Can device featuring advanced can filtering and message acceptance

[0002] This application claims the full benefit and priority of U.S. Provisional Application Serial No. 60/154,022, filed on Sep. 15, 1999, the disclosure of which is fully incorporated herein for all purposes.

[0003] The present invention relates generally to the field of data communications, and more particularly, to the field of serial communications bus controllers and microcontrollers that incorporate the same.

[0004] CAN (Control Area Network) is an industry-standard, two-wire serial communications bus that is widely used in automotive and industrial control applications, as well as in medical devices, avionics, office automation equipment, consumer appliances, and many other products and applications. CAN controllers are currently available either as stand-alone devices adapted to interface with a microcontroller or as circuitry integrated into or modules embedded in a microcontroller chip. Since 1986, CAN users (software programmers) have developed numerous high-level CAN Application Layers (CALs) which extend the capabilities of the CAN while employing the CAN physical layer and the CAN frame format, and adhering to the CAN specification. CALs have heretofore been implemented primarily in software, with very little hardware CAL support. Consequently, CALs have heretofore required a great deal of host CPU intervention, thereby increasing the processing overhead and diminishing the performance of the host CPU.

[0005] Thus, there is a need in the art for a CAN hardware implementation of CAL functions normally implemented in software in order to offload these tasks from the host CPU to the CAN hardware, thereby enabling a great savings in host CPU processing resources and a commensurate improvement in host CPU performance. One of the most demanding and CPU resource-intensive CAL functions is message management, which entails the handling, storage, and processing of incoming CAL/CAN messages received over the CAN serial communications bus and/or outgoing CAL/CAN messages transmitted over the CAN serial communications bus. CAL protocols, such as DeviceNet, CANopen, and OSEK, deliver long messages distributed over many CAN frames, which methodology is sometimes referred to as “fragmented” or “segmented” messaging. The process of assembling such fragmented, multi-frame messages has heretofore required a great deal of host CPU intervention. In particular, CAL software running on the host CPU actively monitors and manages the buffering and processing of the message data, in order to facilitate the assembly of the message fragments or segments into complete messages.

[0006] Based on the above and foregoing, it can be appreciated that there presently exists a need in the art for a hardware implementation of CAL functions normally implemented in software in order to offload these tasks from the host CPU, thereby enabling a great savings in host CPU processing resources and a commensurate improvement in host CPU performance.

[0007] The assignee of the present invention has recently developed a new microcontroller product, designated “XA-C3”, that fulfills this need in the art. The XA-C3 is the newest member of the Philips XA (eXtended Architecture) family of high performance 16-bit single-chip microcontrollers. It is believed that the XA-C3 is the first chip that features hardware CAL support.

[0008] The XA-C3 is a CMOS 16-bit CAL/CAN 2.0B microcontroller that incorporates a number of different inventions, including the present invention. These inventions include novel techniques and hardware for filtering, buffering, handling, and processing CAL/CAN messages, including the automatic assembly of multi-frame fragmented messages with minimal CPU intervention, as well as for managing the storage and retrieval of the message data, and the memory resources utilized therefor.

[0009] The present invention relates to a CAN microcontroller that supports a plurality (e.g., 32) of message objects, each one of which is assigned a respective message buffer within an on-chip and/or off-chip portion of the overall data memory space of the CAN microcontroller. The location and size of each of the message buffers can be reconfigured by the user (programmer) by simple programming of memory-mapped registers provided for this purpose. The message buffers are used to store incoming (receive) messages and to stage outgoing (transmit) messages. With the XA-C3 microcontroller that constitutes a presently preferred implementation of the present invention, Direct Memory Access (DMA) is employed to enable the XA-C3 CAN module to directly access any of the 32 message buffers without interrupting the processor core. This message storage scheme provides a great deal of flexibility to the user, as the user is free to use as much or as little message storage area as an application requires, and is also free to position the message buffers wherever it is most convenient.

[0010] This message storage scheme is a key element of the unique “message management” capabilities of the XA-C3 CAN microcontroller, as this scheme enables the XA-C3 CAN/CAL module to concurrently assemble many (up to 32) incoming, fragmented messages of varying lengths, and, at the same time, stage multiple outgoing messages for tansmission. Since incoming message assembly is handled entirely in hardware, the processor is free to perform other tasks, typically until a complete message is received and ready for processing.

[0011] All CAN devices have the ability to perform receive acceptance filtering or screening. Acceptance filtering is typically accomplished by comparing information in the header portion of a received CAN message to be filtered or screened (and, in some cases, also the first and/or second byte in the data field of the received CAN message) to one or more pre-established values, sometimes called acceptance filters or screeners. Software running on the processor core can perform the acceptance filtering at the expense of performance, because software requires an ample amount of CPU resources and time to perform acceptance filtering. Therefore, hardware filtering is desired in order to reduce the burden or overhead processing load on the CPU (processor core). However, with the presently available technology, hardware filtering is far less flexible than software filtering, and greatly increases the required die area for implementation of the CAN device.

[0012] The presently available acceptance filtering techniques include the use of a match field and a mask field to create what is sometimes referred to as a “match and mask” filter. Each combination of available match and mask fields constitutes a separate filter or filter object. A basic CAN device will typically have one or two filters with full match and mask capabilities. A whole family of messages will typically be accepted. Software or hardware can be employed to actually perform the filtering. However, with the presently available technology, software must decide where to store each individual message. This type of filtering is generally the most flexible; however, it consumes the most CPU resources.

[0013] A full CAN device is one in which the hardware filters a message and matches it to one of a number of screeners to decide where to store the message. Each filter is linked to a message buffer. Upon a successful comparison of the incoming message header to one of the screeners, the message is stored in the message buffer associated with the matching screener. Generally, only “matching” and not “masking” is performed by each filter. However, there could be a “global mask” that is applied to all filters, or one filter could be designated to provide the “mask”.

[0014] Based on the above and foregoing, it can be appreciated that there presently exists a need in the art for a much more flexible and powerful technique for message acceptance filtering in a CAN device, e.g., a CAN microcontroller, that overcomes the above-discussed limitations and shortcomings of the presently available technology, particularly such a technique that does not explode the required die area to implement the CAN device, and which is performed in hardware, rather than software, to thereby minimize loading of the processor (CPU) core, and thereby dramatically improve system performance.

