.
1364 Send Sequence Space
1366 1 2 3 4
1367 ----------|----------|----------|----------
1368 SND.UNA SND.NXT SND.UNA
1369 +SND.WND
1371 1 - old sequence numbers which have been acknowledged
1372 2 - sequence numbers of unacknowledged data
1373 3 - sequence numbers allowed for new data transmission
1374 4 - future sequence numbers which are not yet allowed
1376 Send Sequence Space
1378 Figure 4.
1382 The send window is the portion of the sequence space labeled 3 in
1383 figure 4.
1385 Receive Sequence Space
1387 1 2 3
1388 ----------|----------|----------
1389 RCV.NXT RCV.NXT
1390 +RCV.WND
1392 1 - old sequence numbers which have been acknowledged
1393 2 - sequence numbers allowed for new reception
1394 3 - future sequence numbers which are not yet allowed
1396 Receive Sequence Space
1398 Figure 5.
1402 The receive window is the portion of the sequence space labeled 2 in
1403 figure 5.
1405 There are also some variables used frequently in the discussion that
1406 take their values from the fields of the current segment.
1411 [Page 20]
1412 \f
1414 September 1981
1415 Transmission Control Protocol
1416 Functional Specification
1420 Current Segment Variables
1422 SEG.SEQ - segment sequence number
1423 SEG.ACK - segment acknowledgment number
1424 SEG.LEN - segment length
1425 SEG.WND - segment window
1426 SEG.UP - segment urgent pointer
1427 SEG.PRC - segment precedence value
1429 A connection progresses through a series of states during its
1430 lifetime. The states are: LISTEN, SYN-SENT, SYN-RECEIVED,
1431 ESTABLISHED, FIN-WAIT-1, FIN-WAIT-2, CLOSE-WAIT, CLOSING, LAST-ACK,
1432 TIME-WAIT, and the fictional state CLOSED. CLOSED is fictional
1433 because it represents the state when there is no TCB, and therefore,
1434 no connection. Briefly the meanings of the states are:
1436 LISTEN - represents waiting for a connection request from any remote
1437 TCP and port.
1439 SYN-SENT - represents waiting for a matching connection request
1440 after having sent a connection request.
1442 SYN-RECEIVED - represents waiting for a confirming connection
1443 request acknowledgment after having both received and sent a
1444 connection request.
1446 ESTABLISHED - represents an open connection, data received can be
1447 delivered to the user. The normal state for the data transfer phase
1448 of the connection.
1450 FIN-WAIT-1 - represents waiting for a connection termination request
1451 from the remote TCP, or an acknowledgment of the connection
1452 termination request previously sent.
1454 FIN-WAIT-2 - represents waiting for a connection termination request
1455 from the remote TCP.
1457 CLOSE-WAIT - represents waiting for a connection termination request
1458 from the local user.
1460 CLOSING - represents waiting for a connection termination request
1461 acknowledgment from the remote TCP.
1463 LAST-ACK - represents waiting for an acknowledgment of the
1464 connection termination request previously sent to the remote TCP
1465 (which includes an acknowledgment of its connection termination
1466 request).
1470 [Page 21]
1471 \f
1473 September 1981
1474 Transmission Control Protocol
1475 Functional Specification
1479 TIME-WAIT - represents waiting for enough time to pass to be sure
1480 the remote TCP received the acknowledgment of its connection
1481 termination request.
1483 CLOSED - represents no connection state at all.
1485 A TCP connection progresses from one state to another in response to
1486 events. The events are the user calls, OPEN, SEND, RECEIVE, CLOSE,
1487 ABORT, and STATUS; the incoming segments, particularly those
1488 containing the SYN, ACK, RST and FIN flags; and timeouts.
1490 The state diagram in figure 6 illustrates only state changes, together
1491 with the causing events and resulting actions, but addresses neither
1492 error conditions nor actions which are not connected with state
1493 changes. In a later section, more detail is offered with respect to
1494 the reaction of the TCP to events.
1496 NOTE BENE: this diagram is only a summary and must not be taken as
1497 the total specification.
1529 [Page 22]
1530 \f
1532 September 1981
1533 Transmission Control Protocol
1534 Functional Specification
1539 +---------+ ---------\ active OPEN
1540 | CLOSED | \ -----------
1541 +---------+<---------\ \ create TCB
1542 | ^ \ \ snd SYN
1543 passive OPEN | | CLOSE \ \
1544 ------------ | | ---------- \ \
1545 create TCB | | delete TCB \ \
1546 V | \ \
1547 +---------+ CLOSE | \
1548 | LISTEN | ---------- | |
1549 +---------+ delete TCB | |
1550 rcv SYN | | SEND | |
1551 ----------- | | ------- | V
1552 +---------+ snd SYN,ACK / \ snd SYN +---------+
1553 | |<----------------- ------------------>| |
1554 | SYN | rcv SYN | SYN |
1555 | RCVD |<-----------------------------------------------| SENT |
1556 | | snd ACK | |
1557 | |------------------ -------------------| |
1558 +---------+ rcv ACK of SYN \ / rcv SYN,ACK +---------+
1559 | -------------- | | -----------
1560 | x | | snd ACK
1561 | V V
1562 | CLOSE +---------+
1563 | ------- | ESTAB |
1564 | snd FIN +---------+
1565 | CLOSE | | rcv FIN
1566 V ------- | | -------
1567 +---------+ snd FIN / \ snd ACK +---------+
1568 | FIN |<----------------- ------------------>| CLOSE |
1569 | WAIT-1 |------------------ | WAIT |
1570 +---------+ rcv FIN \ +---------+
1571 | rcv ACK of FIN ------- | CLOSE |
1572 | -------------- snd ACK | ------- |
1573 V x V snd FIN V
1574 +---------+ +---------+ +---------+
1575 |FINWAIT-2| | CLOSING | | LAST-ACK|
1576 +---------+ +---------+ +---------+
1577 | rcv ACK of FIN | rcv ACK of FIN |
1578 | rcv FIN -------------- | Timeout=2MSL -------------- |
1579 | ------- x V ------------ x V
1580 \ snd ACK +---------+delete TCB +---------+
1581 ------------------------>|TIME WAIT|------------------>| CLOSED |
1582 +---------+ +---------+
1584 TCP Connection State Diagram
1585 Figure 6.
1588 [Page 23]
1589 \f
1591 September 1981
1592 Transmission Control Protocol
1593 Functional Specification
1597 3.3. Sequence Numbers
1599 A fundamental notion in the design is that every octet of data sent
1600 over a TCP connection has a sequence number. Since every octet is
1601 sequenced, each of them can be acknowledged. The acknowledgment
1602 mechanism employed is cumulative so that an acknowledgment of sequence
1603 number X indicates that all octets up to but not including X have been
1604 received. This mechanism allows for straight-forward duplicate
1605 detection in the presence of retransmission. Numbering of octets
1606 within a segment is that the first data octet immediately following
1607 the header is the lowest numbered, and the following octets are
1608 numbered consecutively.
1610 It is essential to remember that the actual sequence number space is
1611 finite, though very large. This space ranges from 0 to 2**32 - 1.
1612 Since the space is finite, all arithmetic dealing with sequence
1613 numbers must be performed modulo 2**32. This unsigned arithmetic
1614 preserves the relationship of sequence numbers as they cycle from
1615 2**32 - 1 to 0 again. There are some subtleties to computer modulo
1616 arithmetic, so great care should be taken in programming the
1617 comparison of such values. The symbol "=<" means "less than or equal"
1618 (modulo 2**32).
1620 The typical kinds of sequence number comparisons which the TCP must
1621 perform include:
1623 (a) Determining that an acknowledgment refers to some sequence
1624 number sent but not yet acknowledged.
1626 (b) Determining that all sequence numbers occupied by a segment
1627 have been acknowledged (e.g., to remove the segment from a
1628 retransmission queue).
1630 (c) Determining that an incoming segment contains sequence numbers
1631 which are expected (i.e., that the segment "overlaps" the
1632 receive window).
1647 [Page 24]
1648 \f
1650 September 1981
1651 Transmission Control Protocol
1652 Functional Specification
1656 In response to sending data the TCP will receive acknowledgments. The
1657 following comparisons are needed to process the acknowledgments.
1659 SND.UNA = oldest unacknowledged sequence number
1661 SND.NXT = next sequence number to be sent
1663 SEG.ACK = acknowledgment from the receiving TCP (next sequence
1664 number expected by the receiving TCP)
1666 SEG.SEQ = first sequence number of a segment
1668 SEG.LEN = the number of octets occupied by the data in the segment
1669 (counting SYN and FIN)
1671 SEG.SEQ+SEG.LEN-1 = last sequence number of a segment
1673 A new acknowledgment (called an "acceptable ack"), is one for which
1674 the inequality below holds:
1676 SND.UNA < SEG.ACK =< SND.NXT
1678 A segment on the retransmission queue is fully acknowledged if the sum
1679 of its sequence number and length is less or equal than the
1680 acknowledgment value in the incoming segment.
1682 When data is received the following comparisons are needed:
1684 RCV.NXT = next sequence number expected on an incoming segments, and
1685 is the left or lower edge of the receive window
1687 RCV.NXT+RCV.WND-1 = last sequence number expected on an incoming
1688 segment, and is the right or upper edge of the receive window
1690 SEG.SEQ = first sequence number occupied by the incoming segment
1692 SEG.SEQ+SEG.LEN-1 = last sequence number occupied by the incoming
1693 segment
1695 A segment is judged to occupy a portion of valid receive sequence
1696 space if
1698 RCV.NXT =< SEG.SEQ < RCV.NXT+RCV.WND
1700 or
1702 RCV.NXT =< SEG.SEQ+SEG.LEN-1 < RCV.NXT+RCV.WND
1706 [Page 25]
1707 \f
1709 September 1981
1710 Transmission Control Protocol
1711 Functional Specification
1715 The first part of this test checks to see if the beginning of the
1716 segment falls in the window, the second part of the test checks to see
1717 if the end of the segment falls in the window; if the segment passes
1718 either part of the test it contains data in the window.
1720 Actually, it is a little more complicated than this. Due to zero
1721 windows and zero length segments, we have four cases for the
1722 acceptability of an incoming segment:
1724 Segment Receive Test
1725 Length Window
1726 ------- ------- -------------------------------------------
1728 0 0 SEG.SEQ = RCV.NXT
1730 0 >0 RCV.NXT =< SEG.SEQ < RCV.NXT+RCV.WND
1732 >0 0 not acceptable
1734 >0 >0 RCV.NXT =< SEG.SEQ < RCV.NXT+RCV.WND
1735 or RCV.NXT =< SEG.SEQ+SEG.LEN-1 < RCV.NXT+RCV.WND
1737 Note that when the receive window is zero no segments should be
1738 acceptable except ACK segments. Thus, it is be possible for a TCP to
1739 maintain a zero receive window while transmitting data and receiving
1740 ACKs. However, even when the receive window is zero, a TCP must
1741 process the RST and URG fields of all incoming segments.
1743 We have taken advantage of the numbering scheme to protect certain
1744 control information as well. This is achieved by implicitly including
1745 some control flags in the sequence space so they can be retransmitted
1746 and acknowledged without confusion (i.e., one and only one copy of the
1747 control will be acted upon). Control information is not physically
1748 carried in the segment data space. Consequently, we must adopt rules
1749 for implicitly assigning sequence numbers to control. The SYN and FIN
1750 are the only controls requiring this protection, and these controls
1751 are used only at connection opening and closing. For sequence number
1752 purposes, the SYN is considered to occur before the first actual data
1753 octet of the segment in which it occurs, while the FIN is considered
1754 to occur after the last actual data octet in a segment in which it
1755 occurs. The segment length (SEG.LEN) includes both data and sequence
1756 space occupying controls. When a SYN is present then SEG.SEQ is the
1757 sequence number of the SYN.
1765 [Page 26]
1766 \f
1768 September 1981
1769 Transmission Control Protocol
1770 Functional Specification
1774 Initial Sequence Number Selection
1776 The protocol places no restriction on a particular connection being
1777 used over and over again. A connection is defined by a pair of
1778 sockets. New instances of a connection will be referred to as
1779 incarnations of the connection. The problem that arises from this is
1780 -- "how does the TCP identify duplicate segments from previous
1781 incarnations of the connection?" This problem becomes apparent if the
1782 connection is being opened and closed in quick succession, or if the
1783 connection breaks with loss of memory and is then reestablished.
1785 To avoid confusion we must prevent segments from one incarnation of a
1786 connection from being used while the same sequence numbers may still
1787 be present in the network from an earlier incarnation. We want to
1788 assure this, even if a TCP crashes and loses all knowledge of the
1789 sequence numbers it has been using. When new connections are created,
1790 an initial sequence number (ISN) generator is employed which selects a
1791 new 32 bit ISN. The generator is bound to a (possibly fictitious) 32
1792 bit clock whose low order bit is incremented roughly every 4
1793 microseconds. Thus, the ISN cycles approximately every 4.55 hours.
1794 Since we assume that segments will stay in the network no more than
1795 the Maximum Segment Lifetime (MSL) and that the MSL is less than 4.55
1796 hours we can reasonably assume that ISN's will be unique.
1798 For each connection there is a send sequence number and a receive
1799 sequence number. The initial send sequence number (ISS) is chosen by
1800 the data sending TCP, and the initial receive sequence number (IRS) is
1801 learned during the connection establishing procedure.
1803 For a connection to be established or initialized, the two TCPs must
1804 synchronize on each other's initial sequence numbers. This is done in
1805 an exchange of connection establishing segments carrying a control bit
1806 called "SYN" (for synchronize) and the initial sequence numbers. As a
1807 shorthand, segments carrying the SYN bit are also called "SYNs".
1808 Hence, the solution requires a suitable mechanism for picking an
1809 initial sequence number and a slightly involved handshake to exchange
1810 the ISN's.
1812 The synchronization requires each side to send it's own initial
1813 sequence number and to receive a confirmation of it in acknowledgment
1814 from the other side. Each side must also receive the other side's
1815 initial sequence number and send a confirming acknowledgment.
1817 1) A --> B SYN my sequence number is X
1818 2) A <-- B ACK your sequence number is X
1819 3) A <-- B SYN my sequence number is Y
1820 4) A --> B ACK your sequence number is Y
1824 [Page 27]
1825 \f
1827 September 1981
1828 Transmission Control Protocol
1829 Functional Specification
1833 Because steps 2 and 3 can be combined in a single message this is
1834 called the three way (or three message) handshake.
1836 A three way handshake is necessary because sequence numbers are not
1837 tied to a global clock in the network, and TCPs may have different
1838 mechanisms for picking the ISN's. The receiver of the first SYN has
1839 no way of knowing whether the segment was an old delayed one or not,
1840 unless it remembers the last sequence number used on the connection
1841 (which is not always possible), and so it must ask the sender to
1842 verify this SYN. The three way handshake and the advantages of a
1843 clock-driven scheme are discussed in [3].
1845 Knowing When to Keep Quiet
1847 To be sure that a TCP does not create a segment that carries a
1848 sequence number which may be duplicated by an old segment remaining in
1849 the network, the TCP must keep quiet for a maximum segment lifetime
1850 (MSL) before assigning any sequence numbers upon starting up or
1851 recovering from a crash in which memory of sequence numbers in use was
1852 lost. For this specification the MSL is taken to be 2 minutes. This
1853 is an engineering choice, and may be changed if experience indicates
1854 it is desirable to do so. Note that if a TCP is reinitialized in some
1855 sense, yet retains its memory of sequence numbers in use, then it need
1856 not wait at all; it must only be sure to use sequence numbers larger
1857 than those recently used.
1859 The TCP Quiet Time Concept
1861 This specification provides that hosts which "crash" without
1862 retaining any knowledge of the last sequence numbers transmitted on
1863 each active (i.e., not closed) connection shall delay emitting any
1864 TCP segments for at least the agreed Maximum Segment Lifetime (MSL)
1865 in the internet system of which the host is a part. In the
1866 paragraphs below, an explanation for this specification is given.
1867 TCP implementors may violate the "quiet time" restriction, but only
1868 at the risk of causing some old data to be accepted as new or new
1869 data rejected as old duplicated by some receivers in the internet
1870 system.
1872 TCPs consume sequence number space each time a segment is formed and
1873 entered into the network output queue at a source host. The
1874 duplicate detection and sequencing algorithm in the TCP protocol
1875 relies on the unique binding of segment data to sequence space to
1876 the extent that sequence numbers will not cycle through all 2**32
1877 values before the segment data bound to those sequence numbers has
1878 been delivered and acknowledged by the receiver and all duplicate
1879 copies of the segments have "drained" from the internet. Without
1880 such an assumption, two distinct TCP segments could conceivably be
1883 [Page 28]
1884 \f
1886 September 1981
1887 Transmission Control Protocol
1888 Functional Specification
1892 assigned the same or overlapping sequence numbers, causing confusion
1893 at the receiver as to which data is new and which is old. Remember
1894 that each segment is bound to as many consecutive sequence numbers
1895 as there are octets of data in the segment.
1897 Under normal conditions, TCPs keep track of the next sequence number
1898 to emit and the oldest awaiting acknowledgment so as to avoid
1899 mistakenly using a sequence number over before its first use has
1900 been acknowledged. This alone does not guarantee that old duplicate
1901 data is drained from the net, so the sequence space has been made
1902 very large to reduce the probability that a wandering duplicate will
1903 cause trouble upon arrival. At 2 megabits/sec. it takes 4.5 hours
1904 to use up 2**32 octets of sequence space. Since the maximum segment
1905 lifetime in the net is not likely to exceed a few tens of seconds,
1906 this is deemed ample protection for foreseeable nets, even if data
1907 rates escalate to l0's of megabits/sec. At 100 megabits/sec, the
1908 cycle time is 5.4 minutes which may be a little short, but still
1909 within reason.
1911 The basic duplicate detection and sequencing algorithm in TCP can be
1912 defeated, however, if a source TCP does not have any memory of the
1913 sequence numbers it last used on a given connection. For example, if
1914 the TCP were to start all connections with sequence number 0, then
1915 upon crashing and restarting, a TCP might re-form an earlier
1916 connection (possibly after half-open connection resolution) and emit
1917 packets with sequence numbers identical to or overlapping with
1918 packets still in the network which were emitted on an earlier
1919 incarnation of the same connection. In the absence of knowledge
1920 about the sequence numbers used on a particular connection, the TCP
1921 specification recommends that the source delay for MSL seconds
1922 before emitting segments on the connection, to allow time for
1923 segments from the earlier connection incarnation to drain from the
1924 system.
1926 Even hosts which can remember the time of day and used it to select
1927 initial sequence number values are not immune from this problem
1928 (i.e., even if time of day is used to select an initial sequence
1929 number for each new connection incarnation).
1931 Suppose, for example, that a connection is opened starting with
1932 sequence number S. Suppose that this connection is not used much
1933 and that eventually the initial sequence number function (ISN(t))
1934 takes on a value equal to the sequence number, say S1, of the last
1935 segment sent by this TCP on a particular connection. Now suppose,
1936 at this instant, the host crashes, recovers, and establishes a new
1937 incarnation of the connection. The initial sequence number chosen is
1938 S1 = ISN(t) -- last used sequence number on old incarnation of
1939 connection! If the recovery occurs quickly enough, any old
1942 [Page 29]
1943 \f
1945 September 1981
1946 Transmission Control Protocol
1947 Functional Specification
1951 duplicates in the net bearing sequence numbers in the neighborhood
1952 of S1 may arrive and be treated as new packets by the receiver of
1953 the new incarnation of the connection.
1955 The problem is that the recovering host may not know for how long it
1956 crashed nor does it know whether there are still old duplicates in
1957 the system from earlier connection incarnations.
1959 One way to deal with this problem is to deliberately delay emitting
1960 segments for one MSL after recovery from a crash- this is the "quite
1961 time" specification. Hosts which prefer to avoid waiting are
1962 willing to risk possible confusion of old and new packets at a given
1963 destination may choose not to wait for the "quite time".
1964 Implementors may provide TCP users with the ability to select on a
1965 connection by connection basis whether to wait after a crash, or may
1966 informally implement the "quite time" for all connections.
1967 Obviously, even where a user selects to "wait," this is not
1968 necessary after the host has been "up" for at least MSL seconds.
1970 To summarize: every segment emitted occupies one or more sequence
1971 numbers in the sequence space, the numbers occupied by a segment are
1972 "busy" or "in use" until MSL seconds have passed, upon crashing a
1973 block of space-time is occupied by the octets of the last emitted
1974 segment, if a new connection is started too soon and uses any of the
1975 sequence numbers in the space-time footprint of the last segment of
1976 the previous connection incarnation, there is a potential sequence
1977 number overlap area which could cause confusion at the receiver.
1979 3.4. Establishing a connection
1981 The "three-way handshake" is the procedure used to establish a
1982 connection. This procedure normally is initiated by one TCP and
1983 responded to by another TCP. The procedure also works if two TCP
1984 simultaneously initiate the procedure. When simultaneous attempt
1985 occurs, each TCP receives a "SYN" segment which carries no
1986 acknowledgment after it has sent a "SYN". Of course, the arrival of
1987 an old duplicate "SYN" segment can potentially make it appear, to the
1988 recipient, that a simultaneous connection initiation is in progress.
1989 Proper use of "reset" segments can disambiguate these cases.
1991 Several examples of connection initiation follow. Although these
1992 examples do not show connection synchronization using data-carrying
1993 segments, this is perfectly legitimate, so long as the receiving TCP
1994 doesn't deliver the data to the user until it is clear the data is
1995 valid (i.e., the data must be buffered at the receiver until the
1996 connection reaches the ESTABLISHED state). The three-way handshake
1997 reduces the possibility of false connections. It is the
2001 [Page 30]
2002 \f
2004 September 1981
2005 Transmission Control Protocol
2006 Functional Specification
2010 implementation of a trade-off between memory and messages to provide
2011 information for this checking.
2013 The simplest three-way handshake is shown in figure 7 below. The
2014 figures should be interpreted in the following way. Each line is
2015 numbered for reference purposes. Right arrows (-->) indicate
2016 departure of a TCP segment from TCP A to TCP B, or arrival of a
2017 segment at B from A. Left arrows (<--), indicate the reverse.
2018 Ellipsis (...) indicates a segment which is still in the network
2019 (delayed). An "XXX" indicates a segment which is lost or rejected.
2020 Comments appear in parentheses. TCP states represent the state AFTER
2021 the departure or arrival of the segment (whose contents are shown in
2022 the center of each line). Segment contents are shown in abbreviated
2023 form, with sequence number, control flags, and ACK field. Other
2024 fields such as window, addresses, lengths, and text have been left out
2025 in the interest of clarity.
2029 TCP A TCP B
2031 1. CLOSED LISTEN
2033 2. SYN-SENT --> <SEQ=100><CTL=SYN> --> SYN-RECEIVED
2035 3. ESTABLISHED <-- <SEQ=300><ACK=101><CTL=SYN,ACK> <-- SYN-RECEIVED
2037 4. ESTABLISHED --> <SEQ=101><ACK=301><CTL=ACK> --> ESTABLISHED
2039 5. ESTABLISHED --> <SEQ=101><ACK=301><CTL=ACK><DATA> --> ESTABLISHED
2041 Basic 3-Way Handshake for Connection Synchronization
2043 Figure 7.
2045 In line 2 of figure 7, TCP A begins by sending a SYN segment
2046 indicating that it will use sequence numbers starting with sequence
2047 number 100. In line 3, TCP B sends a SYN and acknowledges the SYN it
2048 received from TCP A. Note that the acknowledgment field indicates TCP
2049 B is now expecting to hear sequence 101, acknowledging the SYN which
2050 occupied sequence 100.
2052 At line 4, TCP A responds with an empty segment containing an ACK for
2053 TCP B's SYN; and in line 5, TCP A sends some data. Note that the
2054 sequence number of the segment in line 5 is the same as in line 4
2055 because the ACK does not occupy sequence number space (if it did, we
2056 would wind up ACKing ACK's!).
2060 [Page 31]
2061 \f
2063 September 1981
2064 Transmission Control Protocol
2065 Functional Specification
2069 Simultaneous initiation is only slightly more complex, as is shown in
2070 figure 8. Each TCP cycles from CLOSED to SYN-SENT to SYN-RECEIVED to
2071 ESTABLISHED.
2075 TCP A TCP B
2077 1. CLOSED CLOSED
2079 2. SYN-SENT --> <SEQ=100><CTL=SYN> ...
2081 3. SYN-RECEIVED <-- <SEQ=300><CTL=SYN> <-- SYN-SENT
2083 4. ... <SEQ=100><CTL=SYN> --> SYN-RECEIVED
2085 5. SYN-RECEIVED --> <SEQ=100><ACK=301><CTL=SYN,ACK> ...
2087 6. ESTABLISHED <-- <SEQ=300><ACK=101><CTL=SYN,ACK> <-- SYN-RECEIVED
2089 7. ... <SEQ=101><ACK=301><CTL=ACK> --> ESTABLISHED
2091 Simultaneous Connection Synchronization
2093 Figure 8.
2095 The principle reason for the three-way handshake is to prevent old
2096 duplicate connection initiations from causing confusion. To deal with
2097 this, a special control message, reset, has been devised. If the
2098 receiving TCP is in a non-synchronized state (i.e., SYN-SENT,
2099 SYN-RECEIVED), it returns to LISTEN on receiving an acceptable reset.
2100 If the TCP is in one of the synchronized states (ESTABLISHED,
2101 FIN-WAIT-1, FIN-WAIT-2, CLOSE-WAIT, CLOSING, LAST-ACK, TIME-WAIT), it
2102 aborts the connection and informs its user. We discuss this latter
2103 case under "half-open" connections below.
2119 [Page 32]
2120 \f
2122 September 1981
2123 Transmission Control Protocol
2124 Functional Specification
2130 TCP A TCP B
2132 1. CLOSED LISTEN
2134 2. SYN-SENT --> <SEQ=100><CTL=SYN> ...
