International Business Machines Corporation

George C Wellwood

Patrick James Meaney

Kevin W Kark

William Wu Shen

Luis A Lastras

Lawrence D Curley

Eldee Stephens

Russell W Lavallee

Kenneth Y Chan

Donald G O'Brien

Michael Rubino

Judy S Chen Johnson

Embodiments relate to stale data detection in a marked channel for a scrub. An aspect includes bringing the marked channel online, wherein the computer comprises a plurality of memory channels comprising the marked channel and a remaining plurality of unmarked channels. Another aspect includes performing a scrub read of an address in the plurality of memory channels. Another aspect includes determining whether data returned by the scrub read from the marked channel is valid or stale based on data returned from the unmarked channels by the scrub read. Another aspect includes based on determining that the data returned by the scrub read from the marked channel is valid, not performing a scrub writeback to the marked channel. Another aspect includes based on determining that the data returned by the scrub read from the marked channel is stale, performing a scrub writeback of corrected data to the marked channel.

Stale data detection in marked channel for scrub

Embodiments include a combined rank and linear memory address incrementing utility. An aspect includes an address incrementing utility suitable for implementation within a memory controller as an integrated subsystem of a central processing unit (CPU) chip. In this type of on-chip embodiment, the address incrementing utility utilizes dedicated hardware, chip-resident firmware, and one or more memory address configuration maps to enhance processing speed, efficiency and accuracy. The combined rank and linear memory address incrementing utility is designed to efficiently increment through all of the individual bit addresses for a large logical memory space divided into a number of ranks on a rank-by-rank basis. The address incrementing utility sequentially generates all of the sequential memory addresses for a selected rank, and then moves to the next rank and sequentially generates all of the memory addresses for that rank, and so forth until of the ranks have been processed.

Combined rank and linear address incrementing utility for computer memory test operations

A self test function in the Memory Controller is utilized to generate unique and continuous data patterns for each of the words which are stored into two consecutive DRAM addresses in two spaced store operations. The self test function then generates fetch commands to read back the unique data patterns from the two DRAM addresses. In the fetch operations, the data transmission for each operation and between both operations is contiguous (no gaps). A self test data comparison function is then used to compare these fetched data words to data patterns which are generated from the self test data generator. Bit error counters from the memory controller keeps track of any miscompares. By reading out a unique signature from these bit counters, it can be determined whether the store path data are misaligned early or late or correct and/or the fetch path data are misaligned early or late or correct. In addition, the exact number of cycles the data are early or late is known. Based on the last results, either or both the store and/or fetch data path capture window parameters are adjusted to correct or early or late bit position.

Data capture window synchronizing method for generating data bit sequences and adjusting capture window on parallel data paths

Memory is scrubbed by an improved non-linear method giving scrubbing preference to the central storage region having the characteristic of a high risk read-only memory such as the CPA region to prevent the accumulation of temporary data errors. The chip row on which the CPA resides is scrubbed after each time the scrubbing of a non-CPA chip row in a PMA completed successfully. The next non-CPA least recently scrubbed chip row would be selected for scrubbing after scrubbing completed on the CPA chip row. This in a first case provides non-linear selection methods of scrubbing central storage of computer systems to more frequently select ('select' herein encompasses the meaning of 'favor') scrub regions having the characteristic of a predominately read-only memory making those regions at a higher risk of failure than those regions having lower risk because of frequent write operations. In a second case, scrub regions having the characteristic of a predominately read-only memory are selected by using a second preferred embodiment selection method which uses the detection of faulty data from normal system accesses to central storage to identify other high risk regions and scrub them before other lower risk regions. In addition, the severity of the detected data error can be used to determine the rate at which scrub commands are sent to the selected region: the higher the severity, the higher the scrub rate.

Method for scrubbing regions in central storage

A memory request protocol allows a memory request to be withdrawn or 'cancelled' without penalty so no memory resource is wasted in doing so during an assigned 'cancel window'. When the memory card starts to process a command from the memory controller, for a predefined number of cycles a period of time is available where the memory card can't accept another command due to a resource conflict. This provides an opportunity to re-balance requests to the memory controller in this period of time or 'cancel window'.

Memory request protocol method

A computer memory maintainence apparatus tests operating system storage and identifies a malfunctioning memory chip in an on-line memory array by detecting and recording all permanent data errors using data comparison along with data complementation and substitutes a spare memory chip for the malfunctioning one for all memory read commands. All write commands are performed on both spare memory and the malfunctioning memory chip. All contents of defective chip are copied to the spare chip. The computer system maintains the scrubbing and a recording counter for each of the data bits in an ECC memory data word. The sparing logic in the memory storage system maintains the bit steering logic and controls for the spare chip. When a counter is incremented above a threshold sparing is invoked to replace the failing bit position. The system writes to the defective and spare chips in parallel even after bit steering is invoked.

Growing community of inventors

Poughkeepsie, NY, United States of America

George C Wellwood