[0015] The present invention fulfills this need in the art. As will become apparent hereinafter, the preferred implementation of the present invention in the XA-C3 microcontroller provides 32 fully programmable match and mask filters, which affords a much more powerful system than is possible with the presently available technology, without loading the processor core. The user can define and dedicate filters to a wide variety of messages without running out. This is especially useful for systems that need to dedicate filters to system management functions, such as unconnected message management, explicit I/O and Dup MAC used in the CAN protocol DeviceNet.

[0016] Most existing CAN devices can filter on an 11-bit CAN identifier (standard CAN frames) or a 29-bit CAN identifier (extended CAN frames). However, each of the 32 filters in the XA-C3 microcontroller can be individually configured to filter on either the 11-bit CAN identifier of standard CAN frames or the 29-bit CAN identifier of extended CAN frames. This allows for a system with mixed CAN devices and provides maximum flexibility.

[0017] Further, the XA-C3 microcontroller provides filtering that goes beyond the CAN specification. More particularly, in addition to filtering on the CAN identifier, the XA-C3 microcontroller permits filtering on the first and/or second data bytes of the data field of incoming CAN frames. This is a very useful feature for higher level CAN protocols such as DeviceNet, CanOpen, and Osek, which have traditionally required software to filter on the first two bytes of the data field.

[0018] Yet further, the XA-C3 microcontroller provides for masking on every bit on every filter (with the exception of the IDE bit). Traditional CAN devices have only provided a subset of this capability, allowing only global masking which is applied to all of the filter objects, or dedicating one or two objects to have the fully masking capability. No presently available device provides for full masking on every filter. This feature is extremely useful to a user that wants to define and receive families of messages under both the standard CAN specification and the higher level protocols such as DeviceNet, CanOpen, and Osek.

[0019] Additionally, the XA-C3 microcontroller provides for masking on the RTR bit in the CAN header, thereby allowing the device to receive and recognize an RTR frame. In this connection, the user can dedicate a filter object exclusively for the purpose of monitoring an RTR frame and/or sending out RTR frames.

[0020] The present invention encompasses, in one of its aspects, a CAN device that supports a plurality n (where n≧3) of message objects, including a plurality of registers associated with each message object, including at least one object match ID register that contains a multi-bit object match ID field, and at least one object mask register that contains a multi-bit object mask field; and, a CAN/CAL module that processes incoming messages. The CAN/CAL module assembles a multi-bit screener ID from selected bits of each incoming message to be acceptance filtered, compares the bits comprising the screener ID with corresponding bits of the object match ID field associated with each of at least designated ones of the plurality n of message objects, disregarding any bits of each object match ID field that are masked by corresponding bits of the associated object mask field, and then determines whether any of the comparisons results in a match. Any selected one or more bits of the object match ID field associated with each of the plurality n of message objects can be set to ‘1’ or ‘0’, and any selected one or more bits of the object mask field associated with each of the plurality n of message objects can be set to ‘1’ or ‘0’ in order to mask any selected one or more bits of the associated object match ID field, whereby the combination of the object match ID field and the object mask field associated with each of the plurality n of message objects constitutes a fully programmable match and mask filter. The CAN/CAL module has the capability to perform acceptance filtering on incoming messages constituting either standard or extended CAN frames. If more than one match is detected, a lowest-numbered (or, alternatively, highest-numbered) one of the message objects is designated to be the matching message object.

[0021] In a presently preferred embodiment, the plurality of registers further include an RTR register associated with each message object. The RTR register associated with each message object contains an RTR handling enable field that is programmable for the purpose of enabling or disabling the associated message object to detect receipt of an incoming message that constitutes an Remote Transmit Request frame and/or to transmit a Remote Transmit Request frame. Preferably, an incoming message for which a match is detected is stored by a DMA engine in the message buffer associated with the matching message object.

[0022] In another of its aspects, the present invention encompasses a CAN microcontroller that includes, in addition to the above-described registers (which are preferably implemented as memory-mapped registers in the microcontroller data memory space) and the above-described CAN/CAL module, a processor core that runs CAN applications, and a DMA engine that enables the CAN/CAL module to directly access the message buffers, e.g., for storage and retrieval of messages.

[0023] In yet another of its aspects, the present invention encompasses, in a CAN device that supports a plurality n (where n≧3) of message objects each of which has an associated message buffer, at least one associated match ID register, and at least one associated mask register, a method for acceptance filtering incoming CAN frames. The method includes the steps of programming the at least one match ID register associated with each of at least designated ones of the message objects by selectively setting each of the bits in a multi-bit match ID field contained therein to ‘1’ or ‘0’; programming the at least one mask register associated with each of the at least designated ones of the message objects by selectively setting each of the bits in a multi-bit mask field contained therein to ‘1’ or ‘0’; extracting a multi-bit screener ID field from a received CAN frame; and, comparing the extracted screener ID field to the multi-bit match ID field stored in the at least one match ID register associated with each of the at least designated ones of the message objects, excluding from the comparison any bits of the match ID field masked by corresponding bits of the associated mask field stored in the at least one associated mask register. If a match if found as a result of the comparing step, the method further includes the step of storing data bytes of the received CAN frame in the message buffer associated with the matching message object.

[0024] These and various other aspects, features, and advantages of the present invention will be readily understood with reference to the following detailed description of the invention read in conjunction with the accompanying drawings, in which:

[0025] FIG. 1 is a diagram illustrating the format of a Standard CAN Frame and the format of an Extended CAN Frame;

[0026] FIG. 2 is a diagram illustrating the interleaving of CAN Data Frames of different, unrelated messages;

[0027] FIG. 3 is a high-level, functional block diagram of the XA-C3 microcontroller;

[0028] FIG. 4 is a table listing all of the Memory Mapped Registers (MMRs) provided by the XA-C3 microcontroller;

[0029] FIG. 5 is a diagram illustrating the mapping of the overall data memory space of the XA-C3 microcontroller;

[0030] FIG. 6 is a diagram illustrating the MMR space contained within the overall data memory space of the XA-C3 microcontroller;

[0031] FIG. 7 is a diagram illustrating formation of the base address of the on-chip XRAM of the XA-C3 microcontroller, with an object n message buffer mapped into off-chip data memory;

[0032] FIG. 8 is a diagram illustrating formation of the base address of the on-chip XRAM of the XA-C3 microcontroller, with an object n message buffer mapped into the on-chip XRAM;

[0033] FIG. 9 is a diagram illustrating the Screener ID Field for a Standard CAN Frame;

[0034] FIG. 10 is a diagram illustrating the Screener ID Field for an Extended CAN Frame;

[0035] FIG. 11 is a diagram illustrating the message storage format for fragmented CAL messages; and,

[0036] FIG. 12 is a diagram illustrating the message storage format for fragmented CAN messages.

