--- /dev/null Thu Jan 01 00:00:00 1970 +0000
+++ b/networkprotocols/iphook/inhook6/include/in_pkt.h Tue Jan 26 15:23:49 2010 +0200
@@ -0,0 +1,713 @@
+// Copyright (c) 2004-2009 Nokia Corporation and/or its subsidiary(-ies).
+// All rights reserved.
+// This component and the accompanying materials are made available
+// under the terms of "Eclipse Public License v1.0"
+// which accompanies this distribution, and is available
+// at the URL "http://www.eclipse.org/legal/epl-v10.html".
+//
+// Initial Contributors:
+// Nokia Corporation - initial contribution.
+//
+// Contributors:
+//
+// Description:
+// in_pkt.h - packet handling routines
+// Generic packet handling utility for mapping packet handling to the RMBufChain.
+//
+
+
+
+/**
+ @file in_pkt.h
+ @publishedAll
+ @released
+*/
+
+#ifndef __IN_PKT_H__
+#define __IN_PKT_H__
+
+#include <nifmbuf.h>
+#include "ip6_hdr.h" // ..should eventually be <inet/ip6_hdr.h>? -- msa
+#include "ip4_hdr.h"
+
+#define TPACKETHEAD_FRAGMENT 1 //< Enable iFragment in TPacketHead
+
+/**
+ TScopeType is only provided so that "magic" constants can be
+ avoided in the source code. However, the max value cannot be changed
+ to anything from 0xF. The scope type is assumed to be 4 bits long
+ in many occasions.
+
+ The value of the scope type is directly bound the the IPv6 Scope
+ level - 1. This can be done, as IPv6 Scope level 0 is not legal
+ (or usable) in any context within the stack.
+ This allows our non-standard network scope (= 0x10) to
+ be coded internally in 4 bits (as 0xF).
+
+ @publishedAll
+ @released
+ @since v7.0s
+*/
+enum TScopeType
+ {
+ EScopeType_IF = 0x0, //< (= #KIp6AddrScopeNodeLocal - 1), id is interface index
+ EScopeType_IAP = 0x1, //< (= #KIp6AddrScopeLinkLocal - 1). id is IAP number
+ EScopeType_GLOBAL = 0xD,//< (= #KIp6AddrScopeGlobal - 1). id is global scope id
+ //
+ // no symbols defined for types 2..14 (they are also valid)
+ //
+ EScopeType_NET = 0xF //< (= #KIp6AddrScopeNetwork - 1), id is network number (must be the last entry)
+ };
+
+//
+// TIpHeader
+// *********
+class TIpHeader
+ /**
+ A simple help class that uses a union to merge handling of either an IPv4 or
+ an IPv6 header.
+ @since v7.0
+ @publishedAll
+ @released
+ */
+ {
+public:
+ /**
+ Gets the minimum header length.
+
+ IPv6 header is longer than minimum IPv4 header, thus
+ returned value is for IPv4. This function only defined
+ because it is required when this class is used as template
+ parameter in TInet6Packet.
+
+ @return Minimum IPv4 header length
+ */
+ inline static TInt MinHeaderLength() {return TInet6HeaderIP4::MinHeaderLength(); }
+ /**
+ Gets the maximum header length.
+
+ IPv6 header always shorter than maximum IPv4 header, thus
+ returned value is for IPv4. This function is only defined
+ because "header mapping" classes are expected to have it.
+
+ @return Maximum IPv4 header length
+ */
+ inline static TInt MaxHeaderLength() {return TInet6HeaderIP4::MaxHeaderLength(); }
+
+ union
+ {
+ TInet6HeaderIP4 ip4;
+ TInet6HeaderIP ip6;
+ };
+ };
+
+
+// RMBufPacketPeek
+// ***************
+class RMBufPacketPeek : public RMBufChain
+ /**
+ Extends RMBufChain to add functions to read packet data as a descriptor
+ and as an IP header.
+
+ The RMBufChain is assumed to contain the raw packet, without
+ the info block prepended (e.g. if this class is used for RMBufPacketBase
+ derived handle, it must be in "unpacked" state).
+
+ @since v7.0
+ @publishedAll
+ @released
+ */
+ {
+public:
+ IMPORT_C TPtr8 Access(TInt aSize, TUint aOffset = 0);
+ IMPORT_C TIpHeader *GetIpHeader();
+ };
+
+// TPacketHead
+// ***********
+class TPacketHead
+ /**
+ Storage for some precomputed information for an outbound packet flow.
+
+ The outbound TPacketHead is part of the flow context (CFlowContext).
+
+ The CFlowContext::Connect initializes the content from the parameters
+ of the flow (TFlowInfo) and runs the connection process.. The connection
+ process (MIp6Hook::OpenL and MFlowHook::ReadyL phases) completes the
+ information. After this, as long as the flow is connected, the content
+ is mostly frozen and <b>must not be modified by anyone</b>.
+
+ When there is a need to change any flow information, the changes must
+ be done to the flow parameters (and not to TPacketHead). The change of
+ flow parameters also sets the CFlowContext::iChanged flag, and this
+ eventually causes a new CFlowContext::Connect, which re-initializes
+ the TPacketHead with the new information.
+
+ For each field in the TPacketHead, the hook writer must follow the
+ basic rule (only for fields that it intends to change):
+
+ - if some field is changed in MIp6Hook::OpenL, then the previous
+ value should be restored in the MFlowHook::ReadyL.
+ - an exeception: the hook must omit the restore, if the
+ previous value was unspecified value (for example, the source
+ address).
+ - the content of #iPacket (and #iOffset) are special: they cannot
+ be modified in the MIp6Hook::OpenL phase. A hook can
+ modify them only in the MFlowHook::ReadyL phase. And, if the hook
+ is adding an IP header for tunneling, it must save the current content
+ of these fields in the ReadyL function, and then clear out the fields
+ (it must make the iPacket empty and zero iOffset). The hook must add
+ the saved iPacket content below the added tunnel header in
+ MFlowHook::ApplyL .
+
+ @since v7.0
+ @publishedAll
+ @released
+ */
+ {
+public:
+ IMPORT_C TBool ExtHdrGet(TInt aType, TInt& aOfs, TInt& aLen);
+ IMPORT_C TBool ExtHdrGetOrPrependL(TInt aType, TInt& aOfs, TInt& aLen);
+ IMPORT_C TBool ExtHdrGetOrAppendL(TInt aType, TInt& aOfs, TInt& aLen);
+ IMPORT_C void AddDestinationOptionL(const TPtrC8& aOption, TUint8 aAlign=0, TUint8 aModulo=4);
+ IMPORT_C void AddDestinationOptionL(const TUint8* aOption, TUint8 aLen, TUint8 aAlign=0, TUint8 aModulo=4);
+
+public:
+ /**
+ "Virtual" IP header. The IPv6 header stucture is used, but the same
+ format is <b>also</b> used for the IPv4 destinations (Version() == 4,
+ even though the header format is still IPv6!)
+
+ This header is initialized in the beginning of the OpenL phase
+ as follows:
+ @li Version = 0
+ @li Traffic Class, copied from the flow iOptions.iTrafficClass
+ @li Flow Label = 0
+ @li Payload Length = 0 (dummy field, not used)
+ @li Next Header, copied from the flow iProtocol
+ @li Hop Limit, copied from the flow iOptions.iHopLimit
+ @li Src Address, copied from the flow Local Address (usually unspecified)
+ @li Dst Address, copied from the flow Remote Address
+
+ At beginning of the ReadyL phase (= at end of OpenL), the destination
+ address (and iDstId) are used to find a route on the interface. Depending
+ on whether this address is IPv4 (mapped) or IPv6, the Version field is set
+ accordingly to either 4 or 6.
+
+ After succesfull completion of the ReadyL, this used for *each* packet
+ which needs an IP header to be generated on send. The Version() determines
+ whether IPv4 or IPv6 frame is to be generated (this is the initial
+ header in the packet, *before* running outbound ApplyL hooks):
+
+ @verbatim
+ IPv6 IPv4
+ Version == 6 ==4
+ Traffic Class used as is used as TOS
+ Flow Label used as is ignored
+ Payload Length ignored ignored
+ Next Header used as is used as Protocol
+ Hop Limit used as is used as TTL
+ Src Address used as is used as IPv4 mapped
+ Dst Address used as is used as IPv4 mapped
+ @endverbatim
+ */
+ TInet6HeaderIP ip6;
+ /**
+ Contains the scope id associated with the destination address
+ which is stored in #ip6 Dst Address. This id and address must
+ always be considered as a unit. Logically, any change changes
+ both values.
