Merge tag 'drm-intel-gt-next-2021-10-21' of git://anongit.freedesktop.org/drm/drm-intel into drm-next

UAPI Changes:

- Expose multi-LRC submission interface

  Similar to the bonded submission interface but simplified.
  Comes with GuC only implementation for now. See kerneldoc
  for more details.

  Userspace changes: https://github.com/intel/media-driver/pull/1252

- Expose logical engine instance to user

  Needed by the multi-LRC submission interface for GuC

  Userspace changes: https://github.com/intel/media-driver/pull/1252

Driver Changes:

- Fix blank screen booting crashes when CONFIG_CC_OPTIMIZE_FOR_SIZE=y (Hugh)
- Add support for multi-LRC submission in the GuC backend (Matt B)
- Add extra cache flushing before making pages userspace visible (Matt A, Thomas)
- Mark internal GPU object pages dirty so they will be flushed properly (Matt A)

- Move remaining debugfs interfaces i915_wedged/i915_forcewake_user into gt (Andi)
- Replace the unconditional clflushes with drm_clflush_virt_range() (Ville)
- Remove IS_ACTIVE macro completely (Lucas)
- Improve kerneldocs for cache_dirty (Matt A)

- Add missing includes (Lucas)
- Selftest improvements (Matt R, Ran, Matt A)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/YXFmLKoq8Fg9JxSd@jlahtine-mobl.ger.corp.intel.com

13 files changed, 947 insertions, 288 deletions

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_busy.c b/drivers/gpu/drm/i915/gem/i915_gem_busy.c
index 6234e17259c1..7358bebef15c 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_busy.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_busy.c
@@ -4,6 +4,8 @@
  * Copyright © 2014-2016 Intel Corporation
  */
 
+#include <linux/dma-fence-array.h>
+
 #include "gt/intel_engine.h"
 
 #include "i915_gem_ioctls.h"
@@ -36,7 +38,7 @@ static __always_inline u32 __busy_write_id(u16 id)
 }
 
 static __always_inline unsigned int
-__busy_set_if_active(const struct dma_fence *fence, u32 (*flag)(u16 id))
+__busy_set_if_active(struct dma_fence *fence, u32 (*flag)(u16 id))
 {
 	const struct i915_request *rq;
 
@@ -46,29 +48,60 @@ __busy_set_if_active(const struct dma_fence *fence, u32 (*flag)(u16 id))
 	 * to eventually flush us, but to minimise latency just ask the
 	 * hardware.
 	 *
-	 * Note we only report on the status of native fences.
+	 * Note we only report on the status of native fences and we currently
+	 * have two native fences:
+	 *
+	 * 1. A composite fence (dma_fence_array) constructed of i915 requests
+	 * created during a parallel submission. In this case we deconstruct the
+	 * composite fence into individual i915 requests and check the status of
+	 * each request.
+	 *
+	 * 2. A single i915 request.
 	 */
-	if (!dma_fence_is_i915(fence))
+	if (dma_fence_is_array(fence)) {
+		struct dma_fence_array *array = to_dma_fence_array(fence);
+		struct dma_fence **child = array->fences;
+		unsigned int nchild = array->num_fences;
+
+		do {
+			struct dma_fence *current_fence = *child++;
+
+			/* Not an i915 fence, can't be busy per above */
+			if (!dma_fence_is_i915(current_fence) ||
+			    !test_bit(I915_FENCE_FLAG_COMPOSITE,
+				      &current_fence->flags)) {
+				return 0;
+			}
+
+			rq = to_request(current_fence);
+			if (!i915_request_completed(rq))
+				return flag(rq->engine->uabi_class);
+		} while (--nchild);
+
+		/* All requests in array complete, not busy */
 		return 0;
+	} else {
+		if (!dma_fence_is_i915(fence))
+			return 0;
 
-	/* opencode to_request() in order to avoid const warnings */
-	rq = container_of(fence, const struct i915_request, fence);
-	if (i915_request_completed(rq))
-		return 0;
+		rq = to_request(fence);
+		if (i915_request_completed(rq))
+			return 0;
 
-	/* Beware type-expansion follies! */
-	BUILD_BUG_ON(!typecheck(u16, rq->engine->uabi_class));
-	return flag(rq->engine->uabi_class);
+		/* Beware type-expansion follies! */
+		BUILD_BUG_ON(!typecheck(u16, rq->engine->uabi_class));
+		return flag(rq->engine->uabi_class);
+	}
 }
 
 static __always_inline unsigned int
-busy_check_reader(const struct dma_fence *fence)
+busy_check_reader(struct dma_fence *fence)
 {
 	return __busy_set_if_active(fence, __busy_read_flag);
 }
 
 static __always_inline unsigned int
-busy_check_writer(const struct dma_fence *fence)
+busy_check_writer(struct dma_fence *fence)
 {
 	if (!fence)
 		return 0;
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index d225d3dd0b40..fb33d0322960 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -556,9 +556,147 @@ set_proto_ctx_engines_bond(struct i915_user_extension __user *base, void *data)
 	return 0;
 }
 
