1 files changed, 112 insertions, 621 deletions

diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
index 2498c55e0ea5..9e42324fdecd 100644
--- a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
+++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
@@ -18,15 +18,6 @@
 #include "i915_drv.h"
 #include "i915_trace.h"
 
-enum {
-	GUC_PREEMPT_NONE = 0,
-	GUC_PREEMPT_INPROGRESS,
-	GUC_PREEMPT_FINISHED,
-};
-#define GUC_PREEMPT_BREADCRUMB_DWORDS	0x8
-#define GUC_PREEMPT_BREADCRUMB_BYTES	\
-	(sizeof(u32) * GUC_PREEMPT_BREADCRUMB_DWORDS)
-
 /**
  * DOC: GuC-based command submission
  *
@@ -36,25 +27,14 @@ enum {
  * code) matches the old submission model and will be updated as part of the
  * upgrade to the new flow.
  *
- * GuC client:
- * A intel_guc_client refers to a submission path through GuC. Currently, there
- * is only one client, which is charged with all submissions to the GuC. This
- * struct is the owner of a doorbell, a process descriptor and a workqueue (all
- * of them inside a single gem object that contains all required pages for these
- * elements).
- *
  * GuC stage descriptor:
  * During initialization, the driver allocates a static pool of 1024 such
- * descriptors, and shares them with the GuC.
- * Currently, there exists a 1:1 mapping between a intel_guc_client and a
- * guc_stage_desc (via the client's stage_id), so effectively only one
- * gets used. This stage descriptor lets the GuC know about the doorbell,
- * workqueue and process descriptor. Theoretically, it also lets the GuC
- * know about our HW contexts (context ID, etc...), but we actually
- * employ a kind of submission where the GuC uses the LRCA sent via the work
- * item instead (the single guc_stage_desc associated to execbuf client
- * contains information about the default kernel context only, but this is
- * essentially unused). This is called a "proxy" submission.
+ * descriptors, and shares them with the GuC. Currently, we only use one
+ * descriptor. This stage descriptor lets the GuC know about the workqueue and
+ * process descriptor. Theoretically, it also lets the GuC know about our HW
+ * contexts (context ID, etc...), but we actually employ a kind of submission
+ * where the GuC uses the LRCA sent via the work item instead. This is called
+ * a "proxy" submission.
  *
  * The Scratch registers:
  * There are 16 MMIO-based registers start from 0xC180. The kernel driver writes
@@ -63,11 +43,6 @@ enum {
  * Firmware writes a success/fail code back to the action register after
  * processes the request. The kernel driver polls waiting for this update and
  * then proceeds.
- * See intel_guc_send()
- *
- * Doorbells:
- * Doorbells are interrupts to uKernel. A doorbell is a single cache line (QW)
- * mapped into process space.
  *
  * Work Items:
  * There are several types of work items that the host may place into a
@@ -84,213 +59,45 @@ static inline struct i915_priolist *to_priolist(struct rb_node *rb)
 	return rb_entry(rb, struct i915_priolist, node);
 }
 
-static inline bool is_high_priority(struct intel_guc_client *client)
+static struct guc_stage_desc *__get_stage_desc(struct intel_guc *guc, u32 id)
 {
-	return (client->priority == GUC_CLIENT_PRIORITY_KMD_HIGH ||
-		client->priority == GUC_CLIENT_PRIORITY_HIGH);
-}
-
-static int reserve_doorbell(struct intel_guc_client *client)
-{
-	unsigned long offset;
-	unsigned long end;
-	u16 id;
-
-	GEM_BUG_ON(client->doorbell_id != GUC_DOORBELL_INVALID);
+	struct guc_stage_desc *base = guc->stage_desc_pool_vaddr;
 
-	/*
-	 * The bitmap tracks which doorbell registers are currently in use.
-	 * It is split into two halves; the first half is used for normal
-	 * priority contexts, the second half for high-priority ones.
-	 */
-	offset = 0;
-	end = GUC_NUM_DOORBELLS / 2;
-	if (is_high_priority(client)) {
-		offset = end;
-		end += offset;
-	}
-
-	id = find_next_zero_bit(client->guc->doorbell_bitmap, end, offset);
-	if (id == end)
-		return -ENOSPC;
-
-	__set_bit(id, client->guc->doorbell_bitmap);
-	client->doorbell_id = id;
-	DRM_DEBUG_DRIVER("client %u (high prio=%s) reserved doorbell: %d\n",
-			 client->stage_id, yesno(is_high_priority(client)),
-			 id);
-	return 0;
+	return &base[id];
 }
 
-static bool has_doorbell(struct intel_guc_client *client)
+static int guc_workqueue_create(struct intel_guc *guc)
 {
-	if (client->doorbell_id == GUC_DOORBELL_INVALID)
-		return false;
-
-	return test_bit(client->doorbell_id, client->guc->doorbell_bitmap);
+	return intel_guc_allocate_and_map_vma(guc, GUC_WQ_SIZE, &guc->workqueue,
+					      &guc->workqueue_vaddr);
 }
 
-static void unreserve_doorbell(struct intel_guc_client *client)
+static void guc_workqueue_destroy(struct intel_guc *guc)
 {
-	GEM_BUG_ON(!has_doorbell(client));
-
-	__clear_bit(client->doorbell_id, client->guc->doorbell_bitmap);
-	client->doorbell_id = GUC_DOORBELL_INVALID;
+	i915_vma_unpin_and_release(&guc->workqueue, I915_VMA_RELEASE_MAP);
 }
 
 /*
- * Tell the GuC to allocate or deallocate a specific doorbell
- */
-
-static int __guc_allocate_doorbell(struct intel_guc *guc, u32 stage_id)
-{
-	u32 action[] = {
-		INTEL_GUC_ACTION_ALLOCATE_DOORBELL,
-		stage_id
-	};
-
-	return intel_guc_send(guc, action, ARRAY_SIZE(action));
-}
-
-static int __guc_deallocate_doorbell(struct intel_guc *guc, u32 stage_id)
-{
-	u32 action[] = {
-		INTEL_GUC_ACTION_DEALLOCATE_DOORBELL,
-		stage_id
-	};
-
-	return intel_guc_send(guc, action, ARRAY_SIZE(action));
-}
-
-static struct guc_stage_desc *__get_stage_desc(struct intel_guc_client *client)
-{
-	struct guc_stage_desc *base = client->guc->stage_desc_pool_vaddr;
-
-	return &base[client->stage_id];
-}
-
-/*
- * Initialise, update, or clear doorbell data shared with the GuC
- *
- * These functions modify shared data and so need access to the mapped
- * client object which contains the page being used for the doorbell
+ * Initialise the process descriptor shared with the GuC firmware.
  */
-
-static void __update_doorbell_desc(struct intel_guc_client *client, u16 new_id)
-{
-	struct guc_stage_desc *desc;
-
-	/* Update the GuC's idea of the doorbell ID */
-	desc = __get_stage_desc(client);
-	desc->db_id = new_id;
-}
-
-static struct guc_doorbell_info *__get_doorbell(struct intel_guc_client *client)
-{
-	return client->vaddr + client->doorbell_offset;
-}
-
-static bool __doorbell_valid(struct intel_guc *guc, u16 db_id)
-{
-	struct intel_uncore *uncore = guc_to_gt(guc)->uncore;
-
-	GEM_BUG_ON(db_id >= GUC_NUM_DOORBELLS);
-	return intel_uncore_read(uncore, GEN8_DRBREGL(db_id)) & GEN8_DRB_VALID;
-}
-
-static void __init_doorbell(struct intel_guc_client *client)
-{
-	struct guc_doorbell_info *doorbell;
-
-	doorbell = __get_doorbell(client);
-	doorbell->db_status = GUC_DOORBELL_ENABLED;
-	doorbell->cookie = 0;
-}
-
-static void __fini_doorbell(struct intel_guc_client *client)
-{
-	struct guc_doorbell_info *doorbell;
-	u16 db_id = client->doorbell_id;
-
-	doorbell = __get_doorbell(client);
-	doorbell->db_status = GUC_DOORBELL_DISABLED;
-
-	/* Doorbell release flow requires that we wait for GEN8_DRB_VALID bit
-	 * to go to zero after updating db_status before we call the GuC to
-	 * release the doorbell
-	 */
-	if (wait_for_us(!__doorbell_valid(client->guc, db_id), 10))
-		WARN_ONCE(true, "Doorbell never became invalid after disable\n");
-}
-
-static int create_doorbell(struct intel_guc_client *client)
+static int guc_proc_desc_create(struct intel_guc *guc)
 {
-	int ret;
-
-	if (WARN_ON(!has_doorbell(client)))
-		return -ENODEV; /* internal setup error, should never happen */
-
-	__update_doorbell_desc(client, client->doorbell_id);
-	__init_doorbell(client);
-
-	ret = __guc_allocate_doorbell(client->guc, client->stage_id);
-	if (ret) {
-		__fini_doorbell(client);
-		__update_doorbell_desc(client, GUC_DOORBELL_INVALID);
-		DRM_DEBUG_DRIVER("Couldn't create client %u doorbell: %d\n",
-				 client->stage_id, ret);
-		return ret;
-	}
+	const u32 size = PAGE_ALIGN(sizeof(struct guc_process_desc));
 
-	return 0;
+	return intel_guc_allocate_and_map_vma(guc, size, &guc->proc_desc,
+					      &guc->proc_desc_vaddr);
 }
 
-static int destroy_doorbell(struct intel_guc_client *client)
+static void guc_proc_desc_destroy(struct intel_guc *guc)
 {
-	int ret;
-
-	GEM_BUG_ON(!has_doorbell(client));
-
-	__fini_doorbell(client);
-	ret = __guc_deallocate_doorbell(client->guc, client->stage_id);
-	if (ret)
-		DRM_ERROR("Couldn't destroy client %u doorbell: %d\n",
-			  client->stage_id, ret);
-
-	__update_doorbell_desc(client, GUC_DOORBELL_INVALID);
-
-	return ret;
+	i915_vma_unpin_and_release(&guc->proc_desc, I915_VMA_RELEASE_MAP);
 }
 
