Digital data processing methods and apparatus for fault detection and fault tolerance

Abstract

A digital data processing device includes a bus for transmitting signals (e.g., data and/or address information) between plural functional units (e.g., a central processing unit and a peripheral controller). A first such unit includes first and second processing sections that concurrently apply to the bus complementary portions of like information signals (e.g., longwords containing data). A fault detection element reads the resultant signal from the bus and compares it with at least portions of the corresponding signals originally generated by the processing sections themselves. If there is discrepancy, the fault-detector signals a fault, e.g., causing the unit to be taken off-line. By use of a redundant unit, processing can continue for fault-tolerant operation.

Description

BACKGROUND OF THE INVENTION

The invention disclosed and claimed herein relates to digital data processing and, particularly, to methods and apparatus for fault-detecting and fault-tolerant computing.

A shortcoming of many conventional digital data processors is their inability to detect and correct a range of data transfer and operational faults. Personal computers and workstations, for example, are typically configured to detect only a single class of data transfer faults, such as, parity errors. Larger computer systems often incorporate error-correcting codes, at least in peripheral device communications, to detect and correct single-bit errors.

Computer systems marketed and described in prior patents to the assignee hereof are capable of detecting and correcting a wider range of faults. U.S. Pat. No. 4,750,177, for example, discloses a fault-tolerant digital data processor having a first functional unit, such as a central processor, with duplicate processing sections coupled to a common system bus for identically processing signals received from other functional units, such as the memory or peripheral device controller. In addition to checking the received data, the first functional unit compares the output generated by the sections while one of them--the so-called "drive" section--transmits processed data to the bus. When the section outputs do not agree, the functional unit drives an error signal onto the bus and takes itself off-line. According to the patent, a functional unit such as a central processor can have a redundant partner that is constructed and operates identically to the original. In such a configuration, if one of the partners is taken off-line due to error, processing continues with the partner.

According to U.S. Pat. No. 4,931,922, also assigned to the assignee hereof, redundant peripheral control units, each with duplicate processing sections, control the latching of data and signaling of errors vis-a-vis data transfers with attached peripheral devices. For this purpose, data signals applied to the peripheral device bus by either the control units or peripheral devices are captured and compared by processing sections within each control unit. The results of those comparisons are shared between the controllers. If the comparisons indicate that the data captured by both control units agrees, then control units generate a "strobe" signal that causes the data to be latched. If the units do not agree after successive retries, the control units withhold issuance of the "strobe" signal and enter an error-handling state for determining the source of the error.

Co-pending, commonly assigned U.S. patent application Ser. No. 07/926,857 discloses, in one aspect, a digital data processing system having dual, redundant processor elements, each including dual, redundant processing sections that operate in lock-step synchronism. Failure detection stages in each section assemble signals generated by that section into first/second and third/fourth groups, respectively. Normally, signals in the first group match those of the third group, while the signals in the second group match those of the fourth group. The groups are routed between the sections along conductor sets separate from the system bus. To detect fault, each stage compares a respective one of the groups with its counterpart. If either stage detects a mismatch, it signals an error to its respective processing section, which can take the corresponding processor element off-line.

Although the methods and apparatus described in the above-mentioned patents and patent applications provide higher degrees of fault detection and fault tolerance than previously attained, still further improvement in this regard is desirable.

An object of this invention, therefore, is to provide digital data processing apparatus and methods having still greater fault-detecting and fault-tolerant capacity than the prior art.

An object of the invention is to provide apparatus and methods for fault detection that may be readily implemented without consuming excessive hardware resources such as chip pins or board conductors.

A related object is to provide such apparatus and methods as can be implemented in high-speed hardware elements, such as ASIC's.

Another object is to provide such apparatus and methods as can be exercised without incurring excessive signal processing delays.

Still another object of the invention is to provide apparatus and methods for fault detection as can be readily used in fault tolerant processing configurations.

SUMMARY OF THE INVENTION

The invention provides, in one aspect, a digital data processing device that includes a bus for transmitting signals (e.g., data and/or address information) between plural functional units (e.g., a central processing unit and a peripheral controller). A first such unit includes first and second processing sections that concurrently apply to the bus complementary portions of like information signals (e.g., longwords containing data). A fault detection element reads the resultant signal from the bus and compares it with at least portions of the corresponding signals originally generated by the processing sections themselves. If there is discrepancy, the fault-detector signals a fault.

