drm/i915: Track gt pm wakerefs

Track every intel_gt_pm_get() until its corresponding release in
intel_gt_pm_put() by returning a cookie to the caller for acquire that
must be passed by on released. When there is an imbalance, we can see who
either tried to free a stale wakeref, or who forgot to free theirs.

v2: track recently added calls in gen8_ggtt_bind_get_ce and
    destroyed_worker_func

Signed-off-by: Andrzej Hajda <andrzej.hajda@intel.com>
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231030-ref_tracker_i915-v1-2-006fe6b96421@intel.com

drivers/gpu/drm/i915/Kconfig.debug

@@ -40,6 +40,7 @@ config DRM_I915_DEBUG
 	select DRM_I915_DEBUG_GEM_ONCE
 	select DRM_I915_DEBUG_MMIO
 	select DRM_I915_DEBUG_RUNTIME_PM
+	select DRM_I915_DEBUG_WAKEREF
 	select DRM_I915_SW_FENCE_DEBUG_OBJECTS
 	select DRM_I915_SELFTEST
 	default n
@@ -244,3 +245,16 @@ config DRM_I915_DEBUG_RUNTIME_PM
 	  Recommended for driver developers only.
 
 	  If in doubt, say "N"
+
+config DRM_I915_DEBUG_WAKEREF
+	bool "Enable extra tracking for wakerefs"
+	depends on DRM_I915
+	select REF_TRACKER
+	select STACKDEPOT
+	select STACKTRACE
+	help
+	  Choose this option to turn on extra state checking and usage
+	  tracking for the wakerefPM functionality. This may introduce
+	  overhead during driver runtime.
+
+	  If in doubt, say "N"

drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c

@@ -253,6 +253,8 @@ struct i915_execbuffer {
 	struct intel_gt *gt; /* gt for the execbuf */
 	struct intel_context *context; /* logical state for the request */
 	struct i915_gem_context *gem_context; /** caller's context */
+	intel_wakeref_t wakeref;
+	intel_wakeref_t wakeref_gt0;
 
 	/** our requests to build */
 	struct i915_request *requests[MAX_ENGINE_INSTANCE + 1];
@@ -2719,13 +2721,13 @@ eb_select_engine(struct i915_execbuffer *eb)
 
 	for_each_child(ce, child)
 		intel_context_get(child);
-	intel_gt_pm_get(gt);
+	eb->wakeref = intel_gt_pm_get(ce->engine->gt);
 	/*
 	 * Keep GT0 active on MTL so that i915_vma_parked() doesn't
 	 * free VMAs while execbuf ioctl is validating VMAs.
 	 */
 	if (gt->info.id)
-		intel_gt_pm_get(to_gt(gt->i915));
+		eb->wakeref_gt0 = intel_gt_pm_get(to_gt(gt->i915));
 
 	if (!test_bit(CONTEXT_ALLOC_BIT, &ce->flags)) {
 		err = intel_context_alloc_state(ce);
@@ -2765,9 +2767,9 @@ eb_select_engine(struct i915_execbuffer *eb)
 
 err:
 	if (gt->info.id)
-		intel_gt_pm_put(to_gt(gt->i915));
+		intel_gt_pm_put(to_gt(gt->i915), eb->wakeref_gt0);
 
-	intel_gt_pm_put(gt);
+	intel_gt_pm_put(ce->engine->gt, eb->wakeref);
 	for_each_child(ce, child)
 		intel_context_put(child);
 	intel_context_put(ce);
@@ -2785,8 +2787,8 @@ eb_put_engine(struct i915_execbuffer *eb)
 	 * i915_vma_parked() from interfering while execbuf validates vmas.
 	 */
 	if (eb->gt->info.id)
-		intel_gt_pm_put(to_gt(eb->gt->i915));
-	intel_gt_pm_put(eb->gt);
+		intel_gt_pm_put(to_gt(eb->gt->i915), eb->wakeref_gt0);
+	intel_gt_pm_put(eb->context->engine->gt, eb->wakeref);
 	for_each_child(eb->context, child)
 		intel_context_put(child);
 	intel_context_put(eb->context);

drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c

@@ -85,6 +85,7 @@ static int cpu_get(struct context *ctx, unsigned long offset, u32 *v)
 
 static int gtt_set(struct context *ctx, unsigned long offset, u32 v)
 {
+	intel_wakeref_t wakeref;
 	struct i915_vma *vma;
 	u32 __iomem *map;
 	int err = 0;
@@ -99,7 +100,7 @@ static int gtt_set(struct context *ctx, unsigned long offset, u32 v)
 	if (IS_ERR(vma))
 		return PTR_ERR(vma);
 
