Programmably Selectable Addresses for Expansion Cards for a Motherboard

1. Field of the Invention

The present invention relates to a computer system and more particularly to a computer system which allows option controller cards for various input/output (I/O) devices to be added on the motherboard at minimum cost.

2. Description of the Prior Art

The modern microcomputer must be capable of supporting a wide audience of users from the hobbyist to the corporate power-user. This requires that the computer must support a wide array of peripherals which might include LANs, modems, disk controllers, video controllers, or multimedia controllers. If the general purpose microcomputer is to address this large audience economically, it must provide these functions as options to a minimally configured basic machine.

One solution to this problem is to provide multiple industry standard bus slots such as ISA, EISA, MCA, or NuBUS, which the user can use to install a third party peripheral controller. The addition of these slots requires the system cabinet to be large and bulky. The power supply must also be capable of supplying the worst case power required for each slot, multiplied by the number of slots available.

With the current level of integration available today these peripheral functions can often be provided using a single LSI device along with several small support devices used for address decoding, resource routing, data buffering and interfacing. This amount of circuitry can easily fit on the smallest, industry standard, bus card and still have room to spare. However, regardless of the minimum physical size of the logic required to implement the function, the user must sacrifice one of the, typically few, standard bus slots designed into the computer. Since the manufacturer of the computer had to provide enough space and power to support the worst case (largest) card that could have been installed into the standard slot, the user ends up paying more than was needed for the extra capability that has not been used.

It is an object of the invention to solve various problems in the prior art.

Briefly, the present invention relates to a computer system which allows a plurality of option controller cards to be added to the motherboard at minimum cost and without the overhead of using bus expansion slots. Various support services, such as address decoding, are provided by the motherboard in order to reduce the size of the option controller cards.

These and other objects and advantages of the present invention will become readily apparent upon consideration of the following detailed description and attached drawing, wherein:

FIG. 1 is a block diagram of a computer system in accordance with the present invention;

FIG. 2 is a schematic diagram of the register file logic illustrated in FIG. 1;

FIG. 3 is a block diagram of the programmable select logic illustrated in FIG. 1;

FIG. 4 is a schematic diagram of the address match logic illustrated in FIG. 3;

FIG. 5 is a schematic diagram of the command match logic illustrated in FIG. 3;

FIG. 6 is a schematic diagram of the index match logic illustrated in FIG. 3;

FIG. 7 is a block/schematic diagram of the direct memory access (DMA) switch register and related circuitry for remapping DMA I/O control lines;

FIG. 8A is a schematic diagram of the demultiplexing logic circuitry used for DMA switching;

FIG. 8B is a truth table corresponding to the logic circuitry of FIG. 8A;

FIG. 9A is a schematic diagram of the multiplexing logic circuitry used for DMA switching;

FIG. 9B is a truth table corresponding to the logic circuitry of FIG. 9A;

FIG. 10 is a schematic diagram of XBUS control logic;

FIG. 11A is a block/schematic diagram of the interrupt request (IRQ) switch register and related circuitry for remapping IRQ I/O control lines;

FIG. 11B is a schematic diagram of the mapping of IRQ inputs to the IRQ outputs;

FIG. 12A is a schematic diagram of the demultiplexing logic circuitry used for IRQ switching;

FIG. 12B is a truth table corresponding to the logic circuitry of FIG. 12A;

FIGS. 13A and 13B are programmable DMA and IRQ mapping tables for recording system I/O configurations;

FIG. 14 is a schematic diagram of the secure-interrupt logic;

FIG. 15 is a schematic diagram of the secure-IOSELECT logic;

FIG. 16 is a schematic diagram of the secure data write logic; and

FIG. 17 is a timing diagram for the XBUS control logic.

A computer system is illustrated in FIG. 1 which includes a system controller 1, a programmable device 2, a system data and address bus 3, an option data bus (XDATA), and an input/output (I/O) option or expansion card 13 for supporting a plurality of I/O devices. Two I/O devices, identified with the reference numerals 11 and 12, are illustrated for simplicity. The I/O expansion card 13 is adapted to be plugged into an edge connector (not shown) on the motherboard; thus conserving AT bus expansion slots. The programmable device 2 provides for the necessary selection, decoding, and resource routing functions to support the I/O devices 11 and 12.

As shown, the expansion card 13 is connected to the option data (XDATA) bus, which, in turn, is connected to the system bus 3 by way of a data transceiver 5. However, it is also contemplated that the expansion card 13 can be directly connected to the system data bus 3. Addressing of the I/O devices 11 and 12 is provided by the programmable device 2 as will be discussed in more detail below.

Various architectures for the system bus 3 are contemplated. For example, the system bus 3 can be an industry standard architecture (ISA) bus, an extended industry standard architecture (EISA) bus, a microchannel architecture (MCA) bus; a video local (VLBUS) bus, or a peripheral component interconnect (PCIBUS) bus.

The option data (XDATA) bus is a multiple bit parallel bus, connected to the system bus 3 by way of the data transceiver 5. The data transceiver 5 allows for data transfers between the system bus 3 and the I/O devices 11 and 12, and between the system data bus 3 and the programmable device 2.

The data transceiver 5 includes read tri-state devices 15 and write tri-state devices 14. These tri-state devices 15 and 14 are enabled by a read signal XREAD.sub.-- EN# and write signal XWRITE.sub.-- EN# from the programmable device 2, as will be discussed in more detail below.

The system controller 1, under the direction of the central processing unit (not shown, CPU), initiates commands on the system bus 3 to transfer data to and from all system peripherals including the I/O devices 11 and 12. These commands include commands to transfer data from I/O devices 11, 12 to the system controller 1 and ultimately to the CPU, as well as WRITE commands to transfer data to the I/O devices 11, 12 from the system controller 1 and the CPU. The system controller 1 also asserts addressing information to the I/O devices 11, 12 to indicate which I/O device 11, 12 is to respond to the command. The addressing information is decoded by the programmable device 2, as will be discussed in more detail below.

Interrupt request lines IRQ allow the I/O devices 11 and 12 to asynchronously notify the system controller 1 when the I/O devices 11 and 12 require service. These interrupt request lines IRQ are prioritized and require coordination between the I/O devices 11 and 12 and the system software to allow the I/O devices 11, 12 interrupt priority level to be changed during installation and to allow the I/O devices to coexist with other similar devices. In known systems, the method of changing the interrupt level is accomplished manually by way of a jumper or shunt; requiring a certain amount of disassembly of the computer system. As will be discussed in more detail below, the interrupt system described herein is programmable.

The system controller 1 also provides to the system bus 3 a number of control signals intended to allow the I/O devices 11 and 12 to request a direct memory access (DMA) by way of DMA request signal DRQ. When a DMA request is received by the system controller 1, it will respond to the I/O devices 11, 12 with a DMA acknowledged signal DACK, assuming that system resources are available. These signals DRQ and DACK are paired together to form a DMA channel. Multiple DMA channels are typically provided in a computer system and are generally allowed to be rerouted from one device to another when the need arises. In known systems, rerouting of DMA channels is similar to remapping of an interrupt request and requires the manual changing of various shunts or jumper. As described below, DMA channel rerouting described herein does not require any disassembly and is accomplished by software commands.