In a related aspect of the invention, the fault-detector performs a "loopback" check that compares portions of the original information signal generated by each processing section with corresponding portions of the resultant signal read from the bus. Where both sections agree that compared portions of their respective original signals differ from the resultant signal, the fault-detector signals a transmission error. On the other hand, where the sections disagree as to which portions of the compared signals differ, the fault-detector signals a processing error.

According to a related aspect of the invention, the first functional unit can respond to a processing error by disabling the first and second processing sections from applying further signals to the bus. The processing sections, moreover, can respond to a transmission error by reapplying respective portions of certain prior information signals to the bus.

In a further aspect, the invention provides a device as describe above in which the first and second processing sections include transceivers that normally route respective portions of the first and second information signals to the bus and, conversely, that route the resultant bus signals to the fault-detector. Loop-check elements in the processing sections respond to a transmission error by generating, as the first and second information signals, error-detecting values (e.g., series of identical digits). The transceivers likewise respond to a transmission error by routing back to the fault-detector, in lieu of any resultant signal received from the bus, the error-detecting values.

Functionality of the type described in the preceding paragraph permits localization of apparent transmission errors. Specifically, it permits determination of whether those errors stem from faults in the bus or, alternatively, from faults in transceiver circuitry interfacing the processing sections and the bus.

In this regard, the fault-detector can compare error-detecting values routed back to it by the transceivers with the original error-detecting values generated by the processing sections. In the event that differences are detected, the functional unit is assumed to be faulty. The processing sections are, accordingly, prevented from applying further signals to the bus.

While a digital data processing device having features of the type described above is fault-detecting, according to further aspects of the invention, the device is fault-tolerant. In this regard, a device as described above can include a partner functional unit that generates redundant information signals substantially identical to those applied to the bus by the first functional unit. During normal operation, the first and partner functional units drive information to the bus during alternating time intervals, or bus cycles. In the event either the first functional unit or its partner is disabled, the other unit alone drives information to the bus in non-alternating cycles.

Still further aspects of the invention embrace fault-detecting and fault-tolerant methods of digital data processing device paralleling the operation of the device described above.

These and other aspects of the invention are evident in the attached drawings and the description that follows.

BRIEF DESCRIPTION OF THE DRAWINGS

A more complete understanding of the invention may be attained by reference to the drawings, in which:

FIG. 1 depicts a digital data processing apparatus 5 defining an environment of the type in which the invention may be practiced;

FIG. 2 depicts a digital logic device according to the invention for detection of faults in a functional unit;

FIG. 3 illustrates a bus transaction in a digital data processing device of the type in conjunction with which the invention may be practiced;

FIG. 4 illustrates the effect of bus errors on the bus cycle in a digital data processing device of the type in conjunction with which the invention may be practiced;

FIG. 5 depicts the signals that are transferred between fault-detection elements in a digital data processor according to the invention.

DETAILED DESCRIPTION OF THE ILLUSTRATED EMBODIMENT

Digital Data Processing Environment

FIG. 1 depicts a digital data processing apparatus 5 defining an environment of the type in which the invention may be practiced. The apparatus 5 includes functional units, such as processor/memory elements 10, 12, peripheral device controllers 14, 16, and communications device 18 connected via a common system bus 20.

The functional units can operated in duplex for lock-step redundant processing. Thus, illustrated processor/memory elements 10, 12 are partnered with one another to respond concurrently and identically to information input from bus 20 to generate further information for output to bus 20. Likewise, illustrated peripheral device controllers 14, 16 are partnered with one another to process and communicate like information between the functional units and peripheral devices (not shown). In the event that one of these functional units 10, 12, 14, 16 is taken off-line--e.g., in response to erroneous operation detected by fault-detection logic described below--its partner can continue processing to provide fault-tolerant operation.

Illustrated communications device 18 is implemented in simplex for non-redundant processing. Device 18 can be taken off-line if its operation is detected as erroneous; however, no partner unit is provided to ensure continuous operation.

Illustrated bus 20 comprises duplicate conductor sets 20a, 20b for normally transferring identical data, addressing and control information between the functional units. A further set of conductors 20c is also provided for transfer of further control and error signals.

Functional units 10-18 comprise dual, redundant processing sections. Thus, for example, peripheral device controller 14 incorporates "drive" and "check" processing sections 14a, 14b, each of which responds in lock-step synchronism with the other to identical information signals input from the buses to generate identical processed information signals for output to the buses.

Further apparatus defining environments in which the invention may be practiced are disclosed in prior patents and patent applications assigned to the assignee hereof, e.g., Baty et al, U.S. Pat. No. 4,931,922, for METHOD AND APPARATUS FOR MONITORING PERIPHERAL DEVICE COMMUNICATIONS (Attorney Docket SCM-005); Reid, U.S. Pat. No. 4,866,604, for DIGITAL DATA PROCESSING APPARATUS WITH PIPELINED MEMORY CYCLES (Attorney Docket SCM-027); Williams, U.S. Pat. No. 4,816,990, for METHOD AND APPARATUS FOR FAULT TOLERANT COMPUTER SYSTEM HAVING EXPANDABLE PROCESSOR (Attorney Docket SCM-030); Long et al, U.S. Pat. No. 4,926,315, for DIGITAL DATA PROCESSOR WITH FAULT TOLERANT PERIPHERAL BUS COMMUNICATIONS (Attorney Docket SCM-041); Clemson et al, U.S. patent application Ser. No. 7/905,120 for COMPUTER APPARATUS AND METHOD FOR DIGITAL PROCESSING SYSTEMS (Attorney Docket SCM-085). The teachings of these patents and patent applications are incorporated herein by reference.

Overview of Fault Detection Logic

FIG. 2 depicts a digital logic device 22 according to the invention for detection of faults in functional unit 14 and, particularly, in the input, processing and output of signals by the lock-step, duplicate drive-side and check-side processing sections 14a, 14b thereof. Those skilled in the art will appreciate that functional unit 14 is referred to by way of example, and that like digital logic, as described in detail below, can be utilized to detect faults in the other functional units having logic for generating redundant bus output signals.

Device 22 includes dual logic checking sections 24a, 24b, each for processing a respective one of the bus outputs of functional unit 14. Device 22 can be implemented in software or hardware, though, it is amenable to implementation in application specific integrated circuits, or ASIC's. For this reason sections 24a, 24b, as well as corresponding logic section in other functional units, are hereinafter referred to as ASIC's 24a, 24b.

ASIC 24a provides an interface between drive-side processing section 14A and bus 20, while ASIC 24b serves as an interface between check-side processing section 14b and bus 20. ASIC's 24a, 24b are constructed and operated identically in lock-step synchronism. A description of the construction and operation of ASIC 24a follows. It will be appreciated that ASIC 24b, as well as corresponding ASIC's in other functional units, is similarly constructed and operated.

ASIC Structure and Operation

ASIC 24a selectively transfers information signals output by drive-side section 14a to bus 20. ASIC 24a likewise transfers information signals received from the bus 20 to that drive side section 14a. ASIC 24a is coupled to its counterpart on functional unit 14, to wit, ASIC 24b, for transfer of fault-detection information over conductors 28.

In the illustrated embodiment, section 14a processes signals that are 32 bits in length. Bus 20, on the other hand, transfers between the functional units 10-18 signals that are 40 bits in length. The additional bits represent error checking and correcting codes (ECC) added to bus output by ASIC 24a or its counterparts. Those skilled in the art will appreciate that ASIC 24a can be constructed and operated in the manner hereof to transfer signals of greater and lesser length, with or without additional ECC bits.

Tracing the output path of ASIC 24a, 32-bit information signals generated by drive-side section 14 are routed through parity generator 26 for addition of eight ECC bits generated in a conventional manner known in the art. Thereafter, a portion of the ECC-incorporating signal is routed to both to bus 20, via transceiver latch 32a, and to loopback latch 30a. Counterpart ASIC 24b simultaneously routes a complementary portion of a normally identical ECC-incorporating signal to the bus 20.

Thus, whereas ASIC 24a routes, e.g., the low-order bits of the ECC-incorporating output signal to bus 20, ASIC 24b routes the complementary high-order bits of the like signal to that bus. In the absence of fault, the resulting signal on the bus 20 identically matches the ECC-incorporating signals generated by each ASIC 24a, 24b.

ASIC 24a routes that portion of its ECC-incorporating signal that was not routed to the bus 20 to loopback latch 30b, as illustrated.

Concurrent with application of the ECC-incorporating signal to bus 20, transceiver latch 32b wire-OR's the non-ECC portion of the entire resultant bus signal (i.e., 32 bits) to loopback line 34. The corresponding latch of ASIC 24b simultaneously wire-OR's the bus signal onto that same line 34, as illustrated.