+static int
+set_proto_ctx_engines_parallel_submit(struct i915_user_extension __user *base,
+				      void *data)
+{
+	struct i915_context_engines_parallel_submit __user *ext =
+		container_of_user(base, typeof(*ext), base);
+	const struct set_proto_ctx_engines *set = data;
+	struct drm_i915_private *i915 = set->i915;
+	u64 flags;
+	int err = 0, n, i, j;
+	u16 slot, width, num_siblings;
+	struct intel_engine_cs **siblings = NULL;
+	intel_engine_mask_t prev_mask;
+
+	/* FIXME: This is NIY for execlists */
+	if (!(intel_uc_uses_guc_submission(&i915->gt.uc)))
+		return -ENODEV;
+
+	if (get_user(slot, &ext->engine_index))
+		return -EFAULT;
+
+	if (get_user(width, &ext->width))
+		return -EFAULT;
+
+	if (get_user(num_siblings, &ext->num_siblings))
+		return -EFAULT;
+
+	if (slot >= set->num_engines) {
+		drm_dbg(&i915->drm, "Invalid placement value, %d >= %d\n",
+			slot, set->num_engines);
+		return -EINVAL;
+	}
+
+	if (set->engines[slot].type != I915_GEM_ENGINE_TYPE_INVALID) {
+		drm_dbg(&i915->drm,
+			"Invalid placement[%d], already occupied\n", slot);
+		return -EINVAL;
+	}
+
+	if (get_user(flags, &ext->flags))
+		return -EFAULT;
+
+	if (flags) {
+		drm_dbg(&i915->drm, "Unknown flags 0x%02llx", flags);
+		return -EINVAL;
+	}
+
+	for (n = 0; n < ARRAY_SIZE(ext->mbz64); n++) {
+		err = check_user_mbz(&ext->mbz64[n]);
+		if (err)
+			return err;
+	}
+
+	if (width < 2) {
+		drm_dbg(&i915->drm, "Width (%d) < 2\n", width);
+		return -EINVAL;
+	}
+
+	if (num_siblings < 1) {
+		drm_dbg(&i915->drm, "Number siblings (%d) < 1\n",
+			num_siblings);
+		return -EINVAL;
+	}
+
+	siblings = kmalloc_array(num_siblings * width,
+				 sizeof(*siblings),
+				 GFP_KERNEL);
+	if (!siblings)
+		return -ENOMEM;
+
+	/* Create contexts / engines */
+	for (i = 0; i < width; ++i) {
+		intel_engine_mask_t current_mask = 0;
+		struct i915_engine_class_instance prev_engine;
+
+		for (j = 0; j < num_siblings; ++j) {
+			struct i915_engine_class_instance ci;
+
+			n = i * num_siblings + j;
+			if (copy_from_user(&ci, &ext->engines[n], sizeof(ci))) {
+				err = -EFAULT;
+				goto out_err;
+			}
+
+			siblings[n] =
+				intel_engine_lookup_user(i915, ci.engine_class,
+							 ci.engine_instance);
+			if (!siblings[n]) {
+				drm_dbg(&i915->drm,
+					"Invalid sibling[%d]: { class:%d, inst:%d }\n",
+					n, ci.engine_class, ci.engine_instance);
+				err = -EINVAL;
+				goto out_err;
+			}
+
+			if (n) {
+				if (prev_engine.engine_class !=
+				    ci.engine_class) {
+					drm_dbg(&i915->drm,
+						"Mismatched class %d, %d\n",
+						prev_engine.engine_class,
+						ci.engine_class);
+					err = -EINVAL;
+					goto out_err;
+				}
+			}
+
+			prev_engine = ci;
+			current_mask |= siblings[n]->logical_mask;
+		}
+
+		if (i > 0) {
+			if (current_mask != prev_mask << 1) {
+				drm_dbg(&i915->drm,
+					"Non contiguous logical mask 0x%x, 0x%x\n",
+					prev_mask, current_mask);
+				err = -EINVAL;
+				goto out_err;
+			}
+		}
+		prev_mask = current_mask;
+	}
+
+	set->engines[slot].type = I915_GEM_ENGINE_TYPE_PARALLEL;
+	set->engines[slot].num_siblings = num_siblings;
+	set->engines[slot].width = width;
+	set->engines[slot].siblings = siblings;
+
+	return 0;
+
+out_err:
+	kfree(siblings);
+
+	return err;
+}
+
 static const i915_user_extension_fn set_proto_ctx_engines_extensions[] = {
 	[I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE] = set_proto_ctx_engines_balance,
 	[I915_CONTEXT_ENGINES_EXT_BOND] = set_proto_ctx_engines_bond,
+	[I915_CONTEXT_ENGINES_EXT_PARALLEL_SUBMIT] =
+		set_proto_ctx_engines_parallel_submit,
 };
 
 static int set_proto_ctx_engines(struct drm_i915_file_private *fpriv,
@@ -794,6 +932,7 @@ static int intel_context_set_gem(struct intel_context *ce,
 	GEM_BUG_ON(rcu_access_pointer(ce->gem_context));
 	RCU_INIT_POINTER(ce->gem_context, ctx);
 
+	GEM_BUG_ON(intel_context_is_pinned(ce));
 	ce->ring_size = SZ_16K;
 
 	i915_vm_put(ce->vm);
@@ -818,6 +957,25 @@ static int intel_context_set_gem(struct intel_context *ce,
 	return ret;
 }
 