The system controller 1 does not form a part of the present invention and includes DMA and IRQ functions as discussed above, as well as bus control functions. This system controller 1 may be formed from a single chip or multiple chips. For example, the DMA and IRQ functions can be provided on a single chip, such as an Intel model number 82357, while the bus control functions may be provided by another chip such as an Intel model 82358.

The programmable device 2 provides the necessary selection, decoding, and resource rerouting Functions necessary to support various I/O devices 11 and 12 on the option card 13, as well as the motherboard itself. In particular, the programmable device 2 may be used to provide all address decoding functions, IRQ, and DMA mapping function for the entire motherboard.

The programmable device includes register file logic 6, a programmable select logic 7, option bus control logic 8, interrupt mapping logic 9, and DMA request/acknowledge mapping logic 10.

The programmable device 2 may be implemented as an application specific integrated circuit (ASIC). The Verilog source code for the ASIC is set forth and incorporated by reference as disclosed in Appendix 1 of U.S. Pat. No. 5,784,642.

The register file logic 6 is illustrated in FIG. 2. The register file logic 6 includes all of the programmable storage elements used to configure the programmable device 2 along with the decode logic to identify accesses to the programmable device 2. As shown, a plurality of internal registers 104 through 114 are shown. These internal registers 104 through 114 are accessed by the system controller 1 by way of an index port 91 and a data port 92. In particular, the index port and the data port can be set at any values, for example, $E0 and $E4. Thus, anytime the system controller 1 writes the address $E0, the index port decode logic 91 will assert the INDEX.sub.-- WRITE signal when the address matches the index port decode address ($E0) during a WRITE operation. The INDEX.sub.-- WRITE signal is utilized by the XBUS control logic 8 (FIG. 1) to cause a XWRITE.sub.-- EN# signal to be asserted to enable the data transceiver 5 to pass system data through to the XDATA bus. The INDEX.sub.-- WRITE signal also enables an 8-bit index register 90 along line 101. The index register 90 is used to select one of the internal registers 104 through 114 as a target register for programming. Thus, once the signals XWRITE.sub.-- EN# and INDEX.sub.-- WRITE are asserted, a data byte from the system controller 1 is written to the index register and latched as the signals XWRITE.sub.-- EN# ENABLE and INDEX.sub.-- WRITE are deasserted. The output of the index register 90 is applied along line 103 to a register select logic block 93, implemented as a demultiplexer, which selects or enables one of the internal registers 104 through 114.

When the system controller 1 asserts the address of the data port (e.g., $E4) and asserts a, WRITE command signal, the data port decode logic 92 asserts a DATA.sub.-- WRITE signal along line 102. The DATA.sub.-- WRITE signal is used in the XBUS control logic 8 to assert an XWRITE.sub.-- EN# signal and to allow system data to be connected to the XDATA buss. The assertion of the DATA.sub.-- WRITE signal at the register select logic 93 asserts one of the WRITE lines 115 to 125 of the internal registers 104 through 114 which corresponds to the target register, thereby loading the data into that register.

The accesses to the data port may be qualified for various reasons, including security. In particular, in order to prevent unauthorized access of the data port 92, accesses to this port 92 may be qualified with a SECURE signal or other signals. As shown in FIG. 1, the SECURE signal may be from a key switch 4 to prevent unauthorized modification of the programmable device 2.

Each of the registers 104 to 114 in the register file logic 6 can be read back by the system controller 1 (FIG. 1). In particular, all eight bits of each of the registers 104 to 114 are connected to a multiplexer 130 along line 135. When the system controller 1 asserts the data port address (e.g., $E4), and additionally asserts a READ command, the data port decode logic 92 will assert a DATA.sub.-- READ signal. The DATA.sub.-- READ signal along with bit 7 from the index register 90 are used to control a tri-state device 129 to enable the multiplexer 130 output to be applied to the XDATA bus. In particular, the DATA.sub.-- READ signal is applied to a noninverting input of the AND gate 133 while bit 7 of the index register 90 is applied to an inverting input of the AND gate 133 along line 134. The output of the AND gate 133 is used to control a tri-state device 129, which, in turn, when enabled, applies the outputs of the selected register 104 to 114 to the XDATA bus. During this condition, the XREAD.sub.-- EN# signal will be enabled from the XBUS control circuit 8 to allow the XDATA bus to be connected to the system bus 3, which, in turn, allows the output of the registers 104 to 114 to be read back by the system controller 1.

Bit 7 of the latch 90 may be used to defeat the read back function by the system controller 1. In particular, if bit 7 is set, then gate 133 will be disabled, which, in turn, will disable the tri-state device 129 to prevent the data at the output of the registers 104 to 114 from being read back by the system controller 1.

A data decode signal DATA.sub.-- DEC is asserted any time the data port decode logic 92 indicates that the data port was addressed by the system controller 1. This data decode signal, available at the output on an OR gate 132, is under the control of the DATA.sub.-- WRITE and DATA.sub.-- READ signals, applied to the inputs of the OR gate 132. This data decode signal DATA.sub.-- DEC is used in other parts of the control logic, including an index match INDX.sub.-- MTCH logic control circuit 162 on FIG. 3 of the drawings.

As mentioned above, the programmable device 2 is accessed by way of the register file logic 6 which includes the internal registers 104 through 114. The internal registers 104 through 107 are used to control each of the IOSEL0# output signals from the programmable select logic 7, used to select an I/O device 11, 12 on the expansion card 13. As discussed below, the registers 108 through 114 are utilized for DMA and IRQ mapping.

The registers 104 through 107 define the address and other qualifying parameters for each output select signal IOSEL. For simplicity, only a single programmable address output signal IOSEL is described.

The registers 104 through 107 represent the command register CMD.sub.-- REG[7:0]; the upper address register UADD.sub.-- REG[7:0]; the lower address register LADD.sub.-- REG[7:0]; and the mask register MASK.sub.-- REG[7:0].

The command register CMD.sub.-- REG[7:0] is used to specify the commands for enabling the I/O select output signal IOSEL and global outputs. The command register CMD.sub.-- REG[7:0] is an 8-bit register. Bit 0, when set, allows the signal I/O 16 to be asserted when the programmable output select signal IOSEL is asserted. The I/O 16 signal is used to indicate to the system bus 3 that the particular I/O device 11, 12 is capable of 16-bit data transfers. Bit 1 is used to qualify the output select signal IOSEL with a PRIVY signal, for example, which may be used to control access to a particular I/O device, such as a hard disk drive (not shown). In particular, as will be discussed in more detail below, when bit 1 is asserted, the programmable output select signal IOSEL will be enabled. Similarly, when bit 1 is disabled, the output select signal IOSEL will be deasserted unless the programmable select logic 7 is in the sticky mode, in which case bit 1 is ignored. Bits 4, 3 are used to qualify the programmable output select signal IOSEL with either an I/O WRITE signal IOW or an I/O READ signal IOR. In particular, when bits 4,3 equal 0,0, neither the I/O READ signal IOR nor the I/O WRITE signal IOW affect the I/O select signal IOSEL. When bits 4,3 equal 0,1, the I/O READ signal IOR is used to qualify or enable the I/O select signal IOSEL. When bits 4,3 equal 1,0, the I/O WRITE signal IOW enables the I/O select signal IOSEL. When bits 4,3 equal 1,1, either the I/O READ signal IOR or the I/O WRITE signal IOW can enable the I/O select signal IOSEL. Bits 5,6 are reserved. Bit 7, identified as INDXD, controls the mode of operation of the programmable select logic 7. In particular, when bit 7 is set, the programmable select logic 7 will be in either the index mode or the sticky mode, depending on whether the sticky bit in the mask register is set.

