Re: [PATCH] fs: handle shrinker registration failure in sget_userns

[Tetsuo Handa]

Michal Hocko wrote:
> On Thu 23-11-17 22:57:06, Tetsuo Handa wrote:
> > Michal Hocko wrote:
> > > > @@ -260,9 +261,8 @@ static struct super_block *alloc_super(struct file_system_type *type, int flags,
> > > >  	s->s_shrink.count_objects = super_cache_count;
> > > >  	s->s_shrink.batch = 1024;
> > > >  	s->s_shrink.flags = SHRINKER_NUMA_AWARE | SHRINKER_MEMCG_AWARE;
> > > > -	INIT_LIST_HEAD(&s->s_shrink.list);
> > > > -	return s;
> > > > -
> > > > +	if (register_shrinker(&s->s_shrink) == 0)
> > > > +		return s;
> > > >  fail:
> > > >  	destroy_unused_super(s);
> > > >  	return NULL;
> > > 
> > > But I am not sure this is correct. So what protects shrinker invocation
> > > while the object is not initialized yet?
> > 
> > Then, what protects shrinker invocation in your patch?
> 
> It is s_umount lock but that one is alreay held at the point where you
> suggested register_shrinker. My bad, I could have noticed that. Feel
> free to take over and send a patch. Considering I've screwed several
> times already I do not feel I am the right one to send the fix.
> 
I will wait for your posting. I feel we want to update the comment block
saying "this object isn't exposed yet", for it is confusing that we
already exposed the shrinker inside the object.