Re: [fsnotify] c738fbabb0: will-it-scale.per_process_ops -9.5% regression

[Jan Kara <jack@suse.cz>]

On Sun 26-07-20 14:52:47, Amir Goldstein wrote:
> On Fri, Jul 24, 2020 at 6:47 AM Amir Goldstein <amir73il@gmail.com> wrote:
> >
> > On Fri, Jul 24, 2020 at 5:45 AM Rong Chen <rong.a.chen@intel.com> wrote:
> > >
> > >
> > >
> > > On 7/21/20 11:59 PM, Amir Goldstein wrote:
> > > > On Tue, Jul 21, 2020 at 3:15 AM kernel test robot <rong.a.chen@intel.com> wrote:
> > > >> Greeting,
> > > >>
> > > >> FYI, we noticed a -9.5% regression of will-it-scale.per_process_ops due to commit:
> > > >>
> > > >>
> > > >> commit: c738fbabb0ff62d0f9a9572e56e65d05a1b34c6a ("fsnotify: fold fsnotify() call into fsnotify_parent()")
> > > > Strange, that's a pretty dumb patch moving some inlined code from one
> > > > function to
> > > > another (assuming there are no fsnotify marks in this test).
> > > >
> > > > Unless I am missing something the only thing that changes slightly is
> > > > an extra d_inode(file->f_path.dentry) deference.
> > > > I can get rid of it.
> > > >
> > > > Is it possible to ask for a re-test with fix patch (attached)?
> > >
> > > Hi Amir,
> > >
> > > We failed to apply this patch, could you tell us the base commit or the
> > > base branch?
> > >
> >
> > Hi Rong,
> >
> > The patch is applied on top of the reported offending commit:
> > c738fbabb0ff62d0f9a9572e56e65d05a1b34c6a ("fsnotify: fold fsnotify()
> > call into fsnotify_parent()")
> >
> > I pushed it to my github:
> > https://github.com/amir73il/linux/commits/for_lkp
> >
> 
> FWIW, I tried reproducing the reported regression on a local machine.
> 
> I ran the test twice on each of the branch commits:
> 
> 26dc3d2bff62 fsnotify: pass inode to fsnotify_parent()
> c738fbabb0ff fsnotify: fold fsnotify() call into fsnotify_parent()
> 71d734103edf fsnotify: Rearrange fast path to minimise overhead when
> there is no watcher
> 47aaabdedf36 fanotify: Avoid softlockups when reading many events
> 
> Not only did I not observe a regression with the reported commit,
> but there was a slight improvement. And then there yet was another
> improvement with the fix commit on top of it.

I suspect this may be closely related to code generation, code cacheline
alignment etc. and thus depends heavily on a particular compiler version
and CPU. I've checked the commit myself and I agree it looks innocent so
for these reasons, I'm not particularly worried about this regression.

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR