[OpenAFS-devel] linux-and-locks-cleanup-20070202 crashes linux kernels older than 2.6.17 (see RT #53457)

Jeffrey Hutzelman jhutz@cmu.edu
Fri, 09 Feb 2007 12:54:31 -0500


On Friday, February 09, 2007 12:35:20 PM -0500 Jeffrey Altman 
<jaltman@secure-endpoints.com> wrote:

> Jeffrey Hutzelman wrote:
>> I think this is rather harder than you or Matt suspect.  The problem
>> with backing local byte-range locks by whole-file locks on the server is
>> that to do it safely, the lock you get on the server has to be at least
>> as good as the local lock, and there's no way to atomically upgrade and
>> downgrade locks on the server.
>
> Hence the need for the upgrade/downgrade proposal I made last April.
> I really should finish coding that.
>
> Of course, that still doesn't address the advisory vs mandatory bit
> and the fact that the file server doesn't actually track who holds locks
> at the moment.

Sure, there are lots of things you can do if you add better lock management 
to the fileserver.  And eventually we'll do that, though I'd be somewhat 
surprised to see that happen before the fileserver supports full byte-range 
locks.  But in the meantime, there is a significant problem with trying to 
handle byte-range locks on a client with existing fileservers.  It's not 
insurmountable, but it's not trivial either.

-- Jeff