Computer system operation with corrected read data function

• US 5,263,032 A
• Filed: 06/27/1991
• Issued: 11/16/1993
• Est. Priority Date: 06/27/1991
• Status: Expired due to Term

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Digital Equipment Corporation (HP Inc.)
• Inventors: Mega, Christopher A., Porter, Brian, Myers, Russell L.
• Primary Examiner(s): Canney, Vincent P.
• Application Number:
  US07/721,959
• Time in Patent Office: 873 Days
• Field of Search: 371/40.2, 371/40.3, 371/40.4, 371/5.1, 371/10.2
• US Class Current: 714/764
• CPC Class Codes: G06F 11/1024 Identification of the type of error
• CPC Class Codes: G06F 11/106 Correcting systematically all correctable errors, i.e. scrubbing

Abstract

A computer system having a memory with an ECC function employs an improved method for handling corrected read data events, so transient errors caused by alpha particle hits in DRAMs may be distinguished from hard errors. When a corrected read data event occurs, a footprint defining its location is compared with previously-stored footprints to determine if this location has failed before. Also, a location showing a corrected read data event is "scrubbed" (data is read, corrected and rewritten) so transient error locations are removed. If another corrected read data event occurs for this same location, after scrubbing, then the location is assumed to have a hard fault, and so the page containing this location is replaced.

Claims

1. A method of operating a computer system comprising the steps of:

a) sending requests by a CPU to a memory to read selected data items stored in said memory with ECC codes;

b) retrieving said selected data items from said memory and checking each said data item using said ECC code, and, if a correctable error is detected, correcting said data item and signalling to said CPU that a corrected read data event has occurred;

c) storing by said CPU an identification print for a corrected read data event;

d) comparing an identification print, before storing, with previously-stored ones of said stored identification prints to see if a match exists.

e) scrubbing a location in said memory for which a corrected read data event occurred; and

f) moving a block of said memory to another location in said memory when a corrected read data event occurs having an identification print matching that of a previous identification print.

2. A method according to claim 1 wherein said CPU maintains a virtual memory system and said block is a page.

3. A method according to claim 2 where said step of moving a page includes marking said page where the corrected read data event occurred as a bad page in a stored list.

4. A method according to claim 3 including the step of checking said page before moving to determine if said page is relocatable.

5. A method according to claim 1 wherein said step of scrubbing is performed by said CPU rewriting data to said memory.

6. A method according to claim 1 wherein said step of scrubbing is performed by said memory without command from said CPU.

7. A method according to claim 1 wherein said identification print includes therewith an indication of whether or not said location has been scrubbed.

8. A method according to claim 1 wherein said identification print includes therewith an indication of how many of said corrected read data events have occurred at said location.

9. A method according to claim 1 including the step of updating by said CPU an error log of said identification prints.

10. A computer system comprising:

a) a CPU and a memory, the CPU having means for sending data items to the memory for storing identified by physical addresses and for sending requests to said memory to read selected ones of said data items;

b) the memory having ECC means for storing said data items along with an ECC code for each item and retrieving said data items and checking said data items against said ECC code to detect and correct any correctable errors occurring;

c) said memory having means for sending to said CPU an identification print of said physical address in memory of each said data item for which said correctable error is detected and corrected;

d) said CPU having means for storing said identification prints and comparing each one of said identification prints representing a correctable error, before storing, with previously-stored ones of said stored identification prints to see if a match exists;

e) means for scrubbing a location in said memory containing each said physical address for which a correctable error was detected; and

f) means for moving each page of said memory to another page location in said memory when a correctable error occurs having an identification print matching that of a previous correctable error.

11. A system according to claim 10 wherein said CPU maintains a virtual memory system and said block is a page.

12. A system according to claim 11 where said means for moving a page includes means for marking said page where the correctable error occurred as a bad page in a bad page list.

13. A system according to claim 12 including means for checking said page before moving to determine if said page is relocatable.

14. A system according to claim 10 wherein said scrubbing is performed by said CPU rewriting data to said memory.

15. A system according to claim 10 wherein said scrubbing is performed by said memory without command from said CPU.

16. A system according to claim 10 wherein said identification print includes therewith an indication of whether or not said location has been scrubbed.

17. A system according to claim 10 wherein said identification print includes therewith an indication of how many of said correctable errors have occurred at said location.

18. A system according to claim 10 including means for updating by said CPU an error log of said identification prints.

19. A method of operating a computer system comprising the steps of:

a) sending data items to memory by a CPU identified by physical addresses;

b) storing said data items in memory along with an ECC code for each item;

c) sending requests by said CPU to said memory to read selected ones of said data items;

d) retrieving said selected ones of said data item from said memory and checking said data items using said ECC code, and correcting each of said data items if a correctable error is detected;

e) storing by said CPU an identification print of said physical address in memory of each said data item for which said correctable error is detected and corrected;

f) comparing each one of said identification prints representing a correctable error, before storing, with previously-stored ones of said stored identification prints to see if a match exists;

e) scrubbing a location in said memory containing each said physical address for which a correctable error was detected; and

f) moving a block of said memory containing the physical address of a correctable error to another location in said memory when a correctable error occurs having an identification print matching that of a previous correctable error.

20. A method according to claim 19 wherein said CPU maintains a virtual memory system and said block is a page.

21. A method according to claim 20 where said step of moving a page includes marking said page where the corrected read data event occurred as a bad page in a stored table.

22. A method according to claim 21 including the step of checking said page before moving to determine if said page is relocatable.