Please help improve the aoe driver.  Contact support@coraid.com or the
aoetools-discuss mailing list (URL below) with feedback.

  http://sourceforge.net/mail/?group_id=130453

Things that need doing:

* Make sure that, for a known AoE target, a previously recorded bufcnt
  is still valid when a new config query response is received.

* Test that Neil Brown's recent changes to the md driver result in
  correct detection of idleness in aoe devices.

  Without that, the user must manually set the raid speed limits in
  /proc so that md idle detection is effectively disabled in order to
  get decent Software RAID initialization rates.

* Test the aoe driver with bonded network interfaces.

  The aoe driver could probably just recognize a bonded network
  interface and ignore it, on the assumption that the component
  network interfaces will simply be used themselves.

* Test OCFS2 on aoe devices.

* Test md over aoe device components that are over 2 terabytes in size
  each.

  (Neil Brown recently added support for large components to mdadm and
  md when the newer version 1 superblocks are used.)

* Add ioctl handling to support smartmontools.

* Find out why striped volume groups in LVM lead to data corruption or
  kernel panics.

  (Maybe this one isn't aoe-specific, but these problems have arisen
  with aoe devices as physical volumes.)

* Add the merging of the aoe to the kernel's history to the wiki at

    http://wiki.kernelnewbies.org/LinuxChanges

* Merge the latest changes from this driver into the aoe driver for
  2.4 kernels.

* Without impairing usability or complicating the user experience,
  sort out the kernel-related variables in Makefile.  For example,
  Tony Wright's patch to depmod only for the installed kernel may not
  work because users override INSTDIR and KDIR (not KVER) so that KVER
  has a stale value.
