Deleting initiator silently deletes target <-> extent mappings

[gcooper]

As noted by creis on #freenas:

16:42 < creis> doh. removing initiators removes the target<->extent mappings...

Verified that this is still the case on r10449.

Expected behavior:

• The user should be warned that their mapping is going to be removed prior to it being removed.

Actual behavior:

See above.

[william]

What if we don't remove target and set the Authorized Initiator to None for the target?

[gcooper]

Replying to william:

What if we don't remove target and set the Authorized Initiator to None for the target?

I'll defer to Josh on this one. I think a warning's ok as I noted in the ticket above to ensure POLA is maintained, but my opinion is just part of the overall picture.

The problem with setting the value to None in my mind is that then you'd have a bunch of references that you'd have to reap later, right, what do you do when values conflict, what do we do when migrating values from one initiator to another, etc? More of what creis was noting in IRC was the fact that he was surprised when it was removed and I agree because the expected behavior isn't documented or intuitive right now.

[william]

This should be fixed in r10930