|
|
// -*- mode: C++; tab-width: 4; indent-tabs-mode: nil -*- (for GNU Emacs)
//
// Copyright (c) 1985-2000 Microsoft Corporation
//
// This file is part of the Microsoft Research IPv6 Network Protocol Stack.
// You should have received a copy of the Microsoft End-User License Agreement
// for this software along with this release; see the file "license.txt".
// If not, please see http://www.research.microsoft.com/msripv6/license.htm,
// or write to Microsoft Research, One Microsoft Way, Redmond, WA 98052-6399.
//
// Abstract:
//
// Transmission Control Protocol definitions.
//
#ifndef _TCP_INCLUDED_
#define _TCP_INCLUDED_
#define IP_PROTOCOL_TCP 6
#define DEFAULT_MSS (IPv6_MINIMUM_MTU - sizeof(IPv6Header) - sizeof(TCPHeader))
// Timer stuff. We keep timers as ticks.
#define MS_PER_TICK 100
#define MS_TO_TICKS(m) ((m) / MS_PER_TICK)
#define MIN_RETRAN_TICKS 3
#define DEL_ACK_TICKS 2
// Define MAX_REXMIT_TO to be number of ticks in 2MSL (=240 seconds)
#define MAX_REXMIT_TO ((ushort)FinWait2TO)
#define SWS_TO MS_TO_TICKS(5000)
#define FIN_WAIT2_TO 240
#define PUSH_TO MS_TO_TICKS(500)
#define TCP_MD5_DATA_LENGTH 44
typedef ulong TCP_TIME;#define MAX_CONN_TO_TICKS 0xffff
#define INFINITE_CONN_TO(t) ((t) == 0)
#define TCP_TIME_TO_TICKS(t) (((t)/MS_PER_TICK)+1)
// Sequence numbers are kept as signed 32 bit quantities, with macros
// defined to do wraparound comparisons on them.
typedef int SeqNum; // A sequence number.
//* Macros for comparions on sequence numbers.
#define SEQ_GT(a, b) (((a) - (b)) > 0)
#define SEQ_GTE(a, b) (((a) - (b)) >= 0)
#define SEQ_LT(a, b) (((a) - (b)) < 0)
#define SEQ_LTE(a, b) (((a) - (b)) <= 0)
#define SEQ_EQ(a, b) ((a) == (b))
// The TCB - transport control block structure. This is the
// structure that contains all of the state for the transport
// connection, including sequence numbers, flow control information,
// pending sends and receives, etc.
#define tcb_signature 0x20424354 // 'TCB '
typedef struct TCB { struct TCB *tcb_next; // Next pointer in TCB table.
#if DBG
ulong tcb_sig; // Debug signature.
#endif
KSPIN_LOCK tcb_lock;
// Send sequence variables.
SeqNum tcb_senduna; // Sequence number of first unack'd data.
SeqNum tcb_sendnext; // Sequence number of next byte to send.
SeqNum tcb_sendmax; // Max value of sendnext this epoch.
uint tcb_sendwin; // Send window.
uint tcb_unacked; // Total number of bytes of unacked data.
uint tcb_maxwin; // Max send window seen.
uint tcb_cwin; // Congestion window.
uint tcb_ssthresh; // Slow start threshold.
uint tcb_phxsum; // Precomputed pseudo-header xsum.
struct TCPSendReq *tcb_cursend; // Current send in use.
PNDIS_BUFFER tcb_sendbuf; // Current buffer chain being sent.
uint tcb_sendofs; // Offset into start of chain.
uint tcb_sendsize; // Number of bytes unsent in current send.
Queue tcb_sendq; // Queue of send requests.
// Receive sequence variables.
SeqNum tcb_rcvnext; // Next byte we expect to receive.
int tcb_rcvwin; // Receive window we're offering.
SeqNum tcb_rcvwinwatch; // Monitors peer's use of our rcv window.
SeqNum tcb_sendwl1; // Window update sequence number.
SeqNum tcb_sendwl2; // Window update ack number.
struct TCPRcvReq *tcb_currcv; // Current receive buffer.
uint tcb_indicated; // Bytes of data indicated.
uint tcb_flags; // Flags for this TCB.
uint tcb_fastchk; // Fast receive path check field.
uint (*tcb_rcvhndlr)(struct TCB *, uint, IPv6Packet *, uint Size);
// Addressing info.
// NOTE: Do not make the [next 6] invariants non-consecutive. That would
// break the MD5 computation.
union { struct { // Addressing info.
IPv6Addr tcb_daddr; // Destination IP address (i.e. our peer's address).
IPv6Addr tcb_saddr; // Source IP address (i.e. one our our addresses).
ulong tcb_dscope_id; // Scope id of destination address (0 if non-scoped).
ulong tcb_sscope_id; // Scope id of source address (0 if non-scoped).
ushort tcb_dport; // Destination port.
ushort tcb_sport; // Source port.
}; uchar tcb_md5data[TCP_MD5_DATA_LENGTH]; };
int tcb_hops; // Hop limit.
uint tcb_refcnt; // Reference count for TCB.
SeqNum tcb_rttseq; // Sequence number being measured for Round Trip Time.
// Retransmit timer information. These are stored as ticks, where by
// default each tick is 100ms.
ushort tcb_smrtt; // Smoothed rtt value.
ushort tcb_delta; // Delta value.
ushort tcb_rexmit; // Retransmit value.
uchar tcb_slowcount; // Count of reasons why we're on the slow path.
uchar tcb_pushtimer; // The 'push' timer.
ushort tcb_mss; // Maximum Segment Size for this connection.
ushort tcb_remmss; // MSS advertised by peer.
// State information.
uchar tcb_state; // State of this TCB.
uchar tcb_rexmitcnt; // Count of rexmits on this TCB.
uchar tcb_pending; // Pending actions on this TCB.
uchar tcb_kacount; // Count of keep alive probes sent.
IP_STATUS tcb_error; // Last error we heard about from IP.
uint tcb_rtt; // Current round trip time TS.
ushort tcb_rexmittimer; // Timer for rexmit.
ushort tcb_delacktimer; // Timer for delayed ack.
uint tcb_defaultwin; // Default rcv. window.
uint tcb_alive; // Keep alive time value.
struct TCPRAHdr *tcb_raq; // Reassembly queue.
struct TCPRcvReq *tcb_rcvhead; // Head of recv. buffer queue.
struct TCPRcvReq *tcb_rcvtail; // Tail of recv. buffer queue.
uint tcb_pendingcnt; // Bytes waiting to be received.
IPv6Packet *tcb_pendhead; // Head of pending receive queue.
IPv6Packet *tcb_pendtail; // Tail of pending receive queue.
struct TCPConnReq *tcb_connreq; // Connection-type request for this connection.
void *tcb_conncontext; // Connection context for this connection.
uint tcb_bcountlow; // Low part of byte count.
uint tcb_bcounthi; // High part of bytecount.
uint tcb_totaltime; // Total number of ticks spent sending.
struct TCB *tcb_aonext; // Next pointer on AddrObj.
struct TCPConn *tcb_conn; // Back pointer to conn for TCB.
Queue tcb_delayq; // Queue linkage for delay queue.
uchar tcb_closereason; // Reason we're closing.
uchar tcb_bhprobecnt; // BH probe count.
ushort tcb_swstimer; // Timer for SWS override.
void *tcb_rcvind; // Receive indication handler.
void *tcb_ricontext; // Receive indication context.
// Miscellaneous info, for IP.
RouteCacheEntry *tcb_rce; // RCE for this connection.
uint tcb_pmtu; // So we know when RCE's PTMU changes.
ulong tcb_security; // So we know when IPsec changes.
struct TCPConnReq *tcb_discwait; // Disc-Wait req., if there is one.
struct TCPRcvReq *tcb_exprcv; // Head of expedited recv. buffer queue.
IPv6Packet *tcb_urgpending; // Urgent data queue.
uint tcb_urgcnt; // Byte count of data on urgent q.
uint tcb_urgind; // Urgent bytes indicated.
SeqNum tcb_urgstart; // Start of urgent data.
SeqNum tcb_urgend; // End of urgent data.
uint tcb_walkcount; // Number of people 'walking' this TCB.
uint tcb_connid; // Cached identifier for this TCB's Conn.
ushort tcb_dupacks; // Number of duplicate acks seen.
ushort tcb_force; // Force send.
} TCB;
//
// Definitions for TCP states.
//
#define TCB_CLOSED 0 // Closed.
#define TCB_LISTEN 1 // Listening.
#define TCB_SYN_SENT 2 // SYN Sent.
#define TCB_SYN_RCVD 3 // SYN received.
#define TCB_ESTAB 4 // Established.
#define TCB_FIN_WAIT1 5 // FIN-WAIT-1
#define TCB_FIN_WAIT2 6 // FIN-WAIT-2
#define TCB_CLOSE_WAIT 7 // Close waiting.
#define TCB_CLOSING 8 // Closing state.
#define TCB_LAST_ACK 9 // Last ack state.
