Mac Developer Library

Developer

IODMACommand Class Reference

Options
Deployment Target:

On This Page
Language:

IODMACommand

Inheritance


Not Applicable

Conforms To


Not Applicable

Import Statement


Not Applicable

Objective-C

@import Kernel;

Availability


Available in OS X v10.4 and later.

A mechanism to convert memory references to I/O bus addresses.

The IODMACommand is supersedes the IOMemoryCursor and greatly enhances the functionality and power of it. The command can be specified to output 64 bit physical addresses and also allows driver writers bypass mapping hardware or get addresses suitable for non-snooped DMA.

The command is designed to be very easily subclassable. Most driver writers need to associate some DMA operations with their memory descriptor and usually use a C structure for that purpose. This structure is often kept in a linked list. This IODMACommand has built it <kern/queue.h> linkage and can be derived and 'public:' variables added, giving the developer a structure that can associate a memory descriptor with a particular dma command but will also allow the developer to generate that command and keep the state necessary for tracking it.

It is envisaged that a pool of IODMACommands will be created at driver initialisation and each command will be kept in an IOCommandPool while not in use. However if developers wishes to maintain their own free lists that is certainly possible. See the <kern/queue.h> and <xnu/iokit/Kernel/IOCommandPool> for sample code on manipulating the command's doubly linked list entries.

The IODMACommand can be used in a 'weak-linked' manner. To do this you must avoid using any static member functions. Use the, much slower but safe, weakWithSpecification function. On success a dma command instance will be returned. This instance can then be used to clone as many commands as is needed. Remember deriving from this class can not be done weakly, that is no weak subclassing!

  • Clears the DMACommand's current memory descriptor

    Declaration

    C++

    virtual IOReturn clearMemoryDescriptor( bool autoComplete = true);

    Parameters

    autoComplete

    An optional boolean variable that will call the complete() function automatically before the memory descriptor is processed. Defaults to true.

    Discussion

    completes and invalidates the cache if the DMA command is currently active, copies all data from bounce buffers if necessary and releases all resources acquired during setMemoryDescriptor.

  • Creates a new command based on the specification of the current one.

    Declaration

    C++

    virtual IODMACommand *cloneCommand( void *refCon = 0);

    Return Value

    Returns a new memory cursor if successfully created and initialised, 0 otherwise.

    Discussion

    Factory function to create and initialise an IODMACommand in one operation. The current command's specification will be duplicated in the new object, but however none of its state will be duplicated. This means that it is safe to clone a command even if it is currently active and running, however you must be certain that the command to be duplicated does have a valid reference for the duration.

  • Complete processing of DMA mappings after an I/O transfer is finished.

    Declaration

    C++

    virtual IOReturn complete( bool invalidateCache = true, bool synchronize = true);

    Parameters

    invalidCache

    Invalidate the caches for the memory descriptor. Defaults to true for kNonCoherent and is ignored by the other types.

    synchronize

    Copy any buffered data back to the target IOMemoryDescriptor. Defaults to true, if synchronize() is being used to explicitly copy data, passing false may avoid an unneeded copy.

    Return Value

    kIOReturnNotReady if not prepared, kIOReturnSuccess otherwise.

    Discussion

    This method should not be called unless a prepare was previously issued; the prepare() and complete() must occur in pairs, before and after an I/O transfer

  • Helper function for a type checked call to genIOVMSegments(qv), for use with an IODMACommand set up with the output function kIODMACommandOutputHost32, kIODMACommandOutputBig32, or kIODMACommandOutputLittle32. If the output function of the IODMACommand is not a 32 bit function, results will be incorrect.

    Declaration

    C++

    inline IOReturn gen32IOVMSegments( UInt64 *offset, Segment32 *segments, UInt32 *numSegments) ;

  • Helper function for a type checked call to genIOVMSegments(qv), for use with an IODMACommand set up with the output function kIODMACommandOutputHost64, kIODMACommandOutputBig64, or kIODMACommandOutputLittle64. If the output function of the IODMACommand is not a 64 bit function, results will be incorrect.