-static unsigned long __select_cacheline(struct intel_guc *guc)
-{
-	unsigned long offset;
-
-	/* Doorbell uses a single cache line within a page */
-	offset = offset_in_page(guc->db_cacheline);
-
-	/* Moving to next cache line to reduce contention */
-	guc->db_cacheline += cache_line_size();
-
-	DRM_DEBUG_DRIVER("reserved cacheline 0x%lx, next 0x%x, linesize %u\n",
-			 offset, guc->db_cacheline, cache_line_size());
-	return offset;
-}
-
-static inline struct guc_process_desc *
-__get_process_desc(struct intel_guc_client *client)
-{
-	return client->vaddr + client->proc_desc_offset;
-}
-
-/*
- * Initialise the process descriptor shared with the GuC firmware.
- */
-static void guc_proc_desc_init(struct intel_guc_client *client)
+static void guc_proc_desc_init(struct intel_guc *guc)
 {
 	struct guc_process_desc *desc;
 
-	desc = memset(__get_process_desc(client), 0, sizeof(*desc));
+	desc = memset(guc->proc_desc_vaddr, 0, sizeof(*desc));
 
 	/*
 	 * XXX: pDoorbell and WQVBaseAddress are pointers in process address
@@ -301,47 +108,27 @@ static void guc_proc_desc_init(struct intel_guc_client *client)
 	desc->wq_base_addr = 0;
 	desc->db_base_addr = 0;
 
-	desc->stage_id = client->stage_id;
 	desc->wq_size_bytes = GUC_WQ_SIZE;
 	desc->wq_status = WQ_STATUS_ACTIVE;
-	desc->priority = client->priority;
+	desc->priority = GUC_CLIENT_PRIORITY_KMD_NORMAL;
 }
 
-static void guc_proc_desc_fini(struct intel_guc_client *client)
+static void guc_proc_desc_fini(struct intel_guc *guc)
 {
-	struct guc_process_desc *desc;
-
-	desc = __get_process_desc(client);
-	memset(desc, 0, sizeof(*desc));
+	memset(guc->proc_desc_vaddr, 0, sizeof(struct guc_process_desc));
 }
 
 static int guc_stage_desc_pool_create(struct intel_guc *guc)
 {
-	struct i915_vma *vma;
-	void *vaddr;
-
-	vma = intel_guc_allocate_vma(guc,
-				     PAGE_ALIGN(sizeof(struct guc_stage_desc) *
-				     GUC_MAX_STAGE_DESCRIPTORS));
-	if (IS_ERR(vma))
-		return PTR_ERR(vma);
-
-	vaddr = i915_gem_object_pin_map(vma->obj, I915_MAP_WB);
-	if (IS_ERR(vaddr)) {
-		i915_vma_unpin_and_release(&vma, 0);
-		return PTR_ERR(vaddr);
-	}
-
-	guc->stage_desc_pool = vma;
-	guc->stage_desc_pool_vaddr = vaddr;
-	ida_init(&guc->stage_ids);
+	u32 size = PAGE_ALIGN(sizeof(struct guc_stage_desc) *
+			      GUC_MAX_STAGE_DESCRIPTORS);
 
-	return 0;
+	return intel_guc_allocate_and_map_vma(guc, size, &guc->stage_desc_pool,
+					      &guc->stage_desc_pool_vaddr);
 }
 
 static void guc_stage_desc_pool_destroy(struct intel_guc *guc)
 {
-	ida_destroy(&guc->stage_ids);
 	i915_vma_unpin_and_release(&guc->stage_desc_pool, I915_VMA_RELEASE_MAP);
 }
 
@@ -349,63 +136,49 @@ static void guc_stage_desc_pool_destroy(struct intel_guc *guc)
  * Initialise/clear the stage descriptor shared with the GuC firmware.
  *
  * This descriptor tells the GuC where (in GGTT space) to find the important
- * data structures relating to this client (doorbell, process descriptor,
- * write queue, etc).
+ * data structures related to work submission (process descriptor, write queue,
+ * etc).
  */
-static void guc_stage_desc_init(struct intel_guc_client *client)
+static void guc_stage_desc_init(struct intel_guc *guc)
 {
-	struct intel_guc *guc = client->guc;
 	struct guc_stage_desc *desc;
-	u32 gfx_addr;
 
-	desc = __get_stage_desc(client);
+	/* we only use 1 stage desc, so hardcode it to 0 */
+	desc = __get_stage_desc(guc, 0);
 	memset(desc, 0, sizeof(*desc));
 
 	desc->attribute = GUC_STAGE_DESC_ATTR_ACTIVE |
 			  GUC_STAGE_DESC_ATTR_KERNEL;
-	if (is_high_priority(client))
-		desc->attribute |= GUC_STAGE_DESC_ATTR_PREEMPT;
-	desc->stage_id = client->stage_id;
-	desc->priority = client->priority;
-	desc->db_id = client->doorbell_id;
 
-	/*
-	 * The doorbell, process descriptor, and workqueue are all parts
-	 * of the client object, which the GuC will reference via the GGTT
-	 */
-	gfx_addr = intel_guc_ggtt_offset(guc, client->vma);
-	desc->db_trigger_phy = sg_dma_address(client->vma->pages->sgl) +
-				client->doorbell_offset;
-	desc->db_trigger_cpu = ptr_to_u64(__get_doorbell(client));
-	desc->db_trigger_uk = gfx_addr + client->doorbell_offset;
-	desc->process_desc = gfx_addr + client->proc_desc_offset;
-	desc->wq_addr = gfx_addr + GUC_DB_SIZE;
-	desc->wq_size = GUC_WQ_SIZE;
+	desc->stage_id = 0;
+	desc->priority = GUC_CLIENT_PRIORITY_KMD_NORMAL;
 
-	desc->desc_private = ptr_to_u64(client);
+	desc->process_desc = intel_guc_ggtt_offset(guc, guc->proc_desc);
+	desc->wq_addr = intel_guc_ggtt_offset(guc, guc->workqueue);
+	desc->wq_size = GUC_WQ_SIZE;
 }
 
-static void guc_stage_desc_fini(struct intel_guc_client *client)
+static void guc_stage_desc_fini(struct intel_guc *guc)
 {
 	struct guc_stage_desc *desc;
 
-	desc = __get_stage_desc(client);
+	desc = __get_stage_desc(guc, 0);
 	memset(desc, 0, sizeof(*desc));
 }
 
 /* Construct a Work Item and append it to the GuC's Work Queue */
-static void guc_wq_item_append(struct intel_guc_client *client,
+static void guc_wq_item_append(struct intel_guc *guc,
 			       u32 target_engine, u32 context_desc,
 			       u32 ring_tail, u32 fence_id)
 {
 	/* wqi_len is in DWords, and does not include the one-word header */
 	const size_t wqi_size = sizeof(struct guc_wq_item);
 	const u32 wqi_len = wqi_size / sizeof(u32) - 1;
-	struct guc_process_desc *desc = __get_process_desc(client);
+	struct guc_process_desc *desc = guc->proc_desc_vaddr;
 	struct guc_wq_item *wqi;
 	u32 wq_off;
 
-	lockdep_assert_held(&client->wq_lock);
+	lockdep_assert_held(&guc->wq_lock);
 
 	/* For now workqueue item is 4 DWs; workqueue buffer is 2 pages. So we
 	 * should not have the case where structure wqi is across page, neither
@@ -425,58 +198,30 @@ static void guc_wq_item_append(struct intel_guc_client *client,
 			      GUC_WQ_SIZE) < wqi_size);
 	GEM_BUG_ON(wq_off & (wqi_size - 1));
 
-	/* WQ starts from the page after doorbell / process_desc */
-	wqi = client->vaddr + wq_off + GUC_DB_SIZE;
-
-	if (I915_SELFTEST_ONLY(client->use_nop_wqi)) {
-		wqi->header = WQ_TYPE_NOOP | (wqi_len << WQ_LEN_SHIFT);
-	} else {
-		/* Now fill in the 4-word work queue item */
-		wqi->header = WQ_TYPE_INORDER |
-			      (wqi_len << WQ_LEN_SHIFT) |
-			      (target_engine << WQ_TARGET_SHIFT) |
-			      WQ_NO_WCFLUSH_WAIT;
-		wqi->context_desc = context_desc;
-		wqi->submit_element_info = ring_tail << WQ_RING_TAIL_SHIFT;
-		GEM_BUG_ON(ring_tail > WQ_RING_TAIL_MAX);
-		wqi->fence_id = fence_id;
-	}
+	wqi = guc->workqueue_vaddr + wq_off;
+
+	/* Now fill in the 4-word work queue item */
+	wqi->header = WQ_TYPE_INORDER |
+		      (wqi_len << WQ_LEN_SHIFT) |
+		      (target_engine << WQ_TARGET_SHIFT) |
+		      WQ_NO_WCFLUSH_WAIT;
+	wqi->context_desc = context_desc;
+	wqi->submit_element_info = ring_tail << WQ_RING_TAIL_SHIFT;
+	GEM_BUG_ON(ring_tail > WQ_RING_TAIL_MAX);
+	wqi->fence_id = fence_id;
 
 	/* Make the update visible to GuC */
 	WRITE_ONCE(desc->tail, (wq_off + wqi_size) & (GUC_WQ_SIZE - 1));
 }
 
-static void guc_ring_doorbell(struct intel_guc_client *client)
-{
-	struct guc_doorbell_info *db;
-	u32 cookie;
-
-	lockdep_assert_held(&client->wq_lock);
-
-	/* pointer of current doorbell cacheline */
-	db = __get_doorbell(client);
-
-	/*
-	 * We're not expecting the doorbell cookie to change behind our back,
-	 * we also need to treat 0 as a reserved value.
-	 */
-	cookie = READ_ONCE(db->cookie);
-	WARN_ON_ONCE(xchg(&db->cookie, cookie + 1 ?: cookie + 2) != cookie);
-
-	/* XXX: doorbell was lost and need to acquire it again */
-	GEM_BUG_ON(db->db_status != GUC_DOORBELL_ENABLED);
-}
-
 static void guc_add_request(struct intel_guc *guc, struct i915_request *rq)
 {
-	struct intel_guc_client *client = guc->execbuf_client;
 	struct intel_engine_cs *engine = rq->engine;
-	u32 ctx_desc = lower_32_bits(rq->hw_context->lrc_desc);
+	u32 ctx_desc = lower_32_bits(rq->context->lrc_desc);
 	u32 ring_tail = intel_ring_set_tail(rq->ring, rq->tail) / sizeof(u64);
 
