Received-SPF: pass (google.com: domain of linux-kernel+bounces-50679-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Precedence: bulk
MIME-Version: 1.0
References: <20240201-b4-zswap-invalidate-entry-v1-0-56ed496b6e55@bytedance.com>
 <20240201-b4-zswap-invalidate-entry-v1-6-56ed496b6e55@bytedance.com>
 <CAKEwX=MOcmUjtrYOJCwh3unRT7OKdrvtR-FFdSevNf5i7z=Q=A@mail.gmail.com> <CAJD7tkb9q5FQqwUTH3kJtEcWB57Kua_RJ7zmDdq9KXgVegs=XQ@mail.gmail.com>
In-Reply-To: <CAJD7tkb9q5FQqwUTH3kJtEcWB57Kua_RJ7zmDdq9KXgVegs=XQ@mail.gmail.com>
From: Nhat Pham <nphamcs@gmail.com>
Date: Fri, 2 Feb 2024 14:44:11 -0800
Message-ID: <CAKEwX=P0Mov3Aqazy8eLkC23wxpocVDf5oyQ6gJA1sK-Zv-BMw@mail.gmail.com>
Subject: Re: [PATCH 6/6] mm/zswap: zswap entry doesn't need refcount anymore
To: Yosry Ahmed <yosryahmed@google.com>
Cc: Chengming Zhou <zhouchengming@bytedance.com>, Johannes Weiner <hannes@cmpxchg.org>, 
	Andrew Morton <akpm@linux-foundation.org>, linux-kernel@vger.kernel.org, 
	linux-mm@kvack.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Fri, Feb 2, 2024 at 2:37=E2=80=AFPM Yosry Ahmed <yosryahmed@google.com> =
wrote:
>
> On Fri, Feb 2, 2024 at 2:33=E2=80=AFPM Nhat Pham <nphamcs@gmail.com> wrot=
e:
> >
> > On Thu, Feb 1, 2024 at 7:50=E2=80=AFAM Chengming Zhou
> > <zhouchengming@bytedance.com> wrote:
> > >
> > > Since we don't need to leave zswap entry on the zswap tree anymore,
> > > we should remove it from tree once we find it from the tree.
> > >
> > > Then after using it, we can directly free it, no concurrent path
> > > can find it from tree. Only the shrinker can see it from lru list,
> > > which will also double check under tree lock, so no race problem.
> > >
> > > So we don't need refcount in zswap entry anymore and don't need to
> > > take the spinlock for the second time to invalidate it.
> > >
> > > The side effect is that zswap_entry_free() maybe not happen in tree
> > > spinlock, but it's ok since nothing need to be protected by the lock.
> > >
> > > Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
> >
> > Oh this is sweet! Fewer things to keep in mind.
> > Reviewed-by: Nhat Pham <nphamcs@gmail.com>
> >
> > > ---
> > >  mm/zswap.c | 63 +++++++++++-----------------------------------------=
----------
> > >  1 file changed, 11 insertions(+), 52 deletions(-)
> > >
> > > diff --git a/mm/zswap.c b/mm/zswap.c
> > > index cbf379abb6c7..cd67f7f6b302 100644
> > > --- a/mm/zswap.c
> > > +++ b/mm/zswap.c
> > > @@ -193,12 +193,6 @@ struct zswap_pool {
> > >   *
> > >   * rbnode - links the entry into red-black tree for the appropriate =
swap type
> > >   * swpentry - associated swap entry, the offset indexes into the red=
-black tree
> > > - * refcount - the number of outstanding reference to the entry. This=
 is needed
> > > - *            to protect against premature freeing of the entry by c=
ode
> > > - *            concurrent calls to load, invalidate, and writeback.  =
The lock
> > > - *            for the zswap_tree structure that contains the entry m=
ust
> > > - *            be held while changing the refcount.  Since the lock m=
ust
> > > - *            be held, there is no reason to also make refcount atom=
ic.
> > >   * length - the length in bytes of the compressed page data.  Needed=
 during
> > >   *          decompression. For a same value filled page length is 0,=
 and both
> > >   *          pool and lru are invalid and must be ignored.
> > > @@ -211,7 +205,6 @@ struct zswap_pool {
> > >  struct zswap_entry {
> > >         struct rb_node rbnode;
> > >         swp_entry_t swpentry;
> > > -       int refcount;
> >
> > Hah this should even make zswap a bit more space-efficient. IIRC Yosry
> > has some analysis regarding how much less efficient zswap will be
> > every time we add a new field to zswap entry - this should go in the
> > opposite direction :)
>
> Unfortunately in this specific case I think it won't change the size
> of the allocation for struct zswap_entry anyway, but it is a step
> nonetheless :)

Ah, is it because of the field alignment requirement? But yeah, one
day we will remove enough of them :)