diff mbox

[14/17,v2] drm/i915: Integrate GuC-based command submission

Message ID 1435243213-22308-15-git-send-email-david.s.gordon@intel.com (mailing list archive)
State New, archived
Headers show

Commit Message

Dave Gordon June 25, 2015, 2:40 p.m. UTC
From: Alex Dai <yu.dai@intel.com>

GuC-based submission is mostly the same as execlist mode, up to
intel_logical_ring_advance_and_submit(), where the context being
dispatched would be added to the execlist queue; at this point
we submit the context to the GuC backend instead.

There are, however, a few other changes also required, notably:
1.  Contexts must be pinned at GGTT addresses accessible by the GuC
    i.e. NOT in the range [0..WOPCM_SIZE), so we have to add the
    PIN_OFFSET_BIAS flag to the relevant GGTT-pinning calls.

2.  The GuC's TLB must be invalidated after a context is pinned at
    a new GGTT address.

3.  GuC firmware uses the one page before Ring Context as shared data.
    Therefore, whenever driver wants to get base address of LRC, we
    will offset one page for it. LRC_PPHWSP_PN is defined as the page
    number of LRCA.

4.  In the work queue used to pass requests to the GuC, the GuC
    firmware requires the ring-tail-offset to be represented as an
    11-bit value, expressed in QWords. Therefore, the ringbuffer
    size must be reduced to the representable range (4 pages).

v2:
    Defer adding #defines until needed [Chris Wilson]
    Rationalise type declarations [Chris Wilson]

Issue: VIZ-4884
Signed-off-by: Alex Dai <yu.dai@intel.com>
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
---
 drivers/gpu/drm/i915/i915_debugfs.c        |    2 +-
 drivers/gpu/drm/i915/i915_guc_submission.c |   48 +++++++++++++++++++++++++---
 drivers/gpu/drm/i915/intel_guc.h           |    1 +
 drivers/gpu/drm/i915/intel_lrc.c           |   48 ++++++++++++++++++++--------
 drivers/gpu/drm/i915/intel_lrc.h           |    6 ++++
 5 files changed, 87 insertions(+), 18 deletions(-)
diff mbox

Patch

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 820e0c5..29f6ccc 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -1957,7 +1957,7 @@  static void i915_dump_lrc_obj(struct seq_file *m,
 		return;
 	}
 
-	page = i915_gem_object_get_page(ctx_obj, 1);
+	page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN);
 	if (!WARN_ON(page == NULL)) {
 		reg_state = kmap_atomic(page);
 
diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c
index 3c28c8b..099f053 100644
--- a/drivers/gpu/drm/i915/i915_guc_submission.c
+++ b/drivers/gpu/drm/i915/i915_guc_submission.c
@@ -348,16 +348,52 @@  static void guc_init_ctx_desc(struct intel_guc *guc,
 {
 	struct guc_context_desc desc;
 	struct sg_table *sg;
+	int i;
 
 	memset(&desc, 0, sizeof(desc));
 
 	desc.attribute = GUC_CTX_DESC_ATTR_ACTIVE | GUC_CTX_DESC_ATTR_KERNEL;
 	desc.context_id = client->ctx_index;
 	desc.priority = client->priority;
-	desc.engines_used = (1 << RCS) | (1 << VCS) | (1 << BCS) |
-			    (1 << VECS) | (1 << VCS2); /* all engines */
 	desc.db_id = client->doorbell_id;
 
