Re: [PATCHSET] idr: implement idr_alloc() and convert existing users

["J. Bruce Fields" <bfields@fieldses.org>]

On Mon, Feb 04, 2013 at 09:10:31AM -0800, Tejun Heo wrote:
> Hey,
> 
> On Sun, Feb 03, 2013 at 07:15:58PM -0500, J. Bruce Fields wrote:
> > On Sun, Feb 03, 2013 at 12:02:41PM -0500, J. Bruce Fields wrote:
> > > On Sat, Feb 02, 2013 at 05:20:01PM -0800, Tejun Heo wrote:
> > > > * Bruce, I couldn't convert nfsd.  Can you please help?  More on it
> > > >   later.
> > > ...
> > > > I converted all in-kernel users except nfsd and staging drivers.  nfsd
> > > > splits preloading and actual id allocation in a way that per-cpu
> > > > preloading can't be used.  I couldn't follow the control flow to
> > > > verify whether the current code is correct either.  I think the best
> > > > way would be allocating ID upfront without installing the handle and
> > > > then later using idr_replace() to install the pointer when the ID
> > > > actually gets used.  Bruce, would something like that be possible?
> > > 
> > > Actually, I'm not even sure if that's necessary, we can probably just
> > > do it all at the start.
> > > 
> > > I'll try to have a patch doing that tomorrow.
> > 
> > So, something like this.
> 
> Yeah, that should be easily convertable to the new interface.  How
> should we route these changes?  Your patch can go through the usual
> nfs channel and the conversion and deprecation patches can be held off
> until after -rc1.  Does that sound okay?

Whatever's easiest for you.  That sounds fine to me.

--b.