Computing module and method for high integrity computing

By using loose lock step logic components between processing channels to achieve task synchronization and data comparison, the problem of error identification and isolation in high-integrity calculations in the existing technology is solved, and more accurate and convenient calculation results are achieved, avoiding false alarms Error, and design is transparent to the application.

CN119003442APending Publication Date: 2024-11-22GE AVIC CIVIL AVIONICS SYST CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310559325.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-17
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

When implementing high-integrity computing, it is difficult to accurately identify and isolate errors between processor cores, and the design depends on CPU vendors and bus architecture, and cannot be effectively used on the latest CPUs, and during multiple executions. The false positive error problem caused by differences in execution time has not been effectively resolved.

Method used

Using at least two processing channels, each processing channel includes a central processing unit and a loose lock step logic component. Task synchronization and data comparison between processing channels are achieved through synchronous logic unit and comparison logic unit to ensure consistency when performing the same task and trigger the exception handling mechanism when inconsistent to avoid false positive errors.

Benefits of technology

More accurate and convenient high integrity calculations are achieved, false positive errors caused by differences in speed and slow execution of processing channels are avoided, and synchronization and comparison of executions are transparent to resident applications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119003442A_ABST
    Figure CN119003442A_ABST
Patent Text Reader

Abstract

The invention discloses a calculation module and method for high integrity calculation. A computing module for high integrity computing includes at least two processing channels each for performing a same task, where a first processing channel includes a first CPU and a first loosen step logic connected thereto. The first unlocking step logic component comprises a first synchronization logic unit and a first comparison logic unit. And the first synchronization logic unit is used for realizing execution synchronization of the first processing channel and the second processing channel on the same task. The first comparison logic unit is used for comparing a first command and first data transmitted to the first unlocking logic component with a second command and second data transmitted to the second unlocking logic component in the second processing channel in response to determining that execution is synchronous, and triggering a comparative exception handling mechanism in response to a result of the comparison being inconsistent.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates generally to the field of computing, and more particularly to high integrity computing. Background Art

[0002] In safety-critical applications such as avionics, high integrity is an important requirement for data calculations because the calculation results output by computers have a significant impact on the safety of the system. In terms of system safety design, in order to provide high integrity at the source, the data processing module is required to promptly identify various errors occurring within itself (for example, identifying computer hardware errors or single-particle reversal events), isolate the errors, handle various errors appropriately, and perform system recovery operations to avoid spreading erroneous calculation results to other modules or components of the system, thereby affecting the safety of the entire system.

[0003] In order to achieve high-integrity computing, it is required that single-point failures cannot cause errors, and the probability of unrecognized errors is extremely unlikely (based on the safety requirements of the resident function). The common design practice is to implement an N (N is an integer ≥ 2) redundant independent computing data processing module, and compare or vote on the output calculation results between the N paths. For example, if N = 2, if the two paths are compared and found to be inconsistent, it means that at least one path has an error. At this time, the calculation results of these two paths are no longer complete and should be discarded and processed later. If N = 3, a 3-choice 2 vote can be performed, so that even if an error occurs in one path, the data processing module can identify and isolate the path where the error occurs, and continue to work to improve system availability.

[0004] In terms of technical implementation, an important solution for achieving high-integrity computing in the past was lock-step technology. Lock-step technology can provide instruction-level granularity comparison, that is, compare instruction fetch transactions and data read and write transactions occurring on the bus to identify errors. This comparison can occur between processing cores of a multi-core general-purpose processing unit (CPU), which is achieved by comparing transactions on the internal bus connecting each core (for example, the Split-Lock technology in the A76AE recently launched by ARM). This comparison can also occur between multiple separate CPUs, by comparing transactions on the front-end bus connecting the CPUs.

[0005] The first technique mentioned above can only identify errors that occur inside the processor core through comparison. However, since the two paths still share components such as memory, the identification of errors that occur in these components still needs to rely on other means. In addition, its design is completely dependent on the implementation and provision of the System on a Chip (SoC) supplier.

[0006] The second technology mentioned above is based on the exposure of the front-end bus of the traditional CPU to the user, so the user can implement the comparison function on the bridge connecting the front-end bus to the memory and other components. This technology is highly dependent on the bus architecture of the CPU. At present, the mainstream high-performance shelf CPU products are all SoC designs, and their processor cores and memory controllers and other components are integrated inside the SoC and are no longer exposed to the user through the connection of the internal bus. This imposes a great design on continuing to use this design on the latest CPU.

