drm/i915: Provide a timeout to i915_gem_wait_for_idle()

Usually we have no idea about the upper bound we need to wait to catch
up with userspace when idling the device, but in a few situations we
know the system was idle beforehand and can provide a short timeout in
order to very quickly catch a failure, long before hangcheck kicks in.

In the following patches, we will use the timeout to curtain two overly
long waits, where we know we can expect the GPU to complete within a
reasonable time or declare it broken.

In particular, with a broken GPU we expect it to fail during the initial
GPU setup where do a couple of context switches to record the defaults.
This is a task that takes a few milliseconds even on the slowest of
devices, but we may have to wait 60s for hangcheck to give in and
declare the machine inoperable. In this a case where any gpu hang is
unacceptable, both from a timeliness and practical standpoint.

The other improvement is that in selftests, we do not need to arm an
independent timer to inject a wedge, as we can just limit the timeout on
the wait directly.

v2: Include the timeout parameter in the trace.

Signed-off-by: Chris Wilson <[e-mail]>
Cc: Joonas Lahtinen <[e-mail]>
Cc: Mika Kuoppala <[e-mail]>
Cc: Tvrtko Ursulin <[e-mail]>
Reviewed-by: Mika Kuoppala <[e-mail]>
Link: https://patchwork.freedesktop.org/patch/msgid/[e-mail]

drivers/gpu/drm/i915/i915_debugfs.c

···
 
 	err = i915_gem_wait_for_idle(i915,
 				     I915_WAIT_LOCKED |
-				     I915_WAIT_INTERRUPTIBLE);
+				     I915_WAIT_INTERRUPTIBLE,
+				     MAX_SCHEDULE_TIMEOUT);
 	if (err)
 		goto err_unlock;
 
···
 		if (val & DROP_ACTIVE)
 			ret = i915_gem_wait_for_idle(dev_priv,
 						     I915_WAIT_INTERRUPTIBLE |
-						     I915_WAIT_LOCKED);
+						     I915_WAIT_LOCKED,
+						     MAX_SCHEDULE_TIMEOUT);
 
 		if (val & DROP_RETIRE)
 			i915_retire_requests(dev_priv);

drivers/gpu/drm/i915/i915_drv.h

···
 void i915_gem_fini(struct drm_i915_private *dev_priv);
 void i915_gem_cleanup_engines(struct drm_i915_private *dev_priv);
 int i915_gem_wait_for_idle(struct drm_i915_private *dev_priv,
-			   unsigned int flags);
+			   unsigned int flags, long timeout);
 int __must_check i915_gem_suspend(struct drm_i915_private *dev_priv);
 void i915_gem_suspend_late(struct drm_i915_private *dev_priv);
 void i915_gem_resume(struct drm_i915_private *dev_priv);

drivers/gpu/drm/i915/i915_gem.c

···
 
 	/* Attempt to reap some mmap space from dead objects */
 	do {
-		err = i915_gem_wait_for_idle(dev_priv, I915_WAIT_INTERRUPTIBLE);
+		err = i915_gem_wait_for_idle(dev_priv,
+					     I915_WAIT_INTERRUPTIBLE,
+					     MAX_SCHEDULE_TIMEOUT);
 		if (err)
 			break;
 
···
 	return ret;
 }
 
-static int wait_for_timeline(struct i915_timeline *tl, unsigned int flags)
+static long wait_for_timeline(struct i915_timeline *tl,
+			      unsigned int flags, long timeout)
 {
 	struct i915_request *rq;
-	long ret;
 
 	rq = i915_gem_active_get_unlocked(&tl->last_request);
 	if (!rq)
-		return 0;
+		return timeout;
 
 	/*
 	 * "Race-to-idle".
···
 	if (flags & I915_WAIT_FOR_IDLE_BOOST)
 		gen6_rps_boost(rq, NULL);
 
-	ret = i915_request_wait(rq, flags, MAX_SCHEDULE_TIMEOUT);
+	timeout = i915_request_wait(rq, flags, timeout);
 	i915_request_put(rq);
 
