Re: [OKS] Module removal

[Benjamin Herrenschmidt]

>> It's not really just the module information. If I can, say, get
>> callbacks from something even after I unregister, I may well
>> have destroyed the data I need to process the callbacks, and
>> oops or worse.
>
>Actually, if module exit synchronizes properly, even the
>return-after-removal case shouldn't exist, because we'd simply
>wait for this call to return.
>
>Hmm, interesting. Did I just make the whole problem go away,
>or is travel fatigue playing tricks on my brain ? :-)

That was one of the solutions proposed by Rusty, that is basically
waiting for all CPUs to have scheduled upon exit from module_exit
and before doing the actual removal.

Ben.