[0007] In order to break the dependence on CPU suppliers and the internal bus architecture of their products, the existing technologies US9170907B2 and CA 2694198 C have proposed some solutions. Their solutions are to implement two or more paths of calculation, each path includes a resident application unit and an input / output (I / O) unit, and the results of the resident application unit are compared in the I / O unit. At the same time, the two paths need to be synchronized and coordinated to ensure that the two or more paths have the same output when there are no errors. However, they do not have a good solution for the deviation of the execution path caused by the difference in execution time between multiple paths when the same program is executed in multiple paths in a partitioned environment. The former has no special design for this aspect (assuming that the transaction synchronization in the I / O unit can avoid this problem). The latter designs a critical region manager (CRM) in the I / O unit to solve the modification synchronization of key data, but this requires the application software to actively identify the key data and trigger the protection of the key data area, which is not transparent to the application.

[0008] Therefore, there is a need in the art for a more accurate and convenient technology for implementing high-integrity computing. Summary of the invention

[0009] The present invention is provided in order to provide a technology for implementing high-integrity computing more accurately and conveniently.

[0010] One aspect of the present invention provides a computing module for high-integrity computing, comprising: at least two processing channels, including a first processing channel and a second processing channel, each processing channel being used to execute the same task, wherein the first processing channel comprises: a first central processing unit CPU; and a first loose-step logic component connected to the first CPU, the first loose-step logic component comprising: a first synchronization logic unit, used to: achieve synchronization of execution of the same task by the first processing channel and the second processing channel; and a first comparison logic unit, used to: in response to determining that the execution of the same task by the first processing channel and the second processing channel is synchronous, respectively compare a first command and first data transmitted to the first loose-step logic component with a second command and second data transmitted to the second loose-step logic component in the second processing channel; in response to the comparison result being inconsistent, triggering a comparison exception handling mechanism.

[0011] The computing module as described above, wherein if the first processing channel executes the same task before the second processing channel, the first CPU switches to a waiting state; if the first processing channel executes the same task after the second processing channel, the second CPU of the second processing channel switches to a waiting state.

[0012] A computing module as described in any of the above items, wherein the first synchronization logic unit is used to: receive an indication that an asynchronous event has occurred; receive second key data associated with the execution of a task by the second processing channel from a second synchronization logic unit of a second loose-step logic component of the second processing channel; compare the first key data associated with the execution of the task by the first processing channel with the second key data; based on the result of the comparison between the first key data and the second key data, adjust the execution of the task by the first processing channel and the second processing channel until the first key data is consistent with the second key data; and indicate that the execution of the same task by the first processing channel and the second processing channel is synchronized.

[0013] A computing module as described in any of the above items, wherein the first key data and the second key data each include: a global variable that plays a key role in execution point synchronization, and a modification or update count of an inter-process communication object.

[0014] A computing module as described in any of the above items, wherein the first processing channel and the second processing channel use the same clock source.

[0015] A computing module as described in any of the above items, wherein the first synchronization logic unit is used to: in response to determining that the first processing channel and the second processing channel are both executing code requesting to read the system time, while the second synchronization logic unit of the second loose-step logic component of the second processing channel provides the time value to the read application programming interface API of the second processing channel, provide the time value to the read API of the first processing channel.

[0016] A computing module as described in any of the above items, wherein the first loose-lock-step logic component also includes: a first input / output I / O controller, connected to the first I / O component; and a first non-volatile storage controller, connected to the first non-volatile storage component, the first command includes: a read and write command from the first CPU to the first I / O component, and / or a read and write command to the first non-volatile storage component, and the first data includes: data read from the first I / O component, and / or data read from the first non-volatile storage component.

[0017] A computing module as described in any of the above items, wherein the first loose lock-step logic component also includes: a first command queue for storing the one or more commands and data to be written associated with a write command in the one or more commands; and a first read data queue for storing results of the read commands in the one or more commands.

[0018] A computing module as described in any of the above items, wherein the first comparison logic unit is used to: compare the one or more commands in the first command queue with the corresponding commands in the second command queue of the second loose lock step logic component of the second processing channel; and in response to determining that the result of the command comparison is inconsistent, trigger the comparison exception handling mechanism.

[0019] A computing module as described in any of the above items, wherein the first comparison logic unit is used to: in response to determining that the result of the command comparison is inconsistent, save the comparison field information and prevent the one or more commands from being executed.

[0020] A computing module as described in any of the above items, wherein the result of the read command includes: successfully read data and a flag indicating that the read command is successfully executed; and a flag indicating that the read command fails to execute and the reason for the failure.

[0021] A computing module as described in any of the above items, wherein the first loose lock-step logic component further includes: a network access logic unit, used to access the first processing channel to a network.

[0022] A computing module as described in any of the above items, wherein the first command includes a first network command to the network from the first CPU, and the first data includes first network input data read from the network, and the first loose lock-step logic component also includes: a first network command queue for storing the first network command; and a first network input data queue for storing the first network input data.

[0023] A computing module as described in any of the above items, wherein the first comparison logic unit is used to: compare the first network command in the first network command queue with the corresponding second network command in the second network command queue of the second loose-step logic component of the second processing channel; compare the first network input data in the first network input data queue with the corresponding second network input data in the second network input data queue of the second loose-step logic component of the second processing channel; and trigger the comparison exception handling mechanism in response to determining that the results of the network command comparison are inconsistent and / or the results of the network input data comparison are inconsistent.

