Add 'zimport' to aid udev driven pool import (WIP)

[prakashsurya]

This change introduces a new userspace command 'zimport' whose intention
is to be used as a udev helper for dynamically import ZFS pools as
devices appear.

Although this is still a work in progress, the command currently
supports a stateful interface for adding and removing devices:

# To add the /dev/sda1 device to the /tmp/zimport.cache file
$ zimport -a /dev/sda1

# To remove the /dev/sda1 device to the /tmp/zimport.cache file
$ zimport -r /dev/sda1

Once the zimport cache file contains all of the vdevs for a given pool,
it will automatically attempt to import the pool. For example, to import
a pool consisting of two devices (sda1 and sdb1) run the following:

$ zimport -a /dev/sda1
$ zimport -a /dev/sdb1

It is also possible to use the '-c' option to specify a custom file name
to use as the cache file. For example, to use the '/var/run/zimport'
instead of the default one can do the following:

$ zimport -a /dev/sda1 -c /var/run/zimport

Work that still needs to be implemented includes the following:

• l2arc devices are not yet supported.

• A udev rules helper needs to be added to make use of the 'zimport'
  command. This helper should run 'zimport -a %k' to add a device as
  they appear and 'zimport -r %k' as they disappear.

• The cache file needs to be cleared and/or devices removed from the
  cache file as pools are successfully imported.

• A configuration interface is needed to specify different policies
  which define the conditions which need to be met for a pool to be
  imported. For example:

  • Which should pools should be imported?
  • Do all vdevs need to be present to import? (e.g. raidz degraded)
  • How much "slack" time between all devices online and import?

  Currently, all vdevs must be present before the pool is imported.
  Also, any pool detected will try to be imported.

• It would also be useful to be able to specify scripts or executables
  to be run at specific points during the device recognition and import
  phase.

Signed-off-by: Prakash Surya [email protected]
Issue #330

[prakashsurya]

This patch is still a work in progress, but it would be nice to get some feedback on what I have so far if anybody has the time.

[behlendorf]

Closing as stale. This is certainly still a goal and something of a work in progress.