drm/i915/gt: Support multiple pinned timelines

We may need to allocate more than one pinned context/timeline for each
engine which can utilise the per-engine HWSP, so we need to give each
a different offset within it.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200730183906.25422-1-chris@chris-wilson.co.ukSigned-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

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

@@ -886,7 +886,7 @@ i915_gem_create_context(struct drm_i915_private *i915, unsigned int flags)
 	if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE) {
 		struct intel_timeline *timeline;
 
-		timeline = intel_timeline_create(&i915->gt, NULL);
+		timeline = intel_timeline_create(&i915->gt);
 		if (IS_ERR(timeline)) {
 			context_close(ctx);
 			return ERR_CAST(timeline);

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

@@ -785,9 +785,11 @@ intel_engine_init_active(struct intel_engine_cs *engine, unsigned int subclass)
 }
 
 static struct intel_context *
-create_kernel_context(struct intel_engine_cs *engine)
+create_pinned_context(struct intel_engine_cs *engine,
+		      unsigned int hwsp,
+		      struct lock_class_key *key,
+		      const char *name)
 {
-	static struct lock_class_key kernel;
 	struct intel_context *ce;
 	int err;
 
@@ -796,6 +798,7 @@ create_kernel_context(struct intel_engine_cs *engine)
 		return ce;
 
 	__set_bit(CONTEXT_BARRIER_BIT, &ce->flags);
+	ce->timeline = page_pack_bits(NULL, hwsp);
 
 	err = intel_context_pin(ce); /* perma-pin so it is always available */
 	if (err) {
@@ -809,11 +812,20 @@ create_kernel_context(struct intel_engine_cs *engine)
 	 * should we need to inject GPU operations during their request
 	 * construction.
 	 */
-	lockdep_set_class(&ce->timeline->mutex, &kernel);
+	lockdep_set_class_and_name(&ce->timeline->mutex, key, name);
 
 	return ce;
 }
 
+static struct intel_context *
+create_kernel_context(struct intel_engine_cs *engine)
+{
+	static struct lock_class_key kernel;
+
+	return create_pinned_context(engine, I915_GEM_HWS_SEQNO_ADDR,
+				     &kernel, "kernel_context");
+}
+
 /**
  * intel_engines_init_common - initialize cengine state which might require hw access
  * @engine: Engine to initialize.

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

@@ -5306,6 +5306,14 @@ populate_lr_context(struct intel_context *ce,
 	return 0;
 }
 
+static struct intel_timeline *pinned_timeline(struct intel_context *ce)
+{
+	struct intel_timeline *tl = fetch_and_zero(&ce->timeline);
+
+	return intel_timeline_create_from_engine(ce->engine,
+						 page_unmask_bits(tl));
+}
+
 static int __execlists_context_alloc(struct intel_context *ce,
 				     struct intel_engine_cs *engine)
 {
@@ -5336,19 +5344,17 @@ static int __execlists_context_alloc(struct intel_context *ce,
 		goto error_deref_obj;
 	}
 
-	if (!ce->timeline) {
+	if (!page_mask_bits(ce->timeline)) {
 		struct intel_timeline *tl;
-		struct i915_vma *hwsp;
 
 		/*
 		 * Use the static global HWSP for the kernel context, and
 		 * a dynamically allocated cacheline for everyone else.
 		 */
-		hwsp = NULL;
-		if (unlikely(intel_context_is_barrier(ce)))
-			hwsp = engine->status_page.vma;
-
-		tl = intel_timeline_create(engine->gt, hwsp);
+		if (unlikely(ce->timeline))
+			tl = pinned_timeline(ce);
+		else
+			tl = intel_timeline_create(engine->gt);
 		if (IS_ERR(tl)) {
 			ret = PTR_ERR(tl);
 			goto error_deref_obj;

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

@@ -1250,7 +1250,8 @@ int intel_ring_submission_setup(struct intel_engine_cs *engine)
 		return -ENODEV;
 	}
 
-	timeline = intel_timeline_create(engine->gt, engine->status_page.vma);
+	timeline = intel_timeline_create_from_engine(engine,
+						     I915_GEM_HWS_SEQNO_ADDR);
 	if (IS_ERR(timeline)) {
 		err = PTR_ERR(timeline);
 		goto err;