[0024] A computing module as described in any one of the above items, wherein the synchronization logic unit and the comparison logic unit of the loose lock-step logic component of each processing channel can be disabled.

[0025] Another aspect of the present invention provides a method for high-integrity computing, comprising: achieving synchronization of execution of the same task by a first processing channel and a second processing channel; in response to determining that the execution of the same task by the first processing channel and the second processing channel is synchronized, comparing a first command and first data transmitted to the first processing channel with a second command and second data transmitted to the second processing channel; in response to the comparison result being inconsistent, triggering a comparison exception handling mechanism.

[0026] The method as described above, wherein achieving synchronization of execution of the same task by the first processing channel and the second processing channel includes: if the execution of the same task by the first processing channel precedes that by the second processing channel, switching the first general processing unit CPU of the first processing channel to a waiting state; if the execution of the same task by the first processing channel follows that by the second processing channel, switching the second CPU of the second processing channel to a waiting state.

[0027] A method as described in any of the above items, wherein achieving synchronization of execution of the same task by a first processing channel and a second processing channel includes: receiving an indication that an asynchronous event has occurred; receiving second key data associated with execution of the task by the second processing channel from the second processing channel; comparing the first key data associated with execution of the task by the first processing channel with the second key data; adjusting execution of the task by the first processing channel and the second processing channel based on a result of the comparison between the first key data and the second key data until the first key data is consistent with the second key data; and indicating that execution of the task by the first processing channel and the second processing channel is synchronized.

[0028] The computing module and method for implementing high-integrity computing of the present invention will not cause false alarm errors due to the difference in speed of executing tasks between different processing channels, and the synchronization and comparison performed are transparent to the resident application. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Embodiments of the present invention are described with reference to the accompanying drawings.

[0030] Figure 1 is a block diagram illustrating a computation module for high integrity computation according to one embodiment of the present invention.

[0031] Figure 2 is a block diagram illustrating a computation module for high integrity computation according to one embodiment of the present invention.

[0032] Figure 3 is a block diagram illustrating a computation module for high integrity computation according to one embodiment of the present invention.

[0033] Figure 4 is a flow chart illustrating a method for high integrity computation according to one embodiment of the present invention.

[0034] Figure 5 is a flow chart illustrating a process for facilitating synchronization of task execution according to one embodiment of the present invention.

[0035] Figure 6 is a flow chart illustrating a process for facilitating synchronization of task execution according to one embodiment of the present invention. DETAILED DESCRIPTION

[0036] In this application, the term "loose lockstep" is relative to conventional instruction-level granularity lockstep (also called "tight lockstep"). The computing module and method according to the embodiments of the present invention can be used in safety-critical application fields involving multitasking, including but not limited to the field of avionics.

[0037] According to one aspect of the present invention, a computing module for high integrity computing is provided.

[0038] Figure 1 is a block diagram illustrating a computing module 100 for high integrity computing according to one embodiment of the present invention.

[0039] Computing module 100 may include processing channel 110 and processing channel 150. Processing channel 110 and processing channel 150 may each be used to perform the same task, so that computing module 100 may be used to perform high integrity computing. In some embodiments, computing module 100 may include more processing channels (in Figure 1 denoted by ellipsis).

[0040] The processing channel 110 may include a CPU 120 and a loose lock-step logic unit 130 connected to the CPU 120. As an example, the loose lock-step logic unit 130 may be connected to the CPU 120 via a Peripheral Component Interconnect Express (PCIe) interface or a high-speed Ethernet interface (e.g., SGMII).

[0041] The processing channel 110 may further include a memory component (eg, a double data rate (DDR) memory) 141 and a read-only memory (ROM) 142, which are connected to the CPU 120. The ROM 142 may be used to store programs and read-only data.

[0042] The processing channel 110 may also include a non-volatile storage component 143 and an I / O component 144, which are connected to the loose lock-step logic component 130. The non-volatile storage component 143 can be used to store data that is not lost when power is lost during the operation of the storage system. As an example, the non-volatile storage component 143 may include a "NOR" flash memory, a "NAND" flash memory, or other storage media.

[0043] Similarly, the processing channel 150 may include a CPU 160 and a loose lock-step logic component 170 connected to the CPU 160. The loose lock-step logic component 170 may be connected to the CPU 160 through a PCIe interface or a high-speed Ethernet interface (eg, SGMII).

[0044] The processing channel 150 may further include a memory component (eg, DDR memory) 181 and a ROM 182, which are connected to the CPU 160. The ROM 182 may be used to store programs and read-only data.

[0045] The processing channel 150 may also include a non-volatile storage component 183 and an I / O component 184, which are connected to the lock-step logic component 170. The non-volatile storage component 183 can be used to store data that is not lost when power is lost during the operation of the storage system. As an example, the non-volatile storage component 183 may include NOR flash memory, NAND flash memory or other storage media.

