drm/i915: Always kick the execlists tasklet after reset

With direct submission being disabled while the reset in progress, we
have a small window where we may forgo the submission of a new request
and not notice its addition during execlists_reset_finish. To close this
window, always schedule the submission tasklet on coming out of reset to
catch any residual work.

<6> [333.144082] i915: Running intel_hangcheck_live_selftests/igt_reset_engines
<3> [333.296927] i915_reset_engine(rcs0:idle): failed to idle after reset
<6> [333.296932] i915 0000:00:02.0: [drm] rcs0
<6> [333.296934] i915 0000:00:02.0: [drm] 	Hangcheck 0:a9ddf7a5 [4157 ms]
<6> [333.296936] i915 0000:00:02.0: [drm] 	Reset count: 36048 (global 754)
<6> [333.296938] i915 0000:00:02.0: [drm] 	Requests:
<6> [333.296997] i915 0000:00:02.0: [drm] 	RING_START: 0x00000000
<6> [333.296999] i915 0000:00:02.0: [drm] 	RING_HEAD:  0x00000000
<6> [333.297001] i915 0000:00:02.0: [drm] 	RING_TAIL:  0x00000000
<6> [333.297003] i915 0000:00:02.0: [drm] 	RING_CTL:   0x00000000
<6> [333.297005] i915 0000:00:02.0: [drm] 	RING_MODE:  0x00000200 [idle]
<6> [333.297007] i915 0000:00:02.0: [drm] 	RING_IMR: fffffeff
<6> [333.297010] i915 0000:00:02.0: [drm] 	ACTHD:  0x00000000_00000000
<6> [333.297012] i915 0000:00:02.0: [drm] 	BBADDR: 0x00000000_00000000
<6> [333.297015] i915 0000:00:02.0: [drm] 	DMA_FADDR: 0x00000000_00000000
<6> [333.297017] i915 0000:00:02.0: [drm] 	IPEIR: 0x00000000
<6> [333.297019] i915 0000:00:02.0: [drm] 	IPEHR: 0x00000000
<6> [333.297021] i915 0000:00:02.0: [drm] 	Execlist status: 0x00000001 00000000
<6> [333.297023] i915 0000:00:02.0: [drm] 	Execlist CSB read 5, write 5 [mmio:7], tasklet queued? no (enabled)
<6> [333.297025] i915 0000:00:02.0: [drm] 		ELSP[0] idle
<6> [333.297027] i915 0000:00:02.0: [drm] 		ELSP[1] idle
<6> [333.297028] i915 0000:00:02.0: [drm] 		HW active? 0x0
<6> [333.297044] i915 0000:00:02.0: [drm] 		Queue priority hint: -8186
<6> [333.297067] i915 0000:00:02.0: [drm] 		Q  2afac:5f2+  prio=-8186 @ 50ms: (null)
<6> [333.297068] i915 0000:00:02.0: [drm] HWSP:
<6> [333.297071] i915 0000:00:02.0: [drm] [0000] 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
<6> [333.297073] i915 0000:00:02.0: [drm] *
<6> [333.297075] i915 0000:00:02.0: [drm] [0040] 00000001 00000000 00000018 00000002 00000001 00000000 00000018 00000000
<6> [333.297077] i915 0000:00:02.0: [drm] [0060] 00000001 00000000 00008002 00000002 00000000 00000000 00000000 00000005
<6> [333.297079] i915 0000:00:02.0: [drm] [0080] 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
<6> [333.297081] i915 0000:00:02.0: [drm] *
<6> [333.297083] i915 0000:00:02.0: [drm] [00c0] 00000000 00000000 00000000 00000000 a9ddf7a5 00000000 00000000 00000000
<6> [333.297085] i915 0000:00:02.0: [drm] [00e0] 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
<6> [333.297087] i915 0000:00:02.0: [drm] *
<6> [333.297089] i915 0000:00:02.0: [drm] Idle? no
<6> [333.297090] i915_reset_engine(rcs0:idle): 3000 resets
<3> [333.297092] i915/intel_hangcheck_live_selftests: igt_reset_engines failed with error -5
<3> [333.455460] i915 0000:00:02.0: Failed to idle engines, declaring wedged!
...
<0> [333.491294] i915_sel-4916    1.... 333262143us : i915_reset_engine: rcs0 flags=4
<0> [333.491328] i915_sel-4916    1.... 333262143us : execlists_reset_prepare: rcs0: depth<-0
<0> [333.491362] i915_sel-4916    1.... 333262143us : intel_engine_stop_cs: rcs0
<0> [333.491396] i915_sel-4916    1d..1 333262144us : process_csb: rcs0 cs-irq head=5, tail=5
<0> [333.491424] i915_sel-4916    1.... 333262145us : intel_gpu_reset: engine_mask=1
<0> [333.491454] kworker/-214     5.... 333262184us : i915_gem_switch_to_kernel_context: awake?=yes
<0> [333.491487] kworker/-214     5.... 333262192us : i915_request_add: rcs0 fence 2afac:1522
<0> [333.491520] kworker/-214     5.... 333262193us : i915_request_add: marking (null) as active
<0> [333.491553] i915_sel-4916    1.... 333262199us : intel_engine_cancel_stop_cs: rcs0
<0> [333.491587] i915_sel-4916    1.... 333262199us : execlists_reset_finish: rcs0: depth->0

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190313162835.30228-1-chris@chris-wilson.co.uk

2 files changed, 9 insertions, 3 deletions

diff --git a/drivers/gpu/drm/i915/i915_gem.h b/drivers/gpu/drm/i915/i915_gem.h
index 74a2ddc1b52f..5c073fe73664 100644
--- a/drivers/gpu/drm/i915/i915_gem.h
+++ b/drivers/gpu/drm/i915/i915_gem.h
@@ -82,7 +82,7 @@ void i915_gem_unpark(struct drm_i915_private *i915);
 
 static inline void __tasklet_disable_sync_once(struct tasklet_struct *t)
 {
-	if (atomic_inc_return(&t->count) == 1)
+	if (!atomic_fetch_inc(&t->count))
 		tasklet_unlock_wait(t);
 }
 
@@ -91,4 +91,9 @@ static inline bool __tasklet_is_enabled(const struct tasklet_struct *t)
 	return !atomic_read(&t->count);
 }
 
+static inline bool __tasklet_enable(struct tasklet_struct *t)
+{
+	return atomic_dec_and_test(&t->count);
+}
+
 #endif /* __I915_GEM_H__ */
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 11b81fd15fab..e54e0064b2d6 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -2024,13 +2024,14 @@ static void execlists_reset_finish(struct intel_engine_cs *engine)
 	 * After a GPU reset, we may have requests to replay. Do so now while
 	 * we still have the forcewake to be sure that the GPU is not allowed
 	 * to sleep before we restart and reload a context.
-	 *
 	 */
 	GEM_BUG_ON(!reset_in_progress(execlists));
 	if (!RB_EMPTY_ROOT(&execlists->queue.rb_root))
 		execlists->tasklet.func(execlists->tasklet.data);
 
-	tasklet_enable(&execlists->tasklet);
+	if (__tasklet_enable(&execlists->tasklet))
+		/* And kick in case we missed a new request submission. */
+		tasklet_hi_schedule(&execlists->tasklet);
 	GEM_TRACE("%s: depth->%d\n", engine->name,
 		  atomic_read(&execlists->tasklet.count));
 }