KVM: PPC: Book3S: PR: Rework irq disabling
authorAlexander Graf <agraf@suse.de>
Sun, 12 Aug 2012 23:04:19 +0000 (01:04 +0200)
committerAlexander Graf <agraf@suse.de>
Fri, 5 Oct 2012 21:38:45 +0000 (23:38 +0200)
commitbd2be6836ee493d41fe42367a2b129aa771185c1
tree6c3f609ad1d1b74c0e08eb616eec0715c6f7679b
parent24afa37b9c8f035d2fe2028e4824bc4e49bafe73
KVM: PPC: Book3S: PR: Rework irq disabling

Today, we disable preemption while inside guest context, because we need
to expose to the world that we are not in a preemptible context. However,
during that time we already have interrupts disabled, which would indicate
that we are in a non-preemptible context.

The reason the checks for irqs_disabled() fail for us though is that we
manually control hard IRQs and ignore all the lazy EE framework. Let's
stop doing that. Instead, let's always use lazy EE to indicate when we
want to disable IRQs, but do a special final switch that gets us into
EE disabled, but soft enabled state. That way when we get back out of
guest state, we are immediately ready to process interrupts.

This simplifies the code drastically and reduces the time that we appear
as preempt disabled.

Signed-off-by: Alexander Graf <agraf@suse.de>
arch/powerpc/include/asm/kvm_ppc.h
arch/powerpc/kvm/book3s_pr.c
arch/powerpc/kvm/book3s_rmhandlers.S
arch/powerpc/kvm/booke.c
arch/powerpc/kvm/powerpc.c