[0046] The data communication between the processing channel 110 and the processing channel 150 may be implemented via a Serdes bus. As an example, the content of the interaction between the processing channel 110 and the processing channel 150 may include a PCIe read / write request command issued by the corresponding CPU 120 or CPU 160, data sent to an I / O interface, a flash memory interface or a network, data read from an I / O interface, a flash memory interface or a network interface, etc.

[0047] In order to achieve high integrity computing, the computing module 100 can be configured to a high integrity working mode, and the same program code and data are stored in the ROM 142 of the processing channel 110 and the ROM 182 of the processing channel 150. When the computing module 100 is started, the processing channel 110 and the processing channel 150 respectively execute the code stored in their respective ROM 142 and ROM 182 to initialize the computing module 100 and perform necessary built-in tests (Built-in Test, BIT). After completing the initialization and passing the BIT and before entering the normal multi-task scheduling environment, the processing channel 110 and the processing channel 150 need to be synchronized to ensure that the processing channel 110 and the processing channel 150 are switched to the scheduling mode at the same time and enter the first scheduled task.

[0048] In some embodiments, the loose lock step logic unit 130 may include a comparison logic unit 131 and a synchronization logic unit 132. The comparison logic unit 131 may be used to compare the calculation results of the CPU 120 and the CPU 160 and the transactions when accessing the external device. As an example, the comparison logic unit 131 may compare the commands and data issued by the corresponding CPU of the corresponding processing channel and transmitted to the corresponding loose lock step logic unit through, for example, a PCIe link. As an example, the command may include a read and write command to the non-volatile storage unit 143, 183, the I / O unit 144, 184 or the end system. As an example, the data may include data sent to the corresponding interface. In some embodiments, each interface may include its own queue (e.g., a buffer queue), and the commands and numbers operating on these interfaces may be placed in these queues. The comparison logic unit 131 may send each item in the queue to the comparison logic unit 171 of the loose lock step logic unit 170 of the processing channel 150, and compare the items in its own queue with the corresponding items sent from the comparison logic unit 171. If the comparison result is inconsistent, the comparison logic unit 131 can save the comparison field information and trigger the comparison exception handling mechanism to capture and report errors occurring in any processing channel.

[0049] In order to ensure that processing channel 110 and processing channel 150 can generate and output the same operation results based on the same input without internal errors, it is necessary to synchronize the execution of tasks by processing channel 110 and processing channel 150. The synchronization has two purposes: first, to ensure that the work completed by processing channel 110 and processing channel 150 is the same when modifying key data items that affect the program execution path; second, to ensure that the inputs obtained by processing channel 110 and processing channel 150 are the same, which may include data stored in non-volatile storage components (e.g., flash memory), I / O data, network messages, system time, etc.

[0050] The synchronization logic unit 132 may be used to implement the above synchronization. In some embodiments, the synchronization logic unit 132 may be used to synchronize the execution of the same task by the processing channel 110 and the processing channel 150. More specifically, the synchronization logic unit 132 may be used to ensure that no errors occur inside the processing channel 110 and the processing channel 150, so that the execution paths and output results of the execution of the same task by the processing channel 110 and the processing channel 150 are consistent. As an example, the synchronization logic unit 132 may be used to synchronize code execution points, synchronize input data reading, synchronize system time reading, synchronize asynchronous event processing, etc. Asynchronous events may include inconsistency in comparison, synchronization timeout, and partition clock interrupts for partition window switching in the ARINC653 time-sharing partition operating system widely used in the avionics field.

[0051] Since the processing channel 110 and the processing channel 150 cannot achieve the lock-step of the instruction-level granularity in the conventional lock-step system when executing tasks, in the computing module 100, the processing channel 110 and the processing channel 150 may have different speeds when executing the same task (program). In a multi-tasking environment, when the task is switched, it is possible that the processing channel 110 has modified some key data shared between tasks while the processing channel 150 has not yet modified these key data. In this case, if the processing channel 110 and the processing channel 150 are switched to a new task at the same time and the new task uses the above-mentioned key data, the inconsistency of the key data between the processing channel 110 and the processing channel 150 will cause the calculation results of the processing channel 110 and the processing channel 150 to execute the new task to be different. This difference in execution results is not due to an error, but due to the different input of the key data of the new task. In the case where the comparison logic unit 131 is used for comparison without synchronization, the inconsistency of the comparison result of the comparison logic unit 131 will be identified as an error and cause a false alarm.

