Wed, Aug 16, 2017 at 04:57:18PM CEST, eric.dumazet@gmail.com wrote:

On Wed, 2017-08-16 at 13:06 +0200, Jiri Pirko wrote:

quoted

Wed, Aug 16, 2017 at 12:58:53PM CEST, eric.dumazet@gmail.com wrote:

quoted

On Wed, 2017-08-16 at 12:53 +0200, Jiri Pirko wrote:

quoted

rhashtable is unnecesary big hammer for this. IDR is nice fit for
this purpose.

Obviously IDR does not fit, since you have to change its ABI.

I don't think it is a problem to adjust something to your needs.
Moreover, if it's API is misdesigned from the beginning. We are just
putting IDR back on track, cleaning it's API. I don't see anything wrong
on that. Everyone would benefit.

Except that your patch is gigantic, and nobody really can review it.

You could define idr_alloc_ext() maybe.

Then provide a patch series grouped so that each maintainer can review
its part.

Or leave legacy code using the old idr_alloc() in place.

Fair. Thanks.