    Declaration

    C++

    inline IOReturn gen64IOVMSegments( UInt64 *offset, Segment64 *segments, UInt32 *numSegments) ;

  • Generates a physical scatter/gather for the current DMA command

    Declaration

    C++

    virtual IOReturn genIOVMSegments( UInt64 *offset, void *segments, UInt32 *numSegments);

    Parameters

    offset

    input/output parameter, defines the starting and ending offset in the memory descriptor, relative to any offset passed to the prepare() method.

    segments

    Void pointer to base of output physical scatter/gather list. Always passed directly onto the SegmentFunction.

    numSegments

    Input/output parameter Number of segments that can fit in the segment array and returns number of segments generated.

    Return Value

    kIOReturnSuccess on success, kIOReturnOverrun if the memory descriptor is exhausted, kIOReturnMessageTooLarge if the output segment function's address bits has insufficient resolution for a segment, kIOReturnNotReady if the DMA command has not be prepared, kIOReturnBadArgument if the DMA command doesn't have a memory descriptor yet or some of the parameters are NULL and kIOReturnNotReady if the DMA command is not prepared.

    Discussion

    Generates a list of physical segments from the given memory descriptor, relative to the current position of the descriptor. The constraints that are set during initialisation will be respected. This function maintains the state across multiple calls for efficiency. However the state is discarded if the new offset is not the expected one.

  • Get the current memory descriptor

    Declaration

    C++

    virtual const IOMemoryDescriptor *getMemoryDescriptor() const;

  • Returns the offset and length into the target IOMemoryDescriptor of a prepared IODDMACommand.

    Declaration

    C++

    virtual IOReturn getPreparedOffsetAndLength( UInt64 *offset, UInt64 *length);

    Parameters

    offset

    returns the starting offset in the memory descriptor the DMA command was prepared with. Pass NULL for don't care.

    length

    returns the length in the memory descriptor the DMA command was prepared with. Pass NULL for don't care.

    Return Value

    An IOReturn code. kIOReturnNotReady if the IODMACommand is not prepared.

    Discussion

    If successfully prepared, returns the offset and length into the IOMemoryDescriptor. Will fail for an unprepared IODMACommand.

  • Primary initializer for the IODMACommand class.

    Declaration

    C++

    virtual bool initWithSpecification( SegmentFunction outSegFunc, UInt8 numAddressBits, UInt64 maxSegmentSize, MappingOptions mappingOptions = kMapped, UInt64 maxTransferSize = 0, UInt32 alignment = 1, IOMapper *mapper = 0, void *refCon = 0);

    Parameters

    outSegFunc

    SegmentFunction to call to output one physical segment. A set of nine commonly required segment functions are provided.

    numAddressBits

    Number of bits that the hardware uses on its internal address bus. Typically 32 but may be more on modern hardware. A 0 implies no-restriction other than that implied by the output segment function.

    maxSegmentSize

    Maximum allowable size for one segment. Defaults to 0 which means any size.

    mappingOptions

    is the type of mapping that is required to translate an IOMemoryDescriptor into the desired number of bits. For instance if your hardware only supports 32 bits but must run on machines with > 4G of RAM some mapping will be required. Number of bits will be specified in numAddressBits, see below.This parameter can take 3 values:- kNonCoherent - used for non-coherent hardware transfers, Mapped - Validate that all I/O bus generated addresses are within the number of addressing bits specified, Bypassed indicates that bypassed addressing is required, this is used when the hardware transferes are into coherent memory but no mapping is required. See also prepare() for failure cases.

    maxTransferSize

    Maximum size of an entire transfer. Defaults to 0 indicating no maximum.

    alignment

    Alignment restriction, in bytes, on I/O bus addresses. Defaults to single byte alignment.

    mapper

    For mapping types kMapped & kBypassed mapper is used to define the hardware that will perform the mapping, defaults to the system mapper.

