Linux filesystem development
 help / color / mirror / Atom feed
From: Christian Brauner <brauner@kernel.org>
To: Christoph Hellwig <hch@lst.de>, Gao Xiang <hsiangkao@linux.alibaba.com>
Cc: syzbot <syzbot+69c477e882e44ce41ad9@syzkaller.appspotmail.com>,
	chao@kernel.org, huyue2@coolpad.com, jack@suse.cz,
	jefflexu@linux.alibaba.com, linkinjeon@kernel.org,
	linux-erofs@lists.ozlabs.org, linux-fsdevel@vger.kernel.org,
	linux-kernel@vger.kernel.org, sj1557.seo@samsung.com,
	syzkaller-bugs@googlegroups.com, xiang@kernel.org
Subject: Re: [syzbot] [erofs?] [fat?] WARNING in erofs_kill_sb
Date: Mon, 31 Jul 2023 18:32:52 +0200	[thread overview]
Message-ID: <20230731-flugbereit-wohnlage-78acdf95ab7e@brauner> (raw)
In-Reply-To: <20230731-deswegen-chatten-4ff6c45563ad@brauner>

On Mon, Jul 31, 2023 at 03:57:57PM +0200, Christian Brauner wrote:
> On Mon, Jul 31, 2023 at 03:53:25PM +0200, Christoph Hellwig wrote:
> > On Mon, Jul 31, 2023 at 03:22:28PM +0200, Christian Brauner wrote:
> > > Uh, no. I vasty underestimated how sensitive that change would be. Plus
> > > arguably ->kill_sb() really should be callable once the sb is visible.
> > > 
> > > Are you looking into this or do you want me to, Christoph?
> > 
> > I'm planning to look into it, but I won't get to it before tomorrow.
> 
> Ok, let me go through the callsites and make sure that all callers are
> safe. I think we should just continue calling deactivate_locked_super()
> exactly the way we do right now but remove shenanigans like the one we
> have in erofs_kill_sb().

The only filesystem that did implicitly rely on fill_super() having been
called was indeed - sorry to single it out - erofs. Everyone else
doesn't have that sort of dependency afaict.

fs/erofs/super.c:       return get_tree_bdev(fc, erofs_fc_fill_super);
=> ->kill_sb() has requirement that ->fill_super() has been called.

So I went and looked at all users of

(1) get_tree_bdev() -> new mount api
(2) mount_bdev      -> old mount api

which are the two functions that were changed in Jan's and Christoph's
patch.

I'm listing the results below. One thing to note is that only 40% (10
out of the 26 I looked at) of block based filesystems reliant on
higher-level fs/super.c helpers directly (e.g., that excludes btrfs)
have converted to the new mount api.

In any case, Gao, can you remove that dependency of erofs_kill_sb() on
erofs_fill_super(), please? Once that hits upstream the syzbot report
will go away.