Within ASIC 24a, these non-ECC portions of the bus signal are routed to loopback check sections 36a, 36b for comparison with the original drive-side section 14a output, as stored in latches 30a, 30b, respectively. Loopback check 36a, thus, compares for identity the signal from latch 30a, representing a portion of the original section 14a output with the corresponding portion of the signal received from the bus 20 (as originally driven to the bus by ASIC 24a) via latch 32b. Loopback check 36b, likewise, compares for identity the other portion of the original section 14a output with the corresponding portion of the signal received from bus 20 (as originally driven to the bus by ASIC 24b) via latch 32b. Parity check circuit 38 confirms that the parity of the non-ECC portion of the resultant bus signal is correct.

In the event of mismatch, the loopback check circuits 36a, 36b route error signals to local error detection logic 40. Error signals are also routed to logic 40 from corresponding loopback check circuits in ASIC 24b as shown. Moreover, parity error signals generated by parity check circuit 38a are routed to local error detection logic 40, although no corresponding signals are received from the parity detection logic of ASIC 24b.

Error Signals

Based on error signals, if any, received from the loopback check and parity check circuits, error detection logic 40 determines whether the signaled error resulted during generation of information signals by processing sections 14a, 14b, or whether it resulted from a transfer of those signals to the bus 20 by ASIC's 24a, 24b. In the event of an error in generation of signals, logic 40 generates a Board Error signal. In the event of error in transmission, error detection logic 40 generates a Bus Error signal. A more complete description of the conditions for generation of bus and board error signals by error detection logic 40 is set forth in the table below.

TABLE 1__________________________________________________________________________ Loopback checksLoopback check results for this reported by other ASIC (e.g.,ASIC (e.g., ASIC 24a) ASIC 24b) Does Info Does Info Driven by this Does Info Driven by this Does Info ASIC to Bus Driven by ASIC to Bus Driven by Match that Other ASIC to Match that Other ASIC toParity Rec'd from Bus Match Rec'd from Bus MatchCorrect? Bus? Expectation? Bus? Expectation? Error Signal__________________________________________________________________________Yes Yes Yes Yes Yes NoneX Yes No Yes No Bus ErrorX No Yes No YesX No No No NoX Yes X No X Board ErrorX No X Yes XX X Yes X NoX X No X YesNo Yes Yes Yes Yes Bus Error__________________________________________________________________________

The table may be summed up as follows: (i) if both ASIC's 24a, 24b get a loopback failure, it is assumed that the bus 20 is bad; (ii) if the ASIC's disagree as to the cause of failure, it is assumed that they are out of sync--either due to an ASIC 24a, 24b failure or a processing section 14a, 14b logic failure.

In response to a Board Error signal generated by ASIC 24a or 24b, the corresponding functional unit 14 can enter an error analysis sequence and/or take itself off-line. In the event that there is a redundant partner, e.g., functional unit 16, that unit may remain on-line to retransmit corrected information onto bus 20 and, thereby, to provide continuous processing.

LoopCheck Phase

In response to the Bus Error, the processing sections 14a, 14b enter a loopcheck phase to determine whether the error resulted from fault in the bus conductors or, rather, in transceiver circuitry used to drive information to the bus.

During the loopcheck phase, the processing section 14a, 14b drive a test pattern to their respective ASIC's 24a, 24b for a short interval. The ASIC's 24a, 24b process and route the test pattern in the manner described above, except that they do not read any resultant signal from that bus. Rather, transceiver latch 32a routes signals that would otherwise be driven to the bus 20 directly back to latch 32b via multiplexor 32c. That rerouted is data checked in the manner described above, i.e., as if it had been routed to, and read from, the bus.

Any discrepancies detected by local loopback check circuits 36a, 36b, during the loopcheck phase necessarily indicate an error in the ASIC and/or its transceiver latches 32a, 32b. In response to such error, the ASIC's 24a, 24b signal a Board Error for processing as above.

ASIC Operation with Redundant Functional Units

In the event exemplary functional unit 14 is provided with a redundant partner 16, as illustrated, operation of the ASIC's of both units 14 and 16 proceed as described above. Except, however, the ASIC's of each functional unit alternately drive signals to bus 20. Thus, for example, ASIC's 24a, 24b can drive their ECC-incorporating information signals to bus 20 during "even" bus cycles or phases, while the ASIC's of functional unit 16 drive ECC-incorporating information signals to bus 20 during odd cycles or phases. Such alternate cycling permits the functional units to localize faults and, thereby, to determine which (if any) is responsible for apparent bus errors.

Those skilled in the art will appreciate that a variety of single node failures can cause multiple outgoing bus signals to differ as between processing section/ASIC pairs. This may be appreciated with respect to operation of the parity generator 26. For example, in the event drive-side section 14a generates an erroneous output signal, e.g., due to a logic failure such as a trace shorted to ground, parity generator 26 will generate an erroneous ECC signal. Thus, the "single node" failure in the drive-side logic 14a effectively results in a "dual node" failure vis-a-vis (i) the erroneous information portion of the outgoing bus signal, and (ii) the erroneous ECC portion of that signal.

With an understanding of this example, it will be appreciated that--unless the bus 20 is alternately driven by the redundant units 14, 16--a situation may develop that could cause the system 5 to fail.

Suppose a hypothetical bit, referred to as bit "a," output from the check-side section 14b to the ASIC 24b, is erroneously flipped from 1 to 0. This causes an ECC bit, referred to as bit "b", that would otherwise have a value of 0, to flip 1. When the resultant ECC-laden signal is wire-OR'ed to bus 20 by ASIC 24b, as well as by its counterpart in functional unit 16, bit a is effectively "corrected." This results from overriding of the erroneous zero-valued bit by the correct one-valued bit output by functional unit 16. The error in bit b, however, is not remedied, since the zero-value output by functional unit 16 cannot override the erroneous one-value output by unit 14.

In such a scenario--again, one in which the partnered functional units do not alternate --ASIC 24b will detect a loopback mismatch on a bit it is driving to the bus, bit a, while ASIC 24a will detect a loopback mismatch on a bit it is driving to the bus, bit b. This information will be signalled between the ASIC's via lines 28 (FIG. 2) and, per Table 1, both ASIC's will agree are that there is a problem on the bus.

As discussed above, this hypothetical situation is remedied by driving the bus 20 with alternating ones of the partnered functional units 14, 16, and by having these functional units otherwise operating independently, including performing all of the same loopback checks and error computations shown in Table 1.

Bus Protocol

Illustrated bus 20 is a synchronous, pipelined, multiplexed address/data, fault-tolerant bus with split-transaction capabilities. Function-code, address and data signals are duplicated and transferred in lock-step on conductors 20a, 20b. Three-way voting is implemented on control lines 20C including error, busy, and arbitration.

The bus 20 has a logical 64-bit wide data/address bus. Processor/memory units 10, 12 connect to all 64 bits of the data/address bus, while controller units 14, 16 connect to only 32 bits of the data/address bus. The bus 20 supports word accesses and has a block transfer capability for support of cache line accesses.

The basic component of transactions on the bus is the "operation." A basic operation is illustrated in FIG. 3. It is composed of 5 phases: ARB, INFO, POST1, POST2 and POST3. During each phase, two information transfers occur; hence, the phases are "double pumped." The double pump frequency is approximately 24 Mhz.

FIG. 3 is a logical diagram which illustrates the logical activity on the bus 20. All information is registered to/from transceivers coupled onto and off of the bus 20.

The phases used are as follows: ARB phase: The functional units 10-18 drive their respective arbitration ID's during both halves (cycles) of the arbitration phase. During the second half of the phase they determine whether they won arbitration and prepare for the INFO phase. INFO phase: For address transfers, units drive the virtual index, function code, and obey state during the first half of the INFO phase and the physical address during the second half. For data transfers, data is driven during both the first and second halves of the cycle. TRID's (transaction ID's) are driven during both halves for all access. Non-cache consistent address transfers need not supply a virtual index, though the driven information must be deterministic and parity will be computed across it. POST1 and POST2 phases: During these phases, functional units 10-18 determine whether any error conditions existed with the INFO phase and whether there is need to BUSY the operation, i.e., to transmit a BUSY signal on the bus and, thereby, to cancel the cycle in its POST3 phase. POST3 phase: Any functional unit detecting an error with the INFO phase drives the error lines during both halves (cycles). If a functional unit 10-18 does get erred, it next goes to the POST4 phase (not illustrated) to perform a loop.sub.-- ck.sub.-- op, as described in detail below, and to redrive the erred data. Any unit detecting the need to BUSY an address/function code driven during the Info phase drives BUSY during the first cycle of this phase. It is also during this phase that accesses are acknowledged. POST4 phase: This phase is only entered as a result of an error asserted in POST3. During this phase, the unit performs a loop.sub.-- ck.sub.-- op and redrives the erred data.