[0052] For the above scenario, in some embodiments, if the execution of the task by the processing channel 110 precedes the execution of the task by the processing channel 150, the CPU 120 of the processing channel 110 may be switched to a waiting state. If the execution of the task by the processing channel 110 is later than the execution of the task by the processing channel 150, the CPU 160 of the processing channel 150 may be switched to a waiting state. For example, when a task is scheduled for execution, until it considers that the task's own work has been completed, it actively switches the CPU of the corresponding processing channel to an idle state, or actively switches the CPU of the corresponding processing channel to a waiting state because the resource it is waiting for is not ready. Thus, by actively triggering the scheduling under the condition of actively giving up the processing resources (e.g., CPU) of the system, the execution point synchronization may be performed during the scheduling, thereby ensuring that the work completed by the processing channel 110 and the processing channel 150 before the task switching is the same. At the same time, the system clock in the operating system may be removed, so as not to switch passively when the system clock is interrupted.

[0053] Some scenarios cannot be converted into the above-mentioned active switching scheduling scenarios, such as exception handling triggered by abnormal events and the resulting task scheduling, partition time window interrupts introduced to achieve time partitioning in a partitioned environment (for example, the ARINC653 operating system partition environment), and rescheduling of tasks within the partition after the interruption.

[0054] For the above scenario, in some embodiments, the synchronization logic unit 132 may receive an indication that an asynchronous event has occurred. The synchronization logic unit 132 may receive key data associated with the processing channel 150 executing the task from the loose lock step logic component 170 of the processing channel 150, and may send the key data associated with the processing channel 110 executing the task to the synchronization logic unit 172 of the loose lock step logic component 170 of the processing channel 150. The synchronization logic unit 132 may compare the key data associated with the processing channel 110 executing the task with the received key data, and based on the result of the comparison, the processing channel 110 and the processing channel 150 adjust the execution of the task until the two compared key data are consistent. The synchronization logic unit 132 may indicate that the execution of the task by the processing channel 110 and the processing channel 150 is synchronized. For this type of passive scheduling, when an asynchronous event arrives, the processing channel 110 and the processing channel 150 respectively send their own key data to each other and wait for synchronization, and determine which party executes faster / slower based on the comparison of the key data. The slower executing party may continue to execute after obtaining the other party's key data until its own key data is consistent with the other party's key data, thereby completing synchronization with the other party, for example, by sending a synchronization signal.

[0055] In some embodiments, the critical data may include global variables that play a key role in execution point synchronization, modification times or update times of inter-process communication objects, etc.

[0056] In addition, in order to ensure that the processing channel 110 and the processing channel 150 can obtain the same system time, in some embodiments, the first processing channel 110 and the processing channel 150 can use the same clock source. In some embodiments, the synchronization logic unit 132 can also provide the same time value to the read application programming interface (API) of the processing channel 110 while the synchronization logic unit 172 of the loose lockstep logic 170 of the processing channel 150 provides the time value to the read API of the processing channel 150 in response to determining that the processing channel 110 and the processing channel 150 are both executing the code requesting to read the system time.

[0057] In some embodiments, the lock-step logic component 130 may include a non-volatile storage controller 133 and an I / O controller 134. The non-volatile storage controller 133 may be connected to the non-volatile storage component 143, and the I / O controller 134 may be connected to the I / O component 144. The non-volatile storage controller 133 may provide support for the CPU 120 to access the non-volatile storage component 143 through a PCIe link. The I / O controller 134 may provide support for the CPU 120 to access the I / O component 144. The I / O controller 134 may provide logic control capabilities such as a serial interface, an Ethernet interface, an SPI interface, an I2C interface, etc.

[0058] Although a non-volatile storage controller and an I / O controller may also exist in the CPU 120, if the non-volatile storage controller and the I / O controller inside the CPU are directly used to connect to the non-volatile storage component 143 and the I / O component 144, the internal states of the non-volatile storage component 143 and the I / O component 144 of the processing channel 110 and the non-volatile storage component 183 and the I / O component 184 of the processing channel 150 may be inconsistent, resulting in inconsistencies in the interaction process between the corresponding non-volatile storage component and the I / O component and the corresponding CPU, which will cause inconsistencies in the execution paths of the CPU 120 and the CPU 160. By setting the non-volatile storage controller 133 and the I / O controller 134 in the loose lock-step logic unit 130 and the non-volatile storage controller 173 and the I / O controller 174 in the loose lock-step logic unit 170, the CPU 120 of the processing channel 110 and the CPU 160 of the processing channel 150 will both issue read and write commands during read and write operations, but only the corresponding processing channels 110 and 150 operate the corresponding non-volatile storage units 143 and 183 and I / O units 144 and 184. The I / O data or non-volatile storage data read in the processing channel 110 will be shared with the processing channel 150, and vice versa.

[0059] In some embodiments, the lock-step logic component 130 may further include a network access logic unit 135. The network access logic unit 135 may be used to access the computing module 100 to a network (e.g., an aviation network). As an example, depending on the type of network, the network access logic unit 135 may be implemented as an end system of ARINC664 or an end system of a TSN network.

[0060] The loose lock-step logic component 170 of the processing channel 150 operates in a similar manner to the loose lock-step logic component 130 .

