Enabling real-time CPU-bound in-kernel workloads to run infinite loops while keeping RCU grace periods finite

• US 10,146,579 B2
• Filed: 02/06/2017
• Issued: 12/04/2018
• Est. Priority Date: 12/11/2016
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: McKenney, Paul E.
• Primary Examiner(s): Patel, Hiren P
• Application Number:
  US15/424,933
• Publication Number:
  US 20180165126A1
• Time in Patent Office: 666 Days
• CPC Class Codes: G06F 9/4887 involving deadlines, e.g. rate based, periodic
• CPC Class Codes: G06F 9/505 considering the load

Abstract

A technique for enabling real-time CPU-bound in-kernel workloads to run infinite loops while keeping read-copy update (RCU) grace periods finite. In an embodiment, a per-CPU indicator may be set to indicate that a CPU running the CPU-bound in-kernel workload has not reported an RCU quiescent state within a defined time. A function may be invoked from within the workload that causes an RCU quiescent state to be reported on behalf of the CPU if the per-CPU indicator is set. If the RCU quiescent state is not reported within a longer defined time, the CPU may be rescheduled.

Claims

1. A method for enabling a real-time CPU-bound in-kernel workload to run an infinite loop while keeping read-copy update (RCU) grace periods finite, said method comprising:

setting a per-CPU indicator to indicate that a CPU running the CPU-bound in-kernel workload has not reported an RCU quiescent state within a first defined time;

invoking from within the workload a function that causes an RCU quiescent state to be reported on behalf of the CPU in response to the per-CPU indicator being set;

rescheduling the CPU in response to the RCU quiescent state being not reported within a longer second defined time that is longer than the first defined time; and

wherein the function that causes the RCU quiescent state to be reported is periodically invoked by the workload at an interval that is shorter than the longer second defined time.

2. The method of claim 1, wherein the indicator is set by an RCU grace period function that handles quiescent states on behalf of CPUs that have failed to report a quiescent state within the first defined time, and which forces the CPUs to be rescheduled in response to the RCU quiescent state being not reported within the longer second defined time.

3. The method of claim 1, wherein the function that causes the RCU quiescent state to be reported is periodically invoked while the workload is active.

4. The method of claim 1, wherein the longer second defined time is based on an RCU CPU stall warning time.

5. The method of claim 1, wherein the longer second defined time represents a time since the beginning of a current grace period that is more than halfway to an RCU CPU stall warning time.

6. The method of claim 1, wherein the function that causes the RCU quiescent state to be checked invokes a flow of control that causes an RCU quiescent state to be reported due to manipulation of an RCU dyntick-idle state of the CPU.