FIG. 4 illustrates the effect of bus errors on the bus cycle. This figure illustrates bus phase transitions. Within each vertical line is a bus phase (approximately 12 Mhz) which is actually composed of two information transfers on the bus. A bus BUSY has no effect on a bus operation during any phase except for POST3, at which time it has the effect of canceling the bus operation. In the event a cycle is both BUSY'ed and ERROR'ed, the ERROR takes precedence.

Fault Detection

As noted above, each processing section, e.g., 14a, 14b, drives half of each field (e.g., data, address or control information) to its ASIC's transceivers and, yet, receives the entire field back from the transceivers. Thus, there are two checks that each ASIC 24a, 24b can do on the data returning from the bus:

1) Does the returning data match the data that was driven from this ASIC? This is the loopback drive check. 2) Does the returning data match the data that was driven by the other ASIC? This is the loopback compare check.

Both of these checks are done for both duplicated bus conductor sets 20a, 20b. Thus, four error signals are generated inside of each ASIC 24a, 24b, to wit, comp.sub.-- err.sub.-- a, comp.sub.-- err.sub.-- b, drive.sub.-- err.sub.-- a, drive.sub.-- err b (where "a" refers to bus 20a and "b" refers to bus 20b.) In order to determine its true state, each ASIC 24a, 24b must monitor the results of the four comparisons it has done, as well as those of the other ASIC. Thus four signals are passed in each direction, as shown in FIG. 5 above, so that each side of the unit may correctly determine the state of the unit.

The error detection logic 40 in each ASIC 24a, 24b determines whether the data on the bus 20a is broken, as well as if the corresponding processing section 14a, 14b is broken. The table that follows shows what combinations of the four loopback signals for each bus indicate that a processing section is broken.

TABLE 2______________________________________ this processingmy my drive.sub.-- err comp.sub.-- err from section/ASICdrive.sub.-- err comp.sub.-- err other ASIC other ASIC broken______________________________________0 0 0 0 00 0 0 1 10 0 1 0 10 0 1 1 10 1 0 0 10 1 0 1 10 1 1 0 00 1 1 1 11 0 0 0 11 0 0 1 01 0 1 0 11 0 1 1 11 1 0 0 11 1 0 1 11 1 1 0 11 1 1 1 0______________________________________

Error detection logic 40 asserts a bus error if a condition exists that should break the corresponding processing section and ASIC. In addition a bus error is asserted for any condition where a compare error (of any type) has occurred and the processing section and/or ASIC has not broken. This error will only be on the bus 20 where the error was detected.

The processing sections 14a, 14b inhibit loopback checking for bus transactions when data is expected to be different from as between the processing sections, as well as when a redundant partner functional unit 16 has actually driven the information on the bus 20.

LoopCheck Phase Revisited

The loop.sub.-- ck.sub.-- op eliminates an additional point of failure. Take, for example, a situation where an outgoing conductor on one of the ASIC's is shorted to ground. This would effectively corrupt all data on the bus, with the result that the loopback circuitry of both ASIC's 24a, 14b would agree that data received from the bus was incorrect and, hence, would incorrectly signal a bus failure.

A loop.sub.-- ck.sub.-- op phase constitutes the following steps: during the bus phase after a signaling of a bus error, functional unit 14 (as well as all other functional units having like fault-detection circuitry 22) put their respective ASIC transceivers into loop-back mode for one bus phase, i.e., two bus cycles. A pattern of 0's and 1's are driven from each ASIC on all duplicated bus 20 out nets (info, trid, func.sub.-- op and parity). If a functional unit's ASIC does not receive the same pattern of 0's and 1's, that functional unit will break, thus detecting any short or open in the looped-back data path.

Multiplexing of Incoming Bus Data

The bus input, as received by each ASIC (e.g., via transceiver 32b) is MUXED and running at twice the speed as the bus 20. This is done to save ASIC pins. Fault-tolerant issues are taken into consideration when defining what bits are muxed together with bus info bits. First, no signal travelling on A BUS 20A should be muxed with a signal travelling on B BUS 20B. If a single trace on an ASIC should fail, this would create a condition where both buses 20a, 20b would have a parity error. This is a non-recoverable condition.

