FreeNAS Code

Rework the samba services to be executed independent of one another and also to honor rc.d better.

The rc.d script reworking was submitted per PR # ports/164010, s.t. we can
now manipulate the states of the individual samba services instead of having
to restart the entire lot and kill services manually (I modified the AD and
CIFS services to follow the new world order, but I didn't change the LDAP code
as I don't have the required infrastructure to test the change, yet). This
saves time when manipulating AD and CIFS now and reduces service interruption
(CIFS shares stay up and running when AD is turned off, and now AD can be on
without having CIFS enabled).

Bring back the rc.d changes for this branch.

Revert to 2c0321d62f5b087ab23ca0b1185bb8222707e10d (accidentally merged the branches incorrectly).

Commit working conversion of samba rc.d scripts into something a bit more
manageable and functional.

Convert notifier over to the new world order.

All that remains is converting the rc.d scripts into a patch that can be
digested by the ports compile.

Push some other iX repo changes out into FreeNAS.

functions.sh:

Some versions of grep had an issue where the "iscompressed" function would fail to discern whether or not a file was compressed based on the file extension. Convert to a case statement instead of using egrep.

notifier.py:

Fix a typo and add an XXX comment in the notifier.sysctl method.

rc.conf.local:

A mostly style change to use $() instead of ``

settings.py:

Update copyright.

create_random_directory_tree.py:

Test tool that helped me reproduce the '/var too small' issue with samba in SF @ r9244.

Other files:

Move hardcoded strings outside in anticipation of pending avatar.conf
infrastructure work where they will no longer be needed.

Do-Not-Merge:

Reloading samba doesn't function as expected, e.g. changing global settings, adding shares, etc don't work as expected. We need to restart nmbd and smbd in order to actually update the configuration.

The unfortunate side-effect is that service can now be interrupted (shouldn't happen unless the serving machine is sufficiently slow), but it's better to be correct, than not.

Clean up cornercase where a zvol, dataset, zpool in use when deleting a zvol would delete the zvol from the GUI/db, but wouldn't actually remove things from the backend in FreeBSD.

Introduce a few new notifier APIs to help ease this pain:

__get_mountpath - get the mountpath for a volume/dataset (name)
__rmdir_mountpoint - remove a mountpoint as a means to determine whether or not the

Split storage.models.Volumes.delete into two functions again to 'clean up' the database after storage.models.Volumes._delete has been run, in such a manner that the mountpoints aren't preemptively removed, even though some of the dataset / zvol objects might have been successfully removed, and instead are removed in the event that the directory, that previously was the mountpoint, is no longer present.

These APIs are coded relatively conservatively for now to avoid putting the user into situations where they need to do database surgery.. they can and should be cleaned up later to be more explicit and straightforward by tossing proper Exceptions up the stack.

This doesn't catch the following cases:

mount -f vol

zpool destroy -f tank

when I'm in /mnt/vol or /mnt/tank (this occurs on 8.0.3 as well and this is sort of by design as they're 'force' operations). This might be a noop, or something of concern.

This change -- along with r9317 and r9318 -- resolves the issue where one couldn't export volumes via the GUI, as explained in ticket # 1126.

Make mount -ru / explicit when modifying root's ssh key is done.

Put comment in the right spot.

Remove always False case, and refactor to be slightly more optimal.