Difference in statvfs(2) block count with ext2 and very recent Linux 3.2.0 kernels

Difference in statvfs(2) block count with ext2 and very recent Linux 3.2.0 kernels

[Richard W.M. Jones]

Hi,

We do some automated testing of the kernel and ext2/3/4 filesystems,
and noticed that the block count returned by statvfs(2) for plain
*ext2* filesystems has changed with the latest 3.2.0rc1 kernel.

This is probably just because of more accurate block accounting, but I
just wanted to check that it isn't a bug.

This posting contains the numbers and a reproducer:

  https://www.redhat.com/archives/libguestfs/2011-November/msg00051.html

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
virt-df lists disk usage of guests without needing to install any
software inside the virtual machine.  Supports Linux and Windows.
http://et.redhat.com/~rjones/virt-df/

Re: Difference in statvfs(2) block count with ext2 and very recent Linux 3.2.0 kernels

[Eric Sandeen]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 11/16/11 3:54 AM, Richard W.M. Jones wrote:
> 
> Hi,
> 
> We do some automated testing of the kernel and ext2/3/4 filesystems,
> and noticed that the block count returned by statvfs(2) for plain
> *ext2* filesystems has changed with the latest 3.2.0rc1 kernel.
> 
> This is probably just because of more accurate block accounting, but I
> just wanted to check that it isn't a bug.
> 
> This posting contains the numbers and a reproducer:
> 
>   https://www.redhat.com/archives/libguestfs/2011-November/msg00051.html
> 
> Rich.
> 

Note that it is a plain *ext2* filesystem run through the *ext4* driver,
in Fedora, and it was tracked down to commit f975d6bcc7a698a10cc755115e27d3612dcfe322
ext4: teach ext4_statfs() to deal with clusters if bigalloc is enabled
according to the above post...

I will try to look at this to see just why the change affects ext2, unless
Ted knows offhand.

- -Eric


-----BEGIN PGP SIGNATURE-----
Version: GnuPG/MacGPG2 v2.0.17 (Darwin)
Comment: GPGTools - http://gpgtools.org
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
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=U1OD
-----END PGP SIGNATURE-----

Re: Difference in statvfs(2) block count with ext2 and very recent Linux 3.2.0 kernels

[Richard W.M. Jones]

On Wed, Nov 16, 2011 at 10:33:19AM -0600, Eric Sandeen wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On 11/16/11 3:54 AM, Richard W.M. Jones wrote:
> > 
> > Hi,
> > 
> > We do some automated testing of the kernel and ext2/3/4 filesystems,
> > and noticed that the block count returned by statvfs(2) for plain
> > *ext2* filesystems has changed with the latest 3.2.0rc1 kernel.
> > 
> > This is probably just because of more accurate block accounting, but I
> > just wanted to check that it isn't a bug.
> > 
> > This posting contains the numbers and a reproducer:
> > 
> >   https://www.redhat.com/archives/libguestfs/2011-November/msg00051.html
> > 
> > Rich.
> > 
> 
> Note that it is a plain *ext2* filesystem run through the *ext4* driver,
> in Fedora,

Correct!  We're using CONFIG_EXT4_USE_FOR_EXT23=y.

> and it was tracked down to commit f975d6bcc7a698a10cc755115e27d3612dcfe322
> ext4: teach ext4_statfs() to deal with clusters if bigalloc is enabled
> according to the above post...

Well, that's what I thought looked likely.  However I didn't bisect
this, so it could be another commit to blame ...

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
virt-df lists disk usage of guests without needing to install any
software inside the virtual machine.  Supports Linux and Windows.
http://et.redhat.com/~rjones/virt-df/

Re: Difference in statvfs(2) block count with ext2 and very recent Linux 3.2.0 kernels

[Eric Sandeen]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 11/16/11 10:43 AM, Richard W.M. Jones wrote:
> On Wed, Nov 16, 2011 at 10:33:19AM -0600, Eric Sandeen wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> On 11/16/11 3:54 AM, Richard W.M. Jones wrote:
>>>
>>> Hi,
>>>
>>> We do some automated testing of the kernel and ext2/3/4 filesystems,
>>> and noticed that the block count returned by statvfs(2) for plain
>>> *ext2* filesystems has changed with the latest 3.2.0rc1 kernel.
>>>
>>> This is probably just because of more accurate block accounting, but I
>>> just wanted to check that it isn't a bug.
>>>
>>> This posting contains the numbers and a reproducer:
>>>
>>>   https://www.redhat.com/archives/libguestfs/2011-November/msg00051.html
>>>
>>> Rich.
>>>
>>
>> Note that it is a plain *ext2* filesystem run through the *ext4* driver,
>> in Fedora,
> 
> Correct!  We're using CONFIG_EXT4_USE_FOR_EXT23=y.
> 
>> and it was tracked down to commit f975d6bcc7a698a10cc755115e27d3612dcfe322
>> ext4: teach ext4_statfs() to deal with clusters if bigalloc is enabled
>> according to the above post...
> 
> Well, that's what I thought looked likely.  However I didn't bisect
> this, so it could be another commit to blame ...

Ok, I will take a look, thanks.

- -Eric
 
> Rich.
> 

-----BEGIN PGP SIGNATURE-----
Version: GnuPG/MacGPG2 v2.0.17 (Darwin)
Comment: GPGTools - http://gpgtools.org
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
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=5ysl
-----END PGP SIGNATURE-----