[0061] In some embodiments, the comparison logic unit (e.g., comparison logic unit 131, 171) and the synchronization logic unit (e.g., synchronization logic unit 132, 172) in the loose lock-step logic component (e.g., loose lock-step logic component 130, 170) in each processing channel (e.g., processing channel 110, 150) can be disabled, so that the comparison and synchronization functions of the loose lock-step logic component are turned off. In this case, each processing channel can be used as a data processing module that runs independently of each other, thereby supporting common integrity calculations. The ability to host more applications can be provided under this configuration.

[0062] Figure 2 is a block diagram illustrating a computation module 200 for high integrity computation according to one embodiment of the present invention. Figure 2 Further details are shown for comparing the read and write commands and data of the corresponding CPU 220 to the non-volatile storage controller 243 and the I / O controller 244. Figure 2 Some components of the loose lockstep logic components 230 , 270 are not shown.

[0063] The computing module 200 may include a processing channel 210 and a processing channel 250. The processing channel 210 may include a CPU 220 and a loose lock-step logic component 230 connected to the CPU 220. The processing channel 210 may also include a memory component 241 and a ROM 242 connected to the CPU 220, and a non-volatile storage component 243 and an I / O component 244 connected to the loose lock-step logic component 230.

[0064] The loose lock-step logic component 230 may include a non-volatile storage controller 233 and an I / O controller 234. In some embodiments, the loose lock-step logic component 230 may also include a command queue 236 and a read data queue 237. The command queue 236 may be used to store read and write commands issued by the CPU 220 and data that needs to be written in the case of a write command. The read data queue 237 may be used to store the results of the read command. As an example, if the read command is executed successfully, the result of the read command may include the data that is successfully read and a flag indicating that the read command is executed successfully; if the read command fails to execute, the result of the read command may include a flag indicating that the read command failed to execute and the reason for the failure. The results of the read command stored in the read data queue 237 may reflect the content read from the non-volatile storage component 243 and the I / O component of the processing channel 210, and may be shared to the read data queue 277 in the loose lock-step component 270 of the processing channel 250. Accordingly, the results of the corresponding read data in the processing channel 250 may be shared with the read data queue 237 by the read data queue 277. In Figure 2 The dashed arrow in FIG. 2 indicates the flow of the read command and the associated data issued by the CPU 220 to the nonvolatile storage component 243 .

[0065] In some embodiments, the comparison logic unit (not shown) of the loose lock-step logic component 230 may compare the commands in the command queue 236 with the commands in the command queue 276 of the loose lock-step logic 270 (e.g., comparing them one by one). If the result of the comparison is inconsistent, the loose lock-step logic component 230 may trigger a comparison exception handling mechanism (e.g., reporting an error). The comparison may be as follows: Figure 2 As indicated by the thick arrow in .

[0066] Figure 2 Also shown are the command queue and read data queue for the I / O controller 234, which are related to the process associated with the CPU 220 issuing read and write operations to the I / O component 244 through the I / O controller 234. This process is similar to the process described above in conjunction with the volatile storage controller 233.

[0067] Other aspects of the computing module 200 may be combined with the above Figure 1 The computing module 100 described is similar.

[0068] Figure 3 is a block diagram illustrating a computation module 300 for high integrity computation according to one embodiment of the present invention. Figure 3 Further details are shown for comparing the read and write commands and data of the corresponding CPU 320, 360 to the network access logic unit 335, 375. Figure 3 Some components of the loose lockstep logic components 330 , 370 are not shown.

[0069] The computing module 300 may include a processing channel 310 and a processing channel 350. The processing channel 310 may include a CPU 320 and a loose lockstep logic component 330 connected to the CPU 320. The processing channel 310 may also include a memory component 341 and a ROM 342 connected to the CPU 320. The processing channel 350 may include a CPU 360 and a loose lockstep logic component 370 connected to the CPU 360. The processing channel 350 may also include a memory component 381 and a ROM 382 connected to the CPU 360.

[0070] The loose lock-step logic component 330 may include a network access logic unit 335. The loose lock-step logic component 370 may include a network access logic unit 375. The network access logic units 335 and 375 may be used to access the computing module 100 to a network (e.g., an aviation network). As an example, depending on the type of network, the network access logic units 335 and 375 may be implemented as a segment system of ARINC664 or an end system of a TSN network.

[0071] The loose lock-step logic component 330 may further include a network command queue 338 and a network input data queue 339. Similarly, the loose lock-step logic component 370 may further include a network command queue 378 and a network input data queue 379. Figure 3 The dotted arrows indicate the flow of read commands and associated data issued by CPU 320 and CPU 360 to the network access logic unit 335, and the thick arrows indicate the comparison of content items in the network command queue 338 with those in the network command queue 378 and the comparison of content items in the network input data queue 339 with those in the network input data queue 379.

