Re: [PATCH RFC 00/10] Introduce lockless shrink_slab()

[Shakeel Butt]

On Tue, Aug 7, 2018 at 6:12 PM Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
>
> Hi Kirill,
>
> On Tue, 07 Aug 2018 18:37:19 +0300 Kirill Tkhai <ktkhai@xxxxxxxxxxxxx> wrote:
> >
> > After bitmaps of not-empty memcg shrinkers were implemented
> > (see "[PATCH v9 00/17] Improve shrink_slab() scalability..."
> > series, which is already in mm tree), all the evil in perf
> > trace has moved from shrink_slab() to down_read_trylock().
> > As reported by Shakeel Butt:
> >
> >      > I created 255 memcgs, 255 ext4 mounts and made each memcg create a
> >      > file containing few KiBs on corresponding mount. Then in a separate
> >      > memcg of 200 MiB limit ran a fork-bomb.
> >      >
> >      > I ran the "perf record -ag -- sleep 60" and below are the results:
> >      > +  47.49%            fb.sh  [kernel.kallsyms]    [k] down_read_trylock
> >      > +  30.72%            fb.sh  [kernel.kallsyms]    [k] up_read
> >      > +   9.51%            fb.sh  [kernel.kallsyms]    [k] mem_cgroup_iter
> >      > +   1.69%            fb.sh  [kernel.kallsyms]    [k] shrink_node_memcg
> >      > +   1.35%            fb.sh  [kernel.kallsyms]    [k] mem_cgroup_protected
> >      > +   1.05%            fb.sh  [kernel.kallsyms]    [k] queued_spin_lock_slowpath
> >      > +   0.85%            fb.sh  [kernel.kallsyms]    [k] _raw_spin_lock
> >      > +   0.78%            fb.sh  [kernel.kallsyms]    [k] lruvec_lru_size
> >      > +   0.57%            fb.sh  [kernel.kallsyms]    [k] shrink_node
> >      > +   0.54%            fb.sh  [kernel.kallsyms]    [k] queue_work_on
> >      > +   0.46%            fb.sh  [kernel.kallsyms]    [k] shrink_slab_memcg
> >
> > The patchset continues to improve shrink_slab() scalability and makes
> > it lockless completely. Here are several steps for that:
>
> So do you have any numbers for after theses changes?
>

I will do the same experiment as before with these patches sometime
this or next week.

BTW Kirill, thanks for pushing this.

regards,
Shakeel