-	intel_gt_pm_get(vma->vm->gt);
+	wakeref = intel_gt_pm_get(vma->vm->gt);
 
 	map = i915_vma_pin_iomap(vma);
 	i915_vma_unpin(vma);
@@ -112,12 +113,13 @@ static int gtt_set(struct context *ctx, unsigned long offset, u32 v)
 	i915_vma_unpin_iomap(vma);
 
 out_rpm:
-	intel_gt_pm_put(vma->vm->gt);
+	intel_gt_pm_put(vma->vm->gt, wakeref);
 	return err;
 }
 
 static int gtt_get(struct context *ctx, unsigned long offset, u32 *v)
 {
+	intel_wakeref_t wakeref;
 	struct i915_vma *vma;
 	u32 __iomem *map;
 	int err = 0;
@@ -132,7 +134,7 @@ static int gtt_get(struct context *ctx, unsigned long offset, u32 *v)
 	if (IS_ERR(vma))
 		return PTR_ERR(vma);
 
-	intel_gt_pm_get(vma->vm->gt);
+	wakeref = intel_gt_pm_get(vma->vm->gt);
 
 	map = i915_vma_pin_iomap(vma);
 	i915_vma_unpin(vma);
@@ -145,7 +147,7 @@ static int gtt_get(struct context *ctx, unsigned long offset, u32 *v)
 	i915_vma_unpin_iomap(vma);
 
 out_rpm:
-	intel_gt_pm_put(vma->vm->gt);
+	intel_gt_pm_put(vma->vm->gt, wakeref);
 	return err;
 }
 

drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c

@@ -630,14 +630,14 @@ static bool assert_mmap_offset(struct drm_i915_private *i915,
 static void disable_retire_worker(struct drm_i915_private *i915)
 {
 	i915_gem_driver_unregister__shrinker(i915);
-	intel_gt_pm_get(to_gt(i915));
+	intel_gt_pm_get_untracked(to_gt(i915));
 	cancel_delayed_work_sync(&to_gt(i915)->requests.retire_work);
 }
 
 static void restore_retire_worker(struct drm_i915_private *i915)
 {
 	igt_flush_test(i915);
-	intel_gt_pm_put(to_gt(i915));
+	intel_gt_pm_put_untracked(to_gt(i915));
 	i915_gem_driver_register__shrinker(i915);
 }
 
@@ -778,6 +778,7 @@ static int igt_mmap_offset_exhaustion(void *arg)
 
 static int gtt_set(struct drm_i915_gem_object *obj)
 {
+	intel_wakeref_t wakeref;
 	struct i915_vma *vma;
 	void __iomem *map;
 	int err = 0;
@@ -786,7 +787,7 @@ static int gtt_set(struct drm_i915_gem_object *obj)
 	if (IS_ERR(vma))
 		return PTR_ERR(vma);
 
-	intel_gt_pm_get(vma->vm->gt);
+	wakeref = intel_gt_pm_get(vma->vm->gt);
 	map = i915_vma_pin_iomap(vma);
 	i915_vma_unpin(vma);
 	if (IS_ERR(map)) {
@@ -798,12 +799,13 @@ static int gtt_set(struct drm_i915_gem_object *obj)
 	i915_vma_unpin_iomap(vma);
 
 out:
-	intel_gt_pm_put(vma->vm->gt);
+	intel_gt_pm_put(vma->vm->gt, wakeref);
 	return err;
 }
 
 static int gtt_check(struct drm_i915_gem_object *obj)
 {
+	intel_wakeref_t wakeref;
 	struct i915_vma *vma;
 	void __iomem *map;
 	int err = 0;
@@ -812,7 +814,7 @@ static int gtt_check(struct drm_i915_gem_object *obj)
 	if (IS_ERR(vma))
 		return PTR_ERR(vma);
 
-	intel_gt_pm_get(vma->vm->gt);
+	wakeref = intel_gt_pm_get(vma->vm->gt);
 	map = i915_vma_pin_iomap(vma);
 	i915_vma_unpin(vma);
 	if (IS_ERR(map)) {
@@ -828,7 +830,7 @@ static int gtt_check(struct drm_i915_gem_object *obj)
 	i915_vma_unpin_iomap(vma);
 
 out:
-	intel_gt_pm_put(vma->vm->gt);
+	intel_gt_pm_put(vma->vm->gt, wakeref);
 	return err;
 }
 

drivers/gpu/drm/i915/gt/intel_breadcrumbs.c

@@ -28,11 +28,14 @@ static void irq_disable(struct intel_breadcrumbs *b)
 
 static void __intel_breadcrumbs_arm_irq(struct intel_breadcrumbs *b)
 {
+	intel_wakeref_t wakeref;
+
 	/*
 	 * Since we are waiting on a request, the GPU should be busy
 	 * and should have its own rpm reference.
 	 */
-	if (GEM_WARN_ON(!intel_gt_pm_get_if_awake(b->irq_engine->gt)))
+	wakeref = intel_gt_pm_get_if_awake(b->irq_engine->gt);
+	if (GEM_WARN_ON(!wakeref))
 		return;
 