+
+ iDstId is initialized from the flow context TFlowInfo::iRemote.Scope() at
+ beginning of the flow connect phase. If application does not define
+ this scope id, then the system will attempt to choose a default value
+ at beginning of the connect phase. If the default cannot be determined,
+ the flow is put into pending state (and no connect happens).
+
+ @par MIp6Hook::OpenL
+ On entry to the OpenL, the iDstId is always non-zero and destination
+ address is specified. If a hook changes the destination address in
+ OpenL method, it must provide the correct id value
+ which goes with the new destination. If it cannot do this, it
+ must either abort the connect by leaving with an error state, or it
+ can leave with PENDING (> 0) status to signal there is no route
+ for the new destination.
+ If the stack cannot find suitable interface for the destination, then
+ it aborts the connect phase, and the flow is placed into holding state.
+
+ @note
+ Only a tunneling hook can safely change the destination
+ address (a use of routing header can also be a kind of
+ tunneling).
+
+ @par MFlowHook::ReadyL
+ If the hook changed the destination address (or id) in the OpenL,
+ the ReadyL must restore the original values back.
+
+ */
+ TUint32 iDstId;
+ /**
+ Contains the scope id associated with the source address
+ which is stored in #ip6 Src address. This is defined when the source
+ address is defined, and otherwise undefined.
+
+ iSrcId is initialized from TFlowInfo::iLocal.Scope() at beginning of the
+ flow connect phase. If application defines the source address,
+ but does not specify this scope id, then the system chooses
+ the id based on the interface defined by the source address.
+ If scope and address are both specified, they must match the
+ selected interface.
+
+ @par MIp6Hook::OpenL
+ On entry to the OpenL, the iSrcId (and source address) may be
+ undefined (#iSourceSet = 0). If defined (iSourceSet = 1), then
+ both address and iSrcId are defined (iSrcId != 0). A hook may
+ force a reselection of the source just by zeroing the
+ iSourceSet.
+
+ @par MFlowHook::ReadyL
+ If the hook changed the source address (or id) in the OpenL,
+ the ReadyL must restore the original values back, but only
+ if the original value was defined (#iSourceSet = 1 in OpenL).
+ */
+ TUint32 iSrcId;
+ /**
+ The source address has been set.
+
+ This bit indicates whether the value stored in #ip6 src field
+ and #iSrcId is to be used as a source address as is.
+
+ Initialized from TFlowInfo::iLocalSet, which tells whether user
+ specified tbe source address or not (e.g used RSocket Bind method).
+ The stack checks the value after each MIp6Hook::OpenL call, and
+ if the flag is set, the source in ip6 is used as is. If the flag
+ is zero, then the stack performs the normal source address selection
+ based on the current destination address (#iSrcId and destination
+ address).
+
+ @par MIp6Hook::OpenL
+ On entry, this flag is always set and source address is defined.
+ A hook may clear this flag, if it wants the
+ stack choose the source address based on current destination.
+ The clearing operation is normally needed only by a tunneling
+ hook.
+
+ @note
+ If the hook specifies the source address, it must be either
+ a valid source address for the interface or unspecified
+ address.
+
+ @par MFlowHook::ReadyL
+ Upon entry to the ReadyL, the source address is always fully
+ known (the hook can assume that #iSrcId and the #ip6 source
+ addresses are valid).
+ If the source address was set before the OpenL, then this
+ must restore the original value (along with the #iSrcId
+ and source address).
+ */
+ TUint iSourceSet:1;
+#ifdef TPACKETHEAD_FRAGMENT
+ /**
+ The fragment processing alredy done.
+
+ This bit is meaningful only in OpenL phase. If already set,
+ then some ealier hook has requested that the packet must
+ be fragmented to fit the mtu.
+
+ A tunneling hook can set this bit in OpenL, if it needs
+ the fragmenting to happen before the ApplyL is called (e.g.
+ the fragments are tunneled instead of fragmenting the
+ tunneling).
+
+ This bit can only be set or left as is. It cannot be cleared
+ once set.
