Re: owner field in `struct fs'

[Alan Cox]

> That's easy to solve by (as posted on netdev) making sure all CPUs are
> running a `freeze' thread on module removal.  If that's *all* it's
> for, it's both overkill for struct fs, and insufficient for other
> congruent cases.

Make sure you cover the hard real time task that wont be pre-empted case.
Your freeze thread also can deadlock very very easily. Suppose your freeze
thread pre-empts a thread holding a resource that has to be released for the
module unload.

Basically that proposal seems to be a walking priority inversion problem

Alan