 	/*
@@ -41,7 +44,7 @@ static void __intel_breadcrumbs_arm_irq(struct intel_breadcrumbs *b)
 	 * which we can add a new waiter and avoid the cost of re-enabling
 	 * the irq.
 	 */
-	WRITE_ONCE(b->irq_armed, true);
+	WRITE_ONCE(b->irq_armed, wakeref);
 
 	/* Requests may have completed before we could enable the interrupt. */
 	if (!b->irq_enabled++ && b->irq_enable(b))
@@ -61,12 +64,14 @@ static void intel_breadcrumbs_arm_irq(struct intel_breadcrumbs *b)
 
 static void __intel_breadcrumbs_disarm_irq(struct intel_breadcrumbs *b)
 {
+	intel_wakeref_t wakeref = b->irq_armed;
+
 	GEM_BUG_ON(!b->irq_enabled);
 	if (!--b->irq_enabled)
 		b->irq_disable(b);
 
-	WRITE_ONCE(b->irq_armed, false);
-	intel_gt_pm_put_async(b->irq_engine->gt);
+	WRITE_ONCE(b->irq_armed, 0);
+	intel_gt_pm_put_async(b->irq_engine->gt, wakeref);
 }
 
 static void intel_breadcrumbs_disarm_irq(struct intel_breadcrumbs *b)

drivers/gpu/drm/i915/gt/intel_breadcrumbs_types.h

@@ -13,6 +13,7 @@
 #include <linux/types.h>
 
 #include "intel_engine_types.h"
+#include "intel_wakeref.h"
 
 /*
  * Rather than have every client wait upon all user interrupts,
@@ -43,7 +44,7 @@ struct intel_breadcrumbs {
 	spinlock_t irq_lock; /* protects the interrupt from hardirq context */
 	struct irq_work irq_work; /* for use from inside irq_lock */
 	unsigned int irq_enabled;
-	bool irq_armed;
+	intel_wakeref_t irq_armed;
 
 	/* Not all breadcrumbs are attached to physical HW */
 	intel_engine_mask_t	engine_mask;

drivers/gpu/drm/i915/gt/intel_context.h

@@ -212,7 +212,7 @@ static inline void intel_context_enter(struct intel_context *ce)
 		return;
 
 	ce->ops->enter(ce);
-	intel_gt_pm_get(ce->vm->gt);
+	ce->wakeref = intel_gt_pm_get(ce->vm->gt);
 }
 
 static inline void intel_context_mark_active(struct intel_context *ce)
@@ -229,7 +229,7 @@ static inline void intel_context_exit(struct intel_context *ce)
 	if (--ce->active_count)
 		return;
 
-	intel_gt_pm_put_async(ce->vm->gt);
+	intel_gt_pm_put_async(ce->vm->gt, ce->wakeref);
 	ce->ops->exit(ce);
 }
 

drivers/gpu/drm/i915/gt/intel_context_types.h

@@ -17,6 +17,7 @@
 #include "i915_utils.h"
 #include "intel_engine_types.h"
 #include "intel_sseu.h"
+#include "intel_wakeref.h"
 
 #include "uc/intel_guc_fwif.h"
 
@@ -112,6 +113,7 @@ struct intel_context {
 	u32 ring_size;
 	struct intel_ring *ring;
 	struct intel_timeline *timeline;
+	intel_wakeref_t wakeref;
 
 	unsigned long flags;
 #define CONTEXT_BARRIER_BIT		0

drivers/gpu/drm/i915/gt/intel_engine_pm.c

@@ -63,7 +63,7 @@ static int __engine_unpark(struct intel_wakeref *wf)
 
 	ENGINE_TRACE(engine, "\n");
 
-	intel_gt_pm_get(engine->gt);
+	engine->wakeref_track = intel_gt_pm_get(engine->gt);
 
 	/* Discard stale context state from across idling */
 	ce = engine->kernel_context;
@@ -122,6 +122,7 @@ __queue_and_release_pm(struct i915_request *rq,
 	 */
 	GEM_BUG_ON(rq->context->active_count != 1);
 	__intel_gt_pm_get(engine->gt);
+	rq->context->wakeref = intel_wakeref_track(&engine->gt->wakeref);
 
