Re: [Qemu-devel] [Qemu-ppc] [PATCH] ppcr: Avoid decrementer related kvm exits

[David Gibson <david@gibson.dropbear.id.au>]

On Fri, Oct 14, 2011 at 05:46:14PM +1100, David Gibson wrote:
> On Fri, Oct 14, 2011 at 08:44:06AM +0200, Alexander Graf wrote:
> > 
> > On 14.10.2011, at 08:36, David Gibson wrote:
> > 
> > > On Fri, Oct 14, 2011 at 07:30:09AM +0200, Alexander Graf wrote:
> > >> 
> > >> On 14.10.2011, at 07:19, David Gibson wrote:
> > >> 
> > >>> In __cpu_ppc_store_decr(), we set up a regular timer used to trigger
> > >>> decrementer interrupts.  This is necessary to implement the decrementer
> > >>> properly under TCG, but is unnecessary under KVM (true for both Book3S-PR
> > >>> and Book3S-HV KVM variants), because the kernel handles generating and
> > >>> delivering decrementer exceptions.
> > >>> 
> > >>> Under kvm, in fact, the timer causes expensive and unnecessary exits from
> > >>> kvm to qemu.  This patch, therefore, disables setting the timer when kvm
> > >>> is in use.
> > >>> 
> > >>> Signed-off-by: Anton Blanchard <anton@au1.ibm.com>
> > >>> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> > >>> ---
> > >>> hw/ppc.c |   25 ++++++++++++++-----------
> > >>> 1 files changed, 14 insertions(+), 11 deletions(-)
> > >>> 
> > >>> diff --git a/hw/ppc.c b/hw/ppc.c
> > >>> index 25b59dd..87aa4e5 100644
> > >>> --- a/hw/ppc.c
> > >>> +++ b/hw/ppc.c
> > >>> @@ -658,21 +658,24 @@ static void __cpu_ppc_store_decr (CPUState *env, uint64_t *nextp,
> > >> 
> > >> Do we ever call store_decr in the kvm case? Isn't that only called
> > >> from emulated mtdec?
> > > 
> > > Yes, from cpu_ppc_set_tb_clk().  Anton observed the kvm exits in the
> > > wild, they're not theoretical.
> > > 
> > > Agh, which reminds me, I forgot to fixup the git author again.  The
> > > patch should show authorship by Anton Blanchard <anton@au1.ibm.com>,
> > > as in the s-o-b.
> > 
> > Wouldn't a simple
> > 
> > if (kvm_enabled()) {
> >     return;
> > }
> > 
> > in the beginning of the function make more sense? There's no code
> > connecting the in-qemu and the in-kvm decrementors atm, so any logic
> > applying to the in-qemu one is moot for kvm.
> 
> Uh.. I guess so.  I wasn't 100% sure the last bit of code in the
> function wouldn't have some effect on kvm.  But I guess it doesn't;
> I'll revise.

Revised patch sent.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson