process shared robust mutexes for tdb

[Jeremy Allison]

On Tue, Dec 25, 2012 at 11:31:57AM +0100, Volker Lendecke wrote:
> 
> There seems to be code in FreeBSD for that as well. Not sure
> if it's released yet though.

Ping Richard Sharpe. I bet he'll be highly motivated in his
new job to make this happen :-).

> This patchset does not do per-record locking. So you will
> still see fcntl for traverse. It carefully looks whether a
> given lock offset/len is a lock for the freelist or
> hashchain and falls back to fcntl if it does not match the
> pattern. For me this is the most important use case.

That looks like a really sensible compromise that changes
the least code - +1.

Jeremy.