On Wed, Oct 04, 2006 at 09:52:39AM -0700, Jason Thorpe wrote:
>
> I'm actually leaning towards the latter now, because I  
> think it's more appropriate for the type of usage that's being  
> suggested.

One option is to make use of the fact that the bottom bits of all
active processes are different.  So with a bit of locking, and care
over the pid_table_doubling code, it shouldn't be too hard to allow
a device driver to have its own 'pid_table' and associated functions.

	David

-- 
David Laight: david@l8s.co.uk