When the programmable select logic 7 is in an address decode mode of operation, the upper address register UADD.sub.-- REG and the lower address register LADD.sub.-- REG are used to define the upper and lower bytes of the programmed address range for the I/O devices 11 and 12 on the expansion card 13. Both the upper address register UADD.sub.-- REG and the lower address register LADD.sub.-- REG are 8-bit registers.

The mask register MASK.sub.-- REG[7:0] is an 8-bit register which allows a range of addresses to be decoded instead of a single address. Each bit in the mask register MASK.sub.-- REG[7:0] corresponds to a bit in the lower address register LADD.sub.-- REG[7:0]. Bit 0 of the mask register MASK.sub.-- REG[7:0] is used to set the "sticky bit" which-enables a sticky mode of operation, as will be discussed below.

The programmable select logic 7 (FIG. 3) is adapted to operate in an address decode mode, an index decode mode, and a sticky mode. The mode of operation of the programmable select logic 7 is under the control of a multiplexer (MUX) 150 and an OR gate 152. The MUX 150 has two selectable inputs A and B which are under the control of a select input S. Bit 7 of the command register CMD.sub.-- REG is applied to the select inputs of the MUX 150 to control the mode of operation. As mentioned above, when bit 7 of the command register CMD.sub.-- REG[7:0] is set, this indicates that the indexed mode of operation has been selected. If bit 0 in the mask register MASK.sub.-- REG[7:0] is also set, the programmable select logic 7 will operate in the sticky mode of operation. Bit 7 of the command register CMD.sub.-- REG[7:0] is applied to the select input S of the MUX 150. When this select input S is high, the MUX 150 selects input B to enable either the index mode of operation or the sticky mode of operation. When the select input S is low, input A is selected for the address mode of operation.

When the address mode of operation is selected, the address select signal ADD.sub.-- SEL is asserted on a successful address decode. This signal ADD.sub.-- SEL is under the control of an AND gate 154. The AND gate 154 is a two-input AND gate. One input to the AND gate 154 is an address match signal ADD.sub.-- MATCH from the ADDMTCH control logic 156. The other signal is an option match signal OPT.sub.-- MTCH from the command match logic CMD.sub.-- MTCH 158. As will be discussed in more detail below, the ADD.sub.-- MTCH signal is asserted when an address matches the programmed address. This address match signal ADD.sub.-- MTCH is qualified with the option match signal OPT.sub.-- MTCH from the command match logic CMD.sub.-- MTCH 158 such that the I/O select output signal IOSEL is only asserted for various programmed commands, such as SECURE, WRITE, READ, or other qualifiers, such as a PRIVY signal, which, in turn, enable the AND gate 154 which, in turn, is applied to the A input of the MUX 150 to provide the I/O select output signal.

If the programmable select logic 7 is in an index mode of operation, bit 7 of the command register CMD.sub.-- REG[7:0] will select input B of the MUX 150 to disable the address mode of operation. Input B of the MUX 150 is used for both the index mode of operation and the sticky mode of operation. Whether the index mode of operation or the sticky mode of operation is selected is under the control of the OR gate 152, whose output is applied to the B input of the MUX 150. In both an index and sticky mode of operation, the B input of the MUX 150 is selected. When the "sticky" bit is set, the particular IOSEL is selected until the "sticky" bit is deasserted.

In an index mode of operation, the IOSEL signal is under the control of the AND gate 160. In a sticky mode of operation, the "sticky" bit (e.g., bit 7 of the command register CMD.sub.-- REG[7:0]) masks the index select signal INDEX.sub.-- SEL such that the IOSEL signal is asserted as long as bit 7 of the command register CMD.sub.-- REG[7:0] is asserted. The index select signal INDEX.sub.-- SEL is under the control of an AND gate 160. One input to the AND gate 160 is the option match signal OPT.sub.-- MTCH from the command match logic CMD.sub.-- MTCH 158. The other input to the AND gate 160 is an index match signal INDEX.sub.-- MTCH from the index control logic INDX.sub.-- MTCH 162. As will be discussed in more detail below, when the index matches the programmed index, the index match INDEX.sub.-- MTCH signal will be asserted to enable the AND gate 160.

The programmable device 2 is thus adapted to decode either physical or indexed addresses. A physical address defines an actual I/O address within the physical address space. An indexed address refers to a method where the address is generated using a base address and an offset. The index values are available at the index register 90 and decoded as discussed below. The indexed decodes allow I/O devices to avoid using any physical address space.

When an address mode is selected, the address can be fully decoded using 16 bits of address or the lower eight address bits can be selectively masked off to enable up to 256 contiguous addresses to match the decode. This allows for global address ranges for option boards if necessary.

As discussed above, the programmable select logic 7 system is also able to provide an ISA bus slave signal I/O 16 which indicates to the system bus 3 that the I/O devices 11, 12 on the expansion card 13 is capable of 16-bit data transfers. This signal I/O 16 is available at the output of an AND gate 164. One input to the AND gate 164 is the address select signal ADD.sub.-- SEL. Bit 0 of the command register is applied to the other input. This bit CMD.sub.-- REG[0] is used to enable or disable the I/O 16 output.

The address match ADD.sub.-- MTCH logic 156 is illustrated in FIG. 4. This logic 156 includes 16 exclusive NOR gates 166 through 196, 8 OR gates 198 to 212, 2 NAND gates 214 and 216, and one NOR gate 218. The upper byte of the system address ADDRESS[15:8] is applied to one input of the exclusive NOR gates 166 through 180. The eight bits of upper address register UADD.sub.-- REG[7:0] are applied to the other inputs. The outputs of the exclusive NOR gates 166 to 180 are applied as inputs to NAND gate 214 whose output is applied to NOR gate 218. The output of the NOR gate 218 is the address match signal ADD.sub.-- MTCH.

The lower byte of the system address ADDRESS[7:0] is applied to one input of the exclusive NOR gates 182 to 196. Bits 0 to 7 of the lower address register LADD.sub.-- REG are applied to the other inputs. The outputs of the exclusive NOR gates 182 to 196 are applied to one input of the OR gates 198 through 212. The MSK.sub.-- REG bits 0 to 7 are applied to the other inputs of the OR gates 198 to 212. The outputs of the OR gates 198 to 212 are applied to the NAND gate 216 whose output is applied to the NOR gate 218.

When the address from the system address bus matches the programmed address in the upper address register UADD.sub.-- REG and the lower address register LADD.sub.-- REG, the outputs of the exclusive NOR gates 166 through 196 will be high. The high output from the exclusive NOR gates 166 to 180 will cause the output of the NAND gate 214 to be low. When the address match logic ADD.sub.-- MTCH 156 is decoding a single address, the bits 0 through 7 of the mask register MASK.sub.-- REG will be deasserted. In this situation, the outputs of the exclusive NOR gates 182 to 196 are used to enable the NOR gates 198 to 212 to cause the output of the NAND gate 216 to be low, which in turn enables the address match signal ADD.sub.-- MTCH.

