The pr implements a global mutex with the help of rados omap, this is needed because there is a concurrency in cephfs when encryption is used. This fixes the issue: #4654.
Is there anything that requires special attention
Due to this kernel requirement the testing is limited to the functionality of the mutex rather than the full feature
Sorry, I had to open this in a work in progress state, since I will be away on vacation the next four weeks and thought in case there is interest in continuing the work this would not go to waste. However I have tested this on a test setup with fscrypt compatible kernel and the concurrency issue is fixed. I will however add some things below that might need some attention still
Checklist:
[ ] Move the rados client from reftracker to more common location so that both reftracker and radosmutex can utilize it nicely
[ ] The LockOwner is not set properly currently
[ ] There is both release and remove lock think if these both are necessary and in that case when can the lock be released
Describe what this PR does
The pr implements a global mutex with the help of rados omap, this is needed because there is a concurrency in cephfs when encryption is used. This fixes the issue: #4654.
Is there anything that requires special attention
Due to this kernel requirement the testing is limited to the functionality of the mutex rather than the full feature
Sorry, I had to open this in a work in progress state, since I will be away on vacation the next four weeks and thought in case there is interest in continuing the work this would not go to waste. However I have tested this on a test setup with fscrypt compatible kernel and the concurrency issue is fixed. I will however add some things below that might need some attention still
Checklist: