Re: [PATCH] slub: make sysfs directories for memcg sub-caches optional

From: Tejun Heo
Date: Tue Feb 07 2017 - 12:10:13 EST


Hello, Andrew.

On Mon, Feb 06, 2017 at 04:22:13PM -0800, Andrew Morton wrote:
> > #ifdef CONFIG_MEMCG
> > - if (is_root_cache(s)) {
> > + if (is_root_cache(s) && memcg_sysfs_enabled) {
>
> This could be turned on and off after bootup but I guess the result
> could be pretty confusing.
>
> However there would be useful use cases? The user would normally have
> this disabled but if he wants to do a bit of debugging then turn this
> on, create a memcg, have a poke around then turn the feature off again.

Hmm... maybe. It's somewhat nasty to do after-the-fact. You would
have to re-walk all the caches to create and destroy the sysfs
entries. Given that it's a pretty fringe debug feature, I'm not sure
the added complexity would be justified.

> > s->memcg_kset = kset_create_and_add("cgroup", NULL, &s->kobj);
> > if (!s->memcg_kset) {
> > err = -ENOMEM;
> > @@ -5673,7 +5695,8 @@ static void sysfs_slab_remove(struct kme
> > return;
> >
> > #ifdef CONFIG_MEMCG
> > - kset_unregister(s->memcg_kset);
> > + if (s->memcg_kset)
> > + kset_unregister(s->memcg_kset);
>
> kset_unregister(NULL) is legal
>
> --- a/mm/slub.c~slub-make-sysfs-directories-for-memcg-sub-caches-optional-fix
> +++ a/mm/slub.c
> @@ -5699,8 +5699,7 @@ static void sysfs_slab_remove(struct kme
> return;
>
> #ifdef CONFIG_MEMCG
> - if (s->memcg_kset)
> - kset_unregister(s->memcg_kset);
> + kset_unregister(s->memcg_kset);
> #endif
> kobject_uevent(&s->kobj, KOBJ_REMOVE);
> kobject_del(&s->kobj);

Ah, of course, looks good to me.

Thanks.

--
tejun