s390/kvm: Fix BUG in include/linux/kvm_host.h:745

commit b080935c8638e08134629d0a9ebdf35669bec14d
    kvm: Directly account vtime to system on guest switch

also removed the irq_disable/enable around kvm guest switch, which
is correct in itself. Unfortunately, there is a BUG ON that (correctly)
checks for preemptible to cover the call to rcu later on.
(Introduced with commit 8fa2206821953a50a3a02ea33fcfb3ced2fd9997
    KVM: make guest mode entry to be rcu quiescent state)

This check might trigger depending on the kernel config.
Lets make sure that no preemption happens during kvm_guest_enter.
We can enable preemption again after the call to
rcu_virt_note_context_switch returns.

Please note that we continue to run s390 guests with interrupts
enabled.

Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
CC: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
index c9011bf..f090e81 100644 (file)
--- a/arch/s390/kvm/kvm-s390.c
+++ b/arch/s390/kvm/kvm-s390.c
@@ -613,7 +613,9 @@ static int __vcpu_run(struct kvm_vcpu *vcpu)
                kvm_s390_deliver_pending_interrupts(vcpu);
 
        vcpu->arch.sie_block->icptcode = 0;
+       preempt_disable();
        kvm_guest_enter();
+       preempt_enable();
        VCPU_EVENT(vcpu, 6, "entering sie flags %x",
                   atomic_read(&vcpu->arch.sie_block->cpuflags));
        trace_kvm_s390_sie_enter(vcpu,