[LU-1070] Inconsistent CLIO lock state maybe cause cl_lock cannot be cached

• Type: Bug
• Resolution: Fixed
• Priority: Blocker
• Fix Version/s: Lustre 2.2.0, Lustre 2.3.0
• Affects Version/s: Lustre 2.2.0
• Story Points: 2
• Severity: 3
• Assignee: nasf (Inactive)
• Reporter: nasf (Inactive)

Description

According to current CLIO lock state machine rules, the last "unuse()" maybe changes the cl_lock state to CLS_CACHED or CLS_NEW, and only the CLS_HELD cl_lock can be changed to CLS_CACHED.

On the other hand, with AGL introduced, the AGL RPC sponsor maybe exits the cl_lock processing without "wait()" called before corresponding OSC lock "upcall()", then it is the last "unuse()" duty to update the cl_lock state according to the low layer osc_lock state. To make the cl_lock interface clean, "wait()" should be called when the last "unuse()" called. Currently, it is done in "lov_lock_unuse()" against toplock. But OSC lock "upcall()" is against sublock. So the sublock state maybe un-updated, and inconsistent with low layer osc_lock state, then when the last "unuse()", its state cannot be changed to CLS_CACHED as expectation.

Issue Links

• Trackbacks: Changelog 2.2 version 2.2.0 Support for networks: o2iblnd OFED 1.5.4 Server support for kernels: 2.6.32220.4.2.el6 (RHEL6) Client support for unpatched kernels: 2.6.18274.18.1.el5 (RHEL5) 2.6.32220.4.2.el6 (RHEL6) 2.6.32.360....