[0072] In some embodiments, a comparison logic unit (not shown) of the lock-step logic component 330 may be configured to trigger a comparison exception handling mechanism in response to any of the above comparison results being inconsistent.

[0073] Other aspects of the computing module 300 may be combined with the above Figure 1 The computing module 100 described is similar.

[0074] According to another aspect of the present invention, a method for high integrity computing is provided.

[0075] Figure 4 4 is a flowchart showing a method 400 for high integrity computing according to an embodiment of the present invention. The method 400 can be, for example, Figure 1-Figure 2 The described computing modules 100, 200, 300 are used to execute.

[0076] Method 400 may include: at block 410, synchronizing the execution of the same task by the first processing channel and the second processing channel. More specifically, this may include synchronizing the execution paths and output results of the execution of the same task by the first processing channel and the second processing channel while ensuring that no errors occur in the first processing channel and the second processing channel. In some embodiments, block 410 may be performed by Figure 1 The synchronization logic unit 132, Figure 2 The synchronization logic unit in Figure 3 The synchronization logic unit in is executed.

[0077] The method 400 may include, at block 420, determining whether execution of the same task by the first processing channel and the second processing channel is synchronized.

[0078] If it is determined at block 420 that execution is synchronized, method 400 may proceed to block 430 by comparing the first command and first data transmitted into the first processing channel with the second command and second data transmitted into the second processing channel, respectively. In some embodiments, block 410 may be performed by Figure 1 The comparison logic unit 131 in Figure 2 The comparison logic unit in Figure 3 The comparison logic unit in is executed.

[0079] If it is determined at block 420 that execution has not yet been synchronized, method 400 may return to block 410 until execution has been synchronized.

[0080] The method 400 may include, at block 440 , determining whether the result of the comparison is consistent.

[0081] If the result of the comparison is determined to be consistent at block 440, method 400 may proceed to block 450: continue to execute the task normally. On the contrary, if the result of the comparison is determined to be inconsistent at block 440, method 400 may proceed to block 460: trigger the comparison exception handling mechanism to capture and report errors occurring in any one processing channel.

[0082] Figure 5 is a flow chart illustrating a process 500 for facilitating synchronization of task execution according to one embodiment of the present invention. Figure 5 The process shown in can be Figure 4 The specific implementation method of box 410 of method 400 in FIG.

[0083] In some embodiments, facilitating synchronization of execution of the same task by a first processing channel and a second processing channel may include: at block 510, determining whether execution of the task by the first processing channel precedes or follows execution of the task by the second processing channel. If execution of the task by the first processing channel precedes execution of the task by the second processing channel, process 500 may proceed to block 520: switching a first CPU of the first processing channel to a wait state. If execution of the task by the first processing channel follows execution of the task by the second processing channel, process 500 may proceed to block 530: switching a second CPU of the second processing channel to a wait state.

[0084] Figure 6 is a flow chart illustrating a process 600 for facilitating synchronization of task execution according to one embodiment of the present invention. Figure 6 The process shown in can be Figure 4 The specific implementation of block 410 of method 400 in Figure 1 The synchronization logic unit 132, Figure 2 The synchronization logic unit in Figure 3 The synchronization logic unit in is executed.

[0085] Process 600 may include, at block 610 , receiving an indication that an asynchronous event has occurred.

[0086] Process 600 may include, at block 630, receiving, from a second processing channel, second critical data associated with a task performed by the second processing channel.

[0087] Process 600 may include, at block 650, comparing first critical data associated with a first processing channel execution task with second critical data.

[0088] Process 600 may include, at block 670, causing the first processing channel and the second processing channel to adjust execution of the task based on a result of the comparison until the first critical data is consistent with the second critical data.

[0089] Process 600 may include, at block 690, indicating that execution of the task by the first processing channel and the second processing channel is synchronized.

[0090] Non-limiting embodiments of the present invention have been described in conjunction with the accompanying drawings. The embodiments are illustrative and non-limiting.

Claims

1. A computing module for high integrity computing, comprising: at least two processing channels, including a first processing channel and a second processing channel, each processing channel is used to perform the same task, Wherein, the first processing channel comprises: a first central processing unit CPU; and A first loose lock-step logic component is connected to the first CPU, and the first loose lock-step logic component includes: A first synchronization logic unit is configured to: Implementing synchronization of execution of the same task by the first processing channel and the second processing channel; and The first comparison logic unit is used for: In response to determining that the first processing channel and the second processing channel are performing the same task synchronously, the first command and the first data transmitted to the first loose lock-step logic component are respectively transmitted to the second processing channel. comparing a second command of a second loose lock-step logic component with second data; In response to the comparison result being inconsistent, a comparison exception handling mechanism is triggered.

2. The computing module according to claim 1, characterized in that: If the first processing channel executes the same task before the second processing channel, the first CPU switches to a waiting state; If the first processing channel executes the same task later than the second processing channel, the second CPU of the second processing channel switches to a waiting state.

3. The computing module according to claim 1, wherein: The first synchronization logic unit is used for: receiving an indication that an asynchronous event has occurred; receiving second critical data associated with a task executed by the second processing channel from a second synchronization logic unit of a second loose lockstep logic component of the second processing channel; comparing first key data associated with the first processing channel execution task with the second key data; Based on the result of comparing the first key data with the second key data, the first processing channel and the second processing channel adjust the execution of the task until the first key data is consistent with the second key data; as well as Indicates that the execution of the same task by the first processing channel and the second processing channel is synchronized.

4. The computing module according to claim 3, characterized in that: The first key data and the second key data each include: a global variable that plays a key role in execution point synchronization, and a modification count or update count of an inter-process communication object.

5. The computing module according to claim 1, wherein: The first processing channel and the second processing channel use the same clock source.

6. The computing module according to claim 1, wherein: The first synchronization logic unit is used for: In response to determining that both the first processing channel and the second processing channel are executing code requesting to read the system time, while the second synchronization logic unit of the second loose lockstep logic component of the second processing channel provides the time value to the read application programming interface API of the second processing channel, the time value is provided to the read API of the first processing channel.

7. The computing module according to claim 1, characterized in that: The first loose lock-step logic component further comprises: A first input / output I / O controller connected to the first I / O component; and A first non-volatile storage controller is connected to the first non-volatile storage component. The first command includes: a read and write command from the first CPU to the first I / O component, and / or a read and write command to the first non-volatile storage component, and The first data includes: data read from the first I / O component and / or data read from the first non-volatile storage component.

8. The computing module according to claim 7, characterized in that: The first loose lock-step logic component further comprises: a first command queue for storing the one or more commands and data to be written associated with a write command among the one or more commands; and The first read data queue is used to store the result of the read command in the one or more commands.

9. The computing module according to claim 8, characterized in that: The first comparison logic unit is used for: comparing the one or more commands in the first command queue with corresponding commands in a second command queue of the second loose lockstep logic component of the second processing channel; as well as In response to determining that the result of the command comparison is inconsistent, the comparison exception handling mechanism is triggered.

10. The computing module according to claim 9, characterized in that: The first comparison logic unit is used for: In response to determining that the result of the command comparison is inconsistent, the comparison context information is saved and the one or more commands are prevented from being executed.

11. The computing module according to claim 8, characterized in that: The results of the read command include: The successfully read data and a flag indicating that the read command was successfully executed; and A flag indicating that the read command failed to execute and the reason for the failure.

12. The computing module according to claim 1, wherein: The first loose lock-step logic component further comprises: A network access logic unit is used to connect the first processing channel to the network.

13. The computing module according to claim 12, characterized in that: The first command comprises a first network command from the first CPU to the network, and The first data comprises first network input data read from the network, The first loose lock-step logic component further comprises: A first network command queue, used to store the first network command; and The first network input data queue is used to store the first network input data.

14. The computing module according to claim 13, characterized in that: The first comparison logic unit is used for: comparing the first network command in the first network command queue with a corresponding second network command in a second network command queue of the second loose lockstep logic component of the second processing channel; comparing the first network input data in the first network input data queue with corresponding second network input data in the second network input data queue of the second loose lock-step logic component of the second processing channel; as well as In response to determining that the results of the network command comparison are inconsistent and / or the results of the network input data comparison are inconsistent, the comparison exception handling mechanism is triggered.

15. The computing module according to claim 1, wherein: Both the synchronization logic unit and the comparison logic unit of the loose lockstep logic unit of each processing channel can be disabled.

16. A method for high integrity computing, comprising: Implementing synchronization of execution of the same task by the first processing channel and the second processing channel; In response to determining that the execution of the same task by the first processing channel and the second processing channel is synchronized, respectively comparing a first command and first data transmitted to the first processing channel with a second command and second data transmitted to the second processing channel; In response to the comparison result being inconsistent, a comparison exception handling mechanism is triggered.

17. The method according to claim 16, characterized in that Implementing synchronization of execution of the same task by the first processing channel and the second processing channel includes: If the first processing channel executes the same task before the second processing channel, switching the first general processing unit CPU of the first processing channel to a waiting state; If the first processing channel executes the same task later than the second processing channel, the second CPU of the second processing channel switches to a waiting state.

18. The method according to claim 16, characterized in that Implementing synchronization of execution of the same task by the first processing channel and the second processing channel includes: receiving an indication that an asynchronous event has occurred; receiving, from the second processing channel, second critical data associated with a task performed by the second processing channel; comparing first key data associated with the first processing channel execution task with the second key data; Based on the result of comparing the first key data with the second key data, causing the first processing channel and the second processing channel to adjust the execution of the task until the first key data is consistent with the second key data; and Indicates that the execution of the task by the first processing channel and the second processing channel is synchronous.