Re: [Gnu-arch-users] Re: [BUG] FEATURE PLANS: revlib locking

[Tom Lord]

    > From: Andrew Suffield <address@hidden>

    > > I understood that

    > >  1. Create nonce file
    > >  2. hardlink lock file to it
    > >  3. check reference count of nonce file

    > > Was a reliable way of locking files across NFS.  Is this not the case?

    > FSVO "reliable". It's non-atomic - it is not the case that precisely
    > one process will get the lock. Specifically, it's possible that no
    > processes will get the lock. 

No it isn't.

    > If you get two clients trying to claim
    > the lock at once, they can get into a mutual busy loop, repeatedly
    > trying and failing, so you have to rely on random wait states to break
    > this loop. This can be quite impressively slow.

While the lock file doesn't exist, the first `link(2)' call made for
it wins the lock.


    > It also requires manual intervention to break stale locks
    > safely.

_That_ is the real problem.   Unless the intervention is able to
include "kill all client processes that might _think_ they hold this
lock" you simply can't break stale locks (of this variety) safely.

-t