[Libguestfs] Should we always do wipefs before mkfs? Discuss ...

[Richard W.M. Jones]

On Tue, Feb 05, 2013 at 10:20:03AM +0100, Karel Zak wrote:
> On Mon, Feb 04, 2013 at 06:23:38PM +0000, Richard W.M. Jones wrote:
> > 
> > https://bugzilla.redhat.com/show_bug.cgi?id=907554
> 
>  All mkfs.<type> should be robust enough to wipe the device. I'm
>  currently working with guys around filesystems to improve mkfs.ext4
>  and mkfs.xfs
> 
>     https://bugzilla.redhat.com/show_bug.cgi?id=902512
> 
>  .. but nothing is perfect so explicitly call wipefs(8) from
>  installers or things like libguestfs is definitely good idea.
> 
> > We could change libguestfs's guestfs_mkfs (internally) so it always
> > does an implicit wipefs on the filesystem.  wipefs is not too onerous --
> > in particular I believe it only writes to a few chosen areas of the
> > disk.  Especially considering that we're about to run mkfs anyway
> > which for some filesystems writes a lot of blocks.
> 
>  wipefs(8) (or blkid_do_wipe() from the library) wipes only magic
>  strings to make the filesystem (raids or partition tables) invisible
>  for libblkid. It means very few bytes.
> 
> > Thoughts?
> 
>  Go ahead :-)

Thanks .  That was my conclusion when I looked at the wipefs code in
blkid last night: it's worth trying it, but we won't fail the whole
operation if for some reason the wipefs fails.

Here is my patch:

https://github.com/libguestfs/libguestfs/commit/72dd398679cd0bb803daf306d12558369615ba70

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Fedora Windows cross-compiler. Compile Windows programs, test, and
build Windows installers. Over 100 libraries supported.
http://fedoraproject.org/wiki/MinGW