Error detection scheme in a multiprocessor environment

• US 5,428,766 A
• Filed: 12/01/1992
• Issued: 06/27/1995
• Est. Priority Date: 12/01/1992
• Status: Expired due to Term

• Current Assignee: Enterasys Networks Incorporated (Extreme Networks, Inc.)
• Original Assignee: Digital Equipment Corporation (HP Inc.)
• Inventors: Seaman, Michael J.
• Primary Examiner(s): Beausoliel, Jr., Robert W.
• Assistant Examiner(s): Palys, Joseph E.
• Application Number:
  US07/983,907
• Time in Patent Office: 938 Days
• Field of Search: 395/575, 395/200, 364/265.1, 364/267, 364/267.1, 364/253, 364/253.1, 364/253.3, 364/264.6, 371/20.6, 371/40.1, 371/51.1, 371/20.1
• US Class Current: 709/215
• CPC Class Codes: G06F 11/0724 in a multiprocessor or a multi-core unit multiprocessors per se G06F15/80
• CPC Class Codes: G06F 11/0751 Error or fault detection not based on redundancy power supply failures G06F1/30; network fault management H04L41/06
• CPC Class Codes: G06F 11/349 for interfaces, buses

Abstract

An error detection scheme to detect a variety of errors, including buffer accesses errors, buffer ownership transfer errors, and address recognition engine access errors, that may occur during the passing of messages between processors in a multi-processor computer system implementing a buffer swapping scheme. The error detection scheme of the present invention provides for the monitoring of bus transactions, maintaining a log of bus activity including buffer access transactions, identifying transactions involving buffer and address recognition operations and checking those operations to insure that they are consistent with the implemented buffer swapping scheme. Upon detection of an error the bus monitoring device asserts an error signal, freezes the log of bus activity and halts buffer swapping activity until the detected error is investigated and dealt with in an appropriate manner.

Claims

1. A computer system, which comprises:

a plurality of source devices, each one of the plurality of source devices being a source of messages;

a plurality of destination devices, each one of the plurality of destination devices being a destination of messages;

a common memory comprising a plurality of buffers for storing messages from any one of the plurality of source devices for passing to any of the plurality of destination devices,

a system bus for coupling each one of the plurality of source and destination devices to each other and to the common memory;

a pool of pointers to buffers of the common memory that are available for use to store messages;

a plurality of pointer memories coupled to each one of the plurality of source and destination devices, each one of the plurality of pointer memories being associated with a particular pair of source and destination devices,

each pointer memory having a set of locations for storing pointers to buffers of the common memory; and

one particular source device of the plurality of source devices transferring a message to one particular destination device of the plurality of destination devices by writing the message to one of the plurality of buffers, and writing the pointer to the one of the plurality of buffers to a preselected location of the pointer memory associated with the particular source and destination device pair,

the particular destination device reading the pointer to the one of the plurality of buffers from the preselected location of the pointer memory;

a buffer interchange observer coupled to the plurality of source and destination devices and to the central memory by the system bus, the buffer interchange observer comprising a pointer memory protocol monitor monitoring accesses to the pointer memories and checking that only the particular source and destination devices associated with a particular pointer memory access the particular pointer memory.

2. The computer system of claim 1, wherein the buffer interchange observer further comprises a master index machine identifying any device initiating a pointer memory access operation and supplies the pointer memory protocol monitor with device identification information identifying the device initiating the pointer memory access operation.

3. The computer system of claim 2, wherein the buffer interchange observer further comprises a memory pointer identification machine including a pointer memory information table containing information on each pointer memory including the particular source and destination device associated with each pointer memory location,

the memory pointer index machine identifying any source or destination device accesses to pointer memory locations and identifying the particular source and destination devices associated with the pointer memory being accessed.

4. The computer system of claim 3 wherein the memory pointer protocol monitor receives the device identification information from the master index machine and the identity of the source and destination device associated with the accessed pointer memory from the memory pointer identification machine, the pointer memory protocol monitor checking that the device identified by the master index machine is either the source or destination processor associated with the particular pointer memory being accessed.

5. The computer system of claim 4, wherein the memory pointer protocol monitor asserts an error signal in response to an access of a particular pointer memory by a device not associated with the particular pointer memory.

6. The computer system of claim 4, wherein each location of the pointer memory includes an ownership field having a first state and a second state, the first state indicating that the corresponding location is owned by the source processor, and stores a pointer to a buffer that is owned by the source device associated with the particular pointer memory, the stored pointer being valid for use by the source device to store a message, and the second state indicating that the corresponding destination device owns the particular pointer memory location and that the corresponding location stores a pointer to a buffer owned by the destination device associated with the particular pointer location, the stored pointer being valid for use by the destination device to read a message at the buffer of the common memory pointed to by the pointer,

the particular source and destination devices associated with the particular pointer memory location operating to change the state of the ownership field as pointers to messages are exchanged between the source and destination devices;

the buffer interchange observer'"'"'s pointer memory protocol monitor checks that each device changes the state of the ownership field associated with a particular pointer memory location only after preforming a read of the particular memory location.

7. The computer system of claim 6, wherein the buffer interchange observer further comprises

a buffer rights table including information on buffer ownership, and

a buffer operations checker, the buffer operations checker receiving processor identification information from the master index machine,

the buffer operations checker using the information contained in the buffer rights table and the information received from the master index machine to check that a particular device accessing a particular buffer in common memory owns the particular buffer.

8. The computer system of claim 7, wherein the buffer operations checker receives pointer memory access information from the pointer memory index machine and checks that transfers of buffer ownership, via the changing of the state of the ownership field associated with a pointer memory location, are initiated only by devices owning the particular buffer at the time of the initiation of the ownership transfer.

9. The computer system of claim 8, wherein the buffer operations checker further comprises a logram logic block maintaining a log of transactions involving the buffers or pointer memories.

10. The computer system of claim 9 wherein the buffer interchange observer further comprises a control logic block freezing the log maintained by the logram logic block upon detection of a check failure.

11. The computer system of claim 2, wherein the buffer interchange observer further comprises

a buffer rights table including information on buffer ownership, and

a buffer operations checker, the buffer operations checker receiving processor identification information from the master index machine,

the buffer operations checker using the information contained in the buffer rights table and the information received from the master index machine to check that a particular device accessing a particular buffer in common memory owns the particular buffer.

12. The computer system of claim 11, further comprising

an access recognition engine coupled to the plurality of source and destination devices by the system bus, the access recognition engine including a request/response RAM for storing access recognition requests and responses to the requests,

the request/response RAM including a plurality of memory locations, each memory location being associated with a particular one of the plurality of source devices;

and

wherein the buffer interchange observer further comprises an access recognition engine checker checking that any device accessing a particular request/response RAM memory location is the particular source device associated with the particular request/response RAM memory location.

13. The computer system of claim 12, wherein the plurality of pointer memories is coupled to each one of the plurality of source and destination devices, the common memory and to the buffer interchange observer by the system bus.

14. The computer system of claim 13, wherein the buffer operations checker further comprises a logram logic block maintaining a log of bus transactions involving the buffer or pointer memory accesses.

15. The computer system of claim 14, wherein the buffer interchange observer further comprises a control logic block freezing the log of the logram logic block upon detection of a failed check.

16. The computer system of claim 15, further comprising a management processor card coupled to the bus, the management processor card initializing the buffer rights table and the memory pointer index machine'"'"'s pointer memory table upon power-up of the computer system.

17. The computer system of claim 16, wherein the buffer interchange monitor asserts an error signal over the bus upon detection of a failed check and the management processor card operates to investigate the cause of the error signal.

18. The computer system of claim 1, wherein the plurality of pointer memories is coupled to each one of the plurality of source and destination devices, the common memory and to the buffer interchange observer by the system bus.

19. The computer system of claim 1, wherein the buffer operations checker further comprises a logram logic block maintaining a log of transactions involving the buffers or pointer memories.

20. The computer system of claim 19 wherein the buffer interchange observer further comprises a control logic block freezing the log maintained by the logram logic block upon detection of a check failure.

21. The computer system of claim 1 further comprising,

an access recognition engine coupled to the plurality of source and destination devices, the access recognition engine including a request/response RAM for storing access recognition requests and responses to the requests,

the request/response RAM including a plurality of memory locations, each memory location being associated with a particular one of the plurality of source devices;

and

wherein the buffer interchange observer further comprises an access recognition engine checker checking that any device accessing a particular request/response RAM memory location is the particular source device associated with the particular request/response RAM memory location.