The mask register MASK.sub.-- REG[7:0] enables a range of addresses to be decoded. In particular, the mask register MASK.sub.-- REG[7:0] controls the outputs of the OR gates 198 through 212. When these bits are enabled, they will force the corresponding outputs of the OR gates 198 to 212 to be high, which, in turn, results in the corresponding address bits being ignored. A summation of all of the mask register bits MASK.sub.-- REG[7:0] being asserted results in the NAND gate 216 being deasserted. When the mask register bits MASK.sub.-- REG[7:0] are low, the lower byte address bits are decoded in the same manner as the upper byte address bits.

The command match CMD.sub.-- MTCH logic 158 is illustrated in FIG. 5 and includes the NAND gates 216 through 228, the AND gate 232, the inverter 230, and the OR gate 231. In an address decode mode of operation and an index decode mode of operation, the command match CMD.sub.-- MTCH logic 158 is used to enable the output signals address select ADD.sub.-- SEL and index select INDX.sub.-- SEL (FIG. 3). As discussed above, additional qualifiers can be added, such that the I/O devices 11 and 12 will only be selected on the expansion card 13 when the system controller 1 matches both the preprogrammed address and the particular command programmed into the command register CMD.sub.-- REG[7:0] for a particular I/O device 11, 12. In particular, bits 0, 2 through 4 from the command register CMD.sub.-- REG[7:0] are applied to the inputs of the NAND gates 216 through 228. Bits 3 and 4 are also applied to the OR gate 231. The WRITE, READ, and other qualifier command signals, such as QUAL.sub.-- A, are applied as inputs of the NAND gates 218, 220, 222, and 228. The SECURE signal is applied to the input of the NAND gate 218 by way of the inverter 230. The outputs of the NAND gates 218 and 228 are applied as inputs to the NAND gates 216 and 226, respectively, along with the bits 2 and 0 of the command register CMD.sub.-- REG. The outputs of the NAND gates 220, 222 and OR gate 231 are applied as inputs to the NAND gate 224. The outputs of the NAND gates 216, 224, and 226 are applied as inputs to the AND gate 232. The output of the AND gate 232 is the option match signal OPT.sub.-- MTCH.

In operation, when a command such as READ or WRITE matches the preprogrammed command qualifier in the command register CMD.sub.-- REG[7:0], the NAND gate 232 will be enabled, which, in turn, will enable the AND gates 154 and 160 (FIG. 3) to enable the I/O select signals IOSEL during an address and index mode of operation. During a sticky bit mode of operation, the command match CMD.sub.-- MTCH logic 158 is ignored.

The index match INDX.sub.-- MTCH logic 162 is illustrated in FIG. 6 and includes the inverters 234 and 236 and the AND gates 238 to 240. The index match logic INDX.sub.-- MTCH 162 is used to enable the AND gate 160 (FIG. 3), which, in turn, is used in an index mode of operation. The value from the index register 90 (FIG. 2) is applied to the inputs of the AND gate 238 with bits 1 and 7 of the index register 90 being inverted by inverters 234 and 236. The output of the AND gate 238 is applied to one input of the AND gate 240 along with the data decode signal DATA.sub.-- DEC. The AND gate 238 decodes the index value written to the index register 90 (FIG. 2).

The signal DATA.sub.-- DEC (FIG. 2) is available at the output of the OR gate 132 and is enabled whenever system controller 1 initiates a read or write access to the port. When both the output of the AND gate 238 is asserted along with the data decode signal DATA.sub.-- DEC, the index match signal INDX.sub.-- MTCH goes high.

To provide flexibility and alleviate the burdens of knowing the details of microcomputer system resources and hardware locations, the embodiment uses a hardware method to provide a system having means for remapping input I/O lines to any one or more output I/O lines under software control. The embodiment includes mapping interrupt requests (IRQs) and direct memory access (DMA) I/O channel control lines to multiple output destinations. Since in many cases the complexity of a microcomputer configuration is directly related to the multiplexing of the interrupt and DMA I/O channel control lines, this complexity may be reduced considerably by eliminating the need for physical jumpers manually placed by the user when configuring the system for use with devices which use the interrupt and DMA I/O channels.

The use of jumpers is burdensome and often confusing to users. In fact, improper jumper configuration often creates system conflicts which, through the eyes of the novice user, often looks like software incompatibility issues when in fact the problem actually only lies in the improper hardware jumper configuration. Microcomputer systems have typically been built with a confusing matrix of these hardware jumpers and complex related documentation which is also confusing. Different motherboards and add-on cards have added to the confusion by placing these hardware jumpers in different locations depending on the particular product. Moreover, if a resource limit is reached because too many different jumper options have been used up, then some add-on boards or system options may not be available to the user. All this has historically left users confused with how interrupts and DMA I/O control lines should be configured. Accordingly, the elimination of physical jumpers as provided by the embodiment considerably reduces the complexity and user confusion associated with interrupt and DMA I/O control line configuration.

The embodiment provides programmable DMA switching of six (6) different pairs of DMA control lines to be mapped to three (3) different sources. The embodiment also provides for the programmable switching of six (6) IRQ IN lines to any of six (6) IRQ OUT lines. Of course, the mapping of three (3) DMA input pairs to six (6) output pairs and the mapping of six (6) IRQ input lines to six (6) IRQ output lines is only an arbitrary choice for the embodiment; any number of input and output I/O control lines could be switched in accordance with the invention (2 channels, e.g., 0,1, are shown in FIG. 1 for the interrupt and DMA control lines). As will be described in more detail below, switching is provided by the programmable device 2 through a series of registers which control multiplexers (MUXs) and demultiplexers (DEMUXs) which carry out the switching of the control lines. The registers reside in the register file 6 as described above. The programmable device 2 is disposed between the system controller 1 via the system bus 3 and the option card 13 to facilitate I/O communications with option card 13 control and support I/O devices 11, 12 over interrupt and DMA I/O channels. DMA request/acknowledging mapping logic 10 (FIG. 1) utilizes MUXs and DEMUXs as well as associated glue logic which is described below. The interrupt mapping logic 9 uses DEMUXs and associated glue logic described below.

In order to make the system as integrated as possible, the programmable device 2 controls all system I/O functions, controls the I/O data paths, and routes signals critical to the DMA and IRQ paths. The mapping control logic is normally programmed by the system firmware during the power-up sequence to industry standard location which may be reprogrammed later under software control. Referring again to FIG. 1, it can be seen that the system address bus and the XDATA bus are both connected to the register file 6 within the programmable device 2. The register file 6 contains numerous READ and WRITE 8-bit registers, six of which are of interest in the present discussion, namely, the three DMA registers and the three IRQ registers. All of these registers are accessed by an indexing mechanism which is controlled through two address locations within the system's real I/O address space, as described above.