+static void __unpin_engines(struct i915_gem_engines *e, unsigned int count)
+{
+	while (count--) {
+		struct intel_context *ce = e->engines[count], *child;
+
+		if (!ce || !test_bit(CONTEXT_PERMA_PIN, &ce->flags))
+			continue;
+
+		for_each_child(ce, child)
+			intel_context_unpin(child);
+		intel_context_unpin(ce);
+	}
+}
+
+static void unpin_engines(struct i915_gem_engines *e)
+{
+	__unpin_engines(e, e->num_engines);
+}
+
 static void __free_engines(struct i915_gem_engines *e, unsigned int count)
 {
 	while (count--) {
@@ -933,6 +1091,40 @@ free_engines:
 	return err;
 }
 
+static int perma_pin_contexts(struct intel_context *ce)
+{
+	struct intel_context *child;
+	int i = 0, j = 0, ret;
+
+	GEM_BUG_ON(!intel_context_is_parent(ce));
+
+	ret = intel_context_pin(ce);
+	if (unlikely(ret))
+		return ret;
+
+	for_each_child(ce, child) {
+		ret = intel_context_pin(child);
+		if (unlikely(ret))
+			goto unwind;
+		++i;
+	}
+
+	set_bit(CONTEXT_PERMA_PIN, &ce->flags);
+
+	return 0;
+
+unwind:
+	intel_context_unpin(ce);
+	for_each_child(ce, child) {
+		if (j++ < i)
+			intel_context_unpin(child);
+		else
+			break;
+	}
+
+	return ret;
+}
+
 static struct i915_gem_engines *user_engines(struct i915_gem_context *ctx,
 					     unsigned int num_engines,
 					     struct i915_gem_proto_engine *pe)
@@ -946,7 +1138,7 @@ static struct i915_gem_engines *user_engines(struct i915_gem_context *ctx,
 	e->num_engines = num_engines;
 
 	for (n = 0; n < num_engines; n++) {
-		struct intel_context *ce;
+		struct intel_context *ce, *child;
 		int ret;
 
 		switch (pe[n].type) {
@@ -956,7 +1148,13 @@ static struct i915_gem_engines *user_engines(struct i915_gem_context *ctx,
 
 		case I915_GEM_ENGINE_TYPE_BALANCED:
 			ce = intel_engine_create_virtual(pe[n].siblings,
-							 pe[n].num_siblings);
+							 pe[n].num_siblings, 0);
+			break;
+
+		case I915_GEM_ENGINE_TYPE_PARALLEL:
+			ce = intel_engine_create_parallel(pe[n].siblings,
+							  pe[n].num_siblings,
+							  pe[n].width);
 			break;
 
 		case I915_GEM_ENGINE_TYPE_INVALID:
@@ -977,6 +1175,30 @@ static struct i915_gem_engines *user_engines(struct i915_gem_context *ctx,
 			err = ERR_PTR(ret);
 			goto free_engines;
 		}
+		for_each_child(ce, child) {
+			ret = intel_context_set_gem(child, ctx, pe->sseu);
+			if (ret) {
+				err = ERR_PTR(ret);
+				goto free_engines;
+			}
+		}
+
+		/*
+		 * XXX: Must be done after calling intel_context_set_gem as that
+		 * function changes the ring size. The ring is allocated when
+		 * the context is pinned. If the ring size is changed after
+		 * allocation we have a mismatch of the ring size and will cause
+		 * the context to hang. Presumably with a bit of reordering we
+		 * could move the perma-pin step to the backend function
+		 * intel_engine_create_parallel.
+		 */
+		if (pe[n].type == I915_GEM_ENGINE_TYPE_PARALLEL) {
+			ret = perma_pin_contexts(ce);
+			if (ret) {
+				err = ERR_PTR(ret);
+				goto free_engines;
+			}
+		}
 	}
 
 	return e;
@@ -1219,6 +1441,7 @@ static void context_close(struct i915_gem_context *ctx)
 
 	/* Flush any concurrent set_engines() */
 	mutex_lock(&ctx->engines_mutex);
+	unpin_engines(__context_engines_static(ctx));
 	engines_idle_release(ctx, rcu_replace_pointer(ctx->engines, NULL, 1));
 	i915_gem_context_set_closed(ctx);
 	mutex_unlock(&ctx->engines_mutex);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
index a627b09c4680..282cdb8a5c5a 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
@@ -78,13 +78,16 @@ enum i915_gem_engine_type {
 
 	/** @I915_GEM_ENGINE_TYPE_BALANCED: A load-balanced engine set */
 	I915_GEM_ENGINE_TYPE_BALANCED,
+
+	/** @I915_GEM_ENGINE_TYPE_PARALLEL: A parallel engine set */
+	I915_GEM_ENGINE_TYPE_PARALLEL,
 };
 
