Uploaded image for project: 'Core Server'
  1. Core Server
  2. SERVER-69461

The behaviour of lock acquisitions differs depending on whether OpContext or Locker is used

XMLWordPrintableJSON

    • Type: Icon: Bug Bug
    • Resolution: Fixed
    • Priority: Icon: Major - P3 Major - P3
    • 7.2.0-rc0
    • Affects Version/s: None
    • Component/s: Concurrency
    • None
    • Storage Execution
    • Fully Compatible
    • ALL
    • Execution Team 2023-06-12, Execution EMEA Team 2023-06-26, Execution EMEA Team 2023-07-10, Execution EMEA Team 2023-07-24, Execution EMEA Team 2023-08-21, Execution EMEA Team 2023-09-04, Execution EMEA Team 2023-09-18, Execution EMEA Team 2023-10-02
    • 146
    • None
    • 3
    • None
    • None
    • None
    • None
    • None
    • None

      There are multiple lock manager RAII classes which take both OperationContext and Locker (or both) and it looks like there exists at least one case where the behaviour of locking differs based on whether locker is passed directly or not.

      For example, if we remove the Locker-only constructor of ResourceLock, this invariant starts getting hit reliably. The condition to check for the invariant doesn't get reached if opCtx is nullptr (which is what happens if we use the locker-only constructor).

      Having such differentiation is dangerous and error-prone and we should get rid of it.

      Architecturally, the OpContext and the Locker are complementary with each other. The OpContext is necessary for the "OS-like" utilities such as interruption, deadline, the operation's settings, transactional state, etc and the Locker is just the container for the currently-acquired locks.

            Assignee:
            yujin.kang@mongodb.com Yujin Kang Park
            Reporter:
            kaloian.manassiev@mongodb.com Kaloian Manassiev
            Votes:
            0 Vote for this issue
            Watchers:
            8 Start watching this issue

              Created:
              Updated:
              Resolved: