Received-SPF: pass (google.com: domain of linux-kernel+bounces-210514-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Precedence: bulk
MIME-Version: 1.0
References: <20240610121820.328876-1-usamaarif642@gmail.com>
 <20240610121820.328876-2-usamaarif642@gmail.com> <CAKEwX=PnwjmZKPLX2=ubD6+-+ZAqpXnczkHe4=1QY1hizOE8WQ@mail.gmail.com>
 <d6088fb2-58d8-4ed1-8d3b-83ea34657db7@gmail.com>
In-Reply-To: <d6088fb2-58d8-4ed1-8d3b-83ea34657db7@gmail.com>
From: Nhat Pham <nphamcs@gmail.com>
Date: Tue, 11 Jun 2024 12:33:29 -0700
Message-ID: <CAKEwX=MOWFQojVgO7u=zEpQSG661j1QHjTGL9m3Gcy47CLNKAg@mail.gmail.com>
Subject: Re: [PATCH v3 1/2] mm: store zero pages to be swapped out in a bitmap
To: Usama Arif <usamaarif642@gmail.com>
Cc: akpm@linux-foundation.org, hannes@cmpxchg.org, david@redhat.com, 
	ying.huang@intel.com, hughd@google.com, willy@infradead.org, 
	yosryahmed@google.com, chengming.zhou@linux.dev, linux-mm@kvack.org, 
	linux-kernel@vger.kernel.org, kernel-team@meta.com, 
	Shakeel Butt <shakeel.butt@linux.dev>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Tue, Jun 11, 2024 at 11:50=E2=80=AFAM Usama Arif <usamaarif642@gmail.com=
> wrote:
> In swap_writepage, with this patch you have:
>
>      if (is_folio_zero_filled(folio)) {
>          swap_zeromap_folio_set(folio);
>          folio_unlock(folio);
>          return 0;
>      }
>      swap_zeromap_folio_clear(folio);
>

I was concerned with the swap slot being freed and reused, without
ever being read :) But looks like it has to be properly reset before
being reused, so all is well on that front.

What about the put_swap_folio() -> swap_free_cluster() case - do we
need to handle zeromap bit clearing here too? Looks like it's clearing
the swap_map (i.e returning it directly to the swapfile, allowing
those slots to be reused) here, and I notice that you clear the
zeromap bitmap wherever the swap_map is cleared as well :)

I jumped around the code a bit - in free_cluster() (called by
swap_free_cluster()), there's this chunk:

if ((si->flags & (SWP_WRITEOK | SWP_PAGE_DISCARD)) =3D=3D
    (SWP_WRITEOK | SWP_PAGE_DISCARD)) {
    swap_cluster_schedule_discard(si, idx);
    return;
}

swap_cluster_schedule_discard() does clear_bit() on the zeromap on the
entire cluster. We also clear_bit() in the work function
swap_do_scheduled_discard() (is this redundant?).

But what if this check is false, i.e the swap device does not have the
SWP_PAGE_DISCARD flag set? Are we not clearing the bits in the zeromap
here?