From: Chris Wilson <chris@chris-wilson.co.uk>
Date: Tue, 24 Apr 2018 15:29:45 +0100
Subject: drm/i915/breadcrumbs: Keep the fake irq armed across reset
Git-commit: 36a501a199336fbf87299ebbe598d0af00922949
Patch-mainline: v4.18-rc1
References: FATE#326289 FATE#326079 FATE#326049 FATE#322398 FATE#326166

Instead of synchronously cancelling the timer and re-enabling it inside
the reset callbacks, keep the timer enabled and let it die on its next
wakeup if no longer required. This allows
intel_engine_reset_breadcrumbs() to be used from an atomic
(timer/softirq) context such as required for resetting an engine.

It also allows us to react better to the user poking around debugfs for
testing missed irqs.

v2: Tighten the order of del_timer_sync as the fake_irq timer
may trigger the hangcheck timer, and so we should cancel it first and
then cancel the hangcheck (Mika)

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180424142945.6787-1-chris@chris-wilson.co.uk

Acked-by: Petr Tesarik <ptesarik@suse.com>
---
 drivers/gpu/drm/i915/intel_breadcrumbs.c |   29 ++++++++++++++++++++---------
 1 file changed, 20 insertions(+), 9 deletions(-)

--- a/drivers/gpu/drm/i915/intel_breadcrumbs.c
+++ b/drivers/gpu/drm/i915/intel_breadcrumbs.c
@@ -130,11 +130,12 @@ static void intel_breadcrumbs_hangcheck(
 
 static void intel_breadcrumbs_fake_irq(struct timer_list *t)
 {
-	struct intel_engine_cs *engine = from_timer(engine, t,
-						    breadcrumbs.fake_irq);
+	struct intel_engine_cs *engine =
+		from_timer(engine, t, breadcrumbs.fake_irq);
 	struct intel_breadcrumbs *b = &engine->breadcrumbs;
 
-	/* The timer persists in case we cannot enable interrupts,
+	/*
+	 * The timer persists in case we cannot enable interrupts,
 	 * or if we have previously seen seqno/interrupt incoherency
 	 * ("missed interrupt" syndrome, better known as a "missed breadcrumb").
 	 * Here the worker will wake up every jiffie in order to kick the
@@ -148,6 +149,12 @@ static void intel_breadcrumbs_fake_irq(s
 	if (!b->irq_armed)
 		return;
 
+	/* If the user has disabled the fake-irq, restore the hangchecking */
+	if (!test_bit(engine->id, &engine->i915->gpu_error.missed_irq_rings)) {
+		mod_timer(&b->hangcheck, wait_timeout());
+		return;
+	}
+
 	mod_timer(&b->fake_irq, jiffies + 1);
 }
 
@@ -831,8 +838,8 @@ static void cancel_fake_irq(struct intel
 {
 	struct intel_breadcrumbs *b = &engine->breadcrumbs;
 
+	del_timer_sync(&b->fake_irq); /* may queue b->hangcheck */
 	del_timer_sync(&b->hangcheck);
-	del_timer_sync(&b->fake_irq);
 	clear_bit(engine->id, &engine->i915->gpu_error.missed_irq_rings);
 }
 
@@ -840,15 +847,22 @@ void intel_engine_reset_breadcrumbs(stru
 {
 	struct intel_breadcrumbs *b = &engine->breadcrumbs;
 
-	cancel_fake_irq(engine);
 	spin_lock_irq(&b->irq_lock);
 
+	/*
+	 * Leave the fake_irq timer enabled (if it is running), but clear the
+	 * bit so that it turns itself off on its next wake up and goes back
+	 * to the long hangcheck interval if still required.
+	 */
+	clear_bit(engine->id, &engine->i915->gpu_error.missed_irq_rings);
+
 	if (b->irq_enabled)
 		irq_enable(engine);
 	else
 		irq_disable(engine);
 
-	/* We set the IRQ_BREADCRUMB bit when we enable the irq presuming the
+	/*
+	 * We set the IRQ_BREADCRUMB bit when we enable the irq presuming the
 	 * GPU is active and may have already executed the MI_USER_INTERRUPT
 	 * before the CPU is ready to receive. However, the engine is currently
 	 * idle (we haven't started it yet), there is no possibility for a
@@ -857,9 +871,6 @@ void intel_engine_reset_breadcrumbs(stru
 	 */
 	clear_bit(ENGINE_IRQ_BREADCRUMB, &engine->irq_posted);
 
-	if (b->irq_armed)
-		enable_fake_irq(b);
-
 	spin_unlock_irq(&b->irq_lock);
 }