REGULATING ATOMIC MEMORY OPERATIONS TO PREVENT DENIAL OF SERVICE ATTACK

• US 20120072984A1
• Filed: 09/22/2010
• Published: 03/22/2012
• Est. Priority Date: 09/22/2010
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Burns, David W., Chappell, Robert S., Raghuvanshi, Pankaj, Raikin, Shlomo, Sade, Raanan, BAIR, MICHAEL S., Math, Prakash, Ong, Leslie A., Tucknott, Michael D., Yanover, Igor
• Granted Patent: US 8,516,577 B2
• Time in Patent Office: Days
• US Class Current: 726/22
• CPC Class Codes: G06F 9/526 Mutual exclusion algorithms

Abstract

In one embodiment, the present invention includes a method for identifying a termination sequence for an atomic memory operation executed by a first thread, associating a timer with the first thread, and preventing the first thread from execution of a memory cluster operation after completion of the atomic memory operation until a prevention window has passed. This method may be executed by regulation logic associated with a memory execution unit of a processor, in some embodiments. Other embodiments are described and claimed.

Claims

1. A processor comprising:

a first core to execute instructions of a first thread and a second thread, the first core having a memory execution unit to handle requests for memory operations by the first and second threads and a regulator logic to prevent the first thread from execution of a second atomic memory operation after completion of a first atomic memory operation until a time window has passed since completion of the first atomic memory operation.

2. The processor of claim 1, wherein the regulator logic further is to allow the first thread to execute the second atomic memory operation after the time window has passed.

3. The processor of claim 1, further comprising at least one lock breaker to execute a lock breaking action when a lock situation has been detected in the first core.

4. The processor of claim 3, wherein the regulator logic is to prevent execution of the second atomic memory operation so that the at least one lock breaker not initiate a lock breaking action responsive to a plurality of atomic memory operations issued by the first thread.

5. The processor of claim 1, wherein the regulator logic includes a plurality of trackers each associated with a thread executing on the first core, and wherein the regulator logic is to prevent issuance of the second atomic memory operation to the memory execution unit if a counter associated with the first thread is in execution of a count operation.

6. The processor of claim 5, wherein the regulator logic is to initiate the counter responsive to detection of a termination sequence of the first atomic memory operation.

7. The processor of claim 5, wherein a length of the counter is programmable by microcode of the processor.

8. The processor of claim 1, wherein the regulator logic is to enable the second thread to make forward progress when a denial of service attack is executed by the first thread.

9. The processor of claim 8, wherein the regulator logic is to lengthen the time window associated with the first thread when the denial of service attack is detected by the regulator logic.

10. The processor of claim 1, wherein the regulator logic is to be disabled responsive to microcode of the processor.

11. A method comprising:

identifying a termination sequence for an atomic memory operation executed by a first thread;

associating a timer with the first thread, the timer to count a period of a prevention window for the first thread; and

preventing the first thread from execution of a memory cluster operation after completion of the atomic memory operation until the prevention window has passed.

12. The method of claim 11, further comprising holding a request for a second atomic memory operation issued by the first thread in a buffer until the prevention window has passed.

13. The method of claim 11, where the atomic memory operation comprises a bus lock operation.

14. The method of claim 11, further comprising controlling a length of the prevention window based at least in part on a number of atomic memory operations requested by the first thread.

15. The method of claim 11, further comprising allowing a second thread to execute an atomic memory operation during the prevention window, and thereafter preventing the second thread from execution of a memory cluster operation until a second prevention window has passed.

16. The method of claim 15, further comprising preventing a denial of service attack executed by the first thread from blocking execution of the second thread, the first and second threads to be executed in a first core of a multicore processor.

17. The method of claim 16, wherein preventing the first thread from execution of the memory cluster operation allows the second thread to make progress without initiation of a lock breaking operation by a lock breaker coupled to the first core.

18. A system comprising:

a processor including at least one core to execute instructions, the at least one core including;

a front end unit to fetch and decode an instruction;

a renamer to associate at least one operand of the instruction with a physical register;

an execution unit coupled to the front end unit to execute the instruction using the at least one operand;

a memory execution unit (MEU) coupled to the execution unit to handle incoming memory requests from the execution unit; and

a regulator coupled to the MEU to delay at least one atomic memory operation requested by a first thread from being provided to the MEU responsive to termination of a prior atomic memory operation requested by the first thread; and

a dynamic random access memory (DRAM) coupled to processor.

19. The system of claim 18, wherein the regulator is to allow an atomic memory operation requested by a second thread to be provided to the MEU while the at least one atomic memory operation requested by the first thread is delayed.

20. The system of claim 19, wherein via the delay of the at least one atomic memory operation, the regulator is to allow the second thread to make forward progress when the first thread is executing a denial of service attack, and further via the delay, at least one lock breaker of the processor does not initiate a lock breaking activity.