    Return Value

    Can fail if the mapping type is not recognised, if one of the 3 mandatory parameters are set to 0, if a 32 bit output function is selected when more than 32 bits of address is required or, if kBypassed is requested on a machine that doesn't support bypassing. Returns true otherwise.

  • Output big-endian Segment32 output segment function.

    Declaration

    C++

    static bool OutputBig32( IODMACommand *target, Segment64 seg, void *segs, UInt32 ind);

  • Output big-endian Segment64 output segment function.

    Declaration

    C++

    static bool OutputBig64( IODMACommand *target, Segment64 seg, void *segs, UInt32 ind);

  • Output host natural Segment32 output segment function.

    Declaration

    C++

    static bool OutputHost32( IODMACommand *target, Segment64 seg, void *segs, UInt32 ind);

  • Output host natural Segment64 output segment function.

    Declaration

    C++

    static bool OutputHost64( IODMACommand *target, Segment64 seg, void *segs, UInt32 ind);

  • Output little-endian Segment32 output segment function.

    Declaration

    C++

    static bool OutputLittle32( IODMACommand *target, Segment64 seg, void *segs, UInt32 ind);

  • Output little-endian Segment64 output segment function.

    Declaration

    C++

    static bool OutputLittle64( IODMACommand *target, Segment64 seg, void *segs, UInt32 ind);

  • Prepare the memory for an I/O transfer.

    Declaration

    C++

    virtual IOReturn prepare( UInt64 offset = 0, UInt64 length = 0, bool flushCache = true, bool synchronize = true);

    Parameters

    offset

    defines the starting offset in the memory descriptor the DMA command will operate on. genIOVMSegments will produce its results based on the offset and length passed to the prepare method.

    length

    defines the ending position in the memory descriptor the DMA command will operate on. genIOVMSegments will produce its results based on the offset and length passed to the prepare method.

    flushCache

    Flush the caches for the memory descriptor and make certain that the memory cycles are complete. Defaults to true for kNonCoherent and is ignored by the other types.

    synchronize

    Copy any buffered data back from the target IOMemoryDescriptor. Defaults to true, if synchronize() is being used to explicitly copy data, passing false may avoid an unneeded copy.

    Return Value

    An IOReturn code.

    Discussion

    Allocate the mapping resources neccessary for this transfer, specifying a sub range of the IOMemoryDescriptor that will be the target of the I/O. The complete() method frees these resources. Data may be copied to buffers for kIODirectionOut memory descriptors, depending on hardware mapping resource availabilty or alignment restrictions. It should be noted that the this function may block and should only be called on the clients context, i.e never call this routine while gated; also the call itself is not thread safe though this should be an issue as each IODMACommand is independant.

  • Prepare the memory for an I/O transfer with a new specification.

    Declaration

    C++

    virtual IOReturn prepareWithSpecification( SegmentFunction outSegFunc, UInt8 numAddressBits, UInt64 maxSegmentSize, MappingOptions mappingOptions = kMapped, UInt64 maxTransferSize = 0, UInt32 alignment = 1, IOMapper *mapper = 0, UInt64 offset = 0, UInt64 length = 0, bool flushCache = true, bool synchronize = true);

    Parameters

    outSegFunc

    SegmentFunction to call to output one physical segment. A set of nine commonly required segment functions are provided.

    numAddressBits

    Number of bits that the hardware uses on its internal address bus. Typically 32 but may be more on modern hardware. A 0 implies no-restriction other than that implied by the output segment function.

    maxSegmentSize

    Maximum allowable size for one segment. Defaults to 0 which means any size.

    mappingOptions

    is the type of mapping that is required to translate an IOMemoryDescriptor into the desired number of bits. For instance if your hardware only supports 32 bits but must run on machines with > 4G of RAM some mapping will be required. Number of bits will be specified in numAddressBits, see below.This parameter can take 3 values:- kNonCoherent - used for non-coherent hardware transfers, Mapped - Validate that all I/O bus generated addresses are within the number of addressing bits specified, Bypassed indicates that bypassed addressing is required, this is used when the hardware transferes are into coherent memory but no mapping is required. See also prepare() for failure cases.

