When we run book3s pr kvm with CONFIG_PREEMPT=y badness occurs. The reason is that we don't expect interrups to occur in the exit path until we hit the point where we enable interrupts (after the svcpu sync). However thanks to our manual interrupt replay logic we do get preempted because the interrupt handler thinks we have interrupts enabled. While debugging this I also stumbled over a register clobbering issue that this patch set also fixes. I would consider all of these patches for 3.13 with CC on stable once they're through review. Alex Alexander Graf (4): KVM: PPC: Book3S: PR: Don't clobber our exit handler id KVM: PPC: Book3S: PR: Export kvmppc_copy_to|from_svcpu KVM: PPC: Book3S: PR: Make svcpu -> vcpu store preempt savvy KVM: PPC: Book3S: PR: Enable interrupts earlier arch/powerpc/include/asm/kvm_book3s.h | 4 ++++ arch/powerpc/include/asm/kvm_book3s_asm.h | 1 + arch/powerpc/kvm/book3s_interrupts.S | 21 ++++++++++++--------- arch/powerpc/kvm/book3s_pr.c | 22 ++++++++++++++++++++++ arch/powerpc/kvm/book3s_rmhandlers.S | 6 +----- 5 files changed, 40 insertions(+), 14 deletions(-) -- 1.8.1.4 -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