+ */
+ TUint iFragment:1;
+#endif
+ /**
+ Selector info, the upper layer protocol.
+
+ iProtocol has the same value as ip6.NextHeader() when iPacket is empty,
+ and otherwise it is the same as NextHeader() of the last extension
+ header in the iPacket.
+
+ The values of the other selector fields: #iIcmpType, #iIcmpCode
+ #iSrcPort and #iDstPort depend on iProtocol. Whenever iProtocol
+ is changed, the other fields must be updated accordingly.
+
+ @par MIp6Hook::OpenL
+ Because iPacket cannot be modified during the OpenL phase, the
+ content of this field and the Next Header (protocol) field in
+ the #ip6 pseudoheader must always be the same. This field should
+ be considered as <b>read-only</b>, unless the hook intends to
+ apply IP-in-IP tunneling, in which case the hook <b>must</b>
+ change the value to the appropriate tunneling protocol
+ (#KProtocolInet6Ipip or #KProtocolInetIpip).
+
+ @par MFlowHook::ReadyL
+ Only a tunneling hook needs to restore the value here to match
+ the original upper layer protocol. See #iPacket for
+ more detailed information.
+ */
+ TUint8 iProtocol;
+ /**
+ Selector field whose value depends on #iProtocol.
+
+ If this field does not have meaning with the protocol,
+ the field content should be set to ZERO.
+ */
+ TUint8 iIcmpType;
+ /**
+ Selector field whose value depends on #iProtocol.
+
+ If this field does not have meaning with the protocol,
+ the field content should be set to ZERO.
+ */
+ TUint8 iIcmpCode;
+ /**
+ Selector field whose value depends on #iProtocol.
+
+ If this field does not have meaning with the protocol,
+ the field content should be set to ZERO.
+ */
+ TUint16 iSrcPort;
+ /**
+ Selector field whose value depends on #iProtocol.
+
+ If this field does not have meaning with the protocol,
+ the field content should be set to ZERO.
+ */
+ TUint16 iDstPort;
+ /**
+ The amount of pre-computed IPv6 extension headers in iPacket which
+ are copied to the beginning of each outgoing packet
+
+ If iOffset > 0, then #iPacket includes that much of extension
+ headers that are copied in front of each packet.
+ */
+ TInt iOffset;
+ /**
+ Pre-computed extension headers for all packets in this flow.
+
+ These can only be added in the ReadyL phase. If any of the
+ ReadyL's adds extension headers into this, it must take care
+ of maintaining the correct Next Header in the virtual IP header
+ (and the original upper layer protocol must be placed in the
+ next header of the last extension header added.
+
+ Stack copies the content of this to each outgoing packet, just below
+ the IP header, before running the ApplyL functions of the outbound
+ flow hooks.
+
+ @par MIp6Hook::OpenL
+ The iPacket <b>must not</b> be modified during the OpenL phase.
+
+ @par MFlowHook::ReadyL
+ A non-tunneling hook may add extension headers into the current
+ iPacket. A tunneling hook has more complex requirements:
+ it must save the current iPacket and #iOffset and initialize
+ iOffset = 0, and iPacket as empty.
+
+ @par MFlowHook::ApplyL
+ When a tunneling hook adds the tunneling IP header, it
+ must also copy the saved iPacket below the added IP header.
+ */
+ RMBufPacketPeek iPacket;
+ /**
+ The received packet which caused an ICMP error reply to be sent.
+
+ This is only used for ICMP error repply flows, and should be
+ ignored by others -- mainly for IPSEC hook. The packet, if
+ present, is in unpacked state.
+ */
+ RMBufPacketBase iIcmp;
+ /**
+ The current destination interface.
+
+ This is ONLY used during connect/OpenL phase.
+
+ The value is maintained by the stack, and is intended as
+ read-only information for the hooks that have a use for
+ it (for example, IPSEC implementing VPN specific policies).
+
+ A hook must not modify this value (the stack will recompute
+ the value after each OpenL, based on the possibly changed
+ address parameters in the TPacketHead)
+
+ @par MIp6Hook::OpenL
+ <b>read-only</b>
+ @par MFlowHook::ReadyL
+ <b>read-only</b>
+ */
+ TUint32 iInterfaceIndex;
+ };
+
+class TInet6PacketBase
+ /**
+ * Thin base class for the TInet6Packet.
+ */
+ {
+public:
+ enum TAlign
+ {
+ EAlign1 = 0, //< Align to byte (no align requirement)
+ EAlign2 = 1, //< Align to 2 byte unit (even address)
+ EAlign4 = 3, //< Align to 4 byte unit
+ EAlign8 = 7 //< Align to 8 byte unit
+ };
+
+ /**
+ Constructor.
+
+ @param aAlign The align requirement.
+ */
+ TInet6PacketBase(TAlign aAlign) : iLength(0), iAlign(aAlign) {}
+
+ /**
+ Length of the mapped region.
+
+ The real mapped length as computed by the Access function.
+ If access returned non-NULL, the following is always TRUE:
+
+ @li aMin <= iLength
+ */
+ TInt iLength;
+
+ IMPORT_C TUint8 *Access(RMBufChain &aPacket, TInt aOffset, TInt aSize, TInt aMin);
+
+ inline void SetAlign(TAlign aAlign)
+ /**
+ * Changes the align requirement.
+ *
+ * @param aAlign The new align requirement.
+ */
+ {
+ iAlign = aAlign;
+ }
+protected:
+ /**
+ The align requirement.
+ */
+ TAlign iAlign;
+ };
+
+// TInet6Packet template
+// *********************
+template <class T>
+class TInet6Packet : public TInet6PacketBase
+ /**
+ Encapsulates an IPv6 packet header as a section of an RMBufChain.
+
+ The T template parameter should represent a packet header type. It should
+ support static functions MaxHeaderLength() and MinHeaderLength() that return
+ TInt values for maximum and minimum header lengths respectively.
+
+ @publishedAll
+ @released
+ @since v7.0
+ */
+ {
+public:
+ TInet6Packet(TAlign aAlign = EAlign4) : TInet6PacketBase(aAlign), iHdr(NULL)
+ /**
+ Default constructor.
+
+ Construct an empty mapping. To be usable, the Set() function
+ must be used.
+ */
+ {}
+ TInet6Packet(RMBufChain &aPacket) : TInet6PacketBase(EAlign4)
+ /**
+ Constructor specifying a RMBufChain object.
+
+ Verify and arrange it so that a class T can be mapped
+ to a contiguous octets from the beginning of the RMBufChain
+ content, and set iHdr to point this area.
+
+ If this is not possible, iHdr is initialized to NULL.
+
+ @param aPacket
+ Packet containing the header T at offset = 0
+ */
+ {
+ iHdr = (T *)Access(aPacket, 0, T::MaxHeaderLength(), T::MinHeaderLength());
+ }
+
+ TInet6Packet(RMBufChain &aPacket, TInt aOffset, TAlign aAlign = EAlign4) : TInet6PacketBase(aAlign)
+ /**
+ Constructor specifying a RMBufChain object and an offset.
+
+ Verify and arrange it so that a class T can be mapped
+ to a contiguous octets starting at specified offset of
+ the RMBufChain content, and set iHdr to point this area.
+
+ If this is not possible, iHdr is initialized to NULL.
+
+ @param aPacket
+ Packet containing the header T at aOffset
+ @param aOffset
+ Offset of the header to be mapped.
+ @param aAlign
+ The alignement requirement.
+ */
+ {
+ iHdr = (T *)Access(aPacket, aOffset, T::MaxHeaderLength(), T::MinHeaderLength());
+ }
+
+ void Set(RMBufChain &aPacket, TInt aOffset, TInt aSize)
+ /**
+ Sets the packet header from a specified RMBufChain object.
+
+ Verify and arrange it so that a aSize octets can be mapped
+ to a contiguous octets starting at specified offset of
+ the RMBufChain content, and set iHdr to point this area.
+
+ If this is not possible, iHdr is initialized to NULL.
+
+ Note that this differs from the contructors: the required
+ size is a parameter, and not determined by the T::MinHeaderLength().
+ However, the "T* iHdr" is set to point the start of the requested
+ area. It's a responsibility of the user of this method to know
+ whether using this pointer is safe with the specified size parameter.