    maxTransferSize

    Maximum size of an entire transfer. Defaults to 0 indicating no maximum.

    alignment

    Alignment restriction, in bytes, on I/O bus addresses. Defaults to single byte alignment.

    mapper

    For mapping types kMapped & kBypassed mapper is used to define the hardware that will perform the mapping, defaults to the system mapper.

    offset

    defines the starting offset in the memory descriptor the DMA command will operate on. genIOVMSegments will produce its results based on the offset and length passed to the prepare method.

    length

    defines the ending position in the memory descriptor the DMA command will operate on. genIOVMSegments will produce its results based on the offset and length passed to the prepare method.

    flushCache

    Flush the caches for the memory descriptor and make certain that the memory cycles are complete. Defaults to true for kNonCoherent and is ignored by the other types.

    synchronize

    Copy any buffered data back from the target IOMemoryDescriptor. Defaults to true, if synchronize() is being used to explicitly copy data, passing false may avoid an unneeded copy.

    Return Value

    An IOReturn code. Can fail if the mapping type is not recognised, if one of the 3 mandatory parameters are set to 0, if a 32 bit output function is selected when more than 32 bits of address is required or, if kBypassed is requested on a machine that doesn't support bypassing.

    Discussion

    Allocate the mapping resources neccessary for this transfer, specifying a sub range of the IOMemoryDescriptor that will be the target of the I/O. The complete() method frees these resources. Data may be copied to buffers for kIODirectionOut memory descriptors, depending on hardware mapping resource availabilty or alignment restrictions. It should be noted that the this function may block and should only be called on the clients context, i.e never call this routine while gated; also the call itself is not thread safe though this should be an issue as each IODMACommand is independant.

  • Copy data from the IODMACommand's buffer to the specified buffer.

    Declaration

    C++

    UInt64 readBytes( UInt64offset, void *bytes, UInt64length);

    Parameters

    offset

    A byte offset into the IODMACommand's memory, relative to the prepared offset.

    bytes

    The caller supplied buffer to copy the data to.

    length

    The length of the data to copy.

    Return Value

    The number of bytes copied, zero will be returned if the specified offset is beyond the prepared length of the IODMACommand.

    Discussion

    This method copies data from the IODMACommand's memory at the given offset, to the caller's buffer. The IODMACommand must be prepared, and the offset is relative to the prepared offset.

  • Sets and resets the DMACommand's current memory descriptor

    Declaration

    C++

    virtual IOReturn setMemoryDescriptor( const IOMemoryDescriptor *mem, bool autoPrepare = true);

    Parameters

    mem

    A pointer to the current I/Os memory descriptor.

    autoPrepare

    An optional boolean variable that will call the prepare() function automatically after the memory descriptor is processed. Defaults to true.

    Return Value

    Returns kIOReturnSuccess, kIOReturnBusy if currently prepared, kIOReturnNoSpace if the length(mem) >= Maximum Transfer Size or the error codes returned by prepare() (qv).

    Discussion

    The DMA command will configure itself based on the information that it finds in the memory descriptor. It looks for things like the direction of the memory descriptor and whether the current memory descriptor is already mapped into some IOMMU. As a programmer convenience it can also prepare the DMA command immediately. See prepare(). Note the IODMACommand is designed to used multiple times with a succession of memory descriptors, making the pooling of commands possible. It is an error though to attempt to reset a currently prepared() DMA command. Warning: This routine may block so never try to autoprepare an IODMACommand while in a gated context, i.e. one of the WorkLoops action call outs.

  • Bring IOMemoryDescriptor and IODMACommand buffers into sync.

