nvic: Don't apply group priority mask to negative priorities
authorPeter Maydell <peter.maydell@linaro.org>
Thu, 14 Sep 2017 17:43:17 +0000 (18:43 +0100)
committerPeter Maydell <peter.maydell@linaro.org>
Thu, 14 Sep 2017 17:43:17 +0000 (18:43 +0100)
In several places we were unconditionally applying the
nvic_gprio_mask() to a priority value. This is incorrect
if the priority is one of the fixed negative priority
values (for NMI and HardFault), so don't do it.

This bug would have caused both NMI and HardFault to be
considered as the same priority and so NMI wouldn't
correctly preempt HardFault.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 1505137930-13255-5-git-send-email-peter.maydell@linaro.org

hw/intc/armv7m_nvic.c

index 1fecfd63776637f68c1e6d1fe311d75109af870c..d3e20561c7910f8493c66549d24190af031ae325 100644 (file)
@@ -152,8 +152,12 @@ static void nvic_recompute_state(NVICState *s)
         }
     }
 
+    if (active_prio > 0) {
+        active_prio &= nvic_gprio_mask(s);
+    }
+
     s->vectpending = pend_irq;
-    s->exception_prio = active_prio & nvic_gprio_mask(s);
+    s->exception_prio = active_prio;
 
     trace_nvic_recompute_state(s->vectpending, s->exception_prio);
 }
@@ -329,7 +333,10 @@ void armv7m_nvic_acknowledge_irq(void *opaque)
     assert(vec->enabled);
     assert(vec->pending);
 
-    pendgroupprio = vec->prio & nvic_gprio_mask(s);
+    pendgroupprio = vec->prio;
+    if (pendgroupprio > 0) {
+        pendgroupprio &= nvic_gprio_mask(s);
+    }
     assert(pendgroupprio < running);
 
     trace_nvic_acknowledge_irq(pending, vec->prio);