Re: BKL removal

[Alan Cox]

> Places that call schedule() explicitly holding the BKL are rare enough
> we can probably handle them.  I have a patch that does so (thus turning
> all cond_resched() calls into no-ops with the preemptive kernel -- my
> goal).  The other implicit situations are near impossible to handle.

There are lots of them hiding 8)

> Summary is, I would love to do things like dismantle the BKLs odd-ball
> features... cleanly and safely.  Good luck ;)

You can actually do it with some testing to catch the missed cases. Move
them to spinlocks, lob a check if the lock is held into the schedule code
run it for a while through its various code paths, then remove the debug
once you trust it