Re: [PATCH] mm: zram: amend SLAB_RECLAIM_ACCOUNT on zspage_cachep

[Zhaoyang Huang]

On Wed, Jun 23, 2021 at 7:38 AM Minchan Kim <minchan@xxxxxxxxxx> wrote:
>
> On Mon, Jun 21, 2021 at 10:35:26AM +0800, Zhaoyang Huang wrote:
> > On Sat, Jun 19, 2021 at 6:02 AM Minchan Kim <minchan@xxxxxxxxxx> wrote:
> > >
> > > On Tue, Jun 08, 2021 at 03:28:17PM +0800, Huangzhaoyang wrote:
> > > > From: Zhaoyang Huang <zhaoyang.huang@xxxxxxxxxx>
> > > >
> > > > Zspage_cachep is found be merged with other kmem cache during test, which
> > > > is not good for debug things(zs_pool->zspage_cachep present to be another
> > > > kmem cache in memory dumpfile). It is also neccessary to do so as shrinker has
> > >
> > > It's not a only problem of zsmalloc because slab want to minimize
> > > fragmentation so try to merge several objects if it's allowed.
> > > So I don't think it's particular problem of zsmalloc.
> > > I guess slub has some option maybe "nomerge" if you want it.
> >
> > >
> > > > been registered for zspage. Amending this flag can help kernel to calculate
> > > > SLAB_RECLAIMBLE correctly.
> > > >
> > > > Signed-off-by: Zhaoyang Huang <zhaoyang.huang@xxxxxxxxxx>
> > > > ---
> > > >  mm/zsmalloc.c | 2 +-
> > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > >
> > > > diff --git a/mm/zsmalloc.c b/mm/zsmalloc.c
> > > > index 19b563b..0b0addd 100644
> > > > --- a/mm/zsmalloc.c
> > > > +++ b/mm/zsmalloc.c
> > > > @@ -328,7 +328,7 @@ static int create_cache(struct zs_pool *pool)
> > > >               return 1;
> > > >
> > > >       pool->zspage_cachep = kmem_cache_create("zspage", sizeof(struct zspage),
> > > > -                                     0, 0, NULL);
> > > > +                                     0, SLAB_RECLAIM_ACCOUNT, NULL);
> > >
> > > How does zspage become SLAB_RECLAIM_ACCOUNT?
> > >
> > > I took the flag as "cacheable" object. IOW, when the shrinker
> > > ask to reclaim the object, it should reclaim(e.g., discarding)
> > > those objects for reclaming. However, that's not the case
> > > in zsmalloc.
> > alloc_slab will take the allocated object into account as
> > SLAB_RECLAIMABLE when this flag set on the kmem_cache
>
> My point is zspage_cachep is not an reclimable slab cache.
> Please describe why you believe it's reclaimable slab.
zspage registered slab shrinker via zs_register_shrinker, in which the
zspage be freed to cache