Received-SPF: pass (google.com: domain of linux-kernel+bounces-133810-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: <00000000000098f75506153551a1@google.com> <0000000000002f2066061539e54b@google.com>
 <CAOQ4uxiS5X19OT2MTo_LnLAx2VL9oA1zBSpbuiWMNy_AyGLDrg@mail.gmail.com>
 <20240404081122.GQ538574@ZenIV> <20240404082110.GR538574@ZenIV>
 <CAOQ4uximHfK78KFabJA3Hf4R0En6-GfJ3eF96Lzmc94PGuGayA@mail.gmail.com> <20240406040923.GX538574@ZenIV>
In-Reply-To: <20240406040923.GX538574@ZenIV>
From: Amir Goldstein <amir73il@gmail.com>
Date: Sat, 6 Apr 2024 08:25:03 +0300
Message-ID: <CAOQ4uxikd7HY3gfvLLJp4VRMEhV0NPOExN+UYQ8dDG=p=f-MRQ@mail.gmail.com>
Subject: Re: [syzbot] [kernfs?] possible deadlock in kernfs_fop_llseek
To: Al Viro <viro@zeniv.linux.org.uk>, Miklos Szeredi <miklos@szeredi.hu>
Cc: syzbot <syzbot+9a5b0ced8b1bfb238b56@syzkaller.appspotmail.com>, 
	gregkh@linuxfoundation.org, linux-fsdevel@vger.kernel.org, 
	linux-kernel@vger.kernel.org, syzkaller-bugs@googlegroups.com, tj@kernel.org, 
	valesini@yandex-team.ru, Christoph Hellwig <hch@lst.de>, 
	Christian Brauner <brauner@kernel.org>, Jan Kara <jack@suse.cz>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Sat, Apr 6, 2024 at 7:09=E2=80=AFAM Al Viro <viro@zeniv.linux.org.uk> wr=
ote:
>
> On Thu, Apr 04, 2024 at 12:33:40PM +0300, Amir Goldstein wrote:
>
> > We do not (anymore) lock ovl inode in ovl_llseek(), see:
> > b1f9d3858f72 ovl: use ovl_inode_lock in ovl_llseek()
> > but ovl inode is held in operations (e.g. ovl_rename)
> > which trigger copy up and call vfs_llseek() on the lower file.
>
> OK, but why do we bother with ovl_inode_lock() there?
> Note that serialization on struct file level is provided
> on syscall level - see call of fdget_pos() in there.
> IOW, which object are you protecting?  If it's struct file
> passed your way, you should already have the serialization.
> If it's underlying file on disk, that's up to vfs_llseek().

You're right.

> Exclusion with copyup by a different operation?

Nah, don't see how this is relevant to file->f_pos.

>
> I'm not saying it's wrong - it's just that the thing is
> tricky enough, so some clarification might be a good idea.

I think I just used inode_lock() in
9e46b840c705 ("ovl: support stacked SEEK_HOLE/SEEK_DATA")
as a common coding pattern in overlayfs when protecting the
"master" copy of overlay inode attributes, but it was not needed
for file->f_pos.

Miklos, please ack that I am not missing anything and that
ovl_inode_lock() is indeed redundant in ovl_llseek().

Anyway, this lock is not part of the lockdep issue that started this thread=
.

Thanks,
Amir.