KVM: x86: Improve thread safety in pit
authorAndy Honig <ahonig@google.com>
Wed, 27 Aug 2014 21:42:54 +0000 (14:42 -0700)
committerPaolo Bonzini <pbonzini@redhat.com>
Fri, 24 Oct 2014 11:21:14 +0000 (13:21 +0200)
There's a race condition in the PIT emulation code in KVM.  In
__kvm_migrate_pit_timer the pit_timer object is accessed without
synchronization.  If the race condition occurs at the wrong time this
can crash the host kernel.

This fixes CVE-2014-3611.

Cc: stable@vger.kernel.org
Signed-off-by: Andrew Honig <ahonig@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/i8254.c

index 518d86471b76f0be7460c6e038755a18083b26b4..298781d4cfb44b7c6d6536d6d2779ada2eeb150a 100644 (file)
@@ -262,8 +262,10 @@ void __kvm_migrate_pit_timer(struct kvm_vcpu *vcpu)
                return;
 
        timer = &pit->pit_state.timer;
+       mutex_lock(&pit->pit_state.lock);
        if (hrtimer_cancel(timer))
                hrtimer_start_expires(timer, HRTIMER_MODE_ABS);
+       mutex_unlock(&pit->pit_state.lock);
 }
 
 static void destroy_pit_timer(struct kvm_pit *pit)