-	return ret < 0 ? ret : 0;
+	return timeout;
 }
 
 static int wait_for_engines(struct drm_i915_private *i915)
···
 	return 0;
 }
 
-int i915_gem_wait_for_idle(struct drm_i915_private *i915, unsigned int flags)
+int i915_gem_wait_for_idle(struct drm_i915_private *i915,
+			   unsigned int flags, long timeout)
 {
-	GEM_TRACE("flags=%x (%s)\n",
-		  flags, flags & I915_WAIT_LOCKED ? "locked" : "unlocked");
+	GEM_TRACE("flags=%x (%s), timeout=%ld%s\n",
+		  flags, flags & I915_WAIT_LOCKED ? "locked" : "unlocked",
+		  timeout, timeout == MAX_SCHEDULE_TIMEOUT ? " (forever)" : "");
 
 	/* If the device is asleep, we have no requests outstanding */
 	if (!READ_ONCE(i915->gt.awake))
···
 		lockdep_assert_held(&i915->drm.struct_mutex);
 
 		list_for_each_entry(tl, &i915->gt.timelines, link) {
-			err = wait_for_timeline(tl, flags);
-			if (err)
-				return err;
+			timeout = wait_for_timeline(tl, flags, timeout);
+			if (timeout < 0)
+				return timeout;
 		}
 
 		err = wait_for_engines(i915);
···
 	} else {
 		struct intel_engine_cs *engine;
 		enum intel_engine_id id;
-		int err;
 
 		for_each_engine(engine, i915, id) {
-			err = wait_for_timeline(&engine->timeline, flags);
-			if (err)
-				return err;
+			struct i915_timeline *tl = &engine->timeline;
+
+			timeout = wait_for_timeline(tl, flags, timeout);
+			if (timeout < 0)
+				return timeout;
 		}
 	}
 
···
 		ret = i915_gem_wait_for_idle(dev_priv,
 					     I915_WAIT_INTERRUPTIBLE |
 					     I915_WAIT_LOCKED |
-					     I915_WAIT_FOR_IDLE_BOOST);
+					     I915_WAIT_FOR_IDLE_BOOST,
+					     MAX_SCHEDULE_TIMEOUT);
 		if (ret && ret != -EIO)
 			goto err_unlock;
 
···
 	if (err)
 		goto err_active;
 
-	err = i915_gem_wait_for_idle(i915, I915_WAIT_LOCKED);
+	err = i915_gem_wait_for_idle(i915,
+				     I915_WAIT_LOCKED,
+				     MAX_SCHEDULE_TIMEOUT);
 	if (err)
 		goto err_active;
 
···
 	if (WARN_ON(i915_gem_switch_to_kernel_context(i915)))
 		goto out_ctx;
 
-	if (WARN_ON(i915_gem_wait_for_idle(i915, I915_WAIT_LOCKED)))
+	if (WARN_ON(i915_gem_wait_for_idle(i915,
+					   I915_WAIT_LOCKED,
+					   MAX_SCHEDULE_TIMEOUT)))
 		goto out_ctx;
 
 	i915_gem_contexts_lost(i915);

drivers/gpu/drm/i915/i915_gem_evict.c

···
 
 	err = i915_gem_wait_for_idle(i915,
 				     I915_WAIT_INTERRUPTIBLE |
-				     I915_WAIT_LOCKED);
+				     I915_WAIT_LOCKED,
+				     MAX_SCHEDULE_TIMEOUT);
 	if (err)
 		return err;
 

drivers/gpu/drm/i915/i915_gem_gtt.c

···
 	struct i915_ggtt *ggtt = &dev_priv->ggtt;
 
 	if (unlikely(ggtt->do_idle_maps)) {
-		if (i915_gem_wait_for_idle(dev_priv, 0)) {
+		if (i915_gem_wait_for_idle(dev_priv, 0, MAX_SCHEDULE_TIMEOUT)) {
 			DRM_ERROR("Failed to wait for idle; VT'd may hang.\n");
 			/* Wait a bit, in hopes it avoids the hang */
 			udelay(10);

