Re: [PATCH] unified device list allocator

[Christoph Hellwig <hch@infradead.org>]

On Mon, Mar 07, 2005 at 09:33:02PM -0800, Andrew Morton wrote:
> Matt Mackall <mpm@selenic.com> wrote:
> >
> > +	/* search for insertion point in reverse for dynamic allocation */
> >  +	list_for_each_prev(l, list) {
> 
> hrmph.  Any time we do anything in O(n) time, some smarty comes along with
> a workload which blows us out of the water.  Although it's hard to think of
> any register_blkdev()-intensive workloads.
> 
> It's not possible to do this with prio-trees?

Andrew, I think in this particular case you suffer of a sporadic severe
condition of overengineeritis.

register_blkdev only happens at module_init time (and in fact should go
away completely, so I'm not happy wit hthe surgey to keep it barely alive
at all)