2136 3. (duplicate) ... <SEQ=90><CTL=SYN> --> SYN-RECEIVED
2138 4. SYN-SENT <-- <SEQ=300><ACK=91><CTL=SYN,ACK> <-- SYN-RECEIVED
2140 5. SYN-SENT --> <SEQ=91><CTL=RST> --> LISTEN
2143 6. ... <SEQ=100><CTL=SYN> --> SYN-RECEIVED
2145 7. SYN-SENT <-- <SEQ=400><ACK=101><CTL=SYN,ACK> <-- SYN-RECEIVED
2147 8. ESTABLISHED --> <SEQ=101><ACK=401><CTL=ACK> --> ESTABLISHED
2149 Recovery from Old Duplicate SYN
2151 Figure 9.
2153 As a simple example of recovery from old duplicates, consider
2154 figure 9. At line 3, an old duplicate SYN arrives at TCP B. TCP B
2155 cannot tell that this is an old duplicate, so it responds normally
2156 (line 4). TCP A detects that the ACK field is incorrect and returns a
2157 RST (reset) with its SEQ field selected to make the segment
2158 believable. TCP B, on receiving the RST, returns to the LISTEN state.
2159 When the original SYN (pun intended) finally arrives at line 6, the
2160 synchronization proceeds normally. If the SYN at line 6 had arrived
2161 before the RST, a more complex exchange might have occurred with RST's
2162 sent in both directions.
2164 Half-Open Connections and Other Anomalies
2166 An established connection is said to be "half-open" if one of the
2167 TCPs has closed or aborted the connection at its end without the
2168 knowledge of the other, or if the two ends of the connection have
2169 become desynchronized owing to a crash that resulted in loss of
2170 memory. Such connections will automatically become reset if an
2171 attempt is made to send data in either direction. However, half-open
2172 connections are expected to be unusual, and the recovery procedure is
2173 mildly involved.
2175 If at site A the connection no longer exists, then an attempt by the
2178 [Page 33]
2179 \f
2181 September 1981
2182 Transmission Control Protocol
2183 Functional Specification
2187 user at site B to send any data on it will result in the site B TCP
2188 receiving a reset control message. Such a message indicates to the
2189 site B TCP that something is wrong, and it is expected to abort the
2190 connection.
2192 Assume that two user processes A and B are communicating with one
2193 another when a crash occurs causing loss of memory to A's TCP.
2194 Depending on the operating system supporting A's TCP, it is likely
2195 that some error recovery mechanism exists. When the TCP is up again,
2196 A is likely to start again from the beginning or from a recovery
2197 point. As a result, A will probably try to OPEN the connection again
2198 or try to SEND on the connection it believes open. In the latter
2199 case, it receives the error message "connection not open" from the
2200 local (A's) TCP. In an attempt to establish the connection, A's TCP
2201 will send a segment containing SYN. This scenario leads to the
2202 example shown in figure 10. After TCP A crashes, the user attempts to
2203 re-open the connection. TCP B, in the meantime, thinks the connection
2204 is open.
2208 TCP A TCP B
2210 1. (CRASH) (send 300,receive 100)
2212 2. CLOSED ESTABLISHED
2214 3. SYN-SENT --> <SEQ=400><CTL=SYN> --> (??)
2216 4. (!!) <-- <SEQ=300><ACK=100><CTL=ACK> <-- ESTABLISHED
2218 5. SYN-SENT --> <SEQ=100><CTL=RST> --> (Abort!!)
2220 6. SYN-SENT CLOSED
2222 7. SYN-SENT --> <SEQ=400><CTL=SYN> -->
2224 Half-Open Connection Discovery
2226 Figure 10.
2228 When the SYN arrives at line 3, TCP B, being in a synchronized state,
2229 and the incoming segment outside the window, responds with an
2230 acknowledgment indicating what sequence it next expects to hear (ACK
2231 100). TCP A sees that this segment does not acknowledge anything it
2232 sent and, being unsynchronized, sends a reset (RST) because it has
2233 detected a half-open connection. TCP B aborts at line 5. TCP A will
2237 [Page 34]
2238 \f
2240 September 1981
2241 Transmission Control Protocol
2242 Functional Specification
2246 continue to try to establish the connection; the problem is now
2247 reduced to the basic 3-way handshake of figure 7.
2249 An interesting alternative case occurs when TCP A crashes and TCP B
2250 tries to send data on what it thinks is a synchronized connection.
2251 This is illustrated in figure 11. In this case, the data arriving at
2252 TCP A from TCP B (line 2) is unacceptable because no such connection
2253 exists, so TCP A sends a RST. The RST is acceptable so TCP B
2254 processes it and aborts the connection.
2258 TCP A TCP B
2260 1. (CRASH) (send 300,receive 100)
2262 2. (??) <-- <SEQ=300><ACK=100><DATA=10><CTL=ACK> <-- ESTABLISHED
2264 3. --> <SEQ=100><CTL=RST> --> (ABORT!!)
2266 Active Side Causes Half-Open Connection Discovery
2268 Figure 11.
2270 In figure 12, we find the two TCPs A and B with passive connections
2271 waiting for SYN. An old duplicate arriving at TCP B (line 2) stirs B
2272 into action. A SYN-ACK is returned (line 3) and causes TCP A to
2273 generate a RST (the ACK in line 3 is not acceptable). TCP B accepts
2274 the reset and returns to its passive LISTEN state.
2278 TCP A TCP B
2280 1. LISTEN LISTEN
2282 2. ... <SEQ=Z><CTL=SYN> --> SYN-RECEIVED
2284 3. (??) <-- <SEQ=X><ACK=Z+1><CTL=SYN,ACK> <-- SYN-RECEIVED
2286 4. --> <SEQ=Z+1><CTL=RST> --> (return to LISTEN!)
2288 5. LISTEN LISTEN
2290 Old Duplicate SYN Initiates a Reset on two Passive Sockets
2292 Figure 12.
2296 [Page 35]
2297 \f
2299 September 1981
2300 Transmission Control Protocol
2301 Functional Specification
2305 A variety of other cases are possible, all of which are accounted for
2306 by the following rules for RST generation and processing.
2308 Reset Generation
2310 As a general rule, reset (RST) must be sent whenever a segment arrives
2311 which apparently is not intended for the current connection. A reset
2312 must not be sent if it is not clear that this is the case.
2314 There are three groups of states:
2316 1. If the connection does not exist (CLOSED) then a reset is sent
2317 in response to any incoming segment except another reset. In
2318 particular, SYNs addressed to a non-existent connection are rejected
2319 by this means.
2321 If the incoming segment has an ACK field, the reset takes its
2322 sequence number from the ACK field of the segment, otherwise the
2323 reset has sequence number zero and the ACK field is set to the sum
2324 of the sequence number and segment length of the incoming segment.
2325 The connection remains in the CLOSED state.
2327 2. If the connection is in any non-synchronized state (LISTEN,
2328 SYN-SENT, SYN-RECEIVED), and the incoming segment acknowledges
2329 something not yet sent (the segment carries an unacceptable ACK), or
2330 if an incoming segment has a security level or compartment which
2331 does not exactly match the level and compartment requested for the
2332 connection, a reset is sent.
2334 If our SYN has not been acknowledged and the precedence level of the
2335 incoming segment is higher than the precedence level requested then
2336 either raise the local precedence level (if allowed by the user and
2337 the system) or send a reset; or if the precedence level of the
2338 incoming segment is lower than the precedence level requested then
2339 continue as if the precedence matched exactly (if the remote TCP
2340 cannot raise the precedence level to match ours this will be
2341 detected in the next segment it sends, and the connection will be
2342 terminated then). If our SYN has been acknowledged (perhaps in this
2343 incoming segment) the precedence level of the incoming segment must
2344 match the local precedence level exactly, if it does not a reset
2345 must be sent.
2347 If the incoming segment has an ACK field, the reset takes its
2348 sequence number from the ACK field of the segment, otherwise the
2349 reset has sequence number zero and the ACK field is set to the sum
2350 of the sequence number and segment length of the incoming segment.
2351 The connection remains in the same state.
2355 [Page 36]
2356 \f
2358 September 1981
2359 Transmission Control Protocol
2360 Functional Specification
2364 3. If the connection is in a synchronized state (ESTABLISHED,
2365 FIN-WAIT-1, FIN-WAIT-2, CLOSE-WAIT, CLOSING, LAST-ACK, TIME-WAIT),
2366 any unacceptable segment (out of window sequence number or
2367 unacceptible acknowledgment number) must elicit only an empty
2368 acknowledgment segment containing the current send-sequence number
2369 and an acknowledgment indicating the next sequence number expected
2370 to be received, and the connection remains in the same state.
2372 If an incoming segment has a security level, or compartment, or
2373 precedence which does not exactly match the level, and compartment,
2374 and precedence requested for the connection,a reset is sent and
2375 connection goes to the CLOSED state. The reset takes its sequence
2376 number from the ACK field of the incoming segment.
2378 Reset Processing
2380 In all states except SYN-SENT, all reset (RST) segments are validated
2381 by checking their SEQ-fields. A reset is valid if its sequence number
2382 is in the window. In the SYN-SENT state (a RST received in response
2383 to an initial SYN), the RST is acceptable if the ACK field
2384 acknowledges the SYN.
2386 The receiver of a RST first validates it, then changes state. If the
2387 receiver was in the LISTEN state, it ignores it. If the receiver was
2388 in SYN-RECEIVED state and had previously been in the LISTEN state,
2389 then the receiver returns to the LISTEN state, otherwise the receiver
2390 aborts the connection and goes to the CLOSED state. If the receiver
2391 was in any other state, it aborts the connection and advises the user
2392 and goes to the CLOSED state.
2394 3.5. Closing a Connection
2396 CLOSE is an operation meaning "I have no more data to send." The
2397 notion of closing a full-duplex connection is subject to ambiguous
2398 interpretation, of course, since it may not be obvious how to treat
2399 the receiving side of the connection. We have chosen to treat CLOSE
2400 in a simplex fashion. The user who CLOSEs may continue to RECEIVE
2401 until he is told that the other side has CLOSED also. Thus, a program
2402 could initiate several SENDs followed by a CLOSE, and then continue to
2403 RECEIVE until signaled that a RECEIVE failed because the other side
2404 has CLOSED. We assume that the TCP will signal a user, even if no
2405 RECEIVEs are outstanding, that the other side has closed, so the user
2406 can terminate his side gracefully. A TCP will reliably deliver all
2407 buffers SENT before the connection was CLOSED so a user who expects no
2408 data in return need only wait to hear the connection was CLOSED
2409 successfully to know that all his data was received at the destination
2410 TCP. Users must keep reading connections they close for sending until
2411 the TCP says no more data.
2414 [Page 37]
2415 \f
2417 September 1981
2418 Transmission Control Protocol
2419 Functional Specification
2423 There are essentially three cases:
2425 1) The user initiates by telling the TCP to CLOSE the connection
2427 2) The remote TCP initiates by sending a FIN control signal
2429 3) Both users CLOSE simultaneously
2431 Case 1: Local user initiates the close
2433 In this case, a FIN segment can be constructed and placed on the
2434 outgoing segment queue. No further SENDs from the user will be
2435 accepted by the TCP, and it enters the FIN-WAIT-1 state. RECEIVEs
2436 are allowed in this state. All segments preceding and including FIN
2437 will be retransmitted until acknowledged. When the other TCP has
2438 both acknowledged the FIN and sent a FIN of its own, the first TCP
2439 can ACK this FIN. Note that a TCP receiving a FIN will ACK but not
2440 send its own FIN until its user has CLOSED the connection also.
2442 Case 2: TCP receives a FIN from the network
2444 If an unsolicited FIN arrives from the network, the receiving TCP
2445 can ACK it and tell the user that the connection is closing. The
2446 user will respond with a CLOSE, upon which the TCP can send a FIN to
2447 the other TCP after sending any remaining data. The TCP then waits
2448 until its own FIN is acknowledged whereupon it deletes the
2449 connection. If an ACK is not forthcoming, after the user timeout
2450 the connection is aborted and the user is told.
2452 Case 3: both users close simultaneously
2454 A simultaneous CLOSE by users at both ends of a connection causes
2455 FIN segments to be exchanged. When all segments preceding the FINs
2456 have been processed and acknowledged, each TCP can ACK the FIN it
2457 has received. Both will, upon receiving these ACKs, delete the
2458 connection.
2473 [Page 38]
2474 \f
2476 September 1981
2477 Transmission Control Protocol
2478 Functional Specification
2484 TCP A TCP B
2486 1. ESTABLISHED ESTABLISHED
2488 2. (Close)
2489 FIN-WAIT-1 --> <SEQ=100><ACK=300><CTL=FIN,ACK> --> CLOSE-WAIT
2491 3. FIN-WAIT-2 <-- <SEQ=300><ACK=101><CTL=ACK> <-- CLOSE-WAIT
2493 4. (Close)
2494 TIME-WAIT <-- <SEQ=300><ACK=101><CTL=FIN,ACK> <-- LAST-ACK
2496 5. TIME-WAIT --> <SEQ=101><ACK=301><CTL=ACK> --> CLOSED
2498 6. (2 MSL)
2499 CLOSED
2501 Normal Close Sequence
2503 Figure 13.
2507 TCP A TCP B
2509 1. ESTABLISHED ESTABLISHED
2511 2. (Close) (Close)
2512 FIN-WAIT-1 --> <SEQ=100><ACK=300><CTL=FIN,ACK> ... FIN-WAIT-1
2513 <-- <SEQ=300><ACK=100><CTL=FIN,ACK> <--
2514 ... <SEQ=100><ACK=300><CTL=FIN,ACK> -->
2516 3. CLOSING --> <SEQ=101><ACK=301><CTL=ACK> ... CLOSING
2517 <-- <SEQ=301><ACK=101><CTL=ACK> <--
2518 ... <SEQ=101><ACK=301><CTL=ACK> -->
2520 4. TIME-WAIT TIME-WAIT
2521 (2 MSL) (2 MSL)
2522 CLOSED CLOSED
2524 Simultaneous Close Sequence
2526 Figure 14.
2532 [Page 39]
2533 \f
2535 September 1981
2536 Transmission Control Protocol
2537 Functional Specification
2541 3.6. Precedence and Security
2543 The intent is that connection be allowed only between ports operating
2544 with exactly the same security and compartment values and at the
2545 higher of the precedence level requested by the two ports.