#define TCB_TIME_WAIT 10 // Time wait state.
#define SYNC_STATE(s) ((s) > TCB_SYN_RCVD)
#define GRACEFUL_CLOSED_STATE(s) ((s) >= TCB_LAST_ACK)
#define DATA_RCV_STATE(s) ((s) >= TCB_ESTAB && (s) <= TCB_FIN_WAIT2)
#define DATA_SEND_STATE(s) ((s) == TCB_ESTAB || (s) == TCB_CLOSE_WAIT)
//
// Definitions for TCB flags.
//
#define WINDOW_SET 0x00000001 // Window explictly set.
#define CLIENT_OPTIONS 0x00000002 // Have client IP options on conn.
#define CONN_ACCEPTED 0x00000004 // Connection was accepted.
#define ACTIVE_OPEN 0x00000008 // Connection came from an active open.
#define DISC_NOTIFIED 0x00000010 // Client's been notified of a disconnect.
#define IN_DELAY_Q 0x00000020 // We're in the delayed action Q.
#define RCV_CMPLTING 0x00000040 // We're completeing rcvs.
#define IN_RCV_IND 0x00000080 // We're calling a rcv. indicate handler.
#define NEED_RCV_CMPLT 0x00000100 // We need to have recvs. completed.
#define NEED_ACK 0x00000200 // We need to send an ACK.
#define NEED_OUTPUT 0x00000400 // We need to output.
#define DELAYED_FLAGS (NEED_RCV_CMPLT | NEED_ACK | NEED_OUTPUT)
#define ACK_DELAYED 0x00000800 // We've delayed sending an ACK.
#define PMTU_BH_PROBE 0x00001000 // We're probing for a PMTU BH.
#define BSD_URGENT 0x00002000 // We're using BSD urgent semantics.
#define IN_DELIV_URG 0x00004000 // We're in the DeliverUrgent routine.
#define URG_VALID 0x00008000 // Seen urgent data, and fields are valid.
#define FIN_NEEDED 0x00010000 // We need to send a FIN.
#define NAGLING 0x00020000 // We are using Nagle's algorithm.
#define IN_TCP_SEND 0x00040000 // We're in TCPSend.
#define FLOW_CNTLD 0x00080000 // We've received a zero window from peer.
#define DISC_PENDING 0x00100000 // A disconnect notification is pending.
#define TW_PENDING 0x00200000 // Waiting to finish going to TIME-WAIT.
#define FORCE_OUTPUT 0x00400000 // Output is being forced.
#define FORCE_OUT_SHIFT 22 // Shift to get FORCE_OUTPUT into low bit.
#define SEND_AFTER_RCV 0x00800000 // Need to send after we get out of recv.
#define GC_PENDING 0x01000000 // A graceful close is pending.
#define KEEPALIVE 0x02000000 // Doing keepalives on this TCB.
#define URG_INLINE 0x04000000 // Urgent data to be processed inline.
#define ACCEPT_PENDING 0x08000000 // Sent SYN-ACK before indicating to ULP.
#define FIN_OUTSTANDING 0x10000000 // We've sent a FIN 'recently', i.e.
// since the last retransmit. When
// this flag is set sendnext == sendmax.
#define FIN_OUTS_SHIFT 28 // Shift to FIN_OUTSTANDING bit into low bit.
#define FIN_SENT 0x20000000 // We've sent a FIN that hasn't been ack'd.
// Once this bit has been turned on in
// FIN-WAIT-1 the sequence number of the
// FIN will be sendmax-1.
#define NEED_RST 0x40000000 // We need to send a RST when closing.
#define IN_TCB_TABLE 0x80000000 // TCB is in the TCB table.
//
// The defintion of the 'slow flags'.
// If any of these flags are set we'll be forced off of the fast path.
#define TCP_SLOW_FLAGS (URG_VALID | FLOW_CNTLD | GC_PENDING | TW_PENDING | \
DISC_NOTIFIED | IN_DELIV_URG | FIN_NEEDED | \ FIN_SENT | FIN_OUTSTANDING | DISC_PENDING | \ PMTU_BH_PROBE)
//
// Close reasons.
//
#define TCB_CLOSE_RST 0x80 // Received a RST segment.
#define TCB_CLOSE_ABORTED 0x40 // Had a local abort.
#define TCB_CLOSE_TIMEOUT 0x20 // Connection timed out.
#define TCB_CLOSE_REFUSED 0x10 // Connect attempt was refused.
#define TCB_CLOSE_UNREACH 0x08 // Remote destination unreachable.
