Is HDB locking safe?

[Peter da Silva]

In article <YZ85N+6 at xds13.ferranti.com> peter at ficc.ferranti.com (Peter da Silva) writes:
> No sleep is ever enough. The system could simply be busier than you ever
> imagined. You don't solve a race problem by narrowing the window: try
> checking the return value of the "unlink": that's the point of failure.

In article <18500 at rpp386.cactus.org> jfh at rpp386.cactus.org (John F. Haugh II) writes:
> There is a big, fat window between the kill() and the unlink() during
> which time the other process and kill() the dead process, unlink() the
> stale lock, and creat() the new lock file.  Your process will then
> unlink() a lock which has just be created.

You're right. Can you suggest a locking protocol that would work here?
I can't imagine that it could be made robust enough to survive a process
crashing during the locking process, but it should be possible to handle
most cases of programs crashing whle locked.

I guess creating a second lock file to be held while deleting the first
would work.
-- 
Peter da Silva.   `-_-'
+1 713 274 5180.   'U`
peter at ferranti.com (currently not working)
peter at hackercorp.com