Ddr memory error recovery

A memory controller, comprising:

• a command queue for receiving and storing memory access requests;
• and
• an arbiter coupled to said command queue for providing a sequence of memory commands to a memory channel;
• a memory interface queue coupled to said command queue for receiving and storing memory access requests;
• and
• a replay queue coupled to said memory interface queue for storing said sequence of memory commands to said memory channel, and continuing to store memory access commands that have not yet received responses from said memory channel, wherein:
• said memory controller remains in a normal state as long as it does not detect any errors, wherein in said normal state said memory interface queue continues to receive commands from said command queue that are picked by said arbiter;
• and
• in response to detecting an error, said memory controller enters a recovery state in which said replay queue replays at least one corresponding memory command by sending said at least one corresponding memory command to said memory interface queue.

A data processing system, comprising:

• a memory accessing agent for providing memory access requests;
• a memory system;
• and
• a memory controller coupled to said memory accessing agent and said memory system, comprising:
• a command queue for receiving and storing memory access requests;
• an arbiter coupled to said command queue for providing a sequence of memory commands to said memory system;
• and
• a replay queue for storing said sequence of memory commands to said memory system, and continuing to store memory access commands that have not yet received responses from said memory system, wherein:
• when a response indicates a completion of a corresponding memory command without any error, said replay queue removes said corresponding memory command without taking further action;
• and
• when a response indicates a completion of said corresponding memory command with an error, said replay queue replays at least said corresponding memory command.

A method, comprising:

• receiving and storing memory access requests;
• providing a sequence of memory commands to a memory channel from stored memory access requests;
• storing said sequence of memory commands to said memory channel, and continuing to store memory commands that have not yet received error-free responses from said memory channel;
• detecting whether an error occurred in responses received from said memory channel;
• remaining in a normal state in response to detecting no memory errors, and while in said normal state continuing to provide commands from among stored memory access requests;
• and
• entering a recovery state in response to detecting an error, and while in said recovery state replaying stored memory commands starting from a corresponding memory command at which said error occurred.