+	for (i = 0; i < I915_NUM_RINGS; i++) {
+		struct guc_execlist_context *lrc = &desc.lrc[i];
+		struct intel_engine_cs *ring;
+		struct drm_i915_gem_object *obj;
+
+		/* TODO: We have a design issue to be solved here. Only when we
+		 * receive the first batch, we know which engine is used by the
+		 * user. But here GuC expects the lrc and ring to be pinned. It
+		 * is not an issue for default context, which is the only one
+		 * for now who owns a GuC client. But for future owner of GuC
+		 * client, need to make sure lrc is pinned prior to enter here.
+		 */
+		obj = client->owner->engine[i].state;
+		if (!obj)
+			break;
+
+		ring = client->owner->engine[i].ringbuf->ring;
+
+		lrc->context_desc = (u32)intel_lr_context_descriptor(ring, obj);
+		/* The state page is after PPHWSP */
+		lrc->ring_lcra = i915_gem_obj_ggtt_offset(obj) +
+				LRC_STATE_PN * PAGE_SIZE;
+		lrc->context_id = (client->ctx_index << GUC_ELC_CTXID_OFFSET) |
+				(ring->id << GUC_ELC_ENGINE_OFFSET);
+
+		obj = client->owner->engine[i].ringbuf->obj;
+
+		lrc->ring_begin = i915_gem_obj_ggtt_offset(obj);
+		lrc->ring_end = lrc->ring_begin + obj->base.size - 1;
+		lrc->ring_next_free_location = lrc->ring_begin;
+		lrc->ring_current_tail_pointer_value = 0;
+
+		desc.engines_used |= (1 << ring->id);
+	}
+
+	WARN_ON(desc.engines_used == 0);
+
 	/*
 	 * The CPU address is only needed at certain points, so kmap_atomic on
 	 * demand instead of storing it in the ctx descriptor.
@@ -634,11 +670,13 @@  static void guc_client_free(struct drm_device *dev,
  * 		The kernel client to replace ExecList submission is created with
  * 		NORMAL priority. Priority of a client for scheduler can be HIGH,
  * 		while a preemption context can use CRITICAL.
+ * @ctx		the context to own the client (we use the default render context)
  *
  * Return:	An i915_guc_client object if success.
  */
 static struct i915_guc_client *guc_client_alloc(struct drm_device *dev,
-						uint32_t priority)
+						uint32_t priority,
+						struct intel_context *ctx)
 {
 	struct i915_guc_client *client;
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -651,6 +689,7 @@  static struct i915_guc_client *guc_client_alloc(struct drm_device *dev,
 
 	client->doorbell_id = GUC_INVALID_DOORBELL_ID;
 	client->priority = priority;
+	client->owner = ctx;
 
 	client->ctx_index = (uint32_t)ida_simple_get(&guc->ctx_ids, 0,
 			GUC_MAX_GPU_CONTEXTS, GFP_KERNEL);
@@ -784,10 +823,11 @@  int i915_guc_submission_enable(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_guc *guc = &dev_priv->guc;
+	struct intel_context *ctx = dev_priv->ring[RCS].default_context;
 	struct i915_guc_client *client;
 
 	/* client for execbuf submission */
-	client = guc_client_alloc(dev, GUC_CTX_PRIORITY_NORMAL);
+	client = guc_client_alloc(dev, GUC_CTX_PRIORITY_NORMAL, ctx);
 	if (!client) {
 		DRM_ERROR("Failed to create execbuf guc_client\n");
 		return -ENOMEM;
diff --git a/drivers/gpu/drm/i915/intel_guc.h b/drivers/gpu/drm/i915/intel_guc.h
index d7ffb0f..b46bd2c 100644
--- a/drivers/gpu/drm/i915/intel_guc.h
+++ b/drivers/gpu/drm/i915/intel_guc.h
@@ -30,6 +30,7 @@ 
 
 struct i915_guc_client {
 	struct drm_i915_gem_object *client_obj;
+	struct intel_context *owner;
 	uint32_t priority;
 	uint32_t ctx_index;
 
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index a77b22d..05ad7f4 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -257,7 +257,8 @@  int intel_sanitize_enable_execlists(struct drm_device *dev, int enable_execlists
  */
 u32 intel_execlists_ctx_id(struct drm_i915_gem_object *ctx_obj)
 {
-	u32 lrca = i915_gem_obj_ggtt_offset(ctx_obj);
+	u32 lrca = i915_gem_obj_ggtt_offset(ctx_obj) +
+			LRC_PPHWSP_PN * PAGE_SIZE;
 