FIG. 2 shows the register file wherein the registers for DMA switching of DMA channel 1 and channel 0, channel 3 and channel 2, and channel 5 and channel 4, DMASW10.sub.-- REG, DMASW32.sub.-- REG, and DMASW54.sub.-- REG, respectively, are programmable registers controlling DMA request acknowledging mapping logic 10. The location of the registers in the embodiment are at index x045, x046, and x047 ports respectively. The registers for interrupt request switching of interrupt channel 1 and channel 0, channel 3 and channel 2, and channel 5 and channel 4, IRQSW10.sub.-- REG, IRQSW32.sub.-- REG, and IRQSW54.sub.-- REG, respectively, are the registers controlling the interrupt mapping logic 9 (FIG. 1) located at index x041, x042, and x043 ports respectively, in the embodiment. The interrupt and DMA I/O channel control registers are written to under program control by indexing in data transfer to the programmable device 2, as described above.

Each of the six interrupt and DMA control registers are 8-bit latches wherein each byte comprises a low nibble and a high nibble, and each nibble controls an I/O channel (DMA control line pair or interrupt control line). The DMA and interrupt control lines are numbered from 0 to 5, wherein each number represents a DMA control line pair comprising a DMA request signal line (DRQ) and a DMA acknowledge signal line (DACK) in the case of a DMA channel, and represents an interrupt request (IRQ) line in the case of an interrupt channel.

FIG. 7 represents the first DMA mapping switch register from the register file 6. As can be seen from FIG. 7, the DMA mapping switch registers control two sets of DMA channel control line pairs each. The DMA switch register shown in FIG. 7 is DMASW10.sub.-- REG[7:0] which controls DMA input channel 0 and DMA input channel 1. The control circuitry shown in FIG. 7 routes either DMA input channel to any one of DMA output channels 0,1, or 2.

The following TABLE 1 represents the programming of the data byte represented by the DMA switch registers.

As can be seen in FIG. 7, the data byte contained in the DMA switch register shown in TABLE 1 is used to configure the multiplexer and demultiplexer circuitry of FIG. 7 as well as the XBUS control circuitry, which includes DATA.sub.-- READ from the OR gate 314 output to DATA.sub.-- READ 818 (FIG. 10). Only the first register (index x045) is illustrated in FIG. 7, however, the other registers are configured similarly for DMA I/O channels 2, 3 and 4, 5.

The DMA switch register for channels 0,1 according to TABLE 1 and FIG. 7 programmed at offset x045 in register file 6 will control two of the six DMA I/O control line pairs (DACK.sub.-- IN0/DRQ.sub.-- OUT0 and DACK.sub.-- IN1/DRQ.sub.-- OUT1). The first two register bits 0,1 are for device select channel 0, DEV.sub.-- SEL0 which controls the demultiplexing at DEMUX 301 of DACK.sub.-- IN0 to DACK.sub.-- OUT0, DACK.sub.-- OUT1, or DACK.sub.-- OUT2. The first two register bits 0,1, DEV.sub.-- SEL0 also control the demultiplexing at MUX 302 of DRQ.sub.-- IN0, DRQ.sub.-- IN1, or DRQ.sub.-- IN2 to DRQ.sub.-- OUT0. DEMUX 301 and MUX 302 are controlled via control lines S0 and S1 from DEV.sub.-- SEL0 in order to multiplex and demultiplex the DMA I/O control line signals. Also, the output of DEMUX 301 and MUX 302 are tri-stated by the control lines S0 and S1 by AND gate 303 by controlling tri-state buffers 304, 305, 306, and 307 as shown in FIG. 7. The DMA switching DEMUX 301 and MUX 302 plus glue logic is represented generally with the dashed line Box 310. Similarly, the upper nibble of the register of TABLE 1 and FIG. 7 controls a DEMUX and MUX with a Box 312 similar to Box 310 to route DMA I/O control line signal paths for DMA channel 1.

The bits of the DMA switch register DMASW10.sub.-- REG[7:0] XBUS.sub.-- CNTL.sub.-- 0 and XBUS.sub.-- CNTL.sub.-- 1 each respectively control the read enables for DMA channel 0 and DMA channel 1 access to the XBUS. An AND gate 308 "ANDS" the XBUS control bit 2, CNTL.sub.-- 0 with a DMA acknowledge input DACK.sub.-- INO to create a read enable signal READ0, which in turn is "ORED" at with a read enable signal READ1 by way of an OR gate 314 from DMA I/O channel 1 to apply a DATA.sub.-- READ signal to a NOR gate 818 (FIG. 10) along line 820 which is described above in conjunction with XBUS control 8.

With DMA switching, in order to allow for six different lines to be mapped to three different sources with the above three registers on the programmable device 2, six DMA acknowledge IN lines are demultiplexed to three DMA acknowledge OUT lines (e.g., DACKIN0 . . . DACKIN5 to DACK.sub.-- OUT0 . . . DACK.sub.-- OUT2) as will become apparent in the discussion that follows.

DMA services provided on the motherboard and on add-on cards may access system resources by becoming a bus user directly or via the DMA controller. The purpose of DMA services is to provide for the transfer of data between I/O and memory using the DMA signals as a request to the DMA controller to obtain the bus and execute the transfer cycles. As one skilled in the art appreciates, each DMA channel has two pairs of signal lines which have already been introduced (1) "DRQ" for DMA request and (2) "DACK" for DMA acknowledge. The "DRQ and DACK" are the "pair" of DMA control signal lines which are switched in the embodiment under program control and without the use of physical jumpers. The DMA request, DRQ signal is driven by the I/O resources to request DMA service from the DMA controller. The DRQ signal will remain active until the controller responds with the appropriate DMA acknowledge, DACK signal.

The programmable DMA channel configuration provided by the embodiment meets the configuration and software requirements for each request/acknowledge pair (DRQ/DACK), providing a very dynamic controllable technique to select and direct where DMA I/O control lines are routed. In order to provide a programmable solution, the embodiment employs the programmable device 2 which provides the required configurable DMA control line pairs for the entire system. Through software control the system interrupt and DMA resources may be mapped or disabled to allow custom configuration of the system. The DRQ.sub.-- IN0-DRQ.sub.-- IN2 input pins may be mapped to any one or more of the DRQ.sub.-- OUT0-DRQ.sub.-- OUT5 output pins. The DRQ.sub.-- OUT pins are tri-stated when not being used by a DRQIN, allowing for an external device on the system side to drive these lines. Since the DMA request works in pairs with the DMA acknowledge lines, the same DMA mapping circuitry also controls the input pin DACK.sub.-- IN0-DACK.sub.-- IN5 to DACK.sub.-- OUT0-DACK.sub.-- OUT2. When a DACK.sub.-- INX from the system side is asserted, the XBUS control block may be used for the cycle, depending on how the DMA switch register DMASWXX.sub.-- REG is configured. The XBUS may be used even though a DACKIN# signal is not being used by any of the device channels.

FIG. 8A represents exemplary logic circuitry for implementing the DMA mapping DEMUX 301, and FIG. 8B shows a truth table relating where the IN signal will be routed according to S0 and S1 to OUT.sub.-- 0, OUT.sub.-- 1, OUT.sub.-- 2, or OUT.sub.-- 3. The circuitry represented by FIG. 8A and FIG. 8B shows how the control lines S0 and S1 affect the DEMUX 301 circuitry. The circuit of FIG. 8A provides for AND gates 356, 358, 360, and 362, each having three inputs and one output. One of the inputs to each of the AND gates 356, 358, 360, and 362 is an IN signal. Inverted and noninverted control signals S0 and S1 are provided to each of the inputs to the AND gates 356, 358, 360, and 362. The control signals S0 and S1 are applied as inputs to each of the AND gates 356, 358, 360 and 362. These signals S0 and S1 are also inverted by way of inverters 352 and 354, respectively, and likewise applied as inputs to each of the AND gates 356, 358, 360 and 362. Accordingly, when S0 and S1 are low for example, the IN signal will be routed to OUT.sub.-- 0 via AND gate 356. The other ways in which the IN signal is routed through the circuitry with control lines S0 and S1 is represented by the truth table shown in FIG. 8B.

FIG. 9A represents exemplary circuitry for DMA mapping MUX 302, and FIG. 9B represents truth table showing outputs indicating which inputs IN.sub.-- 0 to IN.sub.-- 2 are routed to the output signal OUT. The MUX 302 circuitry shown in FIG. 9A illustrates how control lines S0 and S1 affect AND gates 406, 408, 410, and 412 to determine which of input signals IN.sub.-- 0, IN.sub.-- 1, IN.sub.-- 2, or IN.sub.-- 3 are routed to the output signal OUT of a four input OR gate 414. The multiplexing is accomplished by sending noninverted versions of S1 to the input of AND gates 410 and 412, and sending inverted versions of S1 to AND gates 406 and 408 via inverter 404. Noninverted S0 control signals are sent to AND gates 408 and 412, and inverted S0 signals are sent to AND gates 406 and 410 via inverter 402. The output mapping is illustrated in the truth table of FIG. 9B.

Turning now to the interrupt request IRQ I/O channel control line switching, FIG. 11A, shows an IRQ switch register and related circuitry for remapping IRQ I/O control lines, and FIG. 11B expands the circuitry of FIG. 11A to provide mapping of m interrupt channels. Having described above the DMA I/O channel switching in detail, one skilled in the art will realize that IRQ switching is merely a subset of the DMA switching. The demultiplexing used for the interrupt mapping logic is much the same as that used for DMA request mapping logic. In fact, IRQ switching is much simpler than DMA switching because only one control signal needs to be routed for IRQ mapping, whereas two control signals (request and acknowledge) had to be mapped for DMA mapping. Moreover, since the IRQ signal is only an output signal being presented to the microcomputer by devices interrupting the microcomputer, only DEMUXs for demultiplexing IRQ control line signals are required for switching. Multiplexing of signals returned from the microcomputer are not required in interrupt mapping because no acknowledge signal is returned. The demultiplexer circuitry performed for IRQ mapping is much like the demultiplexing circuitry provided by DEMUX 301 described above in the context of DMA mapping.

The IRQ mapping switch registers from the register file 6 controls two (2) sets of IRQ input lines each, thus three (3) registers are required to control six (6) IRQ input lines. The six registers utilized are IRQSW10.sub.-- REG, IRQSW32.sub.-- REG, and IRQSW54.sub.-- REG located at data offset x041, x042, and x043, respectively. The six IRQ channel inputs controlled by the three registers are 0,1 and 2,3, and 4,5, respectively.

The following TABLE 2 represents the programming byte of the IRQ.sub.-- IN.sub.-- SWITCH register.

The register shown in TABLE 2 indicates the data byte programming of any of the three above-described IRQ.sub.-- IN.sub.-- SWITCH registers which may be used for IRQ input channel mapping in accordance with the embodiment. The register shown in TABLE 2 indicates the data byte programming of any of the three above-described IRQ.sub.-- IN.sub.-- SWITCH registers which may be used for IRQ input channel mapping in accordance with the embodiment. The data byte of TABLE 2 is shown in FIG. 11A as IRQ.sub.-- IN.sub.-- SWITCH.sub.-- 10 register controlling two demultiplexer circuits DEMUX 502 and DEMUX 506 showing control of IRQ.sub.-- OUT.sub.-- 0 via NOR gate 504 and IRQ.sub.-- OUT.sub.-- 1 via NOR gate 508. The IRQ mapping of FIG. 11A may be expanded to an arbitrary number m outputs in accordance with the invention as illustrated in FIG. 11B. <Disable, Select> correspond to bits in switch register nibbles for programming DEMUXs 502, 506, and 512. Additional OR gates 510 and 514 are shown for added IRQ.sub.-- OUT.sub.-- X . . . IRQ.sub.-- OUT.sub.-- m control lines.

As TABLE 2 and FIG. 11A illustrate, the 8-bit byte of an IRQSWXX.sub.-- REG register from register file 6 has an upper and lower nibble, each nibble controlling an IRQ I/O channel. As shown, control signals S0, S1, and DISABLE control the routing of an IRQ input signal to an IRQ output determined by the demultiplexers DEMUX 502 and DEMUX 506, etc. The IRQ mapping is programmed in register file 6 as described earlier and will control two of the six IRQ input lines per register. Multiple inputs may be programmed for a single output allowing the sharing of multiple IRQ input to enable a single IRQ output.

FIG. 12A shows circuitry for implementing DEMUX 502 or DEMUX 506. The demultiplexer logic circuitry in FIG. 12A is used to route the IN signal to one of four OUT signals, OUT.sub.-- 0', OUT.sub.-- 1', OUT.sub.-- 2', or OUT.sub.-- 3'. The demultiplexing circuitry of FIG. 12A is not unlike that of FIG. 8A for the DMA control lines. However, the circuitry of FIG. 12A also includes a disable which is inverted via inverter 564 and presented as an input to each of four AND gates 556, 558, 560, and 562. When disable is high its inverted signal, a low signal, is presented to the AND gates, thus disabling their outputs by making them always low. Select signals S0' and S1' control the demultiplexing of the IN signal through the AND gates 556, 558, 560, and 562. Selection control is provided by presenting the noninverted S0' signal to AND gate 558 and AND gate 562 while presenting an inverted S0' to AND gate 556 and AND gate 560 via inverter 552. The noninverted S1' is presented to the input of AND gate 560 and AND gate 562 while inverted S1' signals are presented to the inputs of AND gate 556 and AND gate 558 via inverter 554. The resulting signal flow of the IN signal through the demultiplexer circuitry of FIG. 12A is represented by the truth table shown in FIG. 12B.

FIGS. 13A and 13B are tables which are useful for programming the above-described hardware for remapping of IRQ and DMA control lines without hardware jumpers. FIG. 13A is used for programming the DMA remapping wherein the six control line pairs (DRQ.sub.-- OUT/DACK.sub.-- IN) for the Bus DMA channel are represented in the first column as 0 through 5. The device DMA channel is recorded in the second column (DRQ.sub.-- IN/DACK.sub.-- OUT). XBUS use is recorded in the third column and indicated as on or off. The fourth column records the DMA switch registers for DMA channels 0,1 or 2,3 or 4,5, which are programmed as discussed above. FIG. 13B is used for remapping of IRQ control lines, wherein system requirements are recorded for the six IRQ channels (0-5) in the first three columns and register values for programming are stored in the last two columns of the IRQ table.

In a computer system with multiple buses, it is advantageous to electrically isolate the buses from each other. In the preferred embodiment, the system bus 3 is isolated from the option bus XDATA by a data transceiver 5. The data transceiver 5 includes a write driver 14 and a read driver 15. The write driver 14 drives data toward the option bus 13 during write operations, and is controlled by a write enable signal XWRITE.sub.-- EN#. The read driver 15 drives data toward the system bus 3 during read operations, and is controlled by a read enable signal XREAD.sub.-- EN#. The write enable signal XWRITE.sub.-- EN# and the read enable signal XREAD.sub.-- EN# are controlled by XBUS control logic 8. The XBUS control logic 8 is shown in greater detail in FIG. 10. It is necessary that write data become valid when, or shortly after, a write signal WRITE# begins and after the write signal WRITE# ends to ensure that devices respond to valid data when receiving the write signal WRITE#. Devices latch write data on the trailing edge of the write signal WRITE#. This makes it critical to maintain valid write data beyond the end of the write signal WRITE#. Therefore, the write enable signal XWRITE.sub.-- EN# must remain active beyond the end of the write signal WRITE#, ensuring the write driver 14 continues to drive the write data beyond the end of the write signal WRITE#.

Referring to FIG. 10, all write signals 802 are provided to the inputs of OR gate 804. The output of OR gate 804 XWRITE is provided to NOR gate 806 and to a group of serially connected inverters 808. The output of the group of serially connected inverters 808 is provided to a group of serially connected inverters 810 and to a MUX 814. The output of the group of serially connected inverters 810 is provided to a group of serially connected inverters 812 and to the MUX 814. The output of the group of serially connected inverters 812 is provided to the MUX 814. Finally, the output of the MUX 814 is provided to a NOR gate 806. The MUX 814 provides the outputs of the groups of serially connected inverters 808, 810, and 812 to the NOR gate 806 in response to control lines 816. Each group of serially connected inverters 808 through 812 delays its output by an incremental period. The duration of the write enable signal XWRITE.sub.-- EN# is controlled by enabling the outputs of the groups of serially connected inverters 808 through 812. An OR of the write signal XWRITE and the output of the group of serially connected inverters 808 will provide a signal of slightly longer duration than the write signal XWRITE itself. An OR of the write signal XWRITE and the output of the group of serially connected inverters 808 and the output of the group of serially connected inverters 810 will provide a signal of longer duration than the combination of the write signal XWRITE and the output of serially connected inverters 808. The combination of the write signal XWRITE, the output of the group of serially connected inverters 808, the output of the group of serially connected inverters 810, and the output of the group of serially connected inverters 812 will provide the longest signal.

The timing diagram FIG. 17A demonstrates the typical timing on the system bus 3 of system data SYSDATA and the write signal WRITE#. The system data SYSDATA must be present and valid shortly after time T1 and be continuously present and valid until after time T2. The write signal WRITE# goes active (low) at time T1 and inactive (high) at time T2. The write signal WRITE# informs a device that valid data is being written to it. The device may accept the data anytime the write signal WRITE# is active and latch write data on the trailing edge of the write signal WRITE#. It is important to note that a device may accept data up to thirty or forty nanoseconds after time T2 (the time the write signal WRITE# goes inactive) due to delays in the device's electrical circuitry.

The timing diagram 17B demonstrates the problems encountered when the write signal WRITE# is used to control the write driver 14 (FIG. 1) without the present invention. The XBUS data XDATA becomes valid shortly after the write enable signal XWRITE.sub.-- EN# becomes active (low). The XBUS data XDATA also becomes invalid when the write enable signal XWRITE.sub.-- EN# becomes inactive (high). If a slow device accepts the XBUS XDATA thirty nanoseconds after the write enable signal XWRITE.sub.-- EN# becomes inactive, the device will accept invalid data.

The timing diagram 17C demonstrates the timing of a system incorporating the present invention. The XBUS data XDATA becomes valid at time T1, or shortly after time T1, when the write enable signal XWRITE.sub.-- EN# becomes active (low). The XBUS data XDATA becomes invalid when the write enable signal XWRITE.sub.-- EN# becomes inactive (high). The write signal XWRITE# informs a device when data is valid. The write enable signal XWRITE.sub.-- EN# is equivalent to write signal XWRITE# except that the time in which it becomes inactive (high) is delayed by the delay circuit of FIG. 10. This ensures that the XBUS data XDATA will remain valid for thirty to forty nanoseconds after the write signal XWRITE# becomes inactive (high).

The duration of the delay between when the write enable signal XWRITE# becomes inactive (high) and the write enable signal XWRITE.sub.-- EN# becomes inactive (high) is determined by the number of inverter gates employed in the delay circuit of FIG. 10. Differing considerations in the environment where the XBUS control 8 is used, determines the delay necessary. By way of example, increased temperature tends to reduce the amount of delay required. A temperature sensor can be used to control the delay setting selected by the control signals 816. Some devices respond more quickly than others to write signals. A longer delay may be needed in addressing slower devices than is needed in addressing faster devices. Also, some computer systems respond more quickly than others. Programmability of the delay device allows the delay device to be utilized in a wider range of computer systems. Some ASICS respond more quickly than others. Even within a given manufacturing lot, ASICS will vary in terms of the response time. An increased delay time can be used with faster responding ASICS. The preceding reasons for changing the program delay of the delay circuit of FIG. 10 are provided by way of illustration. Numerous other factors in a complex computer environment will affect the optimum delay of the write enable signal.

The I/O device 11 on option bus includes a write only port. The write only port includes signals such as a signal to enable flash ROM programming, a reset of the numeric coprocessor, a change of the speed mode of the processor, and changes to the memory map. It is often advantageous for the BIOS to be able to read the state of these signals, but they cannot be read directly from the write only port in the I/O device 11. The signals on the port in the I/O device 11 can only be changed when the system is in the index mode. The write only port on the I/O device 11 is a transparent write only latch. A typical example of a transparent write only latch is the 74X373 series from Texas Instruments. The present invention provides a shadow register, or a register that contains the same data as the write only port on the I/O device 11 while the system is in the index mode. The register 105 in FIG. 2 is normally used as the lower address decode register. However, the lower address decode register is not required when the system is in the index mode. Since the register 105 is not needed as a lower address decode register in the index mode, it is advantageous to use the register 105 as a shadow register for the write only port in the I/O device 11. This avoids the necessity of additional hardware. Further, access to the register 105 is controlled by the index register and by the SECURE signal. This allows only the BIOS, or those programs authorized by the BIOS, to either read from or write to the register 105. When data is to be written to the write only port on the I/O device 11, the data, the address of the write only port on the I/O device 11, and the write signal WRITE# are asserted on the system bus 3 by system controller 1. The data is provided via the system bus 3 to the data transceiver 5, and from the data transceiver 5 to the register file 6 and to the option bus. The data is then presented to the programmable select logic 7, and thereby, to registers 104 through 114. The combination of the write signal WRITE#, the address of the register 105, and the SECURE signal to the data port decode logic 92, signals the register select logic 93 to allow data to be written to the register 105.