[0037] The present invention is described below in the context of a particular implementation thereof, i.e., in the context of the XA-C3 microcontroller manufactured by Philips Semiconductors. Of course, it should be clearly understood that the present invention is not limited to this particular implementation, as any one or more of the various aspects and features of the present invention disclosed herein can be utilized either individually or any combination thereof, and in any desired application, e.g., in a stand-alone CAN controller device or as part of any other microcontroller or system.

[0038] The following terms used herein in the context of describing the preferred embodiment of the present invention (i.e., the XA-C3 microcontroller) are defined as follows:

[0039] With reference now to FIG. 3, there can be seen a high-level block diagram of the XA-C3 microcontroller 20. The XA-C3 microcontroller 20 includes the following functional blocks that are fabricated on a single integrated circuit (IC) chip packaged in a 44-pin PLCC or a 44-pin LQFP package:

[0040] an XA CPU Core 22, that is currently implemented as a 16-bit fully static CPU with 24-bit program and data address range, that is upwardly compatible with the 80C51 architecture, and that has an operating frequency of up to 30 MHz;

[0041] a program or code memory 24 that is currently implemented as a 32K ROM/EPROM, and that is bi-directionally coupled to the XA CPU Core 22 via an internal Program bus 25. A map of the code memory space is depicted in FIG. 4;

[0042] a Data RAM 26 (internal or scratch pad data memory) that is currently implemented as a 1024 Byte portion of the overall XA-C3 data memory space, and that is bi-directionally coupled to the XA CPU Core 22 via an internal DATA bus 27;

[0043] an on-chip message buffer RAM or XRAM 28 that is currently implemented as a 512 Byte portion of the overall XA-C3 data memory space which may contain part or all of the CAN/CAL (Transmit & Receive Object) message buffers;

[0044] a Memory Interface (MIF) unit 30 that provides interfaces to generic memory devices such as SRAM, DRAM, flash, ROM, and EPROM memory devices via an external address/data bus 32, via an internal Core Data bus 34, and via an internal MMR bus 36;

[0045] a DMA engine 38 that provides 32 CAL DMA Channels;

[0046] a plurality of on-chip Memory Mapped Registers (MMRs) 40 that are mapped to the overall XA-C3 data memory space—a 4K Byte portion of the overall XA-C3 data memory space is reserved for MMRs. These MMRs include 32 (Message) Object or Address Pointers and 32 ID Screeners or Match IDs, corresponding to the 32 CAL Message Objects. A complete listing of all MMRs is provided in the Table depicted in FIG. 5;

[0047] a 2.0B CAN/DLL Core 42 that is the CAN Controller Core from the Philips SJA1000 CAN (2.0A/B) Data Link Layer (CDLL) device (hereinafter referred to as the “CAN Core Block” (CCB)); and,

[0048] an array of standard microcontroller peripherals that are bi-directionally coupled to the XA CPU Core 22 via a Special Function Register (SFR) bus 43. These standard microcontroller peripherals include Universal Asynchronous Receiver Transmitter (UART) 49, an SPI serial interface (port) 51, three standard timers/counters with toggle output capability, namely, Timer 0 & Timer 1 included in Timer block 53, and Timer 2 included in Timer block 54, a Watchdog Timer 55, and four 8-bit I/O ports, namely, Ports 0-3 included in block 61, each of which has 4 programmable output configurations.

[0049] The DMA engine 38, the MMRs 40, and the CCB 42 can collectively be considered to constitute a CAN/CAL module 77, and will be referred to as such at various times throughout the following description. Further, the particular logic elements within the CAN/CAL module 77 that perform “message management” and “message handling” functions will sometimes be referred to as the “message management engine” and the “message handler”, respectively, at various times throughout the following description. Other nomenclature will be defined as it introduced throughout the following description.

[0050] As previously mentioned, the XA-C3 microcontroller 20 automatically implements, in hardware, many message management and other functions that were previously only implemented in software running on the host CPU (or not implemented at all), including transparent, automatic re-assembly of up to 32 concurrent, interleaved, multi-flame, fragmented CAL messages. For each application that is installed to run on the host CPU (i.e., the XA CPU Core 22), the user (software programmer) must set-up the hardware for performing these functions by programming certain ones of the MMRs and SFRs in the manner set forth in the XA-C3 Functional Specification and XA-C3 CAN Transport Layer Controller User Manual. The register programming procedures that are most relevant to an understanding of the present invention are described below, followed by a description of the various message management and other functions that are automatically performed by the CAL/CAN module 77 during operation of the XA-C3 microcontroller 20 after it has been properly set-up by the user. Following these sections, a more detailed description of the particular invention to which this application is directed is provided.

[0051] Set-up/Programming Procedures

[0052] As an initial matter, the user must map the overall XA-C3 data memory space, as illustrated in FIG. 5. In particular, subject to certain constraints, the user must specify the starting or base address of the XRAM 28 and the starting or base address of the MMRs 40. The base address of the MMRs 40 can be specified by appropriately programming Special Function Registers (SFRs) MRBL and MRBH. The base address of the XRAM 28 can be specified by appropriately programming the MMRs designated MBXSR and XRAMB (see FIG. 4).