fs/exfat/super.c:         return get_tree_bdev(fc, exfat_fill_super);
fs/ntfs3/super.c:         return get_tree_bdev(fc, ntfs_fill_super);
fs/squashfs/super.c:      return get_tree_bdev(fc, squashfs_fill_super);
fs/ext4/super.c:          return get_tree_bdev(fc, ext4_fill_super);
fs/xfs/xfs_super.c:       return get_tree_bdev(fc, xfs_fs_fill_super);
fs/adfs/super.c:          return mount_bdev(fs_type, flags, dev_name, data, adfs_fill_super);
fs/befs/linuxvfs.c:       return mount_bdev(fs_type, flags, dev_name, data, befs_fill_super);
fs/bfs/inode.c:           return mount_bdev(fs_type, flags, dev_name, data, bfs_fill_super);
fs/ext2/super.c:          return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
fs/fat/namei_msdos.c:     return mount_bdev(fs_type, flags, dev_name, data, msdos_fill_super);
fs/fat/namei_vfat.c:      return mount_bdev(fs_type, flags, dev_name, data, vfat_fill_super);
fs/freevxfs/vxfs_super.c: return mount_bdev(fs_type, flags, dev_name, data, vxfs_fill_super);
fs/hfs/super.c:           return mount_bdev(fs_type, flags, dev_name, data, hfs_fill_super);
fs/hfsplus/super.c:       return mount_bdev(fs_type, flags, dev_name, data, hfsplus_fill_super);
fs/hpfs/super.c:          return mount_bdev(fs_type, flags, dev_name, data, hpfs_fill_super);
fs/isofs/inode.c:         return mount_bdev(fs_type, flags, dev_name, data, isofs_fill_super);
fs/jfs/super.c:           return mount_bdev(fs_type, flags, dev_name, data, jfs_fill_super);
fs/minix/inode.c:         return mount_bdev(fs_type, flags, dev_name, data, minix_fill_super);
fs/ntfs/super.c:          return mount_bdev(fs_type, flags, dev_name, data, ntfs_fill_super);
fs/ocfs2/super.c:         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
fs/omfs/inode.c:          return mount_bdev(fs_type, flags, dev_name, data, omfs_fill_super);
fs/qnx6/inode.c:          return mount_bdev(fs_type, flags, dev_name, data, qnx6_fill_super);
fs/sysv/super.c:          return mount_bdev(fs_type, flags, dev_name, data, sysv_fill_super);
fs/udf/super.c:           return mount_bdev(fs_type, flags, dev_name, data, udf_fill_super);
fs/ufs/super.c:           return mount_bdev(fs_type, flags, dev_name, data, ufs_fill_super);
=> no custom ->kill_sb() method.

fs/cramfs/inode.c:        ret = get_tree_bdev(fc, cramfs_blkdev_fill_super);
fs/fuse/inode.c:          err = get_tree_bdev(fsc, fuse_fill_super);
fs/gfs2/ops_fstype.c:     error = get_tree_bdev(fc, gfs2_fill_super);
fs/romfs/super.c:         ret = get_tree_bdev(fc, romfs_fill_super);
fs/affs/super.c:          return mount_bdev(fs_type, flags, dev_name, data, affs_fill_super);
fs/efs/super.c:           return mount_bdev(fs_type, flags, dev_name, data, efs_fill_super);
fs/f2fs/super.c:          return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
fs/qnx4/inode.c:          return mount_bdev(fs_type, flags, dev_name, data, qnx4_fill_super);
fs/reiserfs/super.c:      return mount_bdev(fs_type, flags, dev_name, data, reiserfs_fill_super);
fs/zonefs/super.c:        return mount_bdev(fs_type, flags, dev_name, data, zonefs_fill_super);
=> ->kill_sb() has no requirement that ->fill_super() has been called.

  reply	other threads:[~2023-07-31 16:33 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-07-31  7:57 [syzbot] [erofs?] [fat?] WARNING in erofs_kill_sb syzbot
2023-07-31  9:37 ` Christoph Hellwig
2023-07-31 10:58   ` Gao Xiang
2023-07-31 11:16     ` Christoph Hellwig
2023-07-31 12:43       ` Christian Brauner
2023-07-31 13:22         ` Christian Brauner
2023-07-31 13:53           ` Christoph Hellwig
2023-07-31 13:57             ` Christian Brauner
2023-07-31 16:32               ` Christian Brauner [this message]
2023-08-01  1:47                 ` [PATCH] erofs: drop unnecessary WARN_ON() in erofs_kill_sb() Gao Xiang
2023-08-01  7:19                   ` Christian Brauner
2023-08-01  7:51                   ` Christoph Hellwig
2023-07-31 13:29         ` [syzbot] [erofs?] [fat?] WARNING in erofs_kill_sb Gao Xiang

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20230731-flugbereit-wohnlage-78acdf95ab7e@brauner \
    --to=brauner@kernel.org \
    --cc=chao@kernel.org \
    --cc=hch@lst.de \
    --cc=hsiangkao@linux.alibaba.com \
    --cc=huyue2@coolpad.com \
    --cc=jack@suse.cz \
    --cc=jefflexu@linux.alibaba.com \
    --cc=linkinjeon@kernel.org \
    --cc=linux-erofs@lists.ozlabs.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=sj1557.seo@samsung.com \
    --cc=syzbot+69c477e882e44ce41ad9@syzkaller.appspotmail.com \
    --cc=syzkaller-bugs@googlegroups.com \
    --cc=xiang@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox