Re: [PATCH -next] cgroup/cpuset: add dec_attach_in_progress helper

[Waiman Long]

On 7/25/24 07:40, chenridong wrote:
> On 2024/7/25 19:01, Kamalesh Babulal wrote:
> > On 7/25/24 7:25 AM, Chen Ridong wrote:
> > > There are several functions to decrease attach_in_progress, and they
> > > will wake up cpuset_attach_wq when attach_in_progress is zero. So,
> > > add a helper to make it concise.
> > >
> > > Signed-off-by: Chen Ridong <chenridong@xxxxxxxxxx>
> > > ---
> > >   kernel/cgroup/cpuset.c | 28 +++++++++++++++-------------
> > >   1 file changed, 15 insertions(+), 13 deletions(-)
> > >
> > > diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
> > > index d4322619e59a..c241845694ac 100644
> > > --- a/kernel/cgroup/cpuset.c
> > > +++ b/kernel/cgroup/cpuset.c
> > > @@ -490,6 +490,17 @@ static inline void 
> > > check_insane_mems_config(nodemask_t *nodes)
> > >       }
> > >   }
> > >   +/*
> > > + * decrease cs->attach_in_progress.
> > > + * wake_up cpuset_attach_wq if cs->attach_in_progress==0.
> >
> > In the description, adding locking constraint of cpuset_mutex would 
> > be helpful.
> > Something like "cpuset_mutex must be held by the caller."
> Thank you, I will to that.
> > > + */
> > > +static inline void dec_attach_in_progress(struct cpuset *cs)
> > > +{
> > > +    cs->attach_in_progress--;
> > > +    if (!cs->attach_in_progress)
> > > +        wake_up(&cpuset_attach_wq);
> > > +}
> > > +

I would suggested a dec_attach_in_progress_locked() and a 
dec_attach_in_progress() helpers. The dec_attach_in_progress() helper 
acquires the cpuset_mutex and call dec_attach_in_progress_locked(). 
Inside the dec_attach_in_progress_locked(), you can either add a comment 
about requiring cpuset_mutex held or add a 
lockdep_assert_held(&cpuset_mutex).

Cheers,
Longman