[0053] The user can place the 4 KByte space reserved for MMRs 40 anywhere within the entire 16 Mbyte data memory space supported by the XA architecture, other than at the very bottom of the memory space (i.e., the first 1 KByte portion, starting address of 000000h), where it would conflict with the on-chip Data RAM 26 that serves as the internal or scratch-pad memory. The 4 KBytes of MMR space will always start at a 4K boundary. The reset values for MRBH and MRBL are 0Fh and F0h, respectively. Therefore, after a reset, the MMR space is mapped to the uppermost 4K Bytes of Data Segment 0Fh, but access to the MMRs 40 is disabled. The first 512 Bytes (offset 000h-1FFh) of MMR space are the Message Object Registers (eight per Message Object) for objects n=0-31, as is shown in FIG. 6.

[0054] The base address of the XRAM 28 is determined by the contents of the MMRs designated MBXSR and XRAMB, as is shown in FIGS. 7 and 8. As previously mentioned, the 512 Byte XRAM 28 is where some (or all) of the 32 (Rx/Tx) message buffers (corresponding to Message Objects n=0-31) reside. The message buffers can be extended off-chip to a maximum of 8 KBytes. This off-chip expansion capability can accommodate up to thirty-two, 256-Byte message buffers. Since the uppermost 8 bits of all message buffer addresses are formed by the contents of the MBXSR register, the XRAM 28 and all 32 message buffers must reside in the same 64K Byte data memory segment. Since the XA-C3 microcontroller 20 only provides address lines A0-A19 for accessing external memory, all external memory addresses must be within the lowest 1MByte of address space. Therefore, if there is external memory in the system into which any of the 32 message buffers will be mapped, then all 32 message buffers and the XRAM 28 must also be mapped entirely into that same 64K Byte segment, which must be below the 1MByte address limit.

[0055] After the memory space has been mapped, the user can set-up or define up to 32 separate Message Objects, each of which can be either a Transmit (Tx) or a Receive (Rx) Message Object. A Rx Message Object can be associated either with a unique CAN ID, or with a set of CAN IDs which share certain ID bit fields. As previously mentioned, each Message Object has its own reserved block of data memory space (up to 256 Bytes), which is referred to as that Message Object's message buffer. As will be seen, both the size and the base address of each Message Object's message buffer is programmable.

[0056] As previously mentioned, each Message Object is associated with a set of eight MMRs 40 dedicated to that Message Object. Some of these registers function differently for Tx Message Objects than they do for Rx Message Objects. These eight MMRs 40 are designated “Message Object Registers” (see FIG. 4).

[0057] The names of these eight MMRs 40 are:

[0058] where n ranges from 0 to 31 (i.e., corresponding to 32 independent Message Objects).

[0059] In general, the user defines or sets up a Message Object by configuring (programming) some or all of the eight MMRs dedicated to that Message Object, as will be described below. Additionally, as will be described below, the user must configure (program) the global GCTL register, whose bits control global parameters that apply to all Message Objects.

[0060] In particular, the user can specify the Match ID value for each Message Object to be compared against the Screener IDs extracted from incoming CAN Frames for Acceptance Filtering. The Match ID value for each Message Object n is specified in the MnMIDH and MnMIDL registers associated with that Message Object n. The user can mask any Screener ID bits which are not intended to be used in Acceptance Filtering, on an object-by-object basis, by writing a logic ‘1’ in the desired (to-be-masked) bit position(s) in the appropriate MnMSKH and/or MnMSKL registers associated with each particular Message Object n. The user is responsible, on set-up, for assigning a unique message buffer location for each Message Object n. In particular, the user can specify the least significant 16 bits of the base address of the message buffer for each particular Message Object n by programming the MnBLR register associated with that Message Object n. The upper 8 bits of the 24-bit address, for all Message Objects, are specified by the contents of the MBXSR register, as previously discussed, so that the message buffers for all Message Objects reside within the same 64KByte memory segment. The user is also responsible, on set-up, for specifying the size of the message buffer for each Message Object n. In particular, the user can specify the size of the message buffer for each particular Message Object n by programming the MnBSZ register associated with that Message Object n. The top location of the message buffer for each Message Object n is determined by the size of that message buffer as specified in the corresponding MnBSZ register.

[0061] The user can configure (program) the MnCTL register associated with each particular Message Object n in order to enable or disable that Message Object n, in order to define or designate that Message Object n as a Tx or Rx Message Object; in order to enable or disable automatic hardware assembly of fragmented Rx messages (i.e., automatic fragmented message handling) for that Message Object n; in order to enable or disable automatic generation of a Message-Complete Interrupt for that Message Object n; and, in order to enable or not enable that Message Object n for Remote Transmit Request (RTR) handling. In CANopen and OSEK systems, the user must also initialize the MnFCR register associated with each Message Object n.

[0062] As previously mentioned, on set-up, the user must configure (program) the global GCTL register, whose bits control global parameters that apply to all Message Objects. In particular, the user can configure (program) the GCTL register in order to specify the high-level CAL protocol (if any) being used (e.g., DeviceNet, CANopen, or OSEK); in order to enable or disable automatic acknowledgment of CANopen Frames (CANopen auto-acknowledge); and, in order to specify which of two transmit (Tx) pre-arbitration schemes/policies is to be utilized (i.e., either Tx pre-arbitration based on CAN ID, with the object number being used as a secondary tie-breaker, or Tx pre-arbitration based on object number only).

[0063] During reception (i.e., when an incoming CAN Frame is being received by the XA-C3 microcontroller 20), the CAN/CAL module 77 will store the incorming CAN Frame in a temporary (13-Byte) buffer, and determine whether a complete, error-free CAN frame has been successfully received. If it is determined that a complete, error-free CAN Frame has been successfully received, then the CAN/CAL module 77 will initiate Acceptance Filtering in order to determine whether to accept and store that CAN Frame, or to ignore/discard that CAN Frame.

[0064] Acceptance Filtering

[0065] In general, because the XA-C3 microcontroller 20 provides the user with the ability to program separate Match ID and Mask fields for each of the 32 independent Message Objects, on an object-by-object basis, as described previously, the Acceptance Filtering process performed by the XA-C3 microcontroller 20 can be characterized as a “match and mask” technique. The basic objective of this Acceptance Filtering process is to determine whether a Screener ID field of the received CAN Frame (excluding the “don't care” bits masked by the Mask field for each Message Object) matches the Match ID of any enabled one of the 32 Message Objects that has been designated a Receive Message Object. If there is a match between the received CAN Frame and more than one Message Object, then the received CAN Frame will be deemed to have matched the Message Object with the lowest object number (n).