drivers/gpu/drm/i915/i915_gem_shrinker.c

···
 	 * we will free as much as we can and hope to get a second chance.
 	 */
 	if (flags & I915_SHRINK_ACTIVE)
-		i915_gem_wait_for_idle(i915, I915_WAIT_LOCKED);
+		i915_gem_wait_for_idle(i915,
+				       I915_WAIT_LOCKED,
+				       MAX_SCHEDULE_TIMEOUT);
 
 	trace_i915_gem_shrink(i915, target, flags);
 	i915_retire_requests(i915);
···
 	unsigned long timeout = jiffies + msecs_to_jiffies_timeout(timeout_ms);
 
 	do {
-		if (i915_gem_wait_for_idle(i915, 0) == 0 &&
+		if (i915_gem_wait_for_idle(i915,
+					   0, MAX_SCHEDULE_TIMEOUT) == 0 &&
 		    shrinker_lock(i915, unlock))
 			break;
 
···
 		return NOTIFY_DONE;
 
 	/* Force everything onto the inactive lists */
-	ret = i915_gem_wait_for_idle(i915, I915_WAIT_LOCKED);
+	ret = i915_gem_wait_for_idle(i915,
+				     I915_WAIT_LOCKED,
+				     MAX_SCHEDULE_TIMEOUT);
 	if (ret)
 		goto out;
 

drivers/gpu/drm/i915/i915_perf.c

···
 	 * So far the best way to work around this issue seems to be draining
 	 * the GPU from any submitted work.
 	 */
-	ret = i915_gem_wait_for_idle(dev_priv, wait_flags);
+	ret = i915_gem_wait_for_idle(dev_priv,
+				     wait_flags,
+				     MAX_SCHEDULE_TIMEOUT);
 	if (ret)
 		goto out;
 

drivers/gpu/drm/i915/i915_request.c

···
 	/* Carefully retire all requests without writing to the rings */
 	ret = i915_gem_wait_for_idle(i915,
 				     I915_WAIT_INTERRUPTIBLE |
-				     I915_WAIT_LOCKED);
+				     I915_WAIT_LOCKED,
+				     MAX_SCHEDULE_TIMEOUT);
 	if (ret)
 		return ret;
 
···
 		/* Ratelimit ourselves to prevent oom from malicious clients */
 		ret = i915_gem_wait_for_idle(i915,
 					     I915_WAIT_LOCKED |
-					     I915_WAIT_INTERRUPTIBLE);
+					     I915_WAIT_INTERRUPTIBLE,
+					     MAX_SCHEDULE_TIMEOUT);
 		if (ret)
 			goto err_unreserve;
 

drivers/gpu/drm/i915/selftests/i915_gem_context.c

···
 		}
 	}
 
-	err = i915_gem_wait_for_idle(i915, I915_WAIT_LOCKED);
+	err = i915_gem_wait_for_idle(i915,
+				     I915_WAIT_LOCKED,
+				     MAX_SCHEDULE_TIMEOUT);
 	if (err)
 		return err;
 

drivers/gpu/drm/i915/selftests/i915_request.c

···
 	t->func = func;
 	t->name = name;
 
-	err = i915_gem_wait_for_idle(i915, I915_WAIT_LOCKED);
+	err = i915_gem_wait_for_idle(i915,
+				     I915_WAIT_LOCKED,
+				     MAX_SCHEDULE_TIMEOUT);
 	if (err) {
 		pr_err("%s(%s): failed to idle before, with err=%d!",
 		       func, name, err);

drivers/gpu/drm/i915/selftests/igt_flush_test.c

···
 	}
 
 	wedge_on_timeout(&w, i915, HZ)
-		i915_gem_wait_for_idle(i915, flags);
+		i915_gem_wait_for_idle(i915, flags, MAX_SCHEDULE_TIMEOUT);
 
 	return i915_terminally_wedged(&i915->gpu_error) ? -EIO : 0;
 }