Second, two bus 20 signals that share the same parity chain should NOT be muxed together. If a signal trace should fail, there would be the possibility of getting good parity across the bus thus having a condition of no parity error with the buses having a compare error. This is an unrecoverable error.

Summary

Described above are methods and apparatus for fault-detection/fault-tolerance preferred meeting the objects set forth above. Those skilled in the art will, of course, appreciate that the embodiments described above are by way of example and that other embodiments incorporating modifications thereto may well fall within the scope of the invention. Thus, for example, an apparatus according to the invention is not dependent upon use of a split-transaction bus, nor a duplicated bus, nor a bus of a certain bit-width.

Claims

1. A fault detecting digital data processing device comprising

A. bus means for transmitting applied signals between plural functional units,

B. a first such functional unit including a first processing section for generating a first signal and for applying a portion thereof to said bus means,

C. said first functional unit further including a second processing section for generating a second signal and for applying a portion thereof to said bus means, said second signal normally being substantially identical to said first signal,

said portion of said second signal applied to said bus means by said second processing section being complementary to said portion of said substantially identical first signal applied to said bus means by the first processing section,

said second processing section applying said portion of said second signal to said bus means substantially concurrently with application of said portion of said first signal to said bus means by said first processing section, said portions of said first and second signals so applied to said bus means defining a third signal and

D. fault-detection means, coupled to said first and second processing sections and to said bus means, for receiving said third signal from said bus means and for comparing at least portions thereof with corresponding portions of said first signal and of said second signal generated by those respective processing sections,

said fault-detection means including means for signaling fault in response to at least selected differences between those compared signal portions.

2. A fault-tolerant device for use with a digital data processing device having bus means for transmitting applied signals, said fault-tolerant device comprising

A. first processing section for generating a first signal and for normally applying a portion thereof to said bus means,

B. second processing section for generating a second signal and for normally applying a portion thereof to said bus means, said second signal being substantially identical to said first signal,

said portion of said second signal applied to said bus means by said second processing section being complementary to said portion of the substantially identical first signal applied to the bus by the first processing section,

said second processing section applying said portion of said second signal to said bus means substantially concurrently with application of said complementary portion of said first signal to said bus means by said first processing section, said portions of said first and second signals so applied to said bus means defining a third signal, and

C. fault-detection means, coupled to said first and second processing sections and to said bus means, for receiving said third signal from said bus means and for comparing at least portions thereof with corresponding portions of said first signal and said second signals generated by those respective processing sections, and for signaling fault in response to at least selected differences between those compared portions.

3. A fault-tolerant device for use with a digital data processing device having bus means for transmitting applied signals between functional units, said fault-tolerate device comprising first and second functional units, each comprising

A. first processing section for generating a first signal and for normally applying a portion thereof to said bus means,

B. second processing section for generating a second signal and for normally applying a portion thereof to said bus means, said second signal being substantially identical to said first signal,

said portion of said second signal applied to said bus means by said second processing section being complementary to said portion of the substantially identical first signal applied to the bus by the first processing section,

said second processing section applying said portion of said second signal to said bus means substantially concurrently with application of said complementary portion of said first signal to said bus means by said first processing section, said portions of said first and second signals so applied to said bus means defining a third signal,

C. fault-detection means, coupled to said first and second processing sections and to said bus means, for receiving said third signal from said bus means and for comparing at least portions thereof with corresponding portions of said first signal and said second signals generated by those respective processing sections,

said fault-detection means comprising means for preventing said first and second processing sections from applying signals to said bus means in response to at least selected differences between those compared portions.

4. A fault detecting method of operating a digital data processing device of the type having bus means for transmitting applied signals between functional units, said method comprising operating a first such functional unit with

A. a first processing step for generating a first signal and for normally applying a portion thereof to said bus means,

B. a second concurrent processing step generating a second signal and for normally applying a portion thereof to said bus means, said second signal being substantially identical to said first signal,

said portion of said second signal applied to said bus means being complementary to said portion of the substantially identical first signal applied to the bus means,

said portion of said second signal being applied to said bus means substantially concurrently with application of said complementary portion of said first signal to said bus means, said portions of said first and second signals so applied to said bus means defining a third signal, and

C. a fault-detection step for receiving said third signal from said bus means and comparing at least portions thereof with corresponding portions of said first signal and said second signals,

said fault-detection step further for signaling fault in response to at least selected differences between those compared portions.