In order to allow the mock breadcrumbs tests to run without device irqs
being enabled, move the intel_irqs_enabled() assert deeper to just
before we commit to enabling the HW irq.
v2: Add a FIXME explaining that placing the assertion so deep is not
ideal, but a compromise for mock breadcrumbs.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171107102003.1802-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
 
 
 static void irq_enable(struct intel_engine_cs *engine)
 {
+       /*
+        * FIXME: Ideally we want this on the API boundary, but for the
+        * sake of testing with mock breadcrumbs (no HW so unable to
+        * enable irqs) we place it deep within the bowels, at the point
+        * of no return.
+        */
+       GEM_BUG_ON(!intel_irqs_enabled(engine->i915));
+
        /* Enabling the IRQ may miss the generation of the interrupt, but
         * we still need to force the barrier before reading the seqno,
         * just in case.
        struct drm_i915_private *i915 = engine->i915;
        bool enabled;
 
-       GEM_BUG_ON(!intel_irqs_enabled(i915));
-
        lockdep_assert_held(&b->irq_lock);
        if (b->irq_armed)
                return false;