-	guc_wq_item_append(client, engine->guc_id, ctx_desc,
+	guc_wq_item_append(guc, engine->guc_id, ctx_desc,
 			   ring_tail, rq->fence.seqno);
-	guc_ring_doorbell(client);
 }
 
 /*
@@ -488,10 +233,9 @@ static void guc_add_request(struct intel_guc *guc, struct i915_request *rq)
  */
 static void flush_ggtt_writes(struct i915_vma *vma)
 {
-	struct drm_i915_private *i915 = vma->vm->i915;
-
 	if (i915_vma_is_map_and_fenceable(vma))
-		intel_uncore_posting_read_fw(&i915->uncore, GUC_STATUS);
+		intel_uncore_posting_read_fw(vma->vm->gt->uncore,
+					     GUC_STATUS);
 }
 
 static void guc_submit(struct intel_engine_cs *engine,
@@ -499,9 +243,8 @@ static void guc_submit(struct intel_engine_cs *engine,
 		       struct i915_request **end)
 {
 	struct intel_guc *guc = &engine->gt->uc.guc;
-	struct intel_guc_client *client = guc->execbuf_client;
 
-	spin_lock(&client->wq_lock);
+	spin_lock(&guc->wq_lock);
 
 	do {
 		struct i915_request *rq = *out++;
@@ -510,7 +253,7 @@ static void guc_submit(struct intel_engine_cs *engine,
 		guc_add_request(guc, rq);
 	} while (out != end);
 
-	spin_unlock(&client->wq_lock);
+	spin_unlock(&guc->wq_lock);
 }
 
 static inline int rq_prio(const struct i915_request *rq)
@@ -529,7 +272,7 @@ static struct i915_request *schedule_in(struct i915_request *rq, int idx)
 	 * required if we generalise the inflight tracking.
 	 */
 
-	intel_gt_pm_get(rq->engine->gt);
+	__intel_gt_pm_get(rq->engine->gt);
 	return i915_request_get(rq);
 }
 
@@ -537,7 +280,7 @@ static void schedule_out(struct i915_request *rq)
 {
 	trace_i915_request_out(rq);
 
-	intel_gt_pm_put(rq->engine->gt);
+	intel_gt_pm_put_async(rq->engine->gt);
 	i915_request_put(rq);
 }
 
@@ -572,7 +315,7 @@ static void __guc_dequeue(struct intel_engine_cs *engine)
 		int i;
 
 		priolist_for_each_request_consume(rq, rn, p, i) {
-			if (last && rq->hw_context != last->hw_context) {
+			if (last && rq->context != last->context) {
 				if (port == last_port)
 					goto done;
 
@@ -631,7 +374,7 @@ static void guc_reset_prepare(struct intel_engine_cs *engine)
 {
 	struct intel_engine_execlists * const execlists = &engine->execlists;
 
-	GEM_TRACE("%s\n", engine->name);
+	ENGINE_TRACE(engine, "\n");
 
 	/*
 	 * Prevent request submission to the hardware until we have
@@ -658,7 +401,7 @@ cancel_port_requests(struct intel_engine_execlists * const execlists)
 		memset(execlists->inflight, 0, sizeof(execlists->inflight));
 }
 
-static void guc_reset(struct intel_engine_cs *engine, bool stalled)
+static void guc_reset_rewind(struct intel_engine_cs *engine, bool stalled)
 {
 	struct intel_engine_execlists * const execlists = &engine->execlists;
 	struct i915_request *rq;
@@ -677,20 +420,20 @@ static void guc_reset(struct intel_engine_cs *engine, bool stalled)
 		stalled = false;
 
 	__i915_request_reset(rq, stalled);
-	intel_lr_context_reset(engine, rq->hw_context, rq->head, stalled);
+	intel_lr_context_reset(engine, rq->context, rq->head, stalled);
 
 out_unlock:
 	spin_unlock_irqrestore(&engine->active.lock, flags);
 }
 
-static void guc_cancel_requests(struct intel_engine_cs *engine)
+static void guc_reset_cancel(struct intel_engine_cs *engine)
 {
 	struct intel_engine_execlists * const execlists = &engine->execlists;
 	struct i915_request *rq, *rn;
 	struct rb_node *rb;
 	unsigned long flags;
 
-	GEM_TRACE("%s\n", engine->name);
+	ENGINE_TRACE(engine, "\n");
 
 	/*
 	 * Before we call engine->cancel_requests(), we should have exclusive
@@ -751,8 +494,8 @@ static void guc_reset_finish(struct intel_engine_cs *engine)
 		/* 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));
+	ENGINE_TRACE(engine, "depth->%d\n",
+		     atomic_read(&execlists->tasklet.count));
 }
 
 /*
@@ -761,213 +504,6 @@ static void guc_reset_finish(struct intel_engine_cs *engine)
  * path of guc_submit() above.
  */
 