 	/*
 	 * We have to serialise all potential retirement paths with our
@@ -285,7 +286,7 @@ static int __engine_park(struct intel_wakeref *wf)
 		engine->park(engine);
 
 	/* While gt calls i915_vma_parked(), we have to break the lock cycle */
-	intel_gt_pm_put_async(engine->gt);
+	intel_gt_pm_put_async(engine->gt, engine->wakeref_track);
 	return 0;
 }
 
@@ -296,7 +297,7 @@ static const struct intel_wakeref_ops wf_ops = {
 
 void intel_engine_init__pm(struct intel_engine_cs *engine)
 {
-	intel_wakeref_init(&engine->wakeref, engine->i915, &wf_ops);
+	intel_wakeref_init(&engine->wakeref, engine->i915, &wf_ops, engine->name);
 	intel_engine_init_heartbeat(engine);
 
 	intel_gsc_idle_msg_enable(engine);

drivers/gpu/drm/i915/gt/intel_engine_types.h

@@ -446,7 +446,9 @@ struct intel_engine_cs {
 	unsigned long serial;
 
 	unsigned long wakeref_serial;
+	intel_wakeref_t wakeref_track;
 	struct intel_wakeref wakeref;
+
 	struct file *default_state;
 
 	struct {

drivers/gpu/drm/i915/gt/intel_execlists_submission.c

@@ -630,7 +630,7 @@ static void __execlists_schedule_out(struct i915_request * const rq,
 	execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_OUT);
 	if (engine->fw_domain && !--engine->fw_active)
 		intel_uncore_forcewake_put(engine->uncore, engine->fw_domain);
-	intel_gt_pm_put_async(engine->gt);
+	intel_gt_pm_put_async_untracked(engine->gt);
 
 	/*
 	 * If this is part of a virtual engine, its next request may

drivers/gpu/drm/i915/gt/intel_ggtt.c

@@ -296,7 +296,7 @@ static bool should_update_ggtt_with_bind(struct i915_ggtt *ggtt)
 	return intel_gt_is_bind_context_ready(gt);
 }
 
-static struct intel_context *gen8_ggtt_bind_get_ce(struct i915_ggtt *ggtt)
+static struct intel_context *gen8_ggtt_bind_get_ce(struct i915_ggtt *ggtt, intel_wakeref_t *wakeref)
 {
 	struct intel_context *ce;
 	struct intel_gt *gt = ggtt->vm.gt;
@@ -313,18 +313,19 @@ static struct intel_context *gen8_ggtt_bind_get_ce(struct i915_ggtt *ggtt)
 	 * would conflict with fs_reclaim trying to allocate memory while
 	 * doing rpm_resume().
 	 */
-	if (!intel_gt_pm_get_if_awake(gt))
+	*wakeref = intel_gt_pm_get_if_awake(gt);
+	if (!*wakeref)
 		return NULL;
 
 	intel_engine_pm_get(ce->engine);
 
 	return ce;
 }
 
-static void gen8_ggtt_bind_put_ce(struct intel_context *ce)
+static void gen8_ggtt_bind_put_ce(struct intel_context *ce, intel_wakeref_t wakeref)
 {
 	intel_engine_pm_put(ce->engine);
-	intel_gt_pm_put(ce->engine->gt);
+	intel_gt_pm_put(ce->engine->gt, wakeref);
 }
 
 static bool gen8_ggtt_bind_ptes(struct i915_ggtt *ggtt, u32 offset,
@@ -337,12 +338,13 @@ static bool gen8_ggtt_bind_ptes(struct i915_ggtt *ggtt, u32 offset,
 	struct sgt_iter iter;
 	struct i915_request *rq;
 	struct intel_context *ce;
+	intel_wakeref_t wakeref;
 	u32 *cs;
 
 	if (!num_entries)
 		return true;
 
-	ce = gen8_ggtt_bind_get_ce(ggtt);
+	ce = gen8_ggtt_bind_get_ce(ggtt, &wakeref);
 	if (!ce)
 		return false;
 
@@ -418,13 +420,13 @@ static bool gen8_ggtt_bind_ptes(struct i915_ggtt *ggtt, u32 offset,
 		offset += n_ptes;
 	}
 
-	gen8_ggtt_bind_put_ce(ce);
+	gen8_ggtt_bind_put_ce(ce, wakeref);
 	return true;
 
 err_rq:
 	i915_request_put(rq);
 put_ce:
-	gen8_ggtt_bind_put_ce(ce);
+	gen8_ggtt_bind_put_ce(ce, wakeref);
 	return false;
 }