From: Eric Sandeen <esandeen@redhat.com>
Subject: Re: [PATCH] xfstests: avoid ext4/306 failures caused by incompatible
 mount options
Date: Wed, 29 Jan 2014 15:43:59 -0600
Message-ID: <52E9761F.7070903@redhat.com>
References: <20140129204502.GC2165@wallace> <20140129213838.GF30419@thunk.org>
Reply-To: sandeen@redhat.com
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Cc: xfs@oss.sgi.com, linux-ext4@vger.kernel.org
To: "Theodore Ts'o" <tytso@mit.edu>, Eric Whitney <enwlinux@gmail.com>
In-Reply-To: <20140129213838.GF30419@thunk.org>
Sender: linux-ext4-owner@vger.kernel.org

On 1/29/14, 3:38 PM, Theodore Ts'o wrote:
> On Wed, Jan 29, 2014 at 03:45:02PM -0500, Eric Whitney wrote:
>> ext4/306 will fail when mounting the ext3 file system it creates if an
>> ext3-incompatible mount option is applied by _scratch_mount.  This can
>> happen if EXT_MOUNT_OPTIONS is defined appropriately in the test
>> environment.  For example, the block_validity option is commonly used
>> to enhance ext4 testing, and it is not supported by ext3.  Fix this by
>> not including any mount options defined by the test environment.
> 
> I'm not sure I understand why the test is insisting that the file
> system be mounted using ext3.  If the file system is created without
> the extents flag, all of the files will be created using indirect
> blocks, and fundamentally what this test is getting at is that after
> we grow the file system using resize2fs, the new blocks are available
> to be allocated and attached to an indirect block file.
> 
> We can do this by using ext4; I'm not sure why this test is trying to
> use ext3 to set up the test flie system.  It might be better to get
> rid of the requirement to create the file system using ext3, since it
> will make the test runnable even if the ext3 file system hasn't been
> configured into the system and CONFIG_EXT23_AS_EXT4 is not enabled.
> 
> IIRC, Eric Sandeen wrote this test --- Eric, am I missing some reason
> why it was necessary to use ext3 here?

Nope.  Tomayto, tomahto - I think the original report had trouble with
an ext3 filesystem, so that's how I wrote the testcase.

It could be fixed either way, I think.

-Eric

> 
> 					- Ted
>