 /**
  * struct i915_gem_proto_engine - prototype engine
  *
  * This struct describes an engine that a context may contain.  Engines
- * have three types:
+ * have four types:
  *
  *  - I915_GEM_ENGINE_TYPE_INVALID: Invalid engines can be created but they
  *    show up as a NULL in i915_gem_engines::engines[i] and any attempt to
@@ -97,6 +100,10 @@ enum i915_gem_engine_type {
  *
  *  - I915_GEM_ENGINE_TYPE_BALANCED: A load-balanced engine set, described
  *    i915_gem_proto_engine::num_siblings and i915_gem_proto_engine::siblings.
+ *
+ *  - I915_GEM_ENGINE_TYPE_PARALLEL: A parallel submission engine set, described
+ *    i915_gem_proto_engine::width, i915_gem_proto_engine::num_siblings, and
+ *    i915_gem_proto_engine::siblings.
  */
 struct i915_gem_proto_engine {
 	/** @type: Type of this engine */
@@ -105,10 +112,13 @@ struct i915_gem_proto_engine {
 	/** @engine: Engine, for physical */
 	struct intel_engine_cs *engine;
 
-	/** @num_siblings: Number of balanced siblings */
+	/** @num_siblings: Number of balanced or parallel siblings */
 	unsigned int num_siblings;
 
-	/** @siblings: Balanced siblings */
+	/** @width: Width of each sibling */
+	unsigned int width;
+
+	/** @siblings: Balanced siblings or num_siblings * width for parallel */
 	struct intel_engine_cs **siblings;
 
 	/** @sseu: Client-set SSEU parameters */
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
index afa34111de02..1adcd8e02d29 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
@@ -232,6 +232,7 @@ struct dma_buf *i915_gem_prime_export(struct drm_gem_object *gem_obj, int flags)
 
 static int i915_gem_object_get_pages_dmabuf(struct drm_i915_gem_object *obj)
 {
+	struct drm_i915_private *i915 = to_i915(obj->base.dev);
 	struct sg_table *pages;
 	unsigned int sg_page_sizes;
 
@@ -242,8 +243,11 @@ static int i915_gem_object_get_pages_dmabuf(struct drm_i915_gem_object *obj)
 	if (IS_ERR(pages))
 		return PTR_ERR(pages);
 
-	sg_page_sizes = i915_sg_dma_sizes(pages->sgl);
+	/* XXX: consider doing a vmap flush or something */
+	if (!HAS_LLC(i915) || i915_gem_object_can_bypass_llc(obj))
+		wbinvd_on_all_cpus();
 
+	sg_page_sizes = i915_sg_dma_sizes(pages->sgl);
 	__i915_gem_object_set_pages(obj, pages, sg_page_sizes);
 
 	return 0;
@@ -301,7 +305,8 @@ struct drm_gem_object *i915_gem_prime_import(struct drm_device *dev,
 	}
 
 	drm_gem_private_object_init(dev, &obj->base, dma_buf->size);
-	i915_gem_object_init(obj, &i915_gem_object_dmabuf_ops, &lock_class, 0);
+	i915_gem_object_init(obj, &i915_gem_object_dmabuf_ops, &lock_class,
+			     I915_BO_ALLOC_USER);
 	obj->base.import_attach = attach;
 	obj->base.resv = dma_buf->resv;
 
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 8b3a25bd93e6..4d7da07442f2 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -246,17 +246,25 @@ struct i915_execbuffer {
 	struct drm_i915_gem_exec_object2 *exec; /** ioctl execobj[] */
 	struct eb_vma *vma;
 
-	struct intel_engine_cs *engine; /** engine to queue the request to */
+	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 */
 
-	struct i915_request *request; /** our request to build */
-	struct eb_vma *batch; /** identity of the batch obj/vma */
+	/** our requests to build */
+	struct i915_request *requests[MAX_ENGINE_INSTANCE + 1];
+	/** identity of the batch obj/vma */
+	struct eb_vma *batches[MAX_ENGINE_INSTANCE + 1];
 	struct i915_vma *trampoline; /** trampoline used for chaining */
 
+	/** used for excl fence in dma_resv objects when > 1 BB submitted */
+	struct dma_fence *composite_fence;
+
 	/** actual size of execobj[] as we may extend it for the cmdparser */
 	unsigned int buffer_count;
 
+	/* number of batches in execbuf IOCTL */
+	unsigned int num_batches;
+
 	/** list of vma not yet bound during reservation phase */
 	struct list_head unbound;
 
@@ -283,7 +291,8 @@ struct i915_execbuffer {
 
 	u64 invalid_flags; /** Set of execobj.flags that are invalid */
 
-	u64 batch_len; /** Length of batch within object */
+	/** Length of batch within object */
+	u64 batch_len[MAX_ENGINE_INSTANCE + 1];
 	u32 batch_start_offset; /** Location within object of batch */
 	u32 batch_flags; /** Flags composed for emit_bb_start() */
 	struct intel_gt_buffer_pool_node *batch_pool; /** pool node for batch buffer */
@@ -301,14 +310,13 @@ struct i915_execbuffer {
 };
 
 static int eb_parse(struct i915_execbuffer *eb);
-static struct i915_request *eb_pin_engine(struct i915_execbuffer *eb,
-					  bool throttle);
+static int eb_pin_engine(struct i915_execbuffer *eb, bool throttle);
 static void eb_unpin_engine(struct i915_execbuffer *eb);
 
 static inline bool eb_use_cmdparser(const struct i915_execbuffer *eb)
 {
-	return intel_engine_requires_cmd_parser(eb->engine) ||
-		(intel_engine_using_cmd_parser(eb->engine) &&
+	return intel_engine_requires_cmd_parser(eb->context->engine) ||
+		(intel_engine_using_cmd_parser(eb->context->engine) &&
 		 eb->args->batch_len);
 }
 
@@ -535,11 +543,21 @@ eb_validate_vma(struct i915_execbuffer *eb,
 	return 0;
 }
 
-static void
+static inline bool
+is_batch_buffer(struct i915_execbuffer *eb, unsigned int buffer_idx)
+{
+	return eb->args->flags & I915_EXEC_BATCH_FIRST ?
+		buffer_idx < eb->num_batches :
+		buffer_idx >= eb->args->buffer_count - eb->num_batches;
+}
+
+static int
 eb_add_vma(struct i915_execbuffer *eb,
-	   unsigned int i, unsigned batch_idx,
+	   unsigned int *current_batch,
+	   unsigned int i,
 	   struct i915_vma *vma)
 {
+	struct drm_i915_private *i915 = eb->i915;
 	struct drm_i915_gem_exec_object2 *entry = &eb->exec[i];
 	struct eb_vma *ev = &eb->vma[i];
 
@@ -566,15 +584,43 @@ eb_add_vma(struct i915_execbuffer *eb,
 	 * Note that actual hangs have only been observed on gen7, but for
 	 * paranoia do it everywhere.
 	 */
-	if (i == batch_idx) {
+	if (is_batch_buffer(eb, i)) {
 		if (entry->relocation_count &&
 		    !(ev->flags & EXEC_OBJECT_PINNED))
 			ev->flags |= __EXEC_OBJECT_NEEDS_BIAS;
 		if (eb->reloc_cache.has_fence)
 			ev->flags |= EXEC_OBJECT_NEEDS_FENCE;
 
-		eb->batch = ev;
+		eb->batches[*current_batch] = ev;
+
+		if (unlikely(ev->flags & EXEC_OBJECT_WRITE)) {
+			drm_dbg(&i915->drm,
+				"Attempting to use self-modifying batch buffer\n");
+			return -EINVAL;
+		}
+
+		if (range_overflows_t(u64,
+				      eb->batch_start_offset,
+				      eb->args->batch_len,
+				      ev->vma->size)) {
+			drm_dbg(&i915->drm, "Attempting to use out-of-bounds batch\n");
+			return -EINVAL;
+		}
+
+		if (eb->args->batch_len == 0)
+			eb->batch_len[*current_batch] = ev->vma->size -
+				eb->batch_start_offset;
+		else
+			eb->batch_len[*current_batch] = eb->args->batch_len;
+		if (unlikely(eb->batch_len[*current_batch] == 0)) { /* impossible! */
+			drm_dbg(&i915->drm, "Invalid batch length\n");
+			return -EINVAL;
+		}
+
+		++*current_batch;
 	}
+
+	return 0;
 }
 
 static inline int use_cpu_reloc(const struct reloc_cache *cache,
@@ -718,14 +764,6 @@ static int eb_reserve(struct i915_execbuffer *eb)
 	} while (1);
 }
 
-static unsigned int eb_batch_index(const struct i915_execbuffer *eb)
-{
-	if (eb->args->flags & I915_EXEC_BATCH_FIRST)
-		return 0;
-	else
-		return eb->buffer_count - 1;
-}
-
 static int eb_select_context(struct i915_execbuffer *eb)
 {
 	struct i915_gem_context *ctx;
@@ -846,9 +884,7 @@ static struct i915_vma *eb_lookup_vma(struct i915_execbuffer *eb, u32 handle)
 
 static int eb_lookup_vmas(struct i915_execbuffer *eb)
 {
-	struct drm_i915_private *i915 = eb->i915;
-	unsigned int batch = eb_batch_index(eb);
-	unsigned int i;
+	unsigned int i, current_batch = 0;
 	int err = 0;
 
 	INIT_LIST_HEAD(&eb->relocs);
@@ -868,7 +904,9 @@ static int eb_lookup_vmas(struct i915_execbuffer *eb)
 			goto err;
 		}
 
-		eb_add_vma(eb, i, batch, vma);
+		err = eb_add_vma(eb, &current_batch, i, vma);
+		if (err)
+			return err;
 
 		if (i915_gem_object_is_userptr(vma->obj)) {
 			err = i915_gem_object_userptr_submit_init(vma->obj);
@@ -891,26 +929,6 @@ static int eb_lookup_vmas(struct i915_execbuffer *eb)
 		}
 	}
 
-	if (unlikely(eb->batch->flags & EXEC_OBJECT_WRITE)) {
-		drm_dbg(&i915->drm,
-			"Attempting to use self-modifying batch buffer\n");
-		return -EINVAL;
-	}
-
-	if (range_overflows_t(u64,
-			      eb->batch_start_offset, eb->batch_len,
-			      eb->batch->vma->size)) {
-		drm_dbg(&i915->drm, "Attempting to use out-of-bounds batch\n");
-		return -EINVAL;
-	}
-
-	if (eb->batch_len == 0)
-		eb->batch_len = eb->batch->vma->size - eb->batch_start_offset;
-	if (unlikely(eb->batch_len == 0)) { /* impossible! */
-		drm_dbg(&i915->drm, "Invalid batch length\n");
-		return -EINVAL;
-	}
-
 	return 0;
 
 err:
@@ -1643,8 +1661,7 @@ static int eb_reinit_userptr(struct i915_execbuffer *eb)
 	return 0;
 }
 