-/* Check that a doorbell register is in the expected state */
-static bool doorbell_ok(struct intel_guc *guc, u16 db_id)
-{
-	bool valid;
-
-	GEM_BUG_ON(db_id >= GUC_NUM_DOORBELLS);
-
-	valid = __doorbell_valid(guc, db_id);
-
-	if (test_bit(db_id, guc->doorbell_bitmap) == valid)
-		return true;
-
-	DRM_DEBUG_DRIVER("Doorbell %u has unexpected state: valid=%s\n",
-			 db_id, yesno(valid));
-
-	return false;
-}
-
-static bool guc_verify_doorbells(struct intel_guc *guc)
-{
-	bool doorbells_ok = true;
-	u16 db_id;
-
-	for (db_id = 0; db_id < GUC_NUM_DOORBELLS; ++db_id)
-		if (!doorbell_ok(guc, db_id))
-			doorbells_ok = false;
-
-	return doorbells_ok;
-}
-
-/**
- * guc_client_alloc() - Allocate an intel_guc_client
- * @guc:	the intel_guc structure
- * @priority:	four levels priority _CRITICAL, _HIGH, _NORMAL and _LOW
- *		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.
- *
- * Return:	An intel_guc_client object if success, else NULL.
- */
-static struct intel_guc_client *
-guc_client_alloc(struct intel_guc *guc, u32 priority)
-{
-	struct intel_guc_client *client;
-	struct i915_vma *vma;
-	void *vaddr;
-	int ret;
-
-	client = kzalloc(sizeof(*client), GFP_KERNEL);
-	if (!client)
-		return ERR_PTR(-ENOMEM);
-
-	client->guc = guc;
-	client->priority = priority;
-	client->doorbell_id = GUC_DOORBELL_INVALID;
-	spin_lock_init(&client->wq_lock);
-
-	ret = ida_simple_get(&guc->stage_ids, 0, GUC_MAX_STAGE_DESCRIPTORS,
-			     GFP_KERNEL);
-	if (ret < 0)
-		goto err_client;
-
-	client->stage_id = ret;
-
-	/* The first page is doorbell/proc_desc. Two followed pages are wq. */
-	vma = intel_guc_allocate_vma(guc, GUC_DB_SIZE + GUC_WQ_SIZE);
-	if (IS_ERR(vma)) {
-		ret = PTR_ERR(vma);
-		goto err_id;
-	}
-
-	/* We'll keep just the first (doorbell/proc) page permanently kmap'd. */
-	client->vma = vma;
-
-	vaddr = i915_gem_object_pin_map(vma->obj, I915_MAP_WB);
-	if (IS_ERR(vaddr)) {
-		ret = PTR_ERR(vaddr);
-		goto err_vma;
-	}
-	client->vaddr = vaddr;
-
-	ret = reserve_doorbell(client);
-	if (ret)
-		goto err_vaddr;
-
-	client->doorbell_offset = __select_cacheline(guc);
-
-	/*
-	 * Since the doorbell only requires a single cacheline, we can save
-	 * space by putting the application process descriptor in the same
-	 * page. Use the half of the page that doesn't include the doorbell.
-	 */
-	if (client->doorbell_offset >= (GUC_DB_SIZE / 2))
-		client->proc_desc_offset = 0;
-	else
-		client->proc_desc_offset = (GUC_DB_SIZE / 2);
-
-	DRM_DEBUG_DRIVER("new priority %u client %p: stage_id %u\n",
-			 priority, client, client->stage_id);
-	DRM_DEBUG_DRIVER("doorbell id %u, cacheline offset 0x%lx\n",
-			 client->doorbell_id, client->doorbell_offset);
-
-	return client;
-
-err_vaddr:
-	i915_gem_object_unpin_map(client->vma->obj);
-err_vma:
-	i915_vma_unpin_and_release(&client->vma, 0);
-err_id:
-	ida_simple_remove(&guc->stage_ids, client->stage_id);
-err_client:
-	kfree(client);
-	return ERR_PTR(ret);
-}
-
-static void guc_client_free(struct intel_guc_client *client)
-{
-	unreserve_doorbell(client);
-	i915_vma_unpin_and_release(&client->vma, I915_VMA_RELEASE_MAP);
-	ida_simple_remove(&client->guc->stage_ids, client->stage_id);
-	kfree(client);
-}
-
-static inline bool ctx_save_restore_disabled(struct intel_context *ce)
-{
-	u32 sr = ce->lrc_reg_state[CTX_CONTEXT_CONTROL + 1];
-
-#define SR_DISABLED \
-	_MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT | \
-			   CTX_CTRL_ENGINE_CTX_SAVE_INHIBIT)
-
-	return (sr & SR_DISABLED) == SR_DISABLED;
-
-#undef SR_DISABLED
-}
-
-static int guc_clients_create(struct intel_guc *guc)
-{
-	struct intel_guc_client *client;
-
-	GEM_BUG_ON(guc->execbuf_client);
-
-	client = guc_client_alloc(guc, GUC_CLIENT_PRIORITY_KMD_NORMAL);
-	if (IS_ERR(client)) {
-		DRM_ERROR("Failed to create GuC client for submission!\n");
-		return PTR_ERR(client);
-	}
-	guc->execbuf_client = client;
-
-	return 0;
-}
-
-static void guc_clients_destroy(struct intel_guc *guc)
-{
-	struct intel_guc_client *client;
-
-	client = fetch_and_zero(&guc->execbuf_client);
-	if (client)
-		guc_client_free(client);
-}
-
-static int __guc_client_enable(struct intel_guc_client *client)
-{
-	int ret;
-
-	guc_proc_desc_init(client);
-	guc_stage_desc_init(client);
-
-	ret = create_doorbell(client);
-	if (ret)
-		goto fail;
-
-	return 0;
-
-fail:
-	guc_stage_desc_fini(client);
-	guc_proc_desc_fini(client);
-	return ret;
-}
-
-static void __guc_client_disable(struct intel_guc_client *client)
-{
-	/*
-	 * By the time we're here, GuC may have already been reset. if that is
-	 * the case, instead of trying (in vain) to communicate with it, let's
-	 * just cleanup the doorbell HW and our internal state.
-	 */
-	if (intel_guc_is_running(client->guc))
-		destroy_doorbell(client);
-	else
-		__fini_doorbell(client);
-
-	guc_stage_desc_fini(client);
-	guc_proc_desc_fini(client);
-}
-
-static int guc_clients_enable(struct intel_guc *guc)
-{
-	return __guc_client_enable(guc->execbuf_client);
-}
-
-static void guc_clients_disable(struct intel_guc *guc)
-{
-	if (guc->execbuf_client)
-		__guc_client_disable(guc->execbuf_client);
-}
-
 /*
  * Set up the memory resources to be shared with the GuC (via the GGTT)
  * at firmware loading time.
@@ -988,13 +524,20 @@ int intel_guc_submission_init(struct intel_guc *guc)
 	 */
 	GEM_BUG_ON(!guc->stage_desc_pool);
 