2547 The precedence and security parameters used in TCP are exactly those
2548 defined in the Internet Protocol (IP) [2]. Throughout this TCP
2549 specification the term "security/compartment" is intended to indicate
2550 the security parameters used in IP including security, compartment,
2551 user group, and handling restriction.
2553 A connection attempt with mismatched security/compartment values or a
2554 lower precedence value must be rejected by sending a reset. Rejecting
2555 a connection due to too low a precedence only occurs after an
2556 acknowledgment of the SYN has been received.
2558 Note that TCP modules which operate only at the default value of
2559 precedence will still have to check the precedence of incoming
2560 segments and possibly raise the precedence level they use on the
2561 connection.
2563 The security paramaters may be used even in a non-secure environment
2564 (the values would indicate unclassified data), thus hosts in
2565 non-secure environments must be prepared to receive the security
2566 parameters, though they need not send them.
2568 3.7. Data Communication
2570 Once the connection is established data is communicated by the
2571 exchange of segments. Because segments may be lost due to errors
2572 (checksum test failure), or network congestion, TCP uses
2573 retransmission (after a timeout) to ensure delivery of every segment.
2574 Duplicate segments may arrive due to network or TCP retransmission.
2575 As discussed in the section on sequence numbers the TCP performs
2576 certain tests on the sequence and acknowledgment numbers in the
2577 segments to verify their acceptability.
2579 The sender of data keeps track of the next sequence number to use in
2580 the variable SND.NXT. The receiver of data keeps track of the next
2581 sequence number to expect in the variable RCV.NXT. The sender of data
2582 keeps track of the oldest unacknowledged sequence number in the
2583 variable SND.UNA. If the data flow is momentarily idle and all data
2584 sent has been acknowledged then the three variables will be equal.
2586 When the sender creates a segment and transmits it the sender advances
2587 SND.NXT. When the receiver accepts a segment it advances RCV.NXT and
2588 sends an acknowledgment. When the data sender receives an
2591 [Page 40]
2592 \f
2594 September 1981
2595 Transmission Control Protocol
2596 Functional Specification
2600 acknowledgment it advances SND.UNA. The extent to which the values of
2601 these variables differ is a measure of the delay in the communication.
2602 The amount by which the variables are advanced is the length of the
2603 data in the segment. Note that once in the ESTABLISHED state all
2604 segments must carry current acknowledgment information.
2606 The CLOSE user call implies a push function, as does the FIN control
2607 flag in an incoming segment.
2609 Retransmission Timeout
2611 Because of the variability of the networks that compose an
2612 internetwork system and the wide range of uses of TCP connections the
2613 retransmission timeout must be dynamically determined. One procedure
2614 for determining a retransmission time out is given here as an
2615 illustration.
2617 An Example Retransmission Timeout Procedure
2619 Measure the elapsed time between sending a data octet with a
2620 particular sequence number and receiving an acknowledgment that
2621 covers that sequence number (segments sent do not have to match
2622 segments received). This measured elapsed time is the Round Trip
2623 Time (RTT). Next compute a Smoothed Round Trip Time (SRTT) as:
2625 SRTT = ( ALPHA * SRTT ) + ((1-ALPHA) * RTT)
2627 and based on this, compute the retransmission timeout (RTO) as:
2629 RTO = min[UBOUND,max[LBOUND,(BETA*SRTT)]]
2631 where UBOUND is an upper bound on the timeout (e.g., 1 minute),
2632 LBOUND is a lower bound on the timeout (e.g., 1 second), ALPHA is
2633 a smoothing factor (e.g., .8 to .9), and BETA is a delay variance
2634 factor (e.g., 1.3 to 2.0).
2636 The Communication of Urgent Information
2638 The objective of the TCP urgent mechanism is to allow the sending user
2639 to stimulate the receiving user to accept some urgent data and to
2640 permit the receiving TCP to indicate to the receiving user when all
2641 the currently known urgent data has been received by the user.
2643 This mechanism permits a point in the data stream to be designated as
2644 the end of urgent information. Whenever this point is in advance of
2645 the receive sequence number (RCV.NXT) at the receiving TCP, that TCP
2646 must tell the user to go into "urgent mode"; when the receive sequence
2647 number catches up to the urgent pointer, the TCP must tell user to go
2650 [Page 41]
2651 \f
2653 September 1981
2654 Transmission Control Protocol
2655 Functional Specification
2659 into "normal mode". If the urgent pointer is updated while the user
2660 is in "urgent mode", the update will be invisible to the user.
2662 The method employs a urgent field which is carried in all segments
2663 transmitted. The URG control flag indicates that the urgent field is
2664 meaningful and must be added to the segment sequence number to yield
2665 the urgent pointer. The absence of this flag indicates that there is
2666 no urgent data outstanding.
2668 To send an urgent indication the user must also send at least one data
2669 octet. If the sending user also indicates a push, timely delivery of
2670 the urgent information to the destination process is enhanced.
2672 Managing the Window
2674 The window sent in each segment indicates the range of sequence
2675 numbers the sender of the window (the data receiver) is currently
2676 prepared to accept. There is an assumption that this is related to
2677 the currently available data buffer space available for this
2678 connection.
2680 Indicating a large window encourages transmissions. If more data
2681 arrives than can be accepted, it will be discarded. This will result
2682 in excessive retransmissions, adding unnecessarily to the load on the
2683 network and the TCPs. Indicating a small window may restrict the
2684 transmission of data to the point of introducing a round trip delay
2685 between each new segment transmitted.
2687 The mechanisms provided allow a TCP to advertise a large window and to
2688 subsequently advertise a much smaller window without having accepted
2689 that much data. This, so called "shrinking the window," is strongly
2690 discouraged. The robustness principle dictates that TCPs will not
2691 shrink the window themselves, but will be prepared for such behavior
2692 on the part of other TCPs.
2694 The sending TCP must be prepared to accept from the user and send at
2695 least one octet of new data even if the send window is zero. The
2696 sending TCP must regularly retransmit to the receiving TCP even when
2697 the window is zero. Two minutes is recommended for the retransmission
2698 interval when the window is zero. This retransmission is essential to
2699 guarantee that when either TCP has a zero window the re-opening of the
2700 window will be reliably reported to the other.
2702 When the receiving TCP has a zero window and a segment arrives it must
2703 still send an acknowledgment showing its next expected sequence number
2704 and current window (zero).
2706 The sending TCP packages the data to be transmitted into segments
2709 [Page 42]
2710 \f
2712 September 1981
2713 Transmission Control Protocol
2714 Functional Specification
2718 which fit the current window, and may repackage segments on the
2719 retransmission queue. Such repackaging is not required, but may be
2720 helpful.
2722 In a connection with a one-way data flow, the window information will
2723 be carried in acknowledgment segments that all have the same sequence
2724 number so there will be no way to reorder them if they arrive out of
2725 order. This is not a serious problem, but it will allow the window
2726 information to be on occasion temporarily based on old reports from
2727 the data receiver. A refinement to avoid this problem is to act on
2728 the window information from segments that carry the highest
2729 acknowledgment number (that is segments with acknowledgment number
2730 equal or greater than the highest previously received).
2732 The window management procedure has significant influence on the
2733 communication performance. The following comments are suggestions to
2734 implementers.
2736 Window Management Suggestions
2738 Allocating a very small window causes data to be transmitted in
2739 many small segments when better performance is achieved using
2740 fewer large segments.
2742 One suggestion for avoiding small windows is for the receiver to
2743 defer updating a window until the additional allocation is at
2744 least X percent of the maximum allocation possible for the
2745 connection (where X might be 20 to 40).
2747 Another suggestion is for the sender to avoid sending small
2748 segments by waiting until the window is large enough before
2749 sending data. If the the user signals a push function then the
2750 data must be sent even if it is a small segment.
2752 Note that the acknowledgments should not be delayed or unnecessary
2753 retransmissions will result. One strategy would be to send an
2754 acknowledgment when a small segment arrives (with out updating the
2755 window information), and then to send another acknowledgment with
2756 new window information when the window is larger.
2758 The segment sent to probe a zero window may also begin a break up
2759 of transmitted data into smaller and smaller segments. If a
2760 segment containing a single data octet sent to probe a zero window
2761 is accepted, it consumes one octet of the window now available.
2762 If the sending TCP simply sends as much as it can whenever the
2763 window is non zero, the transmitted data will be broken into
2764 alternating big and small segments. As time goes on, occasional
2765 pauses in the receiver making window allocation available will
2768 [Page 43]
2769 \f
2771 September 1981
2772 Transmission Control Protocol
2773 Functional Specification
2777 result in breaking the big segments into a small and not quite so
2778 big pair. And after a while the data transmission will be in
2779 mostly small segments.
2781 The suggestion here is that the TCP implementations need to
2782 actively attempt to combine small window allocations into larger
2783 windows, since the mechanisms for managing the window tend to lead
2784 to many small windows in the simplest minded implementations.
2786 3.8. Interfaces
2788 There are of course two interfaces of concern: the user/TCP interface
2789 and the TCP/lower-level interface. We have a fairly elaborate model
2790 of the user/TCP interface, but the interface to the lower level
2791 protocol module is left unspecified here, since it will be specified
2792 in detail by the specification of the lowel level protocol. For the
2793 case that the lower level is IP we note some of the parameter values
2794 that TCPs might use.
2796 User/TCP Interface
2798 The following functional description of user commands to the TCP is,
2799 at best, fictional, since every operating system will have different
2800 facilities. Consequently, we must warn readers that different TCP
2801 implementations may have different user interfaces. However, all
2802 TCPs must provide a certain minimum set of services to guarantee
2803 that all TCP implementations can support the same protocol
2804 hierarchy. This section specifies the functional interfaces
2805 required of all TCP implementations.
2807 TCP User Commands
2809 The following sections functionally characterize a USER/TCP
2810 interface. The notation used is similar to most procedure or
2811 function calls in high level languages, but this usage is not
2812 meant to rule out trap type service calls (e.g., SVCs, UUOs,
2813 EMTs).
2815 The user commands described below specify the basic functions the
2816 TCP must perform to support interprocess communication.
2817 Individual implementations must define their own exact format, and
2818 may provide combinations or subsets of the basic functions in
2819 single calls. In particular, some implementations may wish to
2820 automatically OPEN a connection on the first SEND or RECEIVE
2821 issued by the user for a given connection.
2827 [Page 44]
2828 \f
2830 September 1981
2831 Transmission Control Protocol
2832 Functional Specification
2836 In providing interprocess communication facilities, the TCP must
2837 not only accept commands, but must also return information to the
2838 processes it serves. The latter consists of:
2840 (a) general information about a connection (e.g., interrupts,
2841 remote close, binding of unspecified foreign socket).
2843 (b) replies to specific user commands indicating success or
2844 various types of failure.
2846 Open
2848 Format: OPEN (local port, foreign socket, active/passive
2849 [, timeout] [, precedence] [, security/compartment] [, options])
2850 -> local connection name
2852 We assume that the local TCP is aware of the identity of the
2853 processes it serves and will check the authority of the process
2854 to use the connection specified. Depending upon the
2855 implementation of the TCP, the local network and TCP identifiers
2856 for the source address will either be supplied by the TCP or the
2857 lower level protocol (e.g., IP). These considerations are the
2858 result of concern about security, to the extent that no TCP be
2859 able to masquerade as another one, and so on. Similarly, no
2860 process can masquerade as another without the collusion of the
2861 TCP.
2863 If the active/passive flag is set to passive, then this is a
2864 call to LISTEN for an incoming connection. A passive open may
2865 have either a fully specified foreign socket to wait for a
2866 particular connection or an unspecified foreign socket to wait
2867 for any call. A fully specified passive call can be made active
2868 by the subsequent execution of a SEND.
2870 A transmission control block (TCB) is created and partially
2871 filled in with data from the OPEN command parameters.
2873 On an active OPEN command, the TCP will begin the procedure to
2874 synchronize (i.e., establish) the connection at once.
2876 The timeout, if present, permits the caller to set up a timeout
2877 for all data submitted to TCP. If data is not successfully
2878 delivered to the destination within the timeout period, the TCP
2879 will abort the connection. The present global default is five
2880 minutes.
2882 The TCP or some component of the operating system will verify
2883 the users authority to open a connection with the specified
2886 [Page 45]
2887 \f
2889 September 1981
2890 Transmission Control Protocol
2891 Functional Specification
2895 precedence or security/compartment. The absence of precedence
2896 or security/compartment specification in the OPEN call indicates
2897 the default values must be used.