 	/* LRCA is required to be 4K aligned so the more significant 20 bits
 	 * are globally unique */
@@ -269,7 +270,8 @@  uint64_t intel_lr_context_descriptor(struct intel_engine_cs *ring,
 {
 	struct drm_device *dev = ring->dev;
 	uint64_t desc;
-	uint64_t lrca = i915_gem_obj_ggtt_offset(ctx_obj);
+	uint64_t lrca = i915_gem_obj_ggtt_offset(ctx_obj) +
+			LRC_PPHWSP_PN * PAGE_SIZE;
 
 	WARN_ON(lrca & 0xFFFFFFFF00000FFFULL);
 
@@ -339,7 +341,7 @@  void intel_lr_context_update(struct drm_i915_gem_object *ctx_obj,
 	struct page *page;
 	uint32_t *reg_state;
 
-	page = i915_gem_object_get_page(ctx_obj, 1);
+	page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN);
 	reg_state = kmap_atomic(page);
 
 	reg_state[CTX_RING_BUFFER_START+1] = i915_gem_obj_ggtt_offset(ring_obj);
@@ -698,13 +700,17 @@  static void
 intel_logical_ring_advance_and_submit(struct drm_i915_gem_request *request)
 {
 	struct intel_engine_cs *ring = request->ring;
+	struct drm_i915_private *dev_priv = request->i915;
 
 	intel_logical_ring_advance(request->ringbuf);
 
 	if (intel_ring_stopped(ring))
 		return;
 
-	execlists_context_queue(request);
+	if (dev_priv->guc.execbuf_client)
+		i915_guc_submit(dev_priv->guc.execbuf_client, request->ctx, ring);
+	else
+		execlists_context_queue(request);
 }
 
 static int logical_ring_wrap_buffer(struct drm_i915_gem_request *req)
@@ -999,18 +1005,23 @@  static int intel_lr_context_pin(struct intel_engine_cs *ring,
 {
 	struct drm_i915_gem_object *ctx_obj = ctx->engine[ring->id].state;
 	struct intel_ringbuffer *ringbuf = ctx->engine[ring->id].ringbuf;
+	struct drm_i915_private *dev_priv = ring->dev->dev_private;
 	int ret = 0;
 
 	WARN_ON(!mutex_is_locked(&ring->dev->struct_mutex));
 	if (ctx->engine[ring->id].pin_count++ == 0) {
-		ret = i915_gem_obj_ggtt_pin(ctx_obj,
-				GEN8_LR_CONTEXT_ALIGN, 0);
+		ret = i915_gem_obj_ggtt_pin(ctx_obj, GEN8_LR_CONTEXT_ALIGN,
+				PIN_OFFSET_BIAS | GUC_WOPCM_SIZE_VALUE);
 		if (ret)
 			goto reset_pin_count;
 
 		ret = intel_pin_and_map_ringbuffer_obj(ring->dev, ringbuf);
 		if (ret)
 			goto unpin_ctx_obj;
+
+		/* Invalidate GuC TLB. */
+		if (i915.enable_guc_submission)
+			I915_WRITE(GEN8_GTCR, GEN8_GTCR_INVALIDATE);
 	}
 
 	return ret;
@@ -1647,8 +1658,13 @@  out:
 
 static int gen8_init_rcs_context(struct drm_i915_gem_request *req)
 {
+	struct drm_i915_private *dev_priv = req->i915;
 	int ret;
 
+	/* Invalidate GuC TLB. */
+	if (i915.enable_guc_submission)
+		I915_WRITE(GEN8_GTCR, GEN8_GTCR_INVALIDATE);
+
 	ret = intel_logical_ring_workarounds_emit(req);
 	if (ret)
 		return ret;
@@ -2005,7 +2021,7 @@  populate_lr_context(struct intel_context *ctx, struct drm_i915_gem_object *ctx_o
 
 	/* The second page of the context object contains some fields which must
 	 * be set up prior to the first execution. */
-	page = i915_gem_object_get_page(ctx_obj, 1);
+	page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN);
 	reg_state = kmap_atomic(page);
 