    Declaration

    C++

    virtual IOReturn synchronize( IOOptionBitsoptions);

    Parameters

    options

    Specifies the direction of the copy: kIODirectionOut copy IOMemoryDesciptor memory to any IODMACommand buffers. By default this action takes place automatically at prepare(). kIODirectionIn copy any IODMACommand buffers back to the IOMemoryDescriptor. By default this action takes place automatically at complete(). kForceDoubleBuffer copy the entire prepared range to a new page aligned buffer.

    Return Value

    kIOReturnNotReady if not prepared, kIOReturnBadArgument if invalid options are passed, kIOReturnSuccess otherwise.

    Discussion

    This method should not be called unless a prepare was previously issued. If needed a caller may synchronize any IODMACommand buffers with the original IOMemoryDescriptor buffers.

  • Creates and initialises an IODMACommand in one operation if this version of the operating system supports it.

    Declaration

    C++

    static inline IOReturn weakWithSpecification ( IODMACommand **newCommand, SegmentFunction outSegFunc, UInt8 numAddressBits, UInt64 maxSegmentSize, MappingOptions mapType = kMapped, UInt64 maxTransferSize = 0, UInt32 alignment = 1, IOMapper *mapper = 0, void *refCon = 0) __attribute__((always_inline));

    Parameters

    newCommand

    Output reference variable of the newly created IODMACommand.

    outSegFunc

    SegmentFunction to call to output one physical segment. A set of nine commonly required segment functions are provided.

    numAddressBits

    Number of bits that the hardware uses on its internal address bus. Typically 32 but may be more on modern hardware. A 0 implies no-restriction other than that implied by the output segment function.

    maxSegmentSize

    Maximum allowable size for one segment. Zero is treated as an unlimited segment size.

    mapType

    is the type of mapping that is required to translate an IOMemoryDescriptor into the desired number of bits. For instance if your hardware only supports 32 bits but must run on machines with > 4G of RAM some mapping will be required. Number of bits will be specified in numAddressBits, see below. This parameter can take 3 values:- kNonCoherent - used for non-coherent hardware transfers, Mapped - Validate that all I/O bus generated addresses are within the number of addressing bits specified, Bypassed indicates that bypassed addressing is required, this is used when the hardware transfers are into coherent memory but no mapping is required. See also prepare() for failure cases.

    maxTransferSize

    Maximum size of an entire transfer. Defaults to 0 indicating no maximum.

    alignment

    Alignment restriction, in bytes, on I/O bus addresses. Defaults to single byte alignment.

    mapper

    For mapping types kMapped & kBypassed mapper is used to define the hardware that will perform the mapping, defaults to the system mapper.

    Return Value

    kIOReturnSuccess if everything is OK, otherwise kIOReturnBadArgument if newCommand is NULL, kIOReturnUnsupported if the kernel doesn't export IODMACommand or IOReturnError if the new command fails to init, q.v. initWithSpecification.

    Discussion

    Factory function to create and initialise an IODMACommand in one operation. The function allows a developer to 'weak' link with IODMACommand. This function will return kIOReturnUnsupported if the IODMACommand is unavailable. This function is actually fairly slow so it will be better to call it once then clone the successfully create command using cloneCommand (q.v.).

  • Creates and initializes an IODMACommand in one operation.

    Declaration

    C++

    static IODMACommand * withSpecification( SegmentFunction outSegFunc, UInt8 numAddressBits, UInt64 maxSegmentSize, MappingOptions mappingOptions = kMapped, UInt64 maxTransferSize = 0, UInt32 alignment = 1, IOMapper *mapper = 0, void *refCon = 0);

    Parameters

    outSegFunc

    SegmentFunction to call to output one physical segment. A set of nine commonly required segment functions are provided.

    numAddressBits

    Number of bits that the hardware uses on its internal address bus. Typically 32 but may be more on modern hardware. A 0 implies no-restriction other than that implied by the output segment function.

    maxSegmentSize

    Maximum allowable size for one segment. If 0 is passed the maximum segment size is unlimited.