2899 TCP will accept incoming requests as matching only if the
2900 security/compartment information is exactly the same and only if
2901 the precedence is equal to or higher than the precedence
2902 requested in the OPEN call.
2904 The precedence for the connection is the higher of the values
2905 requested in the OPEN call and received from the incoming
2906 request, and fixed at that value for the life of the
2907 connection.Implementers may want to give the user control of
2908 this precedence negotiation. For example, the user might be
2909 allowed to specify that the precedence must be exactly matched,
2910 or that any attempt to raise the precedence be confirmed by the
2911 user.
2913 A local connection name will be returned to the user by the TCP.
2914 The local connection name can then be used as a short hand term
2915 for the connection defined by the <local socket, foreign socket>
2916 pair.
2918 Send
2920 Format: SEND (local connection name, buffer address, byte
2921 count, PUSH flag, URGENT flag [,timeout])
2923 This call causes the data contained in the indicated user buffer
2924 to be sent on the indicated connection. If the connection has
2925 not been opened, the SEND is considered an error. Some
2926 implementations may allow users to SEND first; in which case, an
2927 automatic OPEN would be done. If the calling process is not
2928 authorized to use this connection, an error is returned.
2930 If the PUSH flag is set, the data must be transmitted promptly
2931 to the receiver, and the PUSH bit will be set in the last TCP
2932 segment created from the buffer. If the PUSH flag is not set,
2933 the data may be combined with data from subsequent SENDs for
2934 transmission efficiency.
2936 If the URGENT flag is set, segments sent to the destination TCP
2937 will have the urgent pointer set. The receiving TCP will signal
2938 the urgent condition to the receiving process if the urgent
2939 pointer indicates that data preceding the urgent pointer has not
2940 been consumed by the receiving process. The purpose of urgent
2941 is to stimulate the receiver to process the urgent data and to
2942 indicate to the receiver when all the currently known urgent
2945 [Page 46]
2946 \f
2948 September 1981
2949 Transmission Control Protocol
2950 Functional Specification
2954 data has been received. The number of times the sending user's
2955 TCP signals urgent will not necessarily be equal to the number
2956 of times the receiving user will be notified of the presence of
2957 urgent data.
2959 If no foreign socket was specified in the OPEN, but the
2960 connection is established (e.g., because a LISTENing connection
2961 has become specific due to a foreign segment arriving for the
2962 local socket), then the designated buffer is sent to the implied
2963 foreign socket. Users who make use of OPEN with an unspecified
2964 foreign socket can make use of SEND without ever explicitly
2965 knowing the foreign socket address.
2967 However, if a SEND is attempted before the foreign socket
2968 becomes specified, an error will be returned. Users can use the
2969 STATUS call to determine the status of the connection. In some
2970 implementations the TCP may notify the user when an unspecified
2971 socket is bound.
2973 If a timeout is specified, the current user timeout for this
2974 connection is changed to the new one.
2976 In the simplest implementation, SEND would not return control to
2977 the sending process until either the transmission was complete
2978 or the timeout had been exceeded. However, this simple method
2979 is both subject to deadlocks (for example, both sides of the
2980 connection might try to do SENDs before doing any RECEIVEs) and
2981 offers poor performance, so it is not recommended. A more
2982 sophisticated implementation would return immediately to allow
2983 the process to run concurrently with network I/O, and,
2984 furthermore, to allow multiple SENDs to be in progress.
2985 Multiple SENDs are served in first come, first served order, so
2986 the TCP will queue those it cannot service immediately.
2988 We have implicitly assumed an asynchronous user interface in
2989 which a SEND later elicits some kind of SIGNAL or
2990 pseudo-interrupt from the serving TCP. An alternative is to
2991 return a response immediately. For instance, SENDs might return
2992 immediate local acknowledgment, even if the segment sent had not
2993 been acknowledged by the distant TCP. We could optimistically
2994 assume eventual success. If we are wrong, the connection will
2995 close anyway due to the timeout. In implementations of this
2996 kind (synchronous), there will still be some asynchronous
2997 signals, but these will deal with the connection itself, and not
2998 with specific segments or buffers.
3000 In order for the process to distinguish among error or success
3001 indications for different SENDs, it might be appropriate for the
3004 [Page 47]
3005 \f
3007 September 1981
3008 Transmission Control Protocol
3009 Functional Specification
3013 buffer address to be returned along with the coded response to
3014 the SEND request. TCP-to-user signals are discussed below,
3015 indicating the information which should be returned to the
3016 calling process.
3018 Receive
3020 Format: RECEIVE (local connection name, buffer address, byte
3021 count) -> byte count, urgent flag, push flag
3023 This command allocates a receiving buffer associated with the
3024 specified connection. If no OPEN precedes this command or the
3025 calling process is not authorized to use this connection, an
3026 error is returned.
3028 In the simplest implementation, control would not return to the
3029 calling program until either the buffer was filled, or some
3030 error occurred, but this scheme is highly subject to deadlocks.
3031 A more sophisticated implementation would permit several
3032 RECEIVEs to be outstanding at once. These would be filled as
3033 segments arrive. This strategy permits increased throughput at
3034 the cost of a more elaborate scheme (possibly asynchronous) to
3035 notify the calling program that a PUSH has been seen or a buffer
3036 filled.
3038 If enough data arrive to fill the buffer before a PUSH is seen,
3039 the PUSH flag will not be set in the response to the RECEIVE.
3040 The buffer will be filled with as much data as it can hold. If
3041 a PUSH is seen before the buffer is filled the buffer will be
3042 returned partially filled and PUSH indicated.
3044 If there is urgent data the user will have been informed as soon
3045 as it arrived via a TCP-to-user signal. The receiving user
3046 should thus be in "urgent mode". If the URGENT flag is on,
3047 additional urgent data remains. If the URGENT flag is off, this
3048 call to RECEIVE has returned all the urgent data, and the user
3049 may now leave "urgent mode". Note that data following the
3050 urgent pointer (non-urgent data) cannot be delivered to the user
3051 in the same buffer with preceeding urgent data unless the
3052 boundary is clearly marked for the user.
3054 To distinguish among several outstanding RECEIVEs and to take
3055 care of the case that a buffer is not completely filled, the
3056 return code is accompanied by both a buffer pointer and a byte
3057 count indicating the actual length of the data received.
3059 Alternative implementations of RECEIVE might have the TCP
3063 [Page 48]
3064 \f
3066 September 1981
3067 Transmission Control Protocol
3068 Functional Specification
3072 allocate buffer storage, or the TCP might share a ring buffer
3073 with the user.
3075 Close
3077 Format: CLOSE (local connection name)
3079 This command causes the connection specified to be closed. If
3080 the connection is not open or the calling process is not
3081 authorized to use this connection, an error is returned.
3082 Closing connections is intended to be a graceful operation in
3083 the sense that outstanding SENDs will be transmitted (and
3084 retransmitted), as flow control permits, until all have been
3085 serviced. Thus, it should be acceptable to make several SEND
3086 calls, followed by a CLOSE, and expect all the data to be sent
3087 to the destination. It should also be clear that users should
3088 continue to RECEIVE on CLOSING connections, since the other side
3089 may be trying to transmit the last of its data. Thus, CLOSE
3090 means "I have no more to send" but does not mean "I will not
3091 receive any more." It may happen (if the user level protocol is
3092 not well thought out) that the closing side is unable to get rid
3093 of all its data before timing out. In this event, CLOSE turns
3094 into ABORT, and the closing TCP gives up.
3096 The user may CLOSE the connection at any time on his own
3097 initiative, or in response to various prompts from the TCP
3098 (e.g., remote close executed, transmission timeout exceeded,
3099 destination inaccessible).
3101 Because closing a connection requires communication with the
3102 foreign TCP, connections may remain in the closing state for a
3103 short time. Attempts to reopen the connection before the TCP
3104 replies to the CLOSE command will result in error responses.
3106 Close also implies push function.
3108 Status
3110 Format: STATUS (local connection name) -> status data
3112 This is an implementation dependent user command and could be
3113 excluded without adverse effect. Information returned would
3114 typically come from the TCB associated with the connection.
3116 This command returns a data block containing the following
3117 information:
3119 local socket,
3122 [Page 49]
3123 \f
3125 September 1981
3126 Transmission Control Protocol
3127 Functional Specification
3131 foreign socket,
3132 local connection name,
3133 receive window,
3134 send window,
3135 connection state,
3136 number of buffers awaiting acknowledgment,
3137 number of buffers pending receipt,
3138 urgent state,
3139 precedence,
3140 security/compartment,
3141 and transmission timeout.
3143 Depending on the state of the connection, or on the
3144 implementation itself, some of this information may not be
3145 available or meaningful. If the calling process is not
3146 authorized to use this connection, an error is returned. This
3147 prevents unauthorized processes from gaining information about a
3148 connection.
3150 Abort
3152 Format: ABORT (local connection name)
3154 This command causes all pending SENDs and RECEIVES to be
3155 aborted, the TCB to be removed, and a special RESET message to
3156 be sent to the TCP on the other side of the connection.
3157 Depending on the implementation, users may receive abort
3158 indications for each outstanding SEND or RECEIVE, or may simply
3159 receive an ABORT-acknowledgment.
3161 TCP-to-User Messages
3163 It is assumed that the operating system environment provides a
3164 means for the TCP to asynchronously signal the user program. When
3165 the TCP does signal a user program, certain information is passed
3166 to the user. Often in the specification the information will be
3167 an error message. In other cases there will be information
3168 relating to the completion of processing a SEND or RECEIVE or
3169 other user call.
3171 The following information is provided:
3173 Local Connection Name Always
3174 Response String Always
3175 Buffer Address Send & Receive
3176 Byte count (counts bytes received) Receive
3177 Push flag Receive
3178 Urgent flag Receive
3181 [Page 50]
3182 \f
3184 September 1981
3185 Transmission Control Protocol
3186 Functional Specification
3190 TCP/Lower-Level Interface
3192 The TCP calls on a lower level protocol module to actually send and
3193 receive information over a network. One case is that of the ARPA
3194 internetwork system where the lower level module is the Internet
3195 Protocol (IP) [2].
3197 If the lower level protocol is IP it provides arguments for a type
3198 of service and for a time to live. TCP uses the following settings
3199 for these parameters:
3201 Type of Service = Precedence: routine, Delay: normal, Throughput:
3202 normal, Reliability: normal; or 00000000.
3204 Time to Live = one minute, or 00111100.
3206 Note that the assumed maximum segment lifetime is two minutes.
3207 Here we explicitly ask that a segment be destroyed if it cannot
3208 be delivered by the internet system within one minute.
3210 If the lower level is IP (or other protocol that provides this
3211 feature) and source routing is used, the interface must allow the
3212 route information to be communicated. This is especially important
3213 so that the source and destination addresses used in the TCP
3214 checksum be the originating source and ultimate destination. It is
3215 also important to preserve the return route to answer connection
3216 requests.
3218 Any lower level protocol will have to provide the source address,
3219 destination address, and protocol fields, and some way to determine
3220 the "TCP length", both to provide the functional equivlent service
3221 of IP and to be used in the TCP checksum.
3240 [Page 51]
3241 \f
3243 September 1981
3244 Transmission Control Protocol
3245 Functional Specification
3249 3.9. Event Processing
3251 The processing depicted in this section is an example of one possible
3252 implementation. Other implementations may have slightly different
3253 processing sequences, but they should differ from those in this
3254 section only in detail, not in substance.
3256 The activity of the TCP can be characterized as responding to events.
3257 The events that occur can be cast into three categories: user calls,
3258 arriving segments, and timeouts. This section describes the
3259 processing the TCP does in response to each of the events. In many
3260 cases the processing required depends on the state of the connection.
3262 Events that occur:
3264 User Calls
3266 OPEN
3267 SEND
3268 RECEIVE
3269 CLOSE
3270 ABORT
3271 STATUS
3273 Arriving Segments
3275 SEGMENT ARRIVES
3277 Timeouts
3279 USER TIMEOUT
3280 RETRANSMISSION TIMEOUT
3281 TIME-WAIT TIMEOUT
3283 The model of the TCP/user interface is that user commands receive an
3284 immediate return and possibly a delayed response via an event or
3285 pseudo interrupt. In the following descriptions, the term "signal"
3286 means cause a delayed response.
3288 Error responses are given as character strings. For example, user
3289 commands referencing connections that do not exist receive "error:
3290 connection not open".
3292 Please note in the following that all arithmetic on sequence numbers,
3293 acknowledgment numbers, windows, et cetera, is modulo 2**32 the size
3294 of the sequence number space. Also note that "=<" means less than or
3295 equal to (modulo 2**32).
3299 [Page 52]
3300 \f
3302 September 1981
3303 Transmission Control Protocol
3304 Functional Specification
3308 A natural way to think about processing incoming segments is to
3309 imagine that they are first tested for proper sequence number (i.e.,
3310 that their contents lie in the range of the expected "receive window"
3311 in the sequence number space) and then that they are generally queued
3312 and processed in sequence number order.
3314 When a segment overlaps other already received segments we reconstruct
3315 the segment to contain just the new data, and adjust the header fields
3316 to be consistent.
3318 Note that if no state change is mentioned the TCP stays in the same
3319 state.
