Ged Murphy wrote:
I've spoken to Mike about this, apparently the
lock is held much
higher in the call chain (EnterUserExclusive), which isn't obvious in
the patch.
Sounds like a pretty inefficient lock though.
Yes and exactly the way it's done
in Windows. All USER code is guarded
by one huge lock.
For now it's fine, maybe once we get SMP ready, we can work out a better
solution.