    mappingOptions

    is the type of mapping that is required to translate an IOMemoryDescriptor into the desired number of bits. For instance if your hardware only supports 32 bits but must run on machines with > 4G of RAM some mapping will be required. Number of bits will be specified in numAddressBits, see below.This parameter can take 3 values:- kNonCoherent - used for non-coherent hardware transfers, Mapped - Validate that all I/O bus generated addresses are within the number of addressing bits specified, Bypassed indicates that bypassed addressing is required, this is used when the hardware transferes are into coherent memory but no mapping is required. See also prepare() for failure cases.

    maxTransferSize

    Maximum size of an entire transfer. Defaults to 0 indicating no maximum.

    alignment

    Alignment restriction, in bytes, on I/O bus addresses. Defaults to single byte alignment.

    mapper

    For mapping types kMapped & kBypassed mapper is used to define the hardware that will perform the mapping, defaults to the system mapper.

    Return Value

    Returns a new memory cursor if successfully created and initialized, 0 otherwise.

    Discussion

    Factory function to create and initialize an IODMACommand in one operation.

  • Copy data to the IODMACommand's buffer from the specified buffer.

    Declaration

    C++

    UInt64 writeBytes( UInt64offset, const void *bytes, UInt64length);

    Parameters

    offset

    A byte offset into the IODMACommand's memory, relative to the prepared offset.

    bytes

    The caller supplied buffer to copy the data from.

    length

    The length of the data to copy.

    Return Value

    The number of bytes copied, zero will be returned if the specified offset is beyond the prepared length of the IODMACommand.

    Discussion

    This method copies data to the IODMACommand's memory at the given offset, from the caller's buffer. The IODMACommand must be prepared, and the offset is relative to the prepared offset.

Callbacks

  • Declaration

    C++

    typedef bool ( *SegmentFunction)( IODMACommand *target, Segment64 segment, void *segments, UInt32 segmentIndex);

    Parameters

    segment

    The 64Bit I/O bus address and length.

    segments

    Base of the output vector of DMA address length pairs.

    segmentIndex

    Index to output 'segment' in the 'segments' array.

    Return Value

    Returns true if segment encoding succeeded. false may be returned if the current segment does not fit in an output segment, i.e. a 38bit address wont fit into a 32 encoding.

    Discussion

    Pointer to a C function that translates a 64 segment and outputs a single desired segment to the array at the requested index. There are a group of pre-implemented SegmentFunctions that may be usefull to the developer below.

    Import Statement

    Objective-C

    #include <IODMACommand.h>;

    Availability

    Available in OS X v10.0 through OS X v10.5.

Constants

  • fMemory

    fMemory

    memory descriptor for current I/O.

  • Declaration

    CPlusPlus

    #define kIODMACommandOutputBig32 (IODMACommand::OutputBig32) #define kIODMACommandOutputBig64 (IODMACommand::OutputBig64) #define kIODMACommandOutputLittle64 (IODMACommand::OutputLittle64) #define kIODMACommandOutputHost32 (IODMACommand::OutputHost32) #define kIODMACommandOutputHost64 (IODMACommand::OutputHost64) #define kIODMACommandOutputLittle32 (IODMACommand::OutputLittle32) #define kIODMACommandOutputBig64 (IODMACommand::OutputBig64) #define kIODMACommandOutputLittle64 (IODMACommand::OutputLittle64)

    Constants

    • kIODMACommandOutputBig32

      kIODMACommandOutputBig32

      Output big-endian Segment32 output segment function.

      Available in OS X v10.4 and later.

    • kIODMACommandOutputBig64

      kIODMACommandOutputBig64

      Output little-endian Segment64 output segment function.

      Available in OS X v10.4 and later.

    • kIODMACommandOutputBig64

      kIODMACommandOutputBig64

      Output big-endian Segment64 output segment function.

      Available in OS X v10.4 and later.

    • kIODMACommandOutputHost32

      kIODMACommandOutputHost32

      Output host natural Segment32 output segment function.