-static noinline int eb_relocate_parse_slow(struct i915_execbuffer *eb,
-					   struct i915_request *rq)
+static noinline int eb_relocate_parse_slow(struct i915_execbuffer *eb)
 {
 	bool have_copy = false;
 	struct eb_vma *ev;
@@ -1660,21 +1677,6 @@ repeat:
 	eb_release_vmas(eb, false);
 	i915_gem_ww_ctx_fini(&eb->ww);
 
-	if (rq) {
-		/* nonblocking is always false */
-		if (i915_request_wait(rq, I915_WAIT_INTERRUPTIBLE,
-				      MAX_SCHEDULE_TIMEOUT) < 0) {
-			i915_request_put(rq);
-			rq = NULL;
-
-			err = -EINTR;
-			goto err_relock;
-		}
-
-		i915_request_put(rq);
-		rq = NULL;
-	}
-
 	/*
 	 * We take 3 passes through the slowpatch.
 	 *
@@ -1701,28 +1703,21 @@ repeat:
 	if (!err)
 		err = eb_reinit_userptr(eb);
 
-err_relock:
 	i915_gem_ww_ctx_init(&eb->ww, true);
 	if (err)
 		goto out;
 
 	/* reacquire the objects */
 repeat_validate:
-	rq = eb_pin_engine(eb, false);
-	if (IS_ERR(rq)) {
-		err = PTR_ERR(rq);
-		rq = NULL;
+	err = eb_pin_engine(eb, false);
+	if (err)
 		goto err;
-	}
-
-	/* We didn't throttle, should be NULL */
-	GEM_WARN_ON(rq);
 
 	err = eb_validate_vmas(eb);
 	if (err)
 		goto err;
 
-	GEM_BUG_ON(!eb->batch);
+	GEM_BUG_ON(!eb->batches[0]);
 
 	list_for_each_entry(ev, &eb->relocs, reloc_link) {
 		if (!have_copy) {
@@ -1786,46 +1781,23 @@ out:
 		}
 	}
 
-	if (rq)
-		i915_request_put(rq);
-
 	return err;
 }
 
 static int eb_relocate_parse(struct i915_execbuffer *eb)
 {
 	int err;
-	struct i915_request *rq = NULL;
 	bool throttle = true;
 
 retry:
-	rq = eb_pin_engine(eb, throttle);
-	if (IS_ERR(rq)) {
-		err = PTR_ERR(rq);
-		rq = NULL;
+	err = eb_pin_engine(eb, throttle);
+	if (err) {
 		if (err != -EDEADLK)
 			return err;
 
 		goto err;
 	}
 
-	if (rq) {
-		bool nonblock = eb->file->filp->f_flags & O_NONBLOCK;
-
-		/* Need to drop all locks now for throttling, take slowpath */
-		err = i915_request_wait(rq, I915_WAIT_INTERRUPTIBLE, 0);
-		if (err == -ETIME) {
-			if (nonblock) {
-				err = -EWOULDBLOCK;
-				i915_request_put(rq);
-				goto err;
-			}
-			goto slow;
-		}
-		i915_request_put(rq);
-		rq = NULL;
-	}
-
 	/* only throttle once, even if we didn't need to throttle */
 	throttle = false;
 
@@ -1865,7 +1837,7 @@ err:
 	return err;
 
 slow:
-	err = eb_relocate_parse_slow(eb, rq);
+	err = eb_relocate_parse_slow(eb);
 	if (err)
 		/*
 		 * If the user expects the execobject.offset and
@@ -1879,11 +1851,40 @@ slow:
 	return err;
 }
 
+/*
+ * Using two helper loops for the order of which requests / batches are created
+ * and added the to backend. Requests are created in order from the parent to
+ * the last child. Requests are added in the reverse order, from the last child
+ * to parent. This is done for locking reasons as the timeline lock is acquired
+ * during request creation and released when the request is added to the
+ * backend. To make lockdep happy (see intel_context_timeline_lock) this must be
+ * the ordering.
+ */
+#define for_each_batch_create_order(_eb, _i) \
+	for ((_i) = 0; (_i) < (_eb)->num_batches; ++(_i))
+#define for_each_batch_add_order(_eb, _i) \
+	BUILD_BUG_ON(!typecheck(int, _i)); \
+	for ((_i) = (_eb)->num_batches - 1; (_i) >= 0; --(_i))
+
+static struct i915_request *
+eb_find_first_request_added(struct i915_execbuffer *eb)
+{
+	int i;
+
+	for_each_batch_add_order(eb, i)
+		if (eb->requests[i])
+			return eb->requests[i];
+
+	GEM_BUG_ON("Request not found");
+
+	return NULL;
+}
+
 static int eb_move_to_gpu(struct i915_execbuffer *eb)
 {
 	const unsigned int count = eb->buffer_count;
 	unsigned int i = count;
-	int err = 0;
+	int err = 0, j;
 
 	while (i--) {
 		struct eb_vma *ev = &eb->vma[i];
@@ -1896,11 +1897,17 @@ static int eb_move_to_gpu(struct i915_execbuffer *eb)
 		if (flags & EXEC_OBJECT_CAPTURE) {
 			struct i915_capture_list *capture;
 