3358 [Page 53]
3359 \f
3361 September 1981
3362 Transmission Control Protocol
3363 Functional Specification
3364 OPEN Call
3368 OPEN Call
3370 CLOSED STATE (i.e., TCB does not exist)
3372 Create a new transmission control block (TCB) to hold connection
3373 state information. Fill in local socket identifier, foreign
3374 socket, precedence, security/compartment, and user timeout
3375 information. Note that some parts of the foreign socket may be
3376 unspecified in a passive OPEN and are to be filled in by the
3377 parameters of the incoming SYN segment. Verify the security and
3378 precedence requested are allowed for this user, if not return
3379 "error: precedence not allowed" or "error: security/compartment
3380 not allowed." If passive enter the LISTEN state and return. If
3381 active and the foreign socket is unspecified, return "error:
3382 foreign socket unspecified"; if active and the foreign socket is
3383 specified, issue a SYN segment. An initial send sequence number
3384 (ISS) is selected. A SYN segment of the form <SEQ=ISS><CTL=SYN>
3385 is sent. Set SND.UNA to ISS, SND.NXT to ISS+1, enter SYN-SENT
3386 state, and return.
3388 If the caller does not have access to the local socket specified,
3389 return "error: connection illegal for this process". If there is
3390 no room to create a new connection, return "error: insufficient
3391 resources".
3393 LISTEN STATE
3395 If active and the foreign socket is specified, then change the
3396 connection from passive to active, select an ISS. Send a SYN
3397 segment, set SND.UNA to ISS, SND.NXT to ISS+1. Enter SYN-SENT
3398 state. Data associated with SEND may be sent with SYN segment or
3399 queued for transmission after entering ESTABLISHED state. The
3400 urgent bit if requested in the command must be sent with the data
3401 segments sent as a result of this command. If there is no room to
3402 queue the request, respond with "error: insufficient resources".
3403 If Foreign socket was not specified, then return "error: foreign
3404 socket unspecified".
3417 [Page 54]
3418 \f
3420 September 1981
3421 Transmission Control Protocol
3422 Functional Specification
3423 OPEN Call
3427 SYN-SENT STATE
3428 SYN-RECEIVED STATE
3429 ESTABLISHED STATE
3430 FIN-WAIT-1 STATE
3431 FIN-WAIT-2 STATE
3432 CLOSE-WAIT STATE
3433 CLOSING STATE
3434 LAST-ACK STATE
3435 TIME-WAIT STATE
3437 Return "error: connection already exists".
3476 [Page 55]
3477 \f
3479 September 1981
3480 Transmission Control Protocol
3481 Functional Specification
3482 SEND Call
3486 SEND Call
3488 CLOSED STATE (i.e., TCB does not exist)
3490 If the user does not have access to such a connection, then return
3491 "error: connection illegal for this process".
3493 Otherwise, return "error: connection does not exist".
3495 LISTEN STATE
3497 If the foreign socket is specified, then change the connection
3498 from passive to active, select an ISS. Send a SYN segment, set
3499 SND.UNA to ISS, SND.NXT to ISS+1. Enter SYN-SENT state. Data
3500 associated with SEND may be sent with SYN segment or queued for
3501 transmission after entering ESTABLISHED state. The urgent bit if
3502 requested in the command must be sent with the data segments sent
3503 as a result of this command. If there is no room to queue the
3504 request, respond with "error: insufficient resources". If
3505 Foreign socket was not specified, then return "error: foreign
3506 socket unspecified".
3508 SYN-SENT STATE
3509 SYN-RECEIVED STATE
3511 Queue the data for transmission after entering ESTABLISHED state.
3512 If no space to queue, respond with "error: insufficient
3513 resources".
3515 ESTABLISHED STATE
3516 CLOSE-WAIT STATE
3518 Segmentize the buffer and send it with a piggybacked
3519 acknowledgment (acknowledgment value = RCV.NXT). If there is
3520 insufficient space to remember this buffer, simply return "error:
3521 insufficient resources".
3523 If the urgent flag is set, then SND.UP <- SND.NXT-1 and set the
3524 urgent pointer in the outgoing segments.
3535 [Page 56]
3536 \f
3538 September 1981
3539 Transmission Control Protocol
3540 Functional Specification
3541 SEND Call
3545 FIN-WAIT-1 STATE
3546 FIN-WAIT-2 STATE
3547 CLOSING STATE
3548 LAST-ACK STATE
3549 TIME-WAIT STATE
3551 Return "error: connection closing" and do not service request.
3594 [Page 57]
3595 \f
3597 September 1981
3598 Transmission Control Protocol
3599 Functional Specification
3600 RECEIVE Call
3604 RECEIVE Call
3606 CLOSED STATE (i.e., TCB does not exist)
3608 If the user does not have access to such a connection, return
3609 "error: connection illegal for this process".
3611 Otherwise return "error: connection does not exist".
3613 LISTEN STATE
3614 SYN-SENT STATE
3615 SYN-RECEIVED STATE
3617 Queue for processing after entering ESTABLISHED state. If there
3618 is no room to queue this request, respond with "error:
3619 insufficient resources".
3621 ESTABLISHED STATE
3622 FIN-WAIT-1 STATE
3623 FIN-WAIT-2 STATE
3625 If insufficient incoming segments are queued to satisfy the
3626 request, queue the request. If there is no queue space to
3627 remember the RECEIVE, respond with "error: insufficient
3628 resources".
3630 Reassemble queued incoming segments into receive buffer and return
3631 to user. Mark "push seen" (PUSH) if this is the case.
3633 If RCV.UP is in advance of the data currently being passed to the
3634 user notify the user of the presence of urgent data.
3636 When the TCP takes responsibility for delivering data to the user
3637 that fact must be communicated to the sender via an
3638 acknowledgment. The formation of such an acknowledgment is
3639 described below in the discussion of processing an incoming
3640 segment.
3653 [Page 58]
3654 \f
3656 September 1981
3657 Transmission Control Protocol
3658 Functional Specification
3659 RECEIVE Call
3663 CLOSE-WAIT STATE
3665 Since the remote side has already sent FIN, RECEIVEs must be
3666 satisfied by text already on hand, but not yet delivered to the
3667 user. If no text is awaiting delivery, the RECEIVE will get a
3668 "error: connection closing" response. Otherwise, any remaining
3669 text can be used to satisfy the RECEIVE.
3671 CLOSING STATE
3672 LAST-ACK STATE
3673 TIME-WAIT STATE
3675 Return "error: connection closing".
3712 [Page 59]
3713 \f
3715 September 1981
3716 Transmission Control Protocol
3717 Functional Specification
3718 CLOSE Call
3722 CLOSE Call
3724 CLOSED STATE (i.e., TCB does not exist)
3726 If the user does not have access to such a connection, return
3727 "error: connection illegal for this process".
3729 Otherwise, return "error: connection does not exist".
3731 LISTEN STATE
3733 Any outstanding RECEIVEs are returned with "error: closing"
3734 responses. Delete TCB, enter CLOSED state, and return.
3736 SYN-SENT STATE
3738 Delete the TCB and return "error: closing" responses to any
3739 queued SENDs, or RECEIVEs.
3741 SYN-RECEIVED STATE
3743 If no SENDs have been issued and there is no pending data to send,
3744 then form a FIN segment and send it, and enter FIN-WAIT-1 state;
3745 otherwise queue for processing after entering ESTABLISHED state.
3747 ESTABLISHED STATE
3749 Queue this until all preceding SENDs have been segmentized, then
3750 form a FIN segment and send it. In any case, enter FIN-WAIT-1
3751 state.
3753 FIN-WAIT-1 STATE
3754 FIN-WAIT-2 STATE
3756 Strictly speaking, this is an error and should receive a "error:
3757 connection closing" response. An "ok" response would be
3758 acceptable, too, as long as a second FIN is not emitted (the first
3759 FIN may be retransmitted though).
3771 [Page 60]
3772 \f
3774 September 1981
3775 Transmission Control Protocol
3776 Functional Specification
3777 CLOSE Call
3781 CLOSE-WAIT STATE
3783 Queue this request until all preceding SENDs have been
3784 segmentized; then send a FIN segment, enter CLOSING state.
3786 CLOSING STATE
3787 LAST-ACK STATE
3788 TIME-WAIT STATE
3790 Respond with "error: connection closing".
3830 [Page 61]
3831 \f
3833 September 1981
3834 Transmission Control Protocol
3835 Functional Specification
3836 ABORT Call
3840 ABORT Call
3842 CLOSED STATE (i.e., TCB does not exist)
3844 If the user should not have access to such a connection, return
3845 "error: connection illegal for this process".
3847 Otherwise return "error: connection does not exist".
3849 LISTEN STATE
3851 Any outstanding RECEIVEs should be returned with "error:
3852 connection reset" responses. Delete TCB, enter CLOSED state, and
3853 return.
3855 SYN-SENT STATE
3857 All queued SENDs and RECEIVEs should be given "connection reset"
3858 notification, delete the TCB, enter CLOSED state, and return.
3860 SYN-RECEIVED STATE
3861 ESTABLISHED STATE
3862 FIN-WAIT-1 STATE
3863 FIN-WAIT-2 STATE
3864 CLOSE-WAIT STATE
3866 Send a reset segment:
3868 <SEQ=SND.NXT><CTL=RST>
3870 All queued SENDs and RECEIVEs should be given "connection reset"
3871 notification; all segments queued for transmission (except for the
3872 RST formed above) or retransmission should be flushed, delete the
3873 TCB, enter CLOSED state, and return.
3875 CLOSING STATE
3876 LAST-ACK STATE
3877 TIME-WAIT STATE
3879 Respond with "ok" and delete the TCB, enter CLOSED state, and
3880 return.
3889 [Page 62]
3890 \f
3892 September 1981
3893 Transmission Control Protocol
3894 Functional Specification
3895 STATUS Call
3899 STATUS Call
3901 CLOSED STATE (i.e., TCB does not exist)
3903 If the user should not have access to such a connection, return
3904 "error: connection illegal for this process".
3906 Otherwise return "error: connection does not exist".
3908 LISTEN STATE
3910 Return "state = LISTEN", and the TCB pointer.
3912 SYN-SENT STATE
3914 Return "state = SYN-SENT", and the TCB pointer.
3916 SYN-RECEIVED STATE
3918 Return "state = SYN-RECEIVED", and the TCB pointer.
3920 ESTABLISHED STATE
3922 Return "state = ESTABLISHED", and the TCB pointer.
3924 FIN-WAIT-1 STATE
3926 Return "state = FIN-WAIT-1", and the TCB pointer.
3928 FIN-WAIT-2 STATE
3930 Return "state = FIN-WAIT-2", and the TCB pointer.
3932 CLOSE-WAIT STATE
3934 Return "state = CLOSE-WAIT", and the TCB pointer.
3936 CLOSING STATE
3938 Return "state = CLOSING", and the TCB pointer.
3940 LAST-ACK STATE
3942 Return "state = LAST-ACK", and the TCB pointer.
3948 [Page 63]
3949 \f
3951 September 1981
3952 Transmission Control Protocol
3953 Functional Specification
3954 STATUS Call
3958 TIME-WAIT STATE
3960 Return "state = TIME-WAIT", and the TCB pointer.
4007 [Page 64]
4008 \f
4010 September 1981
4011 Transmission Control Protocol
4012 Functional Specification
4013 SEGMENT ARRIVES
4017 SEGMENT ARRIVES
4019 If the state is CLOSED (i.e., TCB does not exist) then
4021 all data in the incoming segment is discarded. An incoming
4022 segment containing a RST is discarded. An incoming segment not
4023 containing a RST causes a RST to be sent in response. The
4024 acknowledgment and sequence field values are selected to make the
4025 reset sequence acceptable to the TCP that sent the offending
4026 segment.
4028 If the ACK bit is off, sequence number zero is used,
4030 <SEQ=0><ACK=SEG.SEQ+SEG.LEN><CTL=RST,ACK>
4032 If the ACK bit is on,
4034 <SEQ=SEG.ACK><CTL=RST>
4036 Return.
4038 If the state is LISTEN then
4040 first check for an RST
4042 An incoming RST should be ignored. Return.
4044 second check for an ACK
4046 Any acknowledgment is bad if it arrives on a connection still in
4047 the LISTEN state. An acceptable reset segment should be formed
4048 for any arriving ACK-bearing segment. The RST should be
4049 formatted as follows:
4051 <SEQ=SEG.ACK><CTL=RST>
4053 Return.
4055 third check for a SYN
4057 If the SYN bit is set, check the security. If the
4058 security/compartment on the incoming segment does not exactly
4059 match the security/compartment in the TCB then send a reset and
4060 return.
4062 <SEQ=SEG.ACK><CTL=RST>
4066 [Page 65]
4067 \f
4069 September 1981
4070 Transmission Control Protocol
4071 Functional Specification
4072 SEGMENT ARRIVES
4076 If the SEG.PRC is greater than the TCB.PRC then if allowed by
4077 the user and the system set TCB.PRC<-SEG.PRC, if not allowed
4078 send a reset and return.