[0066] Message Storage:

[0067] Each incoming (received) CAN Frame that passes Acceptance Filtering, will be automatically stored, via the DMA engine 38, into the message buffer for the Receive Message Object that particular CAN Frame was found to have matched. In an exemplary implementation, the message buffers for all Message Objects are contained in the XRAM 28.

[0068] Message Assembly:

[0069] In general, the DMA engine 38 will transfer each accepted CAN Frame from the 13-byte pre-buffer to the appropriate message buffer (e.g., in the XRAM 28), one word at a time, starting from the address pointed to by the contents of the MBXSR and MnBLR registers. Every time the DMA engine 38 transfers a byte or a word, it has to re quest the bus. In this regard, the MIF unit 30 arbitrates between accesses from the XA CPU Core 22 and from the DMA engine 38. In general, bus arbitration is done on an “alternate” policy. After a DMA bus access, the XA CPU Core 22 will be granted bus access, if requested. After an XA CPU bus access, the DMA engine 38 will be granted bus access, if requested. (However, a burst access by the XA CPU Core 22 cannot be interrupted by a DMA bus access).

[0070] Once bus access is granted by the MIF unit 30, the DMA engine 38 will write data from the 13-byte pre-buffer to the appropriate message buffer location . The DMA engine 38 will keep requesting the bus, writing message data sequentially to the appropriate message buffer location until the whole accepted CAN Frame is transferred. After the DMA engine 38 has successfully transferred an accepted CAN Frame to the appropriate message buffer location , the contents of the message buffer will depend upon whether the message that the CAN Frame belongs to is a non-fragmented (single frame) message or a fragmented message. Each case is described below:

[0071] Non-Fragmented Message Assembly:

[0072] For Message Objects that have been set up with automatic fragmented message handling disabled (not enabled—i.e., the FRAG bit in the MnCTL register for that Message Object is set to ‘0’), the complete CAN ID of the accepted CAN Frame (which is either 11 or 29 bits, depending on whether the accepted CAN Frame is a Standard or Extended CAN Frame) is written into the MnMIDH and MnMIDL registers associated with the Message Object that has been deemed to constitute a match, once the DMA engine 38 has successfully transferred the accepted CAN Frame to the message buffer associated with that Message Object. This will permit the user application to see the exact CAN ID which resulted in the match, even if a portion of the CAN ID was masked for Acceptance Filtering. As a result of this mechanism, the contents of the MnMIDH and MnMIDL registers can change every time an incoming CAN Frame is accepted. Since the incoming CAN Frame must pass through the Acceptance Filter before it can be accepted, only the bits that are masked out will change. Therefore, the criteria for match and mask Acceptance Filtering will not change as a result of the contents of the MnMIDH and MnMIDL registers being changed in response to an accepted incoming CAN Frame being transferred to the appropriate message buffer.

[0073] Fragmented Message Assembly:

[0074] For Message Objects that have been set up with automatic fragmented message handling enabled (i.e., with the FRAG bit in the MnCTL register for that Message Object set to ‘1’), masking of the 11/29 bit CAN ID field is disallowed. As such, the CAN ID of the accepted CAN Frame is known unambiguously, and is contained in the MnMIDH and MnMIDL registers associated with the Message Object that has been deemed to constitute a match. Therefore, there is no need to write the CAN ID of the accepted CAN Frame into the MnMIDH and MnMIDL registers associated with the Message Object that has been deemed to constitute a match.

[0075] As subsequent CAN Frames of a fragmented message are received, the new data bytes are appended to the end of the previously received and stored data bytes. This process continues until a complete multi-frame message has been received and stored in the appropriate message buffer.

[0076] Under CAL protocols DeviceNet, CANopen, and OSEK, if a Message Object is an enabled Receive Message Object, and its associated MnCTL register has its FRAG bit set to ‘1’ (i.e., automatic fragmented message assembly is enabled for that particular Receive Message Object), then the first data byte (Data Byte 1) of each received CAN Frame that matches that particular Receive Message Object will be used to encode fragmentation information only, and thus, will not be stored in the message buffer for that particular Receive Message Object. Thus, message storage for such “FRAG-enabled” Receive Message Objects will start with the second data byte (Data Byte 2) and proceed in the previously-described manner until a complete multi-frame message has been received and stored in the appropriate message buffer. This message storage format is illustrated in FIG. 11. The message handler hardware will use the fragmentation information contained in Data Byte 1 of each CAN Frame to facilitate this process.

[0077] Under the CAN protocol, if a Message Object is an enabled Receive Message Object, and its associated MnCTL register has its FRAG bit set to ‘1’ (i.e., automatic fragmented message assembly is enabled for that particular Receive Message Object), then the CAN Frames that match that particular Receive Message Object will be stored sequentially in the message buffer for that particular Receive Message Object using the format shown in FIG. 12.

[0078] When writing message data into a message buffer associated with a Message Object n, the DMA engine 38 will generate addresses automatically starting from the base address of that message buffer (as specified in the MnBLR register associated with that Message Object n). Since the size of that message buffer is specified in the MnBSZ register associated with that Message Object n, the DMA engine 38 can determine when it has reached the top location of that message buffer. If the DMA engine 38 determines that it has reached the top location of that message buffer, and that the message being written into that message buffer has not been completely transferred yet, the DMA engine 38 will wrap around by generating addresses starting from the base address of that message buffer again. Some time before this happens, a warning interrupt will be generated so that the user application can take the necessary action to prevent data loss.

[0079] The message handler will keep track of the current address location of the message buffer being written to by the DMA engine 38, and the number of bytes of each CAL message as it is being assembled in the designated message buffer. After an “End of Message” for a CAL message is decoded, the message handler will finish moving the complete CAL message and the Byte Count into the designated message buffer via the DMA engine 38, and then generate an interrupt to the XA CPU Core 22 indicating that a complete message has been received.

[0080] Since Data Byte 1 of each CAN Frame contains the fragmentation information, it will never be stored in the designated message buffer for that CAN Frame. Thus, up to seven data bytes of each CAN Frame will be stored. After the entire message has been stored, the designated message buffer will contain all of the actual informational data bytes received (exclusive of fragmentation information bytes) plus the Byte Count at location 00 which will contain the total number of informational data bytes stored.

[0081] It is noted that there are several specific user set-up/programming procedures that must be followed when invoking automatic hardware assembly of fragmented OSEK and CANopen messages. These and other particulars can be found in the XA-C3 CAN Transport Layer Controller User Manual that is part of the parent Provisional Application Serial No. 60/154,022, the disclosure of which has been fully incorporated herein for all purposes.

[0082] In order to transmit a message, the XA application program must first assemble the complete message and store it in the designated message buffer for the appropriate Transmit Message Object n. The message header (CAN ID and Frame Information) must be written into the MnMIDH, MnMIDL, and MnMSKH registers associated with that Transmit Message Object n. After these steps are completed, the XA application is ready to transmit the message. To initiate a transmission, the object enable bit (OBJ_EN bit) of the MnCTL register associated with that Transmit Message Object n must be set, except when transmitting an Auto-Acknowledge Frame in CANopen. This will allow this ready-to-transmit message to participate in the pre-arbitration process. In this connection, if more than one message is ready to be transmitted (i.e., if more than one Transmit Message Object is enabled), a Tx Pre-Arbitration process will be performed to determine which enabled Transmit Message Object will be selected for transmission. There are two Tx Pre-Arbitration policies which the user can choose between by setting or clearing the Pre_Arb bit in the GCTL register.

[0083] After a Tx Message Complete interrupt is generated in response to a determination being made by the message handler that a completed message has been successfully transmitted, the Tx Pre-Arbitration process is “reset”, and begins again. Also, if the “winning” Transmit Message Object subsequently loses arbitration on the CAN bus, the Tx Pre-Arbitration process gets reset and begins again. If there is only one Transmit Message Object whose OBJ_EN bit is set, it will be selected regardless of the Tx Pre-Arbitration policy selected.

[0084] Once an enabled Transmit Message Object has been selected for transmission, the DMA engine 38 will begin retrieving the transmit message data from the message buffer associated with that Transmit Message Object, and will begin transferring the retrieved transmit message data to the CCB 42 for transmission. The same DMA engine and address pointer logic is used for message retrieval of transmit messages as is used for message storage of receive messages, as described previously. Further, message buffer location and size information is specified in the same way, as described previously. In short, when a transmit message is retrieved, it will be written by the DMA engine 38 to the CCB 42 sequentially. During this process, the DMA engine 38 will keep requesting the bus; when bus access is granted, the DMA engine 38 will sequentially read the transmit message data from the location in the message buffer currently pointed to by the address pointer logic; and, the DMA engine 38 will sequentially write the retrieved transmit message data to the CCB 42. It is noted that when preparing a message for transmission, the user application must not include the CAN ID and Frame Information fields in the transmit message data written into the designated message buffer, since the Transmit (Tx) logic will retrieve this information directly from the appropriate MnMIDH, MnMIDL, and MnMSKH registers.

[0085] The XA-C3 microcontroller 20 does not handle the transmission of fragmented messages in hardware. It is the user's responsibility to write each CAN Frame of a fragmented message to the appropriate message buffer, enable the associated Transmit Message Object for transmission, and wait for a completion before writing the next CAN Frame of that fragmented message to the appropriate message buffer. The user application must therefore transmit multiple CAN Frames one at a time until the whole multi-frame, fragmented transmit message is successfully transmitted. However, by using multiple Transmit Message Objects whose object numbers increase sequentially, and whose CAN IDs have been configured identically, several CAN Frames of a fragmented transmit message can be queued up and enabled, and then transmitted in order.

[0086] To avoid data corruption when transmitting messages, there are three possible approaches:

[0087] 1. If the Tx Message Complete interrupt is enabled for the transmit message, the user application would write the next transmit message to the designated transmit message buffer upon receipt of the Tx Message Complete interrupt. Once the interrupt flag is set, it is known for certain that the pending transmit message has already been transmitted.

[0088] 2. Wait until the OBJ_EN bit of the MnCTL register of the associated Transmit Message Object clears before writing to the associated transmit message buffer. This can be accomplished by polling the OBJ_EN bit of the MnCTL register of the associated Transmit Message Object.

[0089] 3. Clear the OBJ_EN bit of the MnCTL register of the associated Transmit Message Object while that Transmit Message Object is still in Tx Pre-Arbitration.

[0090] In the first two cases above, the pending transmit message will be transmitted completely before the next transmit message gets transmitted. For the third case above, the transmit message will not be transmitted. Instead, a transmit message with new content will enter Tx Pre-Arbitration.

[0091] There is an additional mechanism that prevents corruption of a message that is being transmitted. In particular, if a transmission is ongoing for a Transmit Message Object, the user will be prevented from clearing the OBJ_EN bit in the MnCTL register associated with that particular Transmit Message Object.

[0092] The CAN/CAL module 77 of the XA-C3 microcontroller 20 is presently configured to generate the following five different Event interrupts to the XA CPU Core 22:

[0093] 1. Rx Message Complete

[0094] 2. Tx Message Complete

[0095] 3. Rx Buffer Full

[0096] 4. Message Error

[0097] 5. Frame Error

[0098] For single-frame messages, the “Message Complete” condition occurs at the end of the single frame. For multi-frame (fragmented) messages, the “Message Complete” condition occurs after the last frame is received and stored. Since the XA-C3 microcontroller 20 hardware does not recognize or handle fragmentation for transmit messages, the Tx Message Complete condition will always be generated at the end of each successfully transmitted frame.

[0099] As previously mentioned, there is a control bit associated with each Message Object indicating whether a Message Complete condition should generate an interrupt, or just set a “Message Complete Status Flag” (for polling) without generating an interrupt. This is the INT_EN bit in the MnCTL register associated with each Message Object n.

[0100] There are two 16-bit MMRs 40, MCPLH and MCPLL, which contain the Message Complete Status Flags for all 32 Message Objects. When a Message Complete (Tx or Rx) condition is detected for a particular Message Object, the corresponding bit in the MCPLH or MCPLL register will be set. This will occur regardless of whether the INT_EN bit is set for that particular Message Object (in its associated MnCTL register), or whether Message Complete Status Flags have already been set for any other Message Objects.