22. A computer system, which comprises:

a plurality of source devices, each one of the plurality of source devices being a source of messages;

a plurality of destination devices, each one of the plurality of destination devices being a destination of messages;

a common memory comprising a plurality of buffers for storing messages from any one of the plurality of source devices for passing to any of the plurality of destination devices,

a system bus for coupling each one of the plurality of source and destination devices to each other and to the common memory;

a pool of pointers to buffers of the common memory that are available for use to store messages;

an access recognition engine coupled to the plurality of source and destination devices by the system bus, the access recognition engine including a request/response RAM for storing access recognition requests and responses to the requests,

the request/response RAM including a plurality of memory locations, each memory location being associated with a particular one of the plurality of source devices;

a plurality of pointer memories coupled to each one of the plurality of source and destination devices, each one of the plurality of pointer memories being associated with a particular pair of source and destination devices,

each pointer memory having a set of locations for storing pointers to buffers of the common memory; and

one particular source device of the plurality of source devices accessing the address recognition engine to determine which one particular destination device of the plurality of destination devices to transfer a message to,

the one particular source device transferring the message to the one particular destination device by writing the message to one of the plurality of buffers, and writing the pointer to the one of the plurality of buffers to a preselected location of the pointer memory associated with the particular source and destination device pair,

the particular destination device reading the pointer to the one of the plurality of buffers from the preselected location of the pointer memory; and

an access recognition engine checker coupled to the plurality of source and destination devices and the common memory by the system bus, the access recognition engine checker checking that any device accessing a particular request/response RAM memory location is the particular source device associated with the particular request/response RAM memory location.

23. A computer system, which comprises:

a plurality of source devices, each one of the plurality of source devices being a source of messages;

a plurality of destination devices, each one of the plurality of destination devices being a destination of messages;

a common memory comprising a plurality of buffers for storing messages from any one of the plurality of source devices for passing to any of the plurality of destination devices,

a system bus for coupling each one of the plurality of source and destination devices to each other and to the common memory;

a pool of pointers to buffers of the common memory that are available for use to store messages;

a plurality of pointer memories coupled to each one of the plurality of source and destination devices, each one of the plurality of pointer memories being associated with a particular pair of source and destination devices,

each pointer memory having a set of locations for storing pointers to buffers of the common memory,

each location of the pointer memory including an ownership field having a first state and a second state, the first state indicating that the corresponding location is owned by the source processor, and stores a pointer to a buffer that is owned by the source device associated with the particular pointer memory, the stored pointer being valid for use by the source device to store a message, and the second state indicating that the corresponding destination device owns the particular pointer memory location and that the corresponding location stores a pointer to a buffer owned by the destination device associated with the particular pointer location, the stored pointer being valid for use by the destination device to read a message at the buffer of the common memory pointed to by the pointer,

the particular source and destination devices associated with the particular pointer memory location operating to change the state of the ownership field as pointers to messages are exchanged between the source and destination devices;

one particular source device of the plurality of source devices transferring a message to one particular destination device of the plurality of destination devices by writing the message to one of the plurality of buffers, and writing the pointer to the one of the plurality of buffers to a preselected location of the pointer memory associated with the particular source and destination device pair,

the particular destination device reading the pointer to the one of the plurality of buffers from the preselected location of the pointer memory;

a buffer interchange observer coupled to the plurality of source and destination devices and to the central memory by the system bus, the buffer interchange observer comprising a buffer rights table including information on buffer ownership, and

a buffer operations checker, the buffer operations checker using the information contained in the buffer rights table and to check that a particular device accessing a particular buffer in common memory owns the particular buffer.

24. A method of detecting errors in a multiprocessor computer system implementing a buffer swapping scheme for communicating messages between a plurality of source and destination devices, the computer system having a common memory comprising a plurality of buffers coupled to the plurality of source and destination devices by a system bus, the method comprising the steps of:

creating a pool of pointers to the buffers in the common memory that are available for use to store messages;

associating each one of a plurality of pointer memories with a particular pair of the plurality of source and destination devices, each pointer memory having a set of locations for storing pointers to the buffers of the common memory;

storing a message from a particular one of the plurality of source devices, for passing to a particular one of the plurality of destination devices, in one of the plurality of buffers of the common memory;

writing the pointer to the one of the plurality of buffers to a preselected location of the pointer memory associated with the particular source and destination device pair; and

monitoring accesses to the pointer memories and checking that only the particular source and destination devices associated with a particular pointer memory access the particular pointer memory.