4080 <SEQ=SEG.ACK><CTL=RST>
4082 If the SEG.PRC is less than the TCB.PRC then continue.
4084 Set RCV.NXT to SEG.SEQ+1, IRS is set to SEG.SEQ and any other
4085 control or text should be queued for processing later. ISS
4086 should be selected and a SYN segment sent of the form:
4088 <SEQ=ISS><ACK=RCV.NXT><CTL=SYN,ACK>
4090 SND.NXT is set to ISS+1 and SND.UNA to ISS. The connection
4091 state should be changed to SYN-RECEIVED. Note that any other
4092 incoming control or data (combined with SYN) will be processed
4093 in the SYN-RECEIVED state, but processing of SYN and ACK should
4094 not be repeated. If the listen was not fully specified (i.e.,
4095 the foreign socket was not fully specified), then the
4096 unspecified fields should be filled in now.
4098 fourth other text or control
4100 Any other control or text-bearing segment (not containing SYN)
4101 must have an ACK and thus would be discarded by the ACK
4102 processing. An incoming RST segment could not be valid, since
4103 it could not have been sent in response to anything sent by this
4104 incarnation of the connection. So you are unlikely to get here,
4105 but if you do, drop the segment, and return.
4107 If the state is SYN-SENT then
4109 first check the ACK bit
4111 If the ACK bit is set
4113 If SEG.ACK =< ISS, or SEG.ACK > SND.NXT, send a reset (unless
4114 the RST bit is set, if so drop the segment and return)
4116 <SEQ=SEG.ACK><CTL=RST>
4118 and discard the segment. Return.
4120 If SND.UNA =< SEG.ACK =< SND.NXT then the ACK is acceptable.
4122 second check the RST bit
4125 [Page 66]
4126 \f
4128 September 1981
4129 Transmission Control Protocol
4130 Functional Specification
4131 SEGMENT ARRIVES
4135 If the RST bit is set
4137 If the ACK was acceptable then signal the user "error:
4138 connection reset", drop the segment, enter CLOSED state,
4139 delete TCB, and return. Otherwise (no ACK) drop the segment
4140 and return.
4142 third check the security and precedence
4144 If the security/compartment in the segment does not exactly
4145 match the security/compartment in the TCB, send a reset
4147 If there is an ACK
4149 <SEQ=SEG.ACK><CTL=RST>
4151 Otherwise
4153 <SEQ=0><ACK=SEG.SEQ+SEG.LEN><CTL=RST,ACK>
4155 If there is an ACK
4157 The precedence in the segment must match the precedence in the
4158 TCB, if not, send a reset
4160 <SEQ=SEG.ACK><CTL=RST>
4162 If there is no ACK
4164 If the precedence in the segment is higher than the precedence
4165 in the TCB then if allowed by the user and the system raise
4166 the precedence in the TCB to that in the segment, if not
4167 allowed to raise the prec then send a reset.
4169 <SEQ=0><ACK=SEG.SEQ+SEG.LEN><CTL=RST,ACK>
4171 If the precedence in the segment is lower than the precedence
4172 in the TCB continue.
4174 If a reset was sent, discard the segment and return.
4176 fourth check the SYN bit
4178 This step should be reached only if the ACK is ok, or there is
4179 no ACK, and it the segment did not contain a RST.
4181 If the SYN bit is on and the security/compartment and precedence
4184 [Page 67]
4185 \f
4187 September 1981
4188 Transmission Control Protocol
4189 Functional Specification
4190 SEGMENT ARRIVES
4194 are acceptable then, RCV.NXT is set to SEG.SEQ+1, IRS is set to
4195 SEG.SEQ. SND.UNA should be advanced to equal SEG.ACK (if there
4196 is an ACK), and any segments on the retransmission queue which
4197 are thereby acknowledged should be removed.
4199 If SND.UNA > ISS (our SYN has been ACKed), change the connection
4200 state to ESTABLISHED, form an ACK segment
4202 <SEQ=SND.NXT><ACK=RCV.NXT><CTL=ACK>
4204 and send it. Data or controls which were queued for
4205 transmission may be included. If there are other controls or
4206 text in the segment then continue processing at the sixth step
4207 below where the URG bit is checked, otherwise return.
4209 Otherwise enter SYN-RECEIVED, form a SYN,ACK segment
4211 <SEQ=ISS><ACK=RCV.NXT><CTL=SYN,ACK>
4213 and send it. If there are other controls or text in the
4214 segment, queue them for processing after the ESTABLISHED state
4215 has been reached, return.
4217 fifth, if neither of the SYN or RST bits is set then drop the
4218 segment and return.
4243 [Page 68]
4244 \f
4246 September 1981
4247 Transmission Control Protocol
4248 Functional Specification
4249 SEGMENT ARRIVES
4253 Otherwise,
4255 first check sequence number
4257 SYN-RECEIVED STATE
4258 ESTABLISHED STATE
4259 FIN-WAIT-1 STATE
4260 FIN-WAIT-2 STATE
4261 CLOSE-WAIT STATE
4262 CLOSING STATE
4263 LAST-ACK STATE
4264 TIME-WAIT STATE
4266 Segments are processed in sequence. Initial tests on arrival
4267 are used to discard old duplicates, but further processing is
4268 done in SEG.SEQ order. If a segment's contents straddle the
4269 boundary between old and new, only the new parts should be
4270 processed.
4272 There are four cases for the acceptability test for an incoming
4273 segment:
4275 Segment Receive Test
4276 Length Window
4277 ------- ------- -------------------------------------------
4279 0 0 SEG.SEQ = RCV.NXT
4281 0 >0 RCV.NXT =< SEG.SEQ < RCV.NXT+RCV.WND
4283 >0 0 not acceptable
4285 >0 >0 RCV.NXT =< SEG.SEQ < RCV.NXT+RCV.WND
4286 or RCV.NXT =< SEG.SEQ+SEG.LEN-1 < RCV.NXT+RCV.WND
4288 If the RCV.WND is zero, no segments will be acceptable, but
4289 special allowance should be made to accept valid ACKs, URGs and
4290 RSTs.
4292 If an incoming segment is not acceptable, an acknowledgment
4293 should be sent in reply (unless the RST bit is set, if so drop
4294 the segment and return):
4296 <SEQ=SND.NXT><ACK=RCV.NXT><CTL=ACK>
4298 After sending the acknowledgment, drop the unacceptable segment
4299 and return.
4302 [Page 69]
4303 \f
4305 September 1981
4306 Transmission Control Protocol
4307 Functional Specification
4308 SEGMENT ARRIVES
4312 In the following it is assumed that the segment is the idealized
4313 segment that begins at RCV.NXT and does not exceed the window.
4314 One could tailor actual segments to fit this assumption by
4315 trimming off any portions that lie outside the window (including
4316 SYN and FIN), and only processing further if the segment then
4317 begins at RCV.NXT. Segments with higher begining sequence
4318 numbers may be held for later processing.
4320 second check the RST bit,
4322 SYN-RECEIVED STATE
4324 If the RST bit is set
4326 If this connection was initiated with a passive OPEN (i.e.,
4327 came from the LISTEN state), then return this connection to
4328 LISTEN state and return. The user need not be informed. If
4329 this connection was initiated with an active OPEN (i.e., came
4330 from SYN-SENT state) then the connection was refused, signal
4331 the user "connection refused". In either case, all segments
4332 on the retransmission queue should be removed. And in the
4333 active OPEN case, enter the CLOSED state and delete the TCB,
4334 and return.
4336 ESTABLISHED
4337 FIN-WAIT-1
4338 FIN-WAIT-2
4339 CLOSE-WAIT
4341 If the RST bit is set then, any outstanding RECEIVEs and SEND
4342 should receive "reset" responses. All segment queues should be
4343 flushed. Users should also receive an unsolicited general
4344 "connection reset" signal. Enter the CLOSED state, delete the
4345 TCB, and return.
4347 CLOSING STATE
4348 LAST-ACK STATE
4349 TIME-WAIT
4351 If the RST bit is set then, enter the CLOSED state, delete the
4352 TCB, and return.
4361 [Page 70]
4362 \f
4364 September 1981
4365 Transmission Control Protocol
4366 Functional Specification
4367 SEGMENT ARRIVES
4371 third check security and precedence
4373 SYN-RECEIVED
4375 If the security/compartment and precedence in the segment do not
4376 exactly match the security/compartment and precedence in the TCB
4377 then send a reset, and return.
4379 ESTABLISHED STATE
4381 If the security/compartment and precedence in the segment do not
4382 exactly match the security/compartment and precedence in the TCB
4383 then send a reset, any outstanding RECEIVEs and SEND should
4384 receive "reset" responses. All segment queues should be
4385 flushed. Users should also receive an unsolicited general
4386 "connection reset" signal. Enter the CLOSED state, delete the
4387 TCB, and return.
4389 Note this check is placed following the sequence check to prevent
4390 a segment from an old connection between these ports with a
4391 different security or precedence from causing an abort of the
4392 current connection.
4394 fourth, check the SYN bit,
4396 SYN-RECEIVED
4397 ESTABLISHED STATE
4398 FIN-WAIT STATE-1
4399 FIN-WAIT STATE-2
4400 CLOSE-WAIT STATE
4401 CLOSING STATE
4402 LAST-ACK STATE
4403 TIME-WAIT STATE
4405 If the SYN is in the window it is an error, send a reset, any
4406 outstanding RECEIVEs and SEND should receive "reset" responses,
4407 all segment queues should be flushed, the user should also
4408 receive an unsolicited general "connection reset" signal, enter
4409 the CLOSED state, delete the TCB, and return.
4411 If the SYN is not in the window this step would not be reached
4412 and an ack would have been sent in the first step (sequence
4413 number check).
4420 [Page 71]
4421 \f
4423 September 1981
4424 Transmission Control Protocol
4425 Functional Specification
4426 SEGMENT ARRIVES
4430 fifth check the ACK field,
4432 if the ACK bit is off drop the segment and return
4434 if the ACK bit is on
4436 SYN-RECEIVED STATE
4438 If SND.UNA =< SEG.ACK =< SND.NXT then enter ESTABLISHED state
4439 and continue processing.
4441 If the segment acknowledgment is not acceptable, form a
4442 reset segment,
4444 <SEQ=SEG.ACK><CTL=RST>
4446 and send it.
4448 ESTABLISHED STATE
4450 If SND.UNA < SEG.ACK =< SND.NXT then, set SND.UNA <- SEG.ACK.
4451 Any segments on the retransmission queue which are thereby
4452 entirely acknowledged are removed. Users should receive
4453 positive acknowledgments for buffers which have been SENT and
4454 fully acknowledged (i.e., SEND buffer should be returned with
4455 "ok" response). If the ACK is a duplicate
4456 (SEG.ACK < SND.UNA), it can be ignored. If the ACK acks
4457 something not yet sent (SEG.ACK > SND.NXT) then send an ACK,
4458 drop the segment, and return.
4460 If SND.UNA < SEG.ACK =< SND.NXT, the send window should be
4461 updated. If (SND.WL1 < SEG.SEQ or (SND.WL1 = SEG.SEQ and
4462 SND.WL2 =< SEG.ACK)), set SND.WND <- SEG.WND, set
4463 SND.WL1 <- SEG.SEQ, and set SND.WL2 <- SEG.ACK.
4465 Note that SND.WND is an offset from SND.UNA, that SND.WL1
4466 records the sequence number of the last segment used to update
4467 SND.WND, and that SND.WL2 records the acknowledgment number of
4468 the last segment used to update SND.WND. The check here
4469 prevents using old segments to update the window.
4479 [Page 72]
4480 \f
4482 September 1981
4483 Transmission Control Protocol
4484 Functional Specification
4485 SEGMENT ARRIVES
4489 FIN-WAIT-1 STATE
4491 In addition to the processing for the ESTABLISHED state, if
4492 our FIN is now acknowledged then enter FIN-WAIT-2 and continue
4493 processing in that state.
4495 FIN-WAIT-2 STATE
4497 In addition to the processing for the ESTABLISHED state, if
4498 the retransmission queue is empty, the user's CLOSE can be
4499 acknowledged ("ok") but do not delete the TCB.
4501 CLOSE-WAIT STATE
4503 Do the same processing as for the ESTABLISHED state.
4505 CLOSING STATE
4507 In addition to the processing for the ESTABLISHED state, if
4508 the ACK acknowledges our FIN then enter the TIME-WAIT state,
4509 otherwise ignore the segment.
4511 LAST-ACK STATE
4513 The only thing that can arrive in this state is an
4514 acknowledgment of our FIN. If our FIN is now acknowledged,
4515 delete the TCB, enter the CLOSED state, and return.
4517 TIME-WAIT STATE
4519 The only thing that can arrive in this state is a
4520 retransmission of the remote FIN. Acknowledge it, and restart
4521 the 2 MSL timeout.
4523 sixth, check the URG bit,
4525 ESTABLISHED STATE
4526 FIN-WAIT-1 STATE
4527 FIN-WAIT-2 STATE
4529 If the URG bit is set, RCV.UP <- max(RCV.UP,SEG.UP), and signal
4530 the user that the remote side has urgent data if the urgent
4531 pointer (RCV.UP) is in advance of the data consumed. If the
4532 user has already been signaled (or is still in the "urgent
4533 mode") for this continuous sequence of urgent data, do not
4534 signal the user again.
4538 [Page 73]
4539 \f
4541 September 1981
4542 Transmission Control Protocol
4543 Functional Specification
4544 SEGMENT ARRIVES
4548 CLOSE-WAIT STATE
4549 CLOSING STATE
4550 LAST-ACK STATE
4551 TIME-WAIT
4553 This should not occur, since a FIN has been received from the
4554 remote side. Ignore the URG.
