MIME-Version: 1.0
References: <IA1PR11MB61714FEEAF2C46639891401A89BC9@IA1PR11MB6171.namprd11.prod.outlook.com>
 <2B9F2C1A-B274-41EF-8ABE-1E660521BCE4@joelfernandes.org> <IA1PR11MB6171C7FEE026F421A3CD6A9689BC9@IA1PR11MB6171.namprd11.prod.outlook.com>
In-Reply-To: <IA1PR11MB6171C7FEE026F421A3CD6A9689BC9@IA1PR11MB6171.namprd11.prod.outlook.com>
From:   Joel Fernandes <joel@joelfernandes.org>
Date:   Thu, 16 Mar 2023 10:57:06 -0400
Message-ID: <CAEXW_YTh18nWTWjLBCRiB2AAH76oD7XrMMMPWZ+9thFSmcPaVg@mail.gmail.com>
Subject: Re: [PATCH 1/1] rcu/rcuscale: Stop kfree_scale_thread thread(s) after
 unloading rcuscale
To:     "Zhuo, Qiuxu" <qiuxu.zhuo@intel.com>
Cc:     "paulmck@kernel.org" <paulmck@kernel.org>,
        Frederic Weisbecker <frederic@kernel.org>,
        Josh Triplett <josh@joshtriplett.org>,
        Neeraj Upadhyay <quic_neeraju@quicinc.com>,
        Davidlohr Bueso <dave@stgolabs.net>,
        Steven Rostedt <rostedt@goodmis.org>,
        Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
        Lai Jiangshan <jiangshanlai@gmail.com>,
        "rcu@vger.kernel.org" <rcu@vger.kernel.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Mar 16, 2023 at 9:53=E2=80=AFAM Zhuo, Qiuxu <qiuxu.zhuo@intel.com> =
wrote:
>
[...]
> > >> From: Paul E. McKenney <paulmck@kernel.org> [...]
> > >>>>
> > >>>> How about to pull the rcu_scale_cleanup() function after
> > >> kfree_scale_cleanup().
> > >>>> This groups kfree_* functions and groups rcu_scale_* functions.
> > >>>> Then the code would look cleaner.
> > >>>> So, do you think the changes below are better?
> > >>>
> > >>> IMHO, I don't think doing such a code move is better. Just add a ne=
w
> > >>> header file and declare the function there. But see what Paul says
> > >>> first.
> > >>
> > >> This situation is likely to be an early hint that the kvfree_rcu()
> > >> testing should be split out from kernel/rcu/rcuscale.c.
> > >
> > > Another is that it's a bit expensive to create a new header file just
> > > for eliminating a function declaration. ;-)
> >
> > What is so expensive about new files? It is a natural organization stru=
cture.
> >
> > > So, if no objections, I'd like to send out the v2 patch with the upda=
tes below:
> > >
> > >   - Move rcu_scale_cleanup() after kfree_scale_cleanup() to eliminate=
 the
> > >     declaration of kfree_scale_cleanup(). Though this makes the patch=
 bigger,
> > >     get the file rcuscale.c much cleaner.
> > >
> > >   - Remove the unnecessary step "modprobe torture" from the commit
> > message.
> > >
> > >   - Add the description for why move rcu_scale_cleanup() after
> > >     kfree_scale_cleanup() to the commit message.
> >
> > Honestly if you are moving so many lines around, you may as well split =
it out
> > into a new module.
> > The kfree stuff being clubbed in the same file has also been a major
> > annoyance.
>
> I'm OK with creating a new kernel module for these kfree stuffs,
> but do we really need to do that?

If it were me doing this, I would try to split it just because in the
long term I may have to maintain or deal with it.

I was also thinking a new scale directory _may_ make sense for
performance tests.

kernel/rcu/scaletests/kfree.c
kernel/rcu/scaletests/core.c
kernel/rcu/scaletests/ref.c

Or something like that.

and then maybe putt common code into: kernel/rcu/scaletests/common.c

 - Joel

>
> @paulmck, what's your suggestion for the next step?
>
> >  - Joel
> >
> >
> > > Thanks!
> > > -Qiuxu
> > >
> > >> [...]