On Mon, May 09, 2016 at 13:21:50 +0200, Paolo Bonzini wrote:
> On 30/04/2016 05:40, Emilio G. Cota wrote:
> >> The tb_flush
> >> > is a fairly rare occurrence its not like its on the critical performance
> >> > path (although of course pathological cases are possible).
> > This is what I thought from the beginning, but wanted to give this
> > alternative a go anyway to see if it was feasible.
> >
> > On my end I won't do any more work on this approach. Will go back
> > to locks, despite Paolo's (justified) dislike for them =)
>
> Which locks? tb_lock during tb_find_fast? The problem with that was
> that it slowed down everything a lot, wasn't it?
By "locks" I meant somehow forcing other threads/vcpus to stop, to then
perform tb_flush. This can be achieved with a bunch of primitives
such as condition variables (*gaaah*) -- these of course involve "locks" :P
The removal of tb_lock() when looking up tb's is orthogonal to this.
> To me, the RCU idea is not really about making tb_flush (the rare case)
> faster; it was more about keeping the rest simple and fast.
Well I agree with this idea; I wanted to see how far we could take it.
E.