[0101] In addition to these 32 Message Complete Status Flags, there is a Tx Message Complete Interrupt Flag and an Rx Message Complete Interrupt Flag, corresponding to bits [1] and [0], respectively, of an MMR 40 designated CANINTFLG, which will generate the actual Event interrupt requests to the XA CPU Core 22. When an End-of-Message condition occurs, at the same moment that the Message Complete Status Flag is set, the appropriate Tx or Rx Message Complete Interrupt flip-flop will be set provided that INT_EN=1 for the associated Message Object, and provided that the interrupt is not already set and pending.

[0102] Further details regarding the generation of interrupts and the associated registers can be found in the XA-C3 Functional Specification and in the XA-C3 CAN Transport Layer Controller User Manual, both of which are part of the parent Provisional Application Serial No. 60/154,022, the disclosure of which has been fully incorporated herein for all purposes.

[0103] As was previously described in detail hereinabove, the XA-C3 microcontroller 20 supports up to 32 separate and independent Message Objects, each of which is set-up or defined by virtue of the user (programmer) configuring (programming) some or all of the eight MMRs 40 dedicated to that Message Object. In the XA-C3 microcontroller 20, each of the 32 Message Objects is assigned its own block of address space in data memory, which serves as its message buffer for data storage. The size and location of each message buffer is programmable, and thus, reconfigurable “on the fly” by the user/programmer. The message buffers can be positioned in any desired location within the overall data memory space addressable by the XA-C3 microcontroller 20, which is presently configured to be a 16 Mbyte overall memory space. These message buffers can be located in the XRAM 28 and/or in any off-chip portion of the overall data memory space.

[0104] The location of the message buffer associated with each Message Object n is established by programming the MMR 40 designated MnBLR associated with that Message Object, i.e., by programming the Message n Buffer Location Register. The size of the message buffer associated with each Message Object is established by programming the MMR 40 designated MnBSZ associated with that Message Object, i.e., by programming the Message n Buffer SiZe Register. In the XA-C3 microcontroller 20, allowable buffer sizes are 2, 4, 8, 16, 32, 64, 128, or 256 bytes. Users can select the size of each message buffer based on the anticipated length of the incoming message, or they can conserve memory by deliberately specifying smaller buffers at the expense of increased processor intervention to handle more frequent buffer-full conditions. In the XA-C3 microcontroller 20, Direct Memory Access (DMA) (i.e., the DMA engine 38) is used to enable the XA-C3 CAN/CAL module 77 to directly access the 32 message buffers without interrupting the XA-C3 processor (CPU) core 22.

[0105] The XA-C3 CAN/CAL module 77 uses the values programmed into the buffer size registers MnBSZ to reserve the designated number of bytes of storage for each Message Object n. For Receive Message Objects, this field is also used by logic in the XA-C3 CAN/CAL module 77 to calculate the total number of bytes that have actually been stored in the message buffers, and to identify when a buffer-full condition is reached. Each time a byte of data is stored in a message buffer associated with a Message Object n, the XA-C3 CAN/CAL module 77 concurrently accesses the MnBSZ and MnBLR registers associated with that Message Object. Logic incorporated within the XA-C3 CAN/CAL module 77 decodes the buffer size for that Message Object and compares the decoded buffer size to the address pointer to determine current byte count and available space left in that Message Object's message buffer.

[0106] The present implementation of the XA-C3 microcontroller 20 requires that all of the 32 message buffers reside within the same 64 Kbyte memory segment (or “page”). The user may position the message buffers within any of the 256 pages in the overall XA-C3 data memory space (i.e., 256×64 Kbytes=16 Mbytes). Programming the locations of the message buffers is accomplished in two steps.

[0107] The first step is to program the page number in which all of the message buffers reside into the MMR 40 designated as the MBXSR register, which is one of the CCB Registers depicted in FIG. 4. As was previously described, the contents of this register are subsequently used as the eight MSBs of address for all DMA accesses to any of the message buffers. This register also establishes the memory page in which the XRAM 28 resides.

[0108] The second step is to program the base address (16 bits) for each individual message buffer into the MnBLR associated with that message buffer. These 16-bit address values initially specified by the user/programmer constitute the base addresses of the 32 respective message buffers within the 64 Kbyte memory page specified in the MBXSR register for all message buffers. It should be noted that the message buffers can be placed apart from one another, as there is no requirement that the message buffer space be continuous (i.e., that the message buffers reside in physically contiguous locations within the data memory space). Further, it should also be noted that some or all of the message buffers can be placed in off-chip memory, and others in the on-chip XRAM 28. In the XA-C3 microcontroller 20, it is required that each message buffer start at a binary boundary for its size (i.e., the 8 LSBs must be zero for a 256-byte message buffer, the 7 LSBs must be zero for a 128-byte message buffer, etc.).

[0109] DMA access to each of the message buffers is achieved by using the 8 bits stored in the MBXSR register as the 8 MSBs of the address of that message buffer, and the 16 bits stored in the MnBLR register for that message buffer as the 16 LSBs of the address of that message buffer. The base address initially programmed by the user into the MnBLR register for that message buffer is the address of the first (bottom) location of that message buffer. When the first frame of a new receive message arrives, the CAN/CAL module 77 hardware writes a semaphore code into this bottom location before beginning to store actual data bytes, starting at the next location in that message buffer. At the end of the new receive message (or when a buffer-full condition is detected), the CAN/CAL module 77 hardware computes the total number of bytes actually stored in that message buffer, and writes this value into the bottom location of that message buffer. The processor (i.e., the XA CPU Core 22) can then read this value and determine precisely how many additional bytes must be read and processed.

[0110] Each time a new byte of data must be written to (for receive messages) or retrieve from (for transmit messages) a message buffer, the DMA engine 38 reads the MnBLR register for that message buffer in order to retrieve the current address pointer for the associated Message Object. The DMA engine 38 concatenates the 8 MSBs stored in the global Message Buffer Segment Register (i.e., the MBXSR register) and the 16 LSBs stored in the MnBLR register for that message buffer to form a complete 24-bit message buffer address. The DMA engine 38 then passes this address to the Memory Interface (MIF) unit 30, along with a flag indicating that the DMA engine 38 requires access to the memory. As soon as the current set of XA-C3 processor memory accesses are completed, the MIF unit 30 will initiate a memory read or write to the address provided by the DMA engine 38, and then permit the DMA engine 38 to perform the required data transfer to/from the desired message buffer. DMA accesses are typically done two bytes at a time (i.e., as a 16-bit operation). However, 8-bit operations are employed when there is only a single byte to be transferred.

