@@ -447,21 +447,19 @@ static void kvm_timer_update_status(struct arch_timer_context *ctx, bool level)
static void kvm_timer_update_irq(struct kvm_vcpu *vcpu, bool new_level,
struct arch_timer_context *timer_ctx)
{
- int ret;
-
kvm_timer_update_status(timer_ctx, new_level);
timer_ctx->irq.level = new_level;
trace_kvm_timer_update_irq(vcpu->vcpu_id, timer_irq(timer_ctx),
timer_ctx->irq.level);
- if (!userspace_irqchip(vcpu->kvm)) {
- ret = kvm_vgic_inject_irq(vcpu->kvm, vcpu,
- timer_irq(timer_ctx),
- timer_ctx->irq.level,
- timer_ctx);
- WARN_ON(ret);
- }
+ if (userspace_irqchip(vcpu->kvm))
+ return;
+
+ kvm_vgic_inject_irq(vcpu->kvm, vcpu,
+ timer_irq(timer_ctx),
+ timer_ctx->irq.level,
+ timer_ctx);
}
/* Only called for a fully emulated timer */
We currently spit out a warning if making a timer interrupt pending fails. But not only this is loud and easy to trigger from userspace, we also fail to do anything useful with that information. Dropping the warning is the easiest thing to do for now. We can always add error reporting if we really want in the future. Reported-by: Alexander Potapenko <glider@google.com> Signed-off-by: Marc Zyngier <maz@kernel.org> --- arch/arm64/kvm/arch_timer.c | 16 +++++++--------- 1 file changed, 7 insertions(+), 9 deletions(-)