Received-SPF: pass (google.com: domain of linux-kernel+bounces-43029-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.199.223 as permitted sender) client-ip=147.75.199.223;
Precedence: bulk
MIME-Version: 1.0
References: <20240117230032.2312067-1-daeho43@gmail.com> <df9645d9-1e9a-4bd2-88bb-26425cf45811@kernel.org>
 <CACOAw_yjEuGSvo_qyoA13U0HwOr3gOzGtNf2Twhes01SNSGQeg@mail.gmail.com>
 <b18c286a-cc72-439c-b373-98e0d6504618@kernel.org> <CACOAw_yqrtEhq4wtJbs7CVn260h7iZyC7koCWH1iMyeQo5140g@mail.gmail.com>
 <e879da72-4c4f-4aed-8081-784f2de5c887@kernel.org>
In-Reply-To: <e879da72-4c4f-4aed-8081-784f2de5c887@kernel.org>
From: Daeho Jeong <daeho43@gmail.com>
Date: Mon, 29 Jan 2024 07:47:49 -0800
Message-ID: <CACOAw_xDDoOQEHOAXkG+8PF8yD0MtUAW4J04tYcPCh3VMp7FGQ@mail.gmail.com>
Subject: Re: [f2fs-dev] [PATCH] f2fs-tools: allocate logs after conventional
 area for HM zoned devices
To: Chao Yu <chao@kernel.org>
Cc: linux-kernel@vger.kernel.org, linux-f2fs-devel@lists.sourceforge.net, 
	kernel-team@android.com, Daeho Jeong <daehojeong@google.com>, 
	Yongpeng Yang <yangyongpeng1@oppo.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Sun, Jan 28, 2024 at 5:27=E2=80=AFPM Chao Yu <chao@kernel.org> wrote:
>
> On 2024/1/27 2:17, Daeho Jeong wrote:
> > On Thu, Jan 25, 2024 at 5:27=E2=80=AFPM Chao Yu <chao@kernel.org> wrote=
:
> >>
> >> On 2024/1/26 0:25, Daeho Jeong wrote:
> >>> On Wed, Jan 24, 2024 at 7:34=E2=80=AFPM Chao Yu <chao@kernel.org> wro=
te:
> >>>>
> >>>> +Cc Yongpeng Yang
> >>>>
> >>>> Daeho,
> >>>>
> >>>> Yongpeng reports a potential issue: if c.devices[0].total_segments i=
s
> >>>> larger than segments of mainarea, c.cur_seg[CURSEG_HOT_NODE] will ex=
ceed
> >>>> end boundary of mainarea. Could you please check that? though it's a=
 corner
> >>>> case.
> >>>
> >>> Can you elaborate more?
> >>
> >> Since c.cur_seg[CURSEG_HOT_NODE] is an offset started from main_blkadd=
r.
> >
> > Oh, Got it.
> > Then, how about this?
> >
> >           c.cur_seg[CURSEG_HOT_NODE] =3D c.zoned_model =3D=3D F2FS_ZONE=
D_HM ?
> >                           (c.devices[1].start_blkaddr -
> > get_sb(main_blkaddr)) / c.blks_per_seg : 0;
>
> Better, but log header should align to start blkaddr of zone?

It's already aligned here.

        if (c.zoned_mode && c.ndevs > 1)
                zone_align_start_offset +=3D
                        (c.devices[0].total_sectors * c.sector_size) %
zone_size_bytes;

..

        for (i =3D 0; i < c.ndevs; i++) {
                if (i =3D=3D 0) {
                        c.devices[i].total_segments =3D
                                (c.devices[i].total_sectors *
                                c.sector_size - zone_align_start_offset) /
                                segment_size_bytes;
                        c.devices[i].start_blkaddr =3D 0;
                        c.devices[i].end_blkaddr =3D c.devices[i].total_seg=
ments *
                                                c.blks_per_seg - 1 +
                                                sb->segment0_blkaddr;
                } else {
                        c.devices[i].total_segments =3D
                                c.devices[i].total_sectors /
                                (c.sectors_per_blk * c.blks_per_seg);
                        c.devices[i].start_blkaddr =3D
                                        c.devices[i - 1].end_blkaddr + 1;

..

        total_meta_zones =3D ZONE_ALIGN(total_meta_segments *
                                                c.blks_per_seg);

        set_sb(main_blkaddr, get_sb(segment0_blkaddr) + total_meta_zones *
                                c.segs_per_zone * c.blks_per_seg);

>
> Thanks,
>
> >
> >> If c.cur_seg[CURSEG_HOT_NODE] was assigned w/ c.devices[0].total_segme=
nts,
> >> and c.devices[0].total_segments is larger than segments of mainare,
> >> c.cur_seg[CURSEG_HOT_NODE] will exceed the end boundary of mainarea.
> >>
> >>          c.cur_seg[CURSEG_HOT_NODE] =3D c.zoned_model =3D=3D F2FS_ZONE=
D_HM ?
> >>                          c.devices[0].total_segments : 0;
> >>
> >>> In the case of F2FS_ZONED_HM, we have the devices[1].
> >>> Do you mean the case we format the filesystem intentionally smaller
> >>> than what devices have?
> >>
> >> I mean blew case:
> >> device[0]: conventional device size =3D 10240 MB
> >> device[1]: zone device size =3D 2 MB
> >>
> >> Thanks,
> >>
> >>>
> >>>>
> >>>> On 2024/1/18 7:00, Daeho Jeong wrote:
> >>>>> From: Daeho Jeong <daehojeong@google.com>
> >>>>>
> >>>>> Make to allocate logs after conventional area for HM zoned devices =
to
> >>>>> spare them for file pinning support.
> >>>>>
> >>>>> Signed-off-by: Daeho Jeong <daehojeong@google.com>
> >>>>> ---
> >>>>>     mkfs/f2fs_format.c | 3 ++-
> >>>>>     1 file changed, 2 insertions(+), 1 deletion(-)
> >>>>>
> >>>>> diff --git a/mkfs/f2fs_format.c b/mkfs/f2fs_format.c
> >>>>> index f2840c8..91a7f4b 100644
> >>>>> --- a/mkfs/f2fs_format.c
> >>>>> +++ b/mkfs/f2fs_format.c
> >>>>> @@ -557,7 +557,8 @@ static int f2fs_prepare_super_block(void)
> >>>>>                 c.cur_seg[CURSEG_COLD_DATA] =3D 0;
> >>>>>                 c.cur_seg[CURSEG_WARM_DATA] =3D next_zone(CURSEG_CO=
LD_DATA);
> >>>>>         } else if (c.zoned_mode) {
> >>>>> -             c.cur_seg[CURSEG_HOT_NODE] =3D 0;
> >>>>> +             c.cur_seg[CURSEG_HOT_NODE] =3D c.zoned_model =3D=3D F=
2FS_ZONED_HM ?
> >>>>> +                     c.devices[0].total_segments : 0;
> >>>>>                 c.cur_seg[CURSEG_WARM_NODE] =3D next_zone(CURSEG_HO=
T_NODE);
> >>>>>                 c.cur_seg[CURSEG_COLD_NODE] =3D next_zone(CURSEG_WA=
RM_NODE);
> >>>>>                 c.cur_seg[CURSEG_HOT_DATA] =3D next_zone(CURSEG_COL=
D_NODE);