 	/* A context is actually a big batch buffer with several MI_LOAD_REGISTER_IMM
@@ -2163,12 +2179,14 @@  static void lrc_setup_hardware_status_page(struct intel_engine_cs *ring,
 		struct drm_i915_gem_object *default_ctx_obj)
 {
 	struct drm_i915_private *dev_priv = ring->dev->dev_private;
+	struct page *page;
 
 	/* The status page is offset 0 from the default context object
 	 * in LRC mode. */
-	ring->status_page.gfx_addr = i915_gem_obj_ggtt_offset(default_ctx_obj);
-	ring->status_page.page_addr =
-			kmap(sg_page(default_ctx_obj->pages->sgl));
+	ring->status_page.gfx_addr = i915_gem_obj_ggtt_offset(default_ctx_obj)
+			+ LRC_PPHWSP_PN * PAGE_SIZE;
+	page = i915_gem_object_get_page(default_ctx_obj, LRC_PPHWSP_PN);
+	ring->status_page.page_addr = kmap(page);
 	ring->status_page.obj = default_ctx_obj;
 
 	I915_WRITE(RING_HWS_PGA(ring->mmio_base),
@@ -2204,6 +2222,9 @@  int intel_lr_context_deferred_create(struct intel_context *ctx,
 
 	context_size = round_up(get_lr_context_size(ring), 4096);
 
+	/* One extra page as the sharing data between driver and GuC */
+	context_size += PAGE_SIZE * LRC_PPHWSP_PN;
+
 	ctx_obj = i915_gem_alloc_object(dev, context_size);
 	if (!ctx_obj) {
 		DRM_DEBUG_DRIVER("Alloc LRC backing obj failed.\n");
@@ -2211,7 +2232,8 @@  int intel_lr_context_deferred_create(struct intel_context *ctx,
 	}
 
 	if (is_global_default_ctx) {
-		ret = i915_gem_obj_ggtt_pin(ctx_obj, GEN8_LR_CONTEXT_ALIGN, 0);
+		ret = i915_gem_obj_ggtt_pin(ctx_obj, GEN8_LR_CONTEXT_ALIGN,
+				PIN_OFFSET_BIAS | GUC_WOPCM_SIZE_VALUE);
 		if (ret) {
 			DRM_DEBUG_DRIVER("Pin LRC backing obj failed: %d\n",
 					ret);
@@ -2230,7 +2252,7 @@  int intel_lr_context_deferred_create(struct intel_context *ctx,
 
 	ringbuf->ring = ring;
 
-	ringbuf->size = 32 * PAGE_SIZE;
+	ringbuf->size = 4 * PAGE_SIZE;
 	ringbuf->effective_size = ringbuf->size;
 	ringbuf->head = 0;
 	ringbuf->tail = 0;
@@ -2330,7 +2352,7 @@  void intel_lr_context_reset(struct drm_device *dev,
 			WARN(1, "Failed get_pages for context obj\n");
 			continue;
 		}
-		page = i915_gem_object_get_page(ctx_obj, 1);
+		page = i915_gem_object_get_page(ctx_obj, LRC_STATE_PN);
 		reg_state = kmap_atomic(page);
 
 		reg_state[CTX_RING_HEAD+1] = 0;
diff --git a/drivers/gpu/drm/i915/intel_lrc.h b/drivers/gpu/drm/i915/intel_lrc.h
index b3659a1..467157c 100644
--- a/drivers/gpu/drm/i915/intel_lrc.h
+++ b/drivers/gpu/drm/i915/intel_lrc.h
@@ -66,6 +66,12 @@  static inline void intel_logical_ring_emit(struct intel_ringbuffer *ringbuf,
 }
 
 /* Logical Ring Contexts */
+
+/* One extra page is added before LRC for GuC as shared data */
+#define LRC_GUCSHR_PN	(0)
+#define LRC_PPHWSP_PN	(LRC_GUCSHR_PN + 1)
+#define LRC_STATE_PN	(LRC_PPHWSP_PN + 1)
+
 void intel_lr_context_free(struct intel_context *ctx);
 int intel_lr_context_deferred_create(struct intel_context *ctx,
 				     struct intel_engine_cs *ring);