4556 seventh, process the segment text,
4558 ESTABLISHED STATE
4559 FIN-WAIT-1 STATE
4560 FIN-WAIT-2 STATE
4562 Once in the ESTABLISHED state, it is possible to deliver segment
4563 text to user RECEIVE buffers. Text from segments can be moved
4564 into buffers until either the buffer is full or the segment is
4565 empty. If the segment empties and carries an PUSH flag, then
4566 the user is informed, when the buffer is returned, that a PUSH
4567 has been received.
4569 When the TCP takes responsibility for delivering the data to the
4570 user it must also acknowledge the receipt of the data.
4572 Once the TCP takes responsibility for the data it advances
4573 RCV.NXT over the data accepted, and adjusts RCV.WND as
4574 apporopriate to the current buffer availability. The total of
4575 RCV.NXT and RCV.WND should not be reduced.
4577 Please note the window management suggestions in section 3.7.
4579 Send an acknowledgment of the form:
4581 <SEQ=SND.NXT><ACK=RCV.NXT><CTL=ACK>
4583 This acknowledgment should be piggybacked on a segment being
4584 transmitted if possible without incurring undue delay.
4597 [Page 74]
4598 \f
4600 September 1981
4601 Transmission Control Protocol
4602 Functional Specification
4603 SEGMENT ARRIVES
4607 CLOSE-WAIT STATE
4608 CLOSING STATE
4609 LAST-ACK STATE
4610 TIME-WAIT STATE
4612 This should not occur, since a FIN has been received from the
4613 remote side. Ignore the segment text.
4615 eighth, check the FIN bit,
4617 Do not process the FIN if the state is CLOSED, LISTEN or SYN-SENT
4618 since the SEG.SEQ cannot be validated; drop the segment and
4619 return.
4621 If the FIN bit is set, signal the user "connection closing" and
4622 return any pending RECEIVEs with same message, advance RCV.NXT
4623 over the FIN, and send an acknowledgment for the FIN. Note that
4624 FIN implies PUSH for any segment text not yet delivered to the
4625 user.
4627 SYN-RECEIVED STATE
4628 ESTABLISHED STATE
4630 Enter the CLOSE-WAIT state.
4632 FIN-WAIT-1 STATE
4634 If our FIN has been ACKed (perhaps in this segment), then
4635 enter TIME-WAIT, start the time-wait timer, turn off the other
4636 timers; otherwise enter the CLOSING state.
4638 FIN-WAIT-2 STATE
4640 Enter the TIME-WAIT state. Start the time-wait timer, turn
4641 off the other timers.
4643 CLOSE-WAIT STATE
4645 Remain in the CLOSE-WAIT state.
4647 CLOSING STATE
4649 Remain in the CLOSING state.
4651 LAST-ACK STATE
4653 Remain in the LAST-ACK state.
4656 [Page 75]
4657 \f
4659 September 1981
4660 Transmission Control Protocol
4661 Functional Specification
4662 SEGMENT ARRIVES
4666 TIME-WAIT STATE
4668 Remain in the TIME-WAIT state. Restart the 2 MSL time-wait
4669 timeout.
4671 and return.
4715 [Page 76]
4716 \f
4718 September 1981
4719 Transmission Control Protocol
4720 Functional Specification
4721 USER TIMEOUT
4725 USER TIMEOUT
4727 For any state if the user timeout expires, flush all queues, signal
4728 the user "error: connection aborted due to user timeout" in general
4729 and for any outstanding calls, delete the TCB, enter the CLOSED
4730 state and return.
4732 RETRANSMISSION TIMEOUT
4734 For any state if the retransmission timeout expires on a segment in
4735 the retransmission queue, send the segment at the front of the
4736 retransmission queue again, reinitialize the retransmission timer,
4737 and return.
4739 TIME-WAIT TIMEOUT
4741 If the time-wait timeout expires on a connection delete the TCB,
4742 enter the CLOSED state and return.
4774 [Page 77]
4775 \f
4777 September 1981
4778 Transmission Control Protocol
4833 [Page 78]
4834 \f
4836 September 1981
4837 Transmission Control Protocol
4841 GLOSSARY
4845 1822
4846 BBN Report 1822, "The Specification of the Interconnection of
4847 a Host and an IMP". The specification of interface between a
4848 host and the ARPANET.
4850 ACK
4851 A control bit (acknowledge) occupying no sequence space, which
4852 indicates that the acknowledgment field of this segment
4853 specifies the next sequence number the sender of this segment
4854 is expecting to receive, hence acknowledging receipt of all
4855 previous sequence numbers.
4857 ARPANET message
4858 The unit of transmission between a host and an IMP in the
4859 ARPANET. The maximum size is about 1012 octets (8096 bits).
4861 ARPANET packet
4862 A unit of transmission used internally in the ARPANET between
4863 IMPs. The maximum size is about 126 octets (1008 bits).
4865 connection
4866 A logical communication path identified by a pair of sockets.
4868 datagram
4869 A message sent in a packet switched computer communications
4870 network.
4872 Destination Address
4873 The destination address, usually the network and host
4874 identifiers.
4876 FIN
4877 A control bit (finis) occupying one sequence number, which
4878 indicates that the sender will send no more data or control
4879 occupying sequence space.
4881 fragment
4882 A portion of a logical unit of data, in particular an internet
4883 fragment is a portion of an internet datagram.
4885 FTP
4886 A file transfer protocol.
4892 [Page 79]
4893 \f
4895 September 1981
4896 Transmission Control Protocol
4897 Glossary
4901 header
4902 Control information at the beginning of a message, segment,
4903 fragment, packet or block of data.
4905 host
4906 A computer. In particular a source or destination of messages
4907 from the point of view of the communication network.
4909 Identification
4910 An Internet Protocol field. This identifying value assigned
4911 by the sender aids in assembling the fragments of a datagram.
4913 IMP
4914 The Interface Message Processor, the packet switch of the
4915 ARPANET.
4917 internet address
4918 A source or destination address specific to the host level.
4920 internet datagram
4921 The unit of data exchanged between an internet module and the
4922 higher level protocol together with the internet header.
4924 internet fragment
4925 A portion of the data of an internet datagram with an internet
4926 header.
4928 IP
4929 Internet Protocol.
4931 IRS
4932 The Initial Receive Sequence number. The first sequence
4933 number used by the sender on a connection.
4935 ISN
4936 The Initial Sequence Number. The first sequence number used
4937 on a connection, (either ISS or IRS). Selected on a clock
4938 based procedure.
4940 ISS
4941 The Initial Send Sequence number. The first sequence number
4942 used by the sender on a connection.
4944 leader
4945 Control information at the beginning of a message or block of
4946 data. In particular, in the ARPANET, the control information
4947 on an ARPANET message at the host-IMP interface.
4951 [Page 80]
4952 \f
4954 September 1981
4955 Transmission Control Protocol
4956 Glossary
4960 left sequence
4961 This is the next sequence number to be acknowledged by the
4962 data receiving TCP (or the lowest currently unacknowledged
4963 sequence number) and is sometimes referred to as the left edge
4964 of the send window.
4966 local packet
4967 The unit of transmission within a local network.
4969 module
4970 An implementation, usually in software, of a protocol or other
4971 procedure.
4973 MSL
4974 Maximum Segment Lifetime, the time a TCP segment can exist in
4975 the internetwork system. Arbitrarily defined to be 2 minutes.
4977 octet
4978 An eight bit byte.
4980 Options
4981 An Option field may contain several options, and each option
4982 may be several octets in length. The options are used
4983 primarily in testing situations; for example, to carry
4984 timestamps. Both the Internet Protocol and TCP provide for
4985 options fields.
4987 packet
4988 A package of data with a header which may or may not be
4989 logically complete. More often a physical packaging than a
4990 logical packaging of data.
4992 port
4993 The portion of a socket that specifies which logical input or
4994 output channel of a process is associated with the data.
4996 process
4997 A program in execution. A source or destination of data from
4998 the point of view of the TCP or other host-to-host protocol.
5000 PUSH
5001 A control bit occupying no sequence space, indicating that
5002 this segment contains data that must be pushed through to the
5003 receiving user.
5005 RCV.NXT
5006 receive next sequence number
5010 [Page 81]
5011 \f
5013 September 1981
5014 Transmission Control Protocol
5015 Glossary
5019 RCV.UP
5020 receive urgent pointer
5022 RCV.WND
5023 receive window
5025 receive next sequence number
5026 This is the next sequence number the local TCP is expecting to
5027 receive.
5029 receive window
5030 This represents the sequence numbers the local (receiving) TCP
5031 is willing to receive. Thus, the local TCP considers that
5032 segments overlapping the range RCV.NXT to
5033 RCV.NXT + RCV.WND - 1 carry acceptable data or control.
5034 Segments containing sequence numbers entirely outside of this
5035 range are considered duplicates and discarded.
5037 RST
5038 A control bit (reset), occupying no sequence space, indicating
5039 that the receiver should delete the connection without further
5040 interaction. The receiver can determine, based on the
5041 sequence number and acknowledgment fields of the incoming
5042 segment, whether it should honor the reset command or ignore
5043 it. In no case does receipt of a segment containing RST give
5044 rise to a RST in response.
5046 RTP
5047 Real Time Protocol: A host-to-host protocol for communication
5048 of time critical information.
5050 SEG.ACK
5051 segment acknowledgment
5053 SEG.LEN
5054 segment length
5056 SEG.PRC
5057 segment precedence value
5059 SEG.SEQ
5060 segment sequence
5062 SEG.UP
5063 segment urgent pointer field
5069 [Page 82]
5070 \f
5072 September 1981
5073 Transmission Control Protocol
5074 Glossary
5078 SEG.WND
5079 segment window field
5081 segment
5082 A logical unit of data, in particular a TCP segment is the
5083 unit of data transfered between a pair of TCP modules.
5085 segment acknowledgment
5086 The sequence number in the acknowledgment field of the
5087 arriving segment.
5089 segment length
5090 The amount of sequence number space occupied by a segment,
5091 including any controls which occupy sequence space.
5093 segment sequence
5094 The number in the sequence field of the arriving segment.
5096 send sequence
5097 This is the next sequence number the local (sending) TCP will
5098 use on the connection. It is initially selected from an
5099 initial sequence number curve (ISN) and is incremented for
5100 each octet of data or sequenced control transmitted.
5102 send window
5103 This represents the sequence numbers which the remote
5104 (receiving) TCP is willing to receive. It is the value of the
5105 window field specified in segments from the remote (data
5106 receiving) TCP. The range of new sequence numbers which may
5107 be emitted by a TCP lies between SND.NXT and
5108 SND.UNA + SND.WND - 1. (Retransmissions of sequence numbers
5109 between SND.UNA and SND.NXT are expected, of course.)
5111 SND.NXT
5112 send sequence
5114 SND.UNA
5115 left sequence
5117 SND.UP
5118 send urgent pointer
5120 SND.WL1
5121 segment sequence number at last window update
5123 SND.WL2
5124 segment acknowledgment number at last window update
5128 [Page 83]
5129 \f
5131 September 1981
5132 Transmission Control Protocol
5133 Glossary
5137 SND.WND
5138 send window
5140 socket
5141 An address which specifically includes a port identifier, that
5142 is, the concatenation of an Internet Address with a TCP port.
5144 Source Address
5145 The source address, usually the network and host identifiers.
5147 SYN
5148 A control bit in the incoming segment, occupying one sequence
5149 number, used at the initiation of a connection, to indicate
5150 where the sequence numbering will start.
5152 TCB
5153 Transmission control block, the data structure that records
5154 the state of a connection.
5156 TCB.PRC
5157 The precedence of the connection.
5159 TCP
5160 Transmission Control Protocol: A host-to-host protocol for
5161 reliable communication in internetwork environments.
5163 TOS
5164 Type of Service, an Internet Protocol field.
5166 Type of Service
5167 An Internet Protocol field which indicates the type of service
5168 for this internet fragment.
5170 URG
5171 A control bit (urgent), occupying no sequence space, used to
5172 indicate that the receiving user should be notified to do
5173 urgent processing as long as there is data to be consumed with
5174 sequence numbers less than the value indicated in the urgent
5175 pointer.
5177 urgent pointer
5178 A control field meaningful only when the URG bit is on. This
5179 field communicates the value of the urgent pointer which
5180 indicates the data octet associated with the sending user's
5181 urgent call.
5187 [Page 84]
5188 \f
5190 September 1981
5191 Transmission Control Protocol
5195 REFERENCES
5199 [1] Cerf, V., and R. Kahn, "A Protocol for Packet Network
5200 Intercommunication", IEEE Transactions on Communications,
5201 Vol. COM-22, No. 5, pp 637-648, May 1974.
5203 [2] Postel, J. (ed.), "Internet Protocol - DARPA Internet Program
5204 Protocol Specification", RFC 791, USC/Information Sciences
5205 Institute, September 1981.
5207 [3] Dalal, Y. and C. Sunshine, "Connection Management in Transport
5208 Protocols", Computer Networks, Vol. 2, No. 6, pp. 454-473,
5209 December 1978.
5211 [4] Postel, J., "Assigned Numbers", RFC 790, USC/Information Sciences
5212 Institute, September 1981.
5246 [Page 85]
5247 \f