-			capture = kmalloc(sizeof(*capture), GFP_KERNEL);
-			if (capture) {
-				capture->next = eb->request->capture_list;
-				capture->vma = vma;
-				eb->request->capture_list = capture;
+			for_each_batch_create_order(eb, j) {
+				if (!eb->requests[j])
+					break;
+
+				capture = kmalloc(sizeof(*capture), GFP_KERNEL);
+				if (capture) {
+					capture->next =
+						eb->requests[j]->capture_list;
+					capture->vma = vma;
+					eb->requests[j]->capture_list = capture;
+				}
 			}
 		}
 
@@ -1915,20 +1922,43 @@ static int eb_move_to_gpu(struct i915_execbuffer *eb)
 		 *   !(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ)
 		 * but gcc's optimiser doesn't handle that as well and emits
 		 * two jumps instead of one. Maybe one day...
+		 *
+		 * FIXME: There is also sync flushing in set_pages(), which
+		 * serves a different purpose(some of the time at least).
+		 *
+		 * We should consider:
+		 *
+		 *   1. Rip out the async flush code.
+		 *
+		 *   2. Or make the sync flushing use the async clflush path
+		 *   using mandatory fences underneath. Currently the below
+		 *   async flush happens after we bind the object.
 		 */
 		if (unlikely(obj->cache_dirty & ~obj->cache_coherent)) {
 			if (i915_gem_clflush_object(obj, 0))
 				flags &= ~EXEC_OBJECT_ASYNC;
 		}
 
+		/* We only need to await on the first request */
 		if (err == 0 && !(flags & EXEC_OBJECT_ASYNC)) {
 			err = i915_request_await_object
-				(eb->request, obj, flags & EXEC_OBJECT_WRITE);
+				(eb_find_first_request_added(eb), obj,
+				 flags & EXEC_OBJECT_WRITE);
 		}
 
-		if (err == 0)
-			err = i915_vma_move_to_active(vma, eb->request,
-						      flags | __EXEC_OBJECT_NO_RESERVE);
+		for_each_batch_add_order(eb, j) {
+			if (err)
+				break;
+			if (!eb->requests[j])
+				continue;
+
+			err = _i915_vma_move_to_active(vma, eb->requests[j],
+						       j ? NULL :
+						       eb->composite_fence ?
+						       eb->composite_fence :
+						       &eb->requests[j]->fence,
+						       flags | __EXEC_OBJECT_NO_RESERVE);
+		}
 	}
 
 #ifdef CONFIG_MMU_NOTIFIER
@@ -1959,11 +1989,16 @@ static int eb_move_to_gpu(struct i915_execbuffer *eb)
 		goto err_skip;
 
 	/* Unconditionally flush any chipset caches (for streaming writes). */
-	intel_gt_chipset_flush(eb->engine->gt);
+	intel_gt_chipset_flush(eb->gt);
 	return 0;
 
 err_skip:
-	i915_request_set_error_once(eb->request, err);
+	for_each_batch_create_order(eb, j) {
+		if (!eb->requests[j])
+			break;
+
+		i915_request_set_error_once(eb->requests[j], err);
+	}
 	return err;
 }
 
@@ -2058,14 +2093,17 @@ static int eb_parse(struct i915_execbuffer *eb)
 	int err;
 
 	if (!eb_use_cmdparser(eb)) {
-		batch = eb_dispatch_secure(eb, eb->batch->vma);
+		batch = eb_dispatch_secure(eb, eb->batches[0]->vma);
 		if (IS_ERR(batch))
 			return PTR_ERR(batch);
 
 		goto secure_batch;
 	}
 
-	len = eb->batch_len;
+	if (intel_context_is_parallel(eb->context))
+		return -EINVAL;
+
+	len = eb->batch_len[0];
 	if (!CMDPARSER_USES_GGTT(eb->i915)) {
 		/*
 		 * ppGTT backed shadow buffers must be mapped RO, to prevent
@@ -2079,11 +2117,11 @@ static int eb_parse(struct i915_execbuffer *eb)
 	} else {
 		len += I915_CMD_PARSER_TRAMPOLINE_SIZE;
 	}
-	if (unlikely(len < eb->batch_len)) /* last paranoid check of overflow */
+	if (unlikely(len < eb->batch_len[0])) /* last paranoid check of overflow */
 		return -EINVAL;
 
 	if (!pool) {
-		pool = intel_gt_get_buffer_pool(eb->engine->gt, len,
+		pool = intel_gt_get_buffer_pool(eb->gt, len,
 						I915_MAP_WB);
 		if (IS_ERR(pool))
 			return PTR_ERR(pool);
@@ -2108,7 +2146,7 @@ static int eb_parse(struct i915_execbuffer *eb)
 		trampoline = shadow;
 
 		shadow = shadow_batch_pin(eb, pool->obj,
-					  &eb->engine->gt->ggtt->vm,
+					  &eb->gt->ggtt->vm,
 					  PIN_GLOBAL);
 		if (IS_ERR(shadow)) {
 			err = PTR_ERR(shadow);
@@ -2130,26 +2168,29 @@ static int eb_parse(struct i915_execbuffer *eb)
 	if (err)
 		goto err_trampoline;
 
