Received-SPF: pass (google.com: best guess record for domain of linux-ext4-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Organization: Red Hat UK Ltd. Registered Address: Red Hat UK Ltd, Amberley
        Place, 107-111 Peascod Street, Windsor, Berkshire, SI4 1TE, United
        Kingdom.
        Registered in England and Wales under Company Registration No. 3798903
From:   David Howells <dhowells@redhat.com>
In-Reply-To: <1015227.1584007677@warthog.procyon.org.uk>
References: <1015227.1584007677@warthog.procyon.org.uk> <969260.1584004779@warthog.procyon.org.uk>
To:     clm@fb.com, josef@toxicpanda.com, dsterba@suse.com
Cc:     dhowells@redhat.com, mbobrowski@mbobrowski.org,
        darrick.wong@oracle.com, jack@suse.cz, hch@lst.de,
        linux-ext4@vger.kernel.org, linux-xfs@vger.kernel.org,
        linux-btrfs@vger.kernel.org, linux-fsdevel@vger.kernel.org
Subject: btrfs may be broken too - Re: Is ext4_dio_read_iter() broken?
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <1016627.1584008770.1@warthog.procyon.org.uk>
Content-Transfer-Encoding: quoted-printable
Date:   Thu, 12 Mar 2020 10:26:10 +0000
Message-ID: <1016628.1584008770@warthog.procyon.org.uk>
Sender: linux-ext4-owner@vger.kernel.org
Precedence: bulk

David Howells <dhowells@redhat.com> wrote:

> > Is ext4_dio_read_iter() broken?  It calls:
> > =

> > 	file_accessed(iocb->ki_filp);
> > =

> > at the end of the function - but surely iocb should be expected to hav=
e been
> > freed when iocb->ki_complete() was called?
> =

> I think it's actually worse than that.  You also can't call
> inode_unlock_shared(inode) because you no longer own a ref on the inode =
since
> ->ki_complete() is expected to call fput() on iocb->ki_filp.
> =

> Yes, you own a shared lock on it, but unless somewhere along the
> fput-dput-iput chain the inode lock is taken exclusively, the inode can =
be
> freed whilst you're still holding the lock.
> =

> Oh - and ext4_dax_read_iter() is also similarly broken.
> =

> And xfs_file_dio_aio_read() appears to be broken as it touches the inode=
 after
> calling iomap_dio_rw() to unlock it.

Seems btrfs_file_write_iter() is also broken:

	if (iocb->ki_flags & IOCB_DIRECT) {
		num_written =3D __btrfs_direct_write(iocb, from);
	} else {
		num_written =3D btrfs_buffered_write(iocb, from);
		if (num_written > 0)
			iocb->ki_pos =3D pos + num_written;
		if (clean_page)
			pagecache_isize_extended(inode, oldsize,
						i_size_read(inode));
	}

	inode_unlock(inode);

But if __btrfs_direct_write() returned -EIOCBQUEUED then inode may have be=
en
deallocated by the point it's calling inode_unlock().  Holding the lock is=
 not
a preventative measure that I can see.

David