Date: Wed, 11 Nov 2009 12:33:40 -0800
From: Stephen Hemminger <shemminger@vyatta.com>
To: Jens Axboe <jens.axboe@oracle.com>
Cc: linux-kernel@vger.kernel.org
Subject: Mount -o sync regression in 2.6.31
Message-ID: <20091111123340.703f5c86@nehalam>
Organization: Vyatta
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1236
Lines: 33

One of the configurations supported in our specialized distribution is using
floppy to store configuration information (and live cd for everything else).
The floppy is mounted with '-o sync' so that if floppy is removed the files
are still updated. The floppy is formatted with ext2 filesystem.

This broke somewhere between 2.6.30 and 2.6.31-rc1, and was caught by our
release regression system (with VMware pseudo-floppys) but it is easily
reproducible with real hardware.

The following sequence demonstrates this:

# mount /dev/fd0 /mnt -o sync,noatime,noexec,nosuid 
# cp /etc/motd /mnt/motd.$$

(Manually eject floppy)

# umount /mnt

(I/O errors occur)

With -o sync, all I/O should have been completed when the copy was done.

This may also effect people using -o sync on USB thumb drives as well.

I tried the obvious bisection, but git bisect seems to wonder off into the
weeds now. There is no reason that bisecting between v2.6.30 and v2.6.31-rc1
should go off into v2.6.30-rc2

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/