Specifically, once the programmable select logic 7 is programmed into the indexed write mode, and the register select decode logic 92 contains the address of the register 105, the system controller 1 generates the address of the write only port on the I/O device 11 which is applied to the register file 6 and the programmable select logic 7 by the system bus 3. The data simultaneously generated by the system controller 1 is applied to the register file 6 through the system bus 3, the data transceiver 5, and the option bus. Simultaneously, the system controller 1 also generates the write signal WRITE# which is applied to the programmable select logic 7. The data port decode logic 92 and the register select logic 93 decode the address as the write signal WRITE# goes active (low). The write signal WRITE# in turn drives the the I/O select signal IOSELO# low. The active I/O select signal IOSELO# puts the write only port on the I/O device 11 into transparent mode to read data. When the system controller 1 deactivates the write signal WRITE# (high), the programmable select logic 7 drives the I/O select signal IOSELO# inactive (high) as well. The data on the XDATA bus will be latched into the write only port on the I/O device 11. Simultaneously, the data is written to the register 105. The data port decode logic 92 and the register select logic 93 address lower address decode register 105 by asserting the target register signal 136. The target register signal 136 prepares the register 105 to latch the XDATA. When the system controller 1 deasserts the write signal WRITE#, the target register signal 136 is deasserted and the register 105 latches the data. Thus, the data latched in the write only port on the I/O device 11 is identical to the data latched in register 105.

Therefore, in response to the write signal WRITE#, the data will be written to both register 105 and the write only port on the I/O device 11. When data is to be read back from the write only port on the I/O device 11, the address of the register 105 is written into the register select logic 93 to select the register 105 as the target of the next I/O cycle. Then a read cycle to the data port is initiated which asserts IOSEL0 to the I/O device 11 and the read signal READ# will be asserted on the system bus 3 by the system controller 1. The write only port on the I/O device 11 will not respond since it is a write only port and the device attempting to read from the write only port on the I/O device 11 will receive a signal of all ones. This is the same signal as would be received if no I/O device 11 exists.

When data is to be read from register 105, the system controller 1 asserts the read signal READ# active (low). Since the I/O select signal IOSELO# is a write only signal, it will remain inactive (high). The system controller 1 asserts an address signal to the data port decode logic 92 via the system bus 3. The data port decode logic 92 in response to the read signal READ#, and the SECURE signal, asserts a DATA.sub.-- READ signal 131 active. The index register 90, having been previously programmed to point to the register 105, asserts a signal to the MUX 130 to select the output of the register 105. The DATA.sub.-- READ signal 131 is also applied to the AND gate 133 along with the value in bit 7 of the index register 90. The output of the AND gate 133 enables a tri-state device 129 to receive data from the MUX 130.

However, if the read signal and the address of the write only port on the I/O device 11 are asserted with the SECURE signal, the data port decode logic 92 and the register select logic 93 will cause the MUX 130 to not present the data in the register 105.

This provides maximum security for the write only port on the I/O device 11 since a device not authorized by the BIOS to read the register 105 cannot even determine if the I/O device 11 exists. The invention is most useful where the values in a write only port are critical to system function. In many cases, a read of a register will cause a momentary fluctuation in the value in that register. If user applications are allowed to read register containing critical system control signals, fluctuations in the output signals could be modified at times which could cause damage to the system. In the present invention, not even reads by the BIOS will cause such fluctuations in a protected write only port since reads are applied to the register 105 rather than the write only port.

Security from data loss to or data modification by peripheral devices is provided by control of the peripheral device interrupts. By reading peripheral devices physically connected to the system and blocking the interrupts, the system can access the peripheral devices but not allow the peripheral devices to take control of or interrupt the system. The secure switch 4 in FIG. 1 is shown as a simple key switch. It should be clear to one of skill in the art that the security switch can be provided for by numerous devices capable of providing security such as a key switch, a magnetic card reader, a personal identification number, an encrypted software key, hand scanner, a voice recognition device, other devices which identify the authority of the user asserting the secure function, on any combination of the above.

The output of the secure switch 4 is applied to the data port decode logic 92. The secure signal must be active before any output on the lines 102 or 131 from the data port decode logic 92 is possible. This prevents any of the registers 104 through 114 in the register file 6 from being read from or written to unless the SECURE signal is deasserted.

Register 114 contains a bit mask identifying interrupts that should be blocked with the SECURE signal. The circuit of FIG. 14 is replicated eight times, one for each bit in the register 114. An interrupt request 652 is the output of an AND gate 654. The AND gate 654 receives input from an interrupt request 656 and an OR gate 658. The OR gate 658 receives its input from the secure switch 4 via an inverter 660 and the relevant bit from the interrupt secure register 114 via an inverter 662.

Thus, output signal OUT 652 is a function of (1) a request to activate that interrupt signal was made, and (2) that the SECURE signal is either inactive, or the input channel control was not programmed to disable the channel when the SECURE signal is active. The SECURE interaction is as fundamental to the assembly as the actual external request signal to generate the interrupt.

Referring to FIG. 15, the select match signal SEL.sub.-- MTCH is the output of an AND gate 704. The AND gate 704 receives input from OR gates 706, 708, and 710. The OR gate 706 receives its input from an AND gate 712 and an AND gate 714. The AND gate 712 receives its input from the add match circuit 106 and the index bit, bit 7, of the command register 107. The AND gate 714 receives its input from the index bit, bit 7, of the register 107 and the index match circuit 112. The OR gate 708 receives its input from the secure switch 4 and the secure bit, bit 2, of the command register 107. The input to the OR gate 708 is NOT secure bit 2 or NOT secure switch 4 AND secure bit 2. The OR gate 710 receives its external input from the PRIVY bit, bit 1 of command register 107 and the privy signal PRIVY. The input of the OR gate 710 is not PRIVY and bit 1 in command register 107, or PRIVY bit 1 in command register 107 and the external PRIVY signal. Thus, the select match signal SEL.sub.-- MTCH is a function of (1) one I/O channel selection indicating either direct or indirect addressing modes, (2) the SECURE signal active or the secure select control is not programmed to disable the channel when the SECURE signal is active, and (3) that the PRIVY signal is either active, or the select control is not programmed to disable the channel when PRIVY is active. Finally, the output select match signal SEL.sub.-- MTCH and I/O select IOSEL, the output of MUX 150, are applied to an AND gate 716 to produce an I/O select signal IOSEL# to the option bus XADATA. Hence, no I/O selection on the option bus XADATA is possible unless the select match SEL.sub.-- MTCH is active.

Referring to FIG. 16, the data write signal DATA.sub.-- WRITE supplied to XBUS control logic 8 is a function of the secure switch 4, the data decode signal DATA.sub.-- DEC, and the I/O write signal IOW. The data write signal DATA.sub.-- WRITE supplied to XBUS control 8 from the output of the AND gate 752. The inputs to the AND gate 752 include the output of inverter 754 (the NOT SECURE signal, data decode signal DATA.sub.-- DEC and I/O write signal IOW). Thus, the data write signal DATA.sub.-- WRITE supplied to XBUS control logic is a function of (1) I/O write signal IOW, (2) the address of the programming data register (DATA.sub.-- DEC), and (3) the SECURE signal being inactive.

Obviously, many modifications and variations of the present invention are possible in light of the above teachings. Thus, it is to be understood that, within the scope of the appended claims, the invention may be practiced otherwise than as specifically described above.