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

@@ -215,7 +215,8 @@ static void cacheline_free(struct intel_timeline_cacheline *cl)
 
 static int intel_timeline_init(struct intel_timeline *timeline,
 			       struct intel_gt *gt,
-			       struct i915_vma *hwsp)
+			       struct i915_vma *hwsp,
+			       unsigned int offset)
 {
 	void *vaddr;
 
@@ -246,8 +247,7 @@ static int intel_timeline_init(struct intel_timeline *timeline,
 
 		vaddr = page_mask_bits(cl->vaddr);
 	} else {
-		timeline->hwsp_offset = I915_GEM_HWS_SEQNO_ADDR;
-
+		timeline->hwsp_offset = offset;
 		vaddr = i915_gem_object_pin_map(hwsp->obj, I915_MAP_WB);
 		if (IS_ERR(vaddr))
 			return PTR_ERR(vaddr);
@@ -297,7 +297,9 @@ static void intel_timeline_fini(struct intel_timeline *timeline)
 }
 
 struct intel_timeline *
-intel_timeline_create(struct intel_gt *gt, struct i915_vma *global_hwsp)
+__intel_timeline_create(struct intel_gt *gt,
+			struct i915_vma *global_hwsp,
+			unsigned int offset)
 {
 	struct intel_timeline *timeline;
 	int err;
@@ -306,7 +308,7 @@ intel_timeline_create(struct intel_gt *gt, struct i915_vma *global_hwsp)
 	if (!timeline)
 		return ERR_PTR(-ENOMEM);
 
-	err = intel_timeline_init(timeline, gt, global_hwsp);
+	err = intel_timeline_init(timeline, gt, global_hwsp, offset);
 	if (err) {
 		kfree(timeline);
 		return ERR_PTR(err);

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

@@ -29,10 +29,27 @@
 
 #include "i915_active.h"
 #include "i915_syncmap.h"
-#include "gt/intel_timeline_types.h"
+#include "intel_timeline_types.h"
 
 struct intel_timeline *
-intel_timeline_create(struct intel_gt *gt, struct i915_vma *global_hwsp);
+__intel_timeline_create(struct intel_gt *gt,
+			struct i915_vma *global_hwsp,
+			unsigned int offset);
+
+static inline struct intel_timeline *
+intel_timeline_create(struct intel_gt *gt)
+{
+	return __intel_timeline_create(gt, NULL, 0);
+}
+
+static inline struct intel_timeline *
+intel_timeline_create_from_engine(struct intel_engine_cs *engine,
+				  unsigned int offset)
+{
+	return __intel_timeline_create(engine->gt,
+				       engine->status_page.vma,
+				       offset);
+}
 
 static inline struct intel_timeline *
 intel_timeline_get(struct intel_timeline *timeline)

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

@@ -153,7 +153,7 @@ static int mock_context_alloc(struct intel_context *ce)
 		return -ENOMEM;
 
 	GEM_BUG_ON(ce->timeline);
-	ce->timeline = intel_timeline_create(ce->engine->gt, NULL);
+	ce->timeline = intel_timeline_create(ce->engine->gt);
 	if (IS_ERR(ce->timeline)) {
 		kfree(ce->engine);
 		return PTR_ERR(ce->timeline);

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

@@ -72,7 +72,7 @@ static int __mock_hwsp_timeline(struct mock_hwsp_freelist *state,
 		unsigned long cacheline;
 		int err;
 
-		tl = intel_timeline_create(state->gt, NULL);
+		tl = intel_timeline_create(state->gt);
 		if (IS_ERR(tl))
 			return PTR_ERR(tl);
 
@@ -487,7 +487,7 @@ checked_intel_timeline_create(struct intel_gt *gt)
 {
 	struct intel_timeline *tl;
 
-	tl = intel_timeline_create(gt, NULL);
+	tl = intel_timeline_create(gt);
 	if (IS_ERR(tl))
 		return tl;
 
@@ -660,7 +660,7 @@ static int live_hwsp_wrap(void *arg)
 	 * foreign GPU references.
 	 */
 
-	tl = intel_timeline_create(gt, NULL);
+	tl = intel_timeline_create(gt);
 	if (IS_ERR(tl))
 		return PTR_ERR(tl);