Re: [RFC PATCH v1 2/3] mm/damon/core: Add damon_start_one()

[Jonghyeon Kim]

On Wed, Feb 23, 2022 at 02:11:13PM +0900, Jonghyeon Kim wrote:
> On Tue, Feb 22, 2022 at 09:53:28AM +0000, SeongJae Park wrote:
> > Hello Jonghyeon,
> > 
> > On Fri, 18 Feb 2022 19:26:10 +0900 Jonghyeon Kim <tome01@xxxxxxxxxx> wrote:
> > 
> > > damon_start() function is designed to start multiple damon monitoring
> > > contexts. But, sometimes we need to start monitoring one context.
> > > Although __damon_start() could be considered to start for one monitoring
> > > context, it seems reasonable to adopt a new function that does not need
> > > to handle 'damon_lock' from the caller.
> > > 
> > > Signed-off-by: Jonghyeon Kim <tome01@xxxxxxxxxx>
> > > ---
> > >  include/linux/damon.h |  1 +
> > >  mm/damon/core.c       | 25 +++++++++++++++++++++++++
> > >  2 files changed, 26 insertions(+)
> > > 
> > > diff --git a/include/linux/damon.h b/include/linux/damon.h
> > > index c0adf1566603..069577477662 100644
> > > --- a/include/linux/damon.h
> > > +++ b/include/linux/damon.h
> > > @@ -511,6 +511,7 @@ int damon_register_ops(struct damon_operations *ops);
> > >  int damon_select_ops(struct damon_ctx *ctx, enum damon_ops_id id);
> > >  
> > >  int damon_start(struct damon_ctx **ctxs, int nr_ctxs);
> > > +int damon_start_one(struct damon_ctx *ctx);
> > >  int damon_stop(struct damon_ctx **ctxs, int nr_ctxs);
> > >  
> > >  #endif	/* CONFIG_DAMON */
> > > diff --git a/mm/damon/core.c b/mm/damon/core.c
> > > index 290c9c0535ee..e43f138a3489 100644
> > > --- a/mm/damon/core.c
> > > +++ b/mm/damon/core.c
> > > @@ -466,6 +466,31 @@ int damon_start(struct damon_ctx **ctxs, int nr_ctxs)
> > >  	return err;
> > >  }
> > >  
> > > +/**
> > > + * damon_start_one() - Starts the monitorings for one context.
> > > + * @ctx:	monitoring context
> > > + *
> > > + * This function starts one monitoring thread for only one monitoring context
> > > + * handling damon_lock.
> > > + *
> > > + * Return: 0 on success, negative error code otherwise.
> > > + */
> > > +int damon_start_one(struct damon_ctx *ctx)
> > > +{
> > > +	int err = 0;
> > > +
> > > +	mutex_lock(&damon_lock);
> > > +	err = __damon_start(ctx);
> > > +	if (err) {
> > > +		mutex_unlock(&damon_lock);
> > > +		return err;
> > > +	}
> > > +	nr_running_ctxs++;
> > > +	mutex_unlock(&damon_lock);
> > > +
> > > +	return err;
> > > +}
> > > +
> > 
> > IMHO, this looks like an unnecessary duplication of code.  Unless this is
> > needed for a real usecase, this change might unnecessarily make the code only a
> > little bit more complicated.  And to my understanding of the next patch, this
> > is not really needed for this patchset.  I will left comments on the patch.  If
> > I'm missing something, please clarify why this is really needed.
> > 
> > Furthermore, damon_start() starts a set of DAMON contexts in exclusive manner,
> > to ensure there will be no interference.  This patch breaks the assumption.
> > That is, contexts that started with damon_start() could be interfered by other
> > contexts that started with damon_start_one().  I have a plan to make
> > damon_start() also work for non-exclusive contexts group[1], though.
> > 
> > [1] https://lore.kernel.org/linux-mm/20220217161938.8874-3-sj@xxxxxxxxxx/
> > 
> > 
> > Thanks,
> > SJ
> > 
> 
> I understand your opinion, and it makes sense. I will drop this patch in the
> next version.
> 
> 
> Thanks,
> Jonghyeon