-	err = intel_engine_cmd_parser(eb->engine,
-				      eb->batch->vma,
+	err = intel_engine_cmd_parser(eb->context->engine,
+				      eb->batches[0]->vma,
 				      eb->batch_start_offset,
-				      eb->batch_len,
+				      eb->batch_len[0],
 				      shadow, trampoline);
 	if (err)
 		goto err_unpin_batch;
 
-	eb->batch = &eb->vma[eb->buffer_count++];
-	eb->batch->vma = i915_vma_get(shadow);
-	eb->batch->flags = __EXEC_OBJECT_HAS_PIN;
+	eb->batches[0] = &eb->vma[eb->buffer_count++];
+	eb->batches[0]->vma = i915_vma_get(shadow);
+	eb->batches[0]->flags = __EXEC_OBJECT_HAS_PIN;
 
 	eb->trampoline = trampoline;
 	eb->batch_start_offset = 0;
 
 secure_batch:
 	if (batch) {
-		eb->batch = &eb->vma[eb->buffer_count++];
-		eb->batch->flags = __EXEC_OBJECT_HAS_PIN;
-		eb->batch->vma = i915_vma_get(batch);
+		if (intel_context_is_parallel(eb->context))
+			return -EINVAL;
+
+		eb->batches[0] = &eb->vma[eb->buffer_count++];
+		eb->batches[0]->flags = __EXEC_OBJECT_HAS_PIN;
+		eb->batches[0]->vma = i915_vma_get(batch);
 	}
 	return 0;
 
@@ -2165,19 +2206,18 @@ err:
 	return err;
 }
 
-static int eb_submit(struct i915_execbuffer *eb, struct i915_vma *batch)
+static int eb_request_submit(struct i915_execbuffer *eb,
+			     struct i915_request *rq,
+			     struct i915_vma *batch,
+			     u64 batch_len)
 {
 	int err;
 
-	if (intel_context_nopreempt(eb->context))
-		__set_bit(I915_FENCE_FLAG_NOPREEMPT, &eb->request->fence.flags);
-
-	err = eb_move_to_gpu(eb);
-	if (err)
-		return err;
+	if (intel_context_nopreempt(rq->context))
+		__set_bit(I915_FENCE_FLAG_NOPREEMPT, &rq->fence.flags);
 
 	if (eb->args->flags & I915_EXEC_GEN7_SOL_RESET) {
-		err = i915_reset_gen7_sol_offsets(eb->request);
+		err = i915_reset_gen7_sol_offsets(rq);
 		if (err)
 			return err;
 	}
@@ -2188,26 +2228,26 @@ static int eb_submit(struct i915_execbuffer *eb, struct i915_vma *batch)
 	 * allows us to determine if the batch is still waiting on the GPU
 	 * or actually running by checking the breadcrumb.
 	 */
-	if (eb->engine->emit_init_breadcrumb) {
-		err = eb->engine->emit_init_breadcrumb(eb->request);
+	if (rq->context->engine->emit_init_breadcrumb) {
+		err = rq->context->engine->emit_init_breadcrumb(rq);
 		if (err)
 			return err;
 	}
 
-	err = eb->engine->emit_bb_start(eb->request,
-					batch->node.start +
-					eb->batch_start_offset,
-					eb->batch_len,
-					eb->batch_flags);
+	err = rq->context->engine->emit_bb_start(rq,
+						 batch->node.start +
+						 eb->batch_start_offset,
+						 batch_len,
+						 eb->batch_flags);
 	if (err)
 		return err;
 
 	if (eb->trampoline) {
+		GEM_BUG_ON(intel_context_is_parallel(rq->context));
 		GEM_BUG_ON(eb->batch_start_offset);
-		err = eb->engine->emit_bb_start(eb->request,
-						eb->trampoline->node.start +
-						eb->batch_len,
-						0, 0);
+		err = rq->context->engine->emit_bb_start(rq,
+							 eb->trampoline->node.start +
+							 batch_len, 0, 0);
 		if (err)
 			return err;
 	}
@@ -2215,6 +2255,27 @@ static int eb_submit(struct i915_execbuffer *eb, struct i915_vma *batch)
 	return 0;
 }
 
+static int eb_submit(struct i915_execbuffer *eb)
+{
+	unsigned int i;
+	int err;
+
+	err = eb_move_to_gpu(eb);
+
+	for_each_batch_create_order(eb, i) {
+		if (!eb->requests[i])
+			break;
+
+		trace_i915_request_queue(eb->requests[i], eb->batch_flags);
+		if (!err)
+			err = eb_request_submit(eb, eb->requests[i],
+						eb->batches[i]->vma,
+						eb->batch_len[i]);
+	}
+
+	return err;
+}
+
 static int num_vcs_engines(const struct drm_i915_private *i915)
 {
 	return hweight_long(VDBOX_MASK(&i915->gt));
@@ -2280,26 +2341,11 @@ static struct i915_request *eb_throttle(struct i915_execbuffer *eb, struct intel
 	return i915_request_get(rq);
 }
 
-static struct i915_request *eb_pin_engine(struct i915_execbuffer *eb, bool throttle)
+static int eb_pin_timeline(struct i915_execbuffer *eb, struct intel_context *ce,
+			   bool throttle)
 {
-	struct