-	WARN_ON(!guc_verify_doorbells(guc));
-	ret = guc_clients_create(guc);
+	ret = guc_workqueue_create(guc);
 	if (ret)
 		goto err_pool;
 
+	ret = guc_proc_desc_create(guc);
+	if (ret)
+		goto err_workqueue;
+
+	spin_lock_init(&guc->wq_lock);
+
 	return 0;
 
+err_workqueue:
+	guc_workqueue_destroy(guc);
 err_pool:
 	guc_stage_desc_pool_destroy(guc);
 	return ret;
@@ -1002,83 +545,37 @@ err_pool:
 
 void intel_guc_submission_fini(struct intel_guc *guc)
 {
-	guc_clients_destroy(guc);
-	WARN_ON(!guc_verify_doorbells(guc));
-
-	if (guc->stage_desc_pool)
+	if (guc->stage_desc_pool) {
+		guc_proc_desc_destroy(guc);
+		guc_workqueue_destroy(guc);
 		guc_stage_desc_pool_destroy(guc);
+	}
 }
 
 static void guc_interrupts_capture(struct intel_gt *gt)
 {
-	struct intel_rps *rps = &gt->rps;
 	struct intel_uncore *uncore = gt->uncore;
-	struct intel_engine_cs *engine;
-	enum intel_engine_id id;
-	int irqs;
+	u32 irqs = GT_CONTEXT_SWITCH_INTERRUPT;
+	u32 dmask = irqs << 16 | irqs;
 
-	/* tell all command streamers to forward interrupts (but not vblank)
-	 * to GuC
-	 */
-	irqs = _MASKED_BIT_ENABLE(GFX_INTERRUPT_STEERING);
-	for_each_engine(engine, gt, id)
-		ENGINE_WRITE(engine, RING_MODE_GEN7, irqs);
-
-	/* route USER_INTERRUPT to Host, all others are sent to GuC. */
-	irqs = GT_RENDER_USER_INTERRUPT << GEN8_RCS_IRQ_SHIFT |
-	       GT_RENDER_USER_INTERRUPT << GEN8_BCS_IRQ_SHIFT;
-	/* These three registers have the same bit definitions */
-	intel_uncore_write(uncore, GUC_BCS_RCS_IER, ~irqs);
-	intel_uncore_write(uncore, GUC_VCS2_VCS1_IER, ~irqs);
-	intel_uncore_write(uncore, GUC_WD_VECS_IER, ~irqs);
+	GEM_BUG_ON(INTEL_GEN(gt->i915) < 11);
 
-	/*
-	 * The REDIRECT_TO_GUC bit of the PMINTRMSK register directs all
-	 * (unmasked) PM interrupts to the GuC. All other bits of this
-	 * register *disable* generation of a specific interrupt.
-	 *
-	 * 'pm_intrmsk_mbz' indicates bits that are NOT to be set when
-	 * writing to the PM interrupt mask register, i.e. interrupts
-	 * that must not be disabled.
-	 *
-	 * If the GuC is handling these interrupts, then we must not let
-	 * the PM code disable ANY interrupt that the GuC is expecting.
-	 * So for each ENABLED (0) bit in this register, we must SET the
-	 * bit in pm_intrmsk_mbz so that it's left enabled for the GuC.
-	 * GuC needs ARAT expired interrupt unmasked hence it is set in
-	 * pm_intrmsk_mbz.
-	 *
-	 * Here we CLEAR REDIRECT_TO_GUC bit in pm_intrmsk_mbz, which will
-	 * result in the register bit being left SET!
-	 */
-	rps->pm_intrmsk_mbz |= ARAT_EXPIRED_INTRMSK;
-	rps->pm_intrmsk_mbz &= ~GEN8_PMINTR_DISABLE_REDIRECT_TO_GUC;
+	/* Don't handle the ctx switch interrupt in GuC submission mode */
+	intel_uncore_rmw(uncore, GEN11_RENDER_COPY_INTR_ENABLE, dmask, 0);
+	intel_uncore_rmw(uncore, GEN11_VCS_VECS_INTR_ENABLE, dmask, 0);
 }
 
 static void guc_interrupts_release(struct intel_gt *gt)
 {
-	struct intel_rps *rps = &gt->rps;
 	struct intel_uncore *uncore = gt->uncore;
-	struct intel_engine_cs *engine;
-	enum intel_engine_id id;
-	int irqs;
+	u32 irqs = GT_CONTEXT_SWITCH_INTERRUPT;
+	u32 dmask = irqs << 16 | irqs;
 