[0111] As soon as the requested DMA operation is completed, the DMA engine 38 increments the 16-bit address value stored in the MnBLR register associated with that message buffer (by one or two, depending upon whether a one byte or two byte access was performed), and writes this value back into the MnBLR register for that message buffer. Thus, the MnBLR registers, along with the associated increment logic within the DMA engine 38, effectively function as a set of 32 binary “counters”. Thus, at any given time, each MnBLR register contains the address which will be used for the next data access to the message buffer associated with the Message Object n. In this manner, the MnBLR register for each message buffer serves as an address-pointer. These address-pointer fields are also readable at any time by the processor under software control.

[0112] The above-described approach to message storage also provides an extremely quick and efficient means of freeing up a message buffer when a message completes or when a message buffer is full. The software can respond to a message-complete interrupt or a buffer-full interrupt by simply repositioning the message-buffer space for that particular Message Object to somewhere else in the message buffer memory space. This is accomplished by performing a single write operation to modify the buffer base-address specified in the appropriate MnBLR register (i.e., “address-pointer”). This is essentially the extent of a very short interrupt handling routine. These interrupts must be handled quickly because the message buffer must be freed-up for subsequent message reception. Interrupt response is particularly critical if many completed messages are stacked up and need to be dealt with at once. Once this buffer repositioning is accomplished, the hardware is immediately ready to receive a new message over that Message Object “channel” (or, the continuation of the current message, in the case of a buffer-full interrupt). The memory space that was previously designated as the message buffer for that Message Object n still contains the previously-received message data, but this space now becomes just part of the long-term data memory space. The message information stored in this long-term data memory space can then be processed by the software at its leisure.

[0113] This same buffer repositioning technique can be employed for Transmit Messages to facilitate fragmentation. Unlike the receive case, the XA-C3 CAN/CAL Module 77 does not automatically assemble fragmented outgoing messages. It is incumbent upon the software to “load” a new message frame each time the previous frame is transmitted. Using the XA-C3 microcontroller 20 message storage scheme, however, the software can construct an entire fragmented message prior to enabling transmission. As each frame is transmitted, the processor (XA CPU Core 22) only needs to reposition the buffer (again, using a single write operation) to point to the location of the next frame. This is much faster than competing devices, which require the processor to move up to 13 bytes of data from memory to a dedicated transmit buffer.

[0114] It will be appreciated that with the above-described message buffer scheme of the present invention, each message buffer can be regarded as a separate FIFO having an independently programmable buffer length, which provides a revolutionary approach to storing sequential messages of varying lengths without any CPU intervention.

[0115] As described hereinabove, because the XA-C3 microcontroller 20 provides the user with the ability to program separate Match ID and Mask fields for each of the 32 independent Message Objects, on an object-by-object basis, the Acceptance Filtering process performed by the XA-C3 microcontroller 20 can be characterized as a “match and mask” technique. The basic objective of this Acceptance Filtering process is to determine whether a Screener ID field of the received CAN Frame (excluding the “don't care” bits masked by the Mask field for each Message Object) matches the Match ID of any enabled one of the 32 Message Objects that has been designated a Receive Message Object. If there is a match between the received CAN Frame and more than one Message Object, then the received CAN Frame will be deemed to have matched the Message Object with the lowest object number (n).

[0116] In accordance with the present invention, Acceptance Filtering is performed as follows by the XA-C3 microcontroller 20:

[0117] (1) A Screener ID field is extracted from the incoming (received) CAN Frame. In this regard, the Screener ID field that is assembled from the incoming bit stream is different for Standard and Extended CAN Frames. In particular, as is illustrated in FIG. 9, the Screener ID field for a Standard CAN Frame is 28 bits, consisting of 11 CAN ID bits extracted from the header of the received CAN Frame+2×8 (16) bits from the first and second data bytes (Data Byte 1 and Data Byte 2) of the received CAN Frame+the IDE bit. Thus, the user is required to set the Msk1 and Msk0 bits in the Mask Field (MnMSKL register) for Standard CAN Frame Message Objects, i.e., to “don't care”. In addition, in many applications based on Standard CAN Frames, either Data Byte 1, Data Byte 2, or both do not participate in Acceptance Filtering. In those applications, the user must also mask out the unused Data Byte(s). The IDE bit is not maskable. As is illustrated in FIG. 10, the Screener ID field for an Extended CAN Frame is 30 bits, consisting of 29 CAN ID bits extracted from the header of the incoming CAN Frame+the IDE bit. Again, the IDE bit is not maskable.

[0118] (2) The assembled Screener ID field of the received CAN Frame is then sequentially compared to the corresponding Match ID values specified in the MnMIDH and MnMIDL registers for all currently enabled Receive Message Objects. Of course, any bits in the Screener ID field that are masked by a particular Message Object are not included in the comparison. That is, if there is a ‘1’ in a bit position of the Mask field specified in the MnMSKH and MnMSKL registers for a particular Message Object, then the corresponding bit position in the Match ID field for that particular Message Object becomes a “don't care”, i.e., always yields a match with the corresponding bit of the Screener ID of the received CAN Frame.

[0119] (3) If the above comparison process yields a match with more than one Message Object, then the received CAN Frame will be deemed to have matched the Message Object having the lowest object number (n).

[0120] As was described hereinabove, each incoming (received) CAN Frame that passes Acceptance Filtering will be automatically stored, via the DMA engine 38, into the message buffer for the Receive Message Object that particular CAN Frame was found to have matched.

[0121] Although the present invention has been described in detail hereinabove in the context of a specific preferred embodiment/implementation, it should be clearly understood that many variations, modifications, and/or alternative embodiments/implementations of the basic inventive concepts taught herein which may appear to those skilled in the pertinent art will still fall within the spirit and scope of the present invention, as defined in the appended claims.