+
+ @param aPacket
+ Packet containing the header T at aOffset
+ @param aOffset
+ Offset (position) of the header to be mapped
+ @param aSize
+ Length of required contiguous memory
+ */
+ {
+ iHdr = (T *)Access(aPacket, aOffset, aSize, aSize);
+ }
+
+ inline T& operator()()
+ {
+ return *iHdr;
+ }
+ /**
+ The pointer to the mapped region (if non-NULL). If NULL,
+ then there is no mapping, and iLength == 0.
+ */
+ T *iHdr;
+ };
+
+
+// TPacketPoker
+// ************
+class TPacketPoker
+ /**
+ Provides a utility for linear scanning of a chain of RMBuf objects (an RMBufChain).
+
+ An object of this type maintains a current point in the RMBufChain. This point
+ can only move forward, and a leave occurs if the point advances beyond the
+ end of the chain.
+
+ Any pointers and aligns arranged before the current point, remain valid: for
+ example, you can save a reference and advance the pointer, and the reference
+ remains usable.
+
+ If instead you need to go to a single specified offset, then use
+ RMBufChain::Goto() or RMBufPacketPeek::Access().
+
+ @post
+ A Generic implementation assert:
+ after construct, iTail == 0 iff iCurrent == 0 (all scanned), or
+ in other words: as long as there are bytes after current point,
+ iTail will be non-zero (and More() returns ETrue).
+ All methods maintain this invariant or leave, if impossible.
+
+ Some other utility methods, not directly related to scanning, are also included.
+ @since v7.0
+ @publishedAll
+ @released
+ */
+ {
+public:
+ IMPORT_C TPacketPoker(RMBufChain &aChain);
+
+ inline void SkipL(TInt aSize)
+ /**
+ Moves the current point forward a specified number of bytes.
+
+ @param aSize Number of bytes to move forward
+ @leave KErrEof
+ if the request cannot be satisfied.
+ */
+ { if (aSize < iTail) { iTail -= aSize; iOffset += aSize; } else OverL(aSize); }
+
+ inline TUint8 *Ptr() const
+ /**
+ Raw pointer to the current point (can be invalid, if iTail = 0).
+
+ @note Internal "unsafe" method
+ */
+ {return iCurrent->Ptr() + iOffset; }
+
+ inline TUint8 *ReferenceL(TInt aSize = 1)
+ /**
+ Gets a pointer to the current point, such that
+ at least the specified minimum number of bytes can be read.
+
+ @param aSize
+ Specified minimum number of bytes to be read through
+ the returned pointer.
+ @return Raw data pointer
+ @leave KErrEof
+ if the request cannot be satisfied.
+ */
+ { if (iTail >= aSize) return Ptr(); else return AdjustL(aSize); }
+
+ inline TUint8 *ReferenceAndSkipL(TInt aSize)
+ /**
+ Gets a pointer to the current point, such that at least the
+ specified minimum number of bytes can be read,
+ and moves the point the specified number of bytes forward.
+
+ @param aSize
+ Specified minimum number of bytes to be read through the returned
+ pointer, and the number of bytes to move forward
+ @return
+ Raw data pointer
+ @leave KErrEof
+ if the request cannot be satisfied.
+ */
+ { TUint8 *x = ReferenceL(aSize); SkipL(aSize); return x; }
+
+ inline TInt Remainder() const
+ /**
+ Gets the length of the contiguous space after the current point.
+
+ @return Length after the current point
+ */
+ { return iTail; }
+
+ inline TBool AtBegin() const
+ /**
+ Tests whether the current point is at the beginning of an RMBuf.
+
+ @return ETrue if current point is at the beginning
+ */
+ { return iOffset == 0; }
+
+ inline TBool More() const
+ /**
+ Tests whether there is more data to scan.
+
+ @return ETrue if there is more data to scan
+ */
+ { return iTail > 0; }
+
+ IMPORT_C static TBool IsExtensionHeader(TInt aProtocolId);
+private:
+ IMPORT_C void OverL(TInt aSize);
+ IMPORT_C TUint8 *AdjustL(TInt aSize);
+ /** The RMBuf of the current point. */
+ RMBuf *iCurrent;
+ /** The offset of the current point in the RMBuf. */
+ TInt iOffset;
+ /** Remaining bytes starting from the current point in the RMBuf. */
+ TInt iTail;
+ };
+
+#endif