Re: [PATCH 0/3] Fiemap, an extent mapping ioctl

[Andreas Dilger <adilger@sun.com>]

On Sep 10, 2008  11:46 -0700, Andrew Morton wrote:
> On Wed, 10 Sep 2008 05:49:34 -0700 Mark Fasheh <mfasheh@suse.com> wrote:
> > 	The following patches are the latest attempt at implementing a
> > fiemap ioctl, which can be used by userspace software to get extent
> > information for an inode in an efficient manner.
> 
> As these are applicable to all filesystems, Cc:ing only linux-ext4 is
> not sufficient.  All filesystem developers (at least) need an
> opportunity to review and understand these changes.

That was an oversight, partly caused because I launched the thread for
this on linux-ext4 after an ext4 concall.  All previous patches have
gone to linux-fsdevel and been through numerous discussions.

> > The fiemap ioctl is an efficient method for userspace to get file
> > extent mappings. Instead of block-by-block mapping (such as bmap), fiemap
> > returns a list of extents.
> 
> The above is, afacit, the only offered rationale for the addition of
> these new feature.  I don't recall seeing anyone complain about bmap()
> inefficiency.  In fact I rarely hear of anyone using bmap() at all.
> 
> This rationale needs expanding, please.  A lot.

There are several reasons for this new API:
- it avoids tools like "filefrag" (which currently use FIBMAP) having to
  do an ioctl for every block in a file, have the kernel map that block
  from an on-disk extent (in most newer filesystems), then re-assemble the
  extents in userspace.
- it works with filesystems that are not block based (e.g. NTFS, btrfs, etc)
  that may align file data on boundaries other than $blocksize boundaries
- it provides a much more rich API for finding out about on-disk allocation,
  such as whether allocated blocks are unwritten (e.g. fallocate), if they
  are packed along with other data, if the data is in the inode, etc.
- it can share existing XFS-specific functionality (which FIEMAP was designed
  to provide a superset of functionality for) with other filesystems.

Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.