Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220327051853.57647-1-songmuchun@bytedance.com>
 <20220327051853.57647-2-songmuchun@bytedance.com> <CANpmjNPA71CyZefox1rb_f8HqEM_R70EgZCX8fHeeAnDyujO8w@mail.gmail.com>
In-Reply-To: <CANpmjNPA71CyZefox1rb_f8HqEM_R70EgZCX8fHeeAnDyujO8w@mail.gmail.com>
From:   Muchun Song <songmuchun@bytedance.com>
Date:   Mon, 28 Mar 2022 09:52:40 +0800
Message-ID: <CAMZfGtXt9xWnVv8hav+zWHYRmOqBGu3WPaasYwGxCb1-MDDwgQ@mail.gmail.com>
Subject: Re: [PATCH 2/2] mm: kfence: fix objcgs vector allocation
To:     Marco Elver <elver@google.com>
Cc:     Linus Torvalds <torvalds@linux-foundation.org>,
        Alexander Potapenko <glider@google.com>,
        Dmitry Vyukov <dvyukov@google.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Christoph Lameter <cl@linux.com>,
        Pekka Enberg <penberg@kernel.org>,
        David Rientjes <rientjes@google.com>,
        Joonsoo Kim <iamjoonsoo.kim@lge.com>,
        Vlastimil Babka <vbabka@suse.cz>,
        Roman Gushchin <roman.gushchin@linux.dev>,
        kasan-dev <kasan-dev@googlegroups.com>,
        Linux Memory Management List <linux-mm@kvack.org>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Mon, Mar 28, 2022 at 1:31 AM Marco Elver <elver@google.com> wrote:
>
> On Sun, 27 Mar 2022 at 07:19, Muchun Song <songmuchun@bytedance.com> wrote:
> >
> > If the kfence object is allocated to be used for objects vector, then
> > this slot of the pool eventually being occupied permanently since
> > the vector is never freed.  The solutions could be 1) freeing vector
> > when the kfence object is freed or 2) allocating all vectors statically.
> > Since the memory consumption of object vectors is low, it is better to
> > chose 2) to fix the issue and it is also can reduce overhead of vectors
> > allocating in the future.
> >
> > Fixes: d3fb45f370d9 ("mm, kfence: insert KFENCE hooks for SLAB")
> > Signed-off-by: Muchun Song <songmuchun@bytedance.com>
> > ---
> >  mm/kfence/core.c   | 3 +++
> >  mm/kfence/kfence.h | 1 +
> >  2 files changed, 4 insertions(+)
>
> Thanks for this -- mostly looks good. Minor comments below + also
> please fix what the test robot reported.

Will do.

>
> > diff --git a/mm/kfence/core.c b/mm/kfence/core.c
> > index 13128fa13062..9976b3f0d097 100644
> > --- a/mm/kfence/core.c
> > +++ b/mm/kfence/core.c
> > @@ -579,9 +579,11 @@ static bool __init kfence_init_pool(void)
> >         }
> >
> >         for (i = 0; i < CONFIG_KFENCE_NUM_OBJECTS; i++) {
> > +               struct slab *slab = virt_to_slab(addr);
> >                 struct kfence_metadata *meta = &kfence_metadata[i];
> >
> >                 /* Initialize metadata. */
> > +               slab->memcg_data = (unsigned long)&meta->objcg | MEMCG_DATA_OBJCGS;
>
> Maybe just move it to kfence_guarded_alloc(), see "/* Set required
> slab fields */", where similar initialization on slab is done.

But slab->memcg_data is special since it is only needed to be
initialized once.  I think it is better move it to the place where
__SetPageSlab(&pages[i]) is.  What do you think?

>
> >                 INIT_LIST_HEAD(&meta->list);
> >                 raw_spin_lock_init(&meta->lock);
> >                 meta->state = KFENCE_OBJECT_UNUSED;
> > @@ -938,6 +940,7 @@ void __kfence_free(void *addr)
> >  {
> >         struct kfence_metadata *meta = addr_to_metadata((unsigned long)addr);
> >
> > +       KFENCE_WARN_ON(meta->objcg);
>
> This holds true for both SLAB and SLUB, right? (I think it does, but
> just double-checking.)

Right.

Thanks.