Back to results
US 7,353,433 B2Granted
Poisoned error signaling for proactive OS recovery
AssigneeIntel Corporation
InventorsKushagra Vaid, Suresh Marisetty, Yaron Shragai, Koichi Yamada, Rajendra Kuramkote, Scott Brenden
Life of this patent
Estimating the term…
Abstract
Use of data poisoning techniques may permit proactive operating system recovery without needing to always bringing down the operating system when uncorrectable errors are encountered.
Claims20
A method of addressing data errors in a computer system, comprising:
- pre-determining a software-programmable data poisoning policy to control actions to be taken based on different classes of data poisoning events by a user;
- error-checking a unit of data by an error-control decoder;
- detecting an uncorrectable error in the unit of data by the error-control decoder;
- if the uncorrectable error is detected in the unit of data, based on the pre-determined data poisoning policy, determining if the detected uncorrectable error is a data poisoning event;
- marking the unit of data containing a data poisoning event with a software-visible bit by the error-control decoder which software-visible bit is a status bit to indicate to an operating system that the data unit contains the data poisoning event;
- determining, based on the pre-determined data poisoning policy, if the unit of data containing the poisoning event is to be acted upon;
- handing over the data units including the detected uncorrectable errors including data poisoning events from the error-control decoder to the operating system;
- detecting by the operating system whether the software-visible bit is present in the data unit;
- and
- based on the detected software-visible bit and the pre-determined data poisoning policy, acting, by the operating system of the computer system, to address the presence of the uncorrectable error in the unit of data, including:
- if the software-visible bit is detected, one of removing the marked data unit from a use by the operating system and recovering the data unit or determining if the unit of data is in user space and one of terminating an application running on the computer system and removing the unit of data from use by the operating system or shutting down the operating system, and
- if the software-visible bit is not detected, determining whether the data unit is in the user space and one of terminating the application running on the computer system and removing data unit from use by the operating system or shutting down the operating system.
9 more elements
A computer system comprising:
- a software-programmable data poisoning policy to control actions to be taken based on different classes of data poisoning events;
- a processor;
- at least one error control decoding implementation including an error-control decoder to implement error-control decoding in conjunction with the processor, adapted to process units of data and to detect if a unit of data contains at least one uncorrectable error;
- and
- a module to run on said processor to determine, based on the pre-determined data poisoning policy, if said uncorrectable error is a data poisoning event, and, if so, to mark unit of data containing said uncorrectable error and hand over the data units including the detected uncorrectable errors including data poisoning events from the error-control decoder to an operating system which runs on said processor, the operating system to determine, based on the pre-determined data poisoning policy, if a particular data poisoning event is to be acted upon or not, the operating system adapted to detect the marked unit of data and to act to mitigate the detected uncorrectable error without always bringing down the operating system upon detection of the marked unit of data.
2 more elements
A machine-accessible storage medium containing software code that, when read by a computer, causes the computer to perform a method comprising:
- by a user, pre-determining a software-programmable data poisoning policy to control actions to be taken based on different classes of data poisoning events;
- error-checking a unit of data by an error-control decoder;
- if an uncorrectable error is detected in the unit of data, based on the pre-determined data poisoning policy, determining if the detected uncorrectable error is a data poisoning event, and if so, marking the unit of data containing a data poisoning event with a software-visible bit by the error-control decoder;
- determining, based on the pre-determined data poisoning policy, if the unit of data containing poisoning event is to be acted upon, and if so, handing over the data units including the detected uncorrectable errors including data poisoning events from the error-control decoder to an operating system;
- detecting, by the operating system of the computer, the software-visible bit in the unit of data;
- and
- in accordance with the detected software-visible control bit and the pre-determined data poisoning policy, acting, by the operating system of the computer to address the presence of the uncorrectable error in the unit of data, wherein the operating system is not always brought down.
4 more elements
A computer system comprising:
- a processor;
- and
- machine-accessible storage medium to be coupled to the processor, the processor to access the machine-accessible storage medium and to execute software code stored on the machine-accessible storage medium, to cause the computer system to perform a method comprising:
- by a user, pre-determining a software-programmable data poisoning policy to control actions to be taken based on different classes of data poisoning events;
- error-checking a unit of data by an error-control decoder;
- if an uncorrectable error is detected in the unit of data, based on the pre-determined data poisoning policy, determining if the detected uncorrectable error is a data poisoning event, and if so, marking the unit of data containing a data poisoning event with a software-visible bit by the error-control decoder;
- determining, based on the pre-determined data poisoning policy, if the unit of data containing poisoning event is to be acted upon, and if so, handing over the data units including the detected uncorrectable errors including data poisoning events from the error-control decoder to an operating system and detecting, by the operating system of the computer system, the software-visible bit in the unit of data;
- and
- in accordance with the detected software-visible control bit and the pre-determined data poisoning policy, acting, by the operating system to address the presence of the uncorrectable error in the unit of data, wherein the operating system is not always brought down.
6 more elements
Dependencies are inferred from the claim text.