Firmware mechanism for correcting soft errors

A method for handling soft errors in a processor capable of operating first and second execution cores in redundant and split modes, the method comprising:

• detecting a soft error when the processor is operating in the redundant mode;
• executing an error recovery routine on each execution core to save uncorrupted data from storage structures associated with the first and second execution cores, wherein the error recovery routine is stored in a non-volatile memory, the error recovery routine includes switching the processor from redundant mode to split mode and saving uncorrupted data from each execution core to an uncacheable portion of a main memory system;
• and
• recovering processor state data from the saved uncorrupted data.

A machine readable medium on which are stored instructions that are executable by a dual execution core processor to implement a method for recovering from soft errors, the method comprising:

• switching the processor to operate the dual execution cores independently when a soft error is detected;
• executing an error recovery routine on each of the execution cores to recover a minimum set of processor state data from uncorrupted processor state data;
• initializing each of the dual execution cores, using the minimum set of processor state data, wherein initializing each of the dual execution cores comprises switching the processor to operate the dual execution cores in lock step and copying the minimum set of processor state data to each of the dual execution cores;
• and
• resuming execution of an interrupted program thread.

A computer system, comprising:

• a processor having a first execution core and a second execution core, the first and second execution cores operate in lock step when the processor is in a redundant execution mode and operate independently when the processor is in a split execution mode;
• and
• a non-volatile memory in which are stored instructions that may be implemented by the processor to implement a method for recovering from soft errors when the processor is in redundant execution mode, the method comprising
• switching the processor to split execution mode;
• saving uncorrupted processor state data from each execution core to a designated memory location;
• recovering a minimal set of processor state data from the saved, uncorrupted processor state data, wherein recovering the minimal set of processor state data comprises replacing corrupted processor state data from one of the execution cores with uncorrupted processor state data saved from the other of the execution cores;
• and
• initializing each execution core with the minimal set of processor state data.

A computer system, comprising:

• a processor having a first execution core and a second execution core, the first and second execution cores operate in lock step when the processor is in a redundant execution mode and operate independently when the processor is in a split execution mode, wherein each execution core includes first and second execution clusters that operate in lock step when the processor is in split mode;
• and
• a non-volatile memory in which are stored instructions that may be implemented by the processor to implement a method for recovering from soft errors when the processor is in redundant execution mode, the method comprising
• switching the processor to split execution mode;
• saving uncorrupted processor state data from each execution core to a designated memory location;
• recovering a minimal set of processor state data from the saved, uncorrupted processor state data;
• and
• initializing each execution core with the minimal set of processor state data.