/*
* There is a race window between reading and incrementing, but we do
* not care about potentially loosing timer events in the !reinject
- * case anyway.
+ * case anyway. Note: KVM_REQ_PENDING_TIMER is implicitly checked
+ * in vcpu_enter_guest.
*/
if (ktimer->reinject || !atomic_read(&ktimer->pending)) {
atomic_inc(&ktimer->pending);