Hello,

On Fri, Dec 09, 2016 at 02:01:40PM -0800, Andrew Morton wrote:

On Thu,  8 Dec 2016 02:22:55 +0100 Rasmus Villemoes [off-list ref] wrote:

quoted

TL;DR: these patches save 250 KB of memory, with more low-hanging
fruit ready to pick.

While browsing through the lib/idr.c code, I noticed that the code at
the end of ida_get_new_above() probably doesn't work as intended: Most
users of ida use it via ida_simple_get(), and that starts by
unconditionally calling ida_pre_get(), ensuring that ida->idr has
8==MAX_IDR_FREE idr_layers in its free list id_free. In the common
case, none (or at most one) of these get used during
ida_get_new_above(), and we only free one, leaving at least 6 (usually
7) idr_layers in the free list.

Please be aware of

http://ozlabs.org/~akpm/mmots/broken-out/reimplement-idr-and-ida-using-the-radix-tree.patch
http://lkml.kernel.org/r/1480369871-5271-68-git-send-email-mawilcox@linuxonhyperv.com

I expect we'll be merging patches 1-32 of that series into 4.10-rc1 and
the above patch (#33) into 4.11-rc1.

Ah, yeah, great to see the silly implementation being replaced the
radix tree.  ida_pre_get() looks suspicious tho.  idr_preload()
immedicately being followed by idr_preload_end() probably is broken.
Maybe what we need is moving ida to idr like preload interface and
then convert it to radix based interface?  ida currently assumes
per-ida preloading.

Thanks.

-- 
tejun