#define TCB_CLOSE_SUCCESS 0x01 // Successfull close.
//
// TCB Timer macros.
//
#define START_TCB_TIMER(t, v) (t) = (v)
#define STOP_TCB_TIMER(t) (t) = 0
#define TCB_TIMER_RUNNING(t) ((t) != 0)
// Macro to compute retransmit timeout.
#define REXMIT_TO(t) ((((t)->tcb_smrtt >> 2) + (t)->tcb_delta) >> 1)
//
// Definitons for pending actions. We define a PENDING_ACTION macro that can
// be used to decide whether or not we can proceed with an activity. The only
// pending action we really care about is DELETE - others are low priority and
// can be put off.
//
#define PENDING_ACTION(t) ((t)->tcb_pending & DEL_PENDING)
#define DEL_PENDING 0x01 // Delete is pending.
#define OPT_PENDING 0x02 // Option set is pending.
// Macro to see if a TCB is closing.
#define CLOSING(t) ((t)->tcb_pending & DEL_PENDING)
//
// Structure of a TCP packet header.
//
typedef struct TCPHeader { ushort tcp_src; // Source port.
ushort tcp_dest; // Destination port.
SeqNum tcp_seq; // Sequence number.
SeqNum tcp_ack; // Ack number.
ushort tcp_flags; // Flags and data offset.
ushort tcp_window; // Window offered.
ushort tcp_xsum; // Checksum.
ushort tcp_urgent; // Urgent pointer.
} TCPHeader;
//
// Definitions for TCP header flags.
//
#define TCP_FLAG_FIN 0x00000100
#define TCP_FLAG_SYN 0x00000200
#define TCP_FLAG_RST 0x00000400
#define TCP_FLAG_PUSH 0x00000800
#define TCP_FLAG_ACK 0x00001000
#define TCP_FLAG_URG 0x00002000
#define TCP_FLAGS_ALL (TCP_FLAG_FIN | TCP_FLAG_SYN | TCP_FLAG_RST | \
TCP_FLAG_ACK | TCP_FLAG_URG)
//
// Flags in the tcb_fastchk field that are not in the TCP header proper.
// Setting these flags forces us off the fast path.
//
#define TCP_FLAG_SLOW 0x00000001 // Need to be on slow path.
#define TCP_FLAG_IN_RCV 0x00000002 // In recv. path already.
#define TCP_OFFSET_MASK 0xf0
#define TCP_HDR_SIZE(t) (uint)(((*(uchar *)&(t)->tcp_flags) & TCP_OFFSET_MASK) >> 2)
#define MAKE_TCP_FLAGS(o, f) ((f) | ((o) << 4))
//
// TCP Option Identifiers.
//
#define TCP_OPT_EOL 0
#define TCP_OPT_NOP 1
#define TCP_OPT_MSS 2
#define MSS_OPT_SIZE 4
//
// Convenient byte swapped structure for receives.
//
typedef struct TCPRcvInfo { SeqNum tri_seq; // Sequence number.
SeqNum tri_ack; // Ack number.
uint tri_window; // Window.
uint tri_urgent; // Urgent pointer.
uint tri_flags; // Flags.
} TCPRcvInfo;
//
// General structure, at the start of all command specific request structures.
//
#define tr_signature 0x20205254 // 'TR '
typedef struct TCPReq { struct Queue tr_q; // Q linkage.
#if DBG
ulong tr_sig;#endif
RequestCompleteRoutine tr_rtn; // Completion routine.
PVOID tr_context; // User context.
int tr_status; // Final complete status.
} TCPReq;
#define TCP6_TAG '6PCT'
#ifdef POOL_TAGGING
#ifdef ExAllocatePool
#undef ExAllocatePool
#endif
#define ExAllocatePool(type, size) ExAllocatePoolWithTag(type, size, TCP6_TAG)
#endif // POOL_TAGGING
//
// TCP endpoint context structure allocated for each open of TCP/UDP.
// A pointer to this structure is stored in FileObject->FsContext.
//
typedef struct _TCP_CONTEXT { union { HANDLE AddressHandle; CONNECTION_CONTEXT ConnectionContext; HANDLE ControlChannel; } Handle; ULONG ReferenceCount; BOOLEAN CancelIrps; KSPIN_LOCK EndpointLock;#if DBG
LIST_ENTRY PendingIrpList; LIST_ENTRY CancelledIrpList;#endif
KEVENT CleanupEvent;} TCP_CONTEXT, *PTCP_CONTEXT;
#include "tcpdeb.h"
#endif // _TCP_INCLUDED_
|