      Available in OS X v10.4 and later.

    • kIODMACommandOutputHost64

      kIODMACommandOutputHost64

      Output host natural Segment64 output segment function.

      Available in OS X v10.4 and later.

    • kIODMACommandOutputLittle32

      kIODMACommandOutputLittle32

      Output little-endian Segment32 output segment function.

      Available in OS X v10.4 and later.

    • kIODMACommandOutputLittle64

      kIODMACommandOutputLittle64

      Output little-endian Segment64 output segment function.

      Available in OS X v10.4 and later.

    • kIODMACommandOutputLittle64

      kIODMACommandOutputLittle64

      Output big-endian Segment64 output segment function.

      Available in OS X v10.4 and later.

  • Mapping types to indicate the desired mapper type for translating memory descriptors into I/O DMA Bus addresses.

    Declaration

    CPlusPlus

    enum MappingOptions { kMapped = 0x00000000, kBypassed = 0x00000001, kNonCoherent = 0x00000002, kTypeMask = 0x0000000f, kNoCacheStore = 0x00000010 , // Memory in descriptor kOnChip = 0x00000020 , // Indicates DMA is on South Bridge kIterateOnly = 0x00000040 // DMACommand will be used as a cursor only };

    Constants

    • kNonCoherent

      kNonCoherent

      Used by drivers for non-coherent transfers, implies unmapped memmory

    • kMapped

      kMapped

      Allow a driver to define addressing size

    • kBypassed

      kBypassed

      Allow drivers to bypass any mapper

    • kMaxMappingOptions

      kMaxMappingOptions

      Internal use only

    Import Statement

    Objective-C

    #include <IODMACommand.h>;

    Availability

    Available in OS X v10.4 through OS X v10.5.

  • Options for the synchronize method.

    Declaration

    CPlusPlus

    enum SynchronizeOptions { kForceDoubleBuffer = 0x01000000 };

    Constants

    • kForceDoubleBuffer

      kForceDoubleBuffer

      Copy the entire prepared range to a new page aligned buffer.

    Import Statement

    Objective-C

    #include <IODMACommand.h>;

    Availability

    Available in OS X v10.4 through OS X v10.5.

  • Data Types

    • Declaration

      C++

      struct Segment32 { UInt32 fIOVMAddr, fLength; };

      Discussion

      A 32 bit I/O bus address/length pair

    • Declaration

      C++

      struct Segment64 { UInt64 fIOVMAddr, fLength; };

      Discussion

      A 64 bit I/O bus address/length pair

    Instance Variables

    • Reserved for future use. (Internal use only)

      Declaration

      C++

      struct IODMACommandInternal * reserved;

    • Reference Constant, client defined publicly avialable

      Declaration

      C++

      void *fRefCon;

    • The action method called when an event has been delivered

      Declaration

      C++

      SegmentFunction fOutSeg;

    • Number of contiguous segments required for the current memory descriptor and desired mapping

      Declaration

      C++

      UInt32 fNumSegments;

    • Number of bits that the hardware can address

      Declaration

      C++

      UInt32 fNumAddressBits;

    • Maximum size of a transfer that this memory cursor is allowed to generate

      Declaration

      C++

      UInt64 fMaxTransferSize;

    • Maximum size of one segment in a scatter/gather list

      Declaration

      C++

      UInt64 fMaxSegmentSize;

    • What type of I/O virtual address mapping is required for this command

      Declaration

      C++

      MappingOptions fMappingOptions;

    • Client defined mapper.

      Declaration

      C++

      IOMapper *fMapper;

    • Mask to be ored into the address to bypass the given iommu's mapping.

      Declaration

      C++

      UInt64 fBypassMask;

    • Alignment restriction mask.

      Declaration

      C++

      UInt32 fAlignMask;

    • fActive indicates that this DMA command is currently prepared and ready to go

      Declaration

      C++

      UInt32 fActive;