-	/*
-	 * tell all command streamers NOT to forward interrupts or vblank
-	 * to GuC.
-	 */
-	irqs = _MASKED_FIELD(GFX_FORWARD_VBLANK_MASK, GFX_FORWARD_VBLANK_NEVER);
-	irqs |= _MASKED_BIT_DISABLE(GFX_INTERRUPT_STEERING);
-	for_each_engine(engine, gt, id)
-		ENGINE_WRITE(engine, RING_MODE_GEN7, irqs);
-
-	/* route all GT interrupts to the host */
-	intel_uncore_write(uncore, GUC_BCS_RCS_IER, 0);
-	intel_uncore_write(uncore, GUC_VCS2_VCS1_IER, 0);
-	intel_uncore_write(uncore, GUC_WD_VECS_IER, 0);
-
-	rps->pm_intrmsk_mbz |= GEN8_PMINTR_DISABLE_REDIRECT_TO_GUC;
-	rps->pm_intrmsk_mbz &= ~ARAT_EXPIRED_INTRMSK;
+	GEM_BUG_ON(INTEL_GEN(gt->i915) < 11);
+
+	/* Handle ctx switch interrupts again */
+	intel_uncore_rmw(uncore, GEN11_RENDER_COPY_INTR_ENABLE, 0, dmask);
+	intel_uncore_rmw(uncore, GEN11_VCS_VECS_INTR_ENABLE, 0, dmask);
 }
 
 static void guc_set_default_submission(struct intel_engine_cs *engine)
@@ -1102,11 +599,10 @@ static void guc_set_default_submission(struct intel_engine_cs *engine)
 	engine->park = engine->unpark = NULL;
 
 	engine->reset.prepare = guc_reset_prepare;
-	engine->reset.reset = guc_reset;
+	engine->reset.rewind = guc_reset_rewind;
+	engine->reset.cancel = guc_reset_cancel;
 	engine->reset.finish = guc_reset_finish;
 
-	engine->cancel_requests = guc_cancel_requests;
-
 	engine->flags &= ~I915_ENGINE_SUPPORTS_STATS;
 	engine->flags |= I915_ENGINE_NEEDS_BREADCRUMB_TASKLET;
 
@@ -1119,16 +615,11 @@ static void guc_set_default_submission(struct intel_engine_cs *engine)
 	GEM_BUG_ON(engine->irq_enable || engine->irq_disable);
 }
 
-int intel_guc_submission_enable(struct intel_guc *guc)
+void intel_guc_submission_enable(struct intel_guc *guc)
 {
 	struct intel_gt *gt = guc_to_gt(guc);
 	struct intel_engine_cs *engine;
 	enum intel_engine_id id;
-	int err;
-
-	err = i915_inject_probe_error(gt->i915, -ENXIO);
-	if (err)
-		return err;
 
 	/*
 	 * We're using GuC work items for submitting work through GuC. Since
@@ -1143,11 +634,8 @@ int intel_guc_submission_enable(struct intel_guc *guc)
 		     sizeof(struct guc_wq_item) *
 		     I915_NUM_ENGINES > GUC_WQ_SIZE);
 
-	GEM_BUG_ON(!guc->execbuf_client);
-
-	err = guc_clients_enable(guc);
-	if (err)
-		return err;
+	guc_proc_desc_init(guc);
+	guc_stage_desc_init(guc);
 
 	/* Take over from manual control of ELSP (execlists) */
 	guc_interrupts_capture(gt);
@@ -1156,8 +644,6 @@ int intel_guc_submission_enable(struct intel_guc *guc)
 		engine->set_default_submission = guc_set_default_submission;
 		engine->set_default_submission(engine);
 	}
-
-	return 0;
 }
 
 void intel_guc_submission_disable(struct intel_guc *guc)
@@ -1166,8 +652,12 @@ void intel_guc_submission_disable(struct intel_guc *guc)
 
 	GEM_BUG_ON(gt->awake); /* GT should be parked first */
 
+	/* Note: By the time we're here, GuC may have already been reset */
+
 	guc_interrupts_release(gt);
-	guc_clients_disable(guc);
+
+	guc_stage_desc_fini(guc);
+	guc_proc_desc_fini(guc);
 }
 
 static bool __guc_submission_support(struct intel_guc *guc)
@@ -1186,6 +676,7 @@ void intel_guc_submission_init_early(struct intel_guc *guc)
 	guc->submission_supported = __guc_submission_support(guc);
 }
 
-#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST)
-#include "selftest_guc.c"
-#endif
+bool intel_engine_in_guc_submission_mode(const struct intel_engine_cs *engine)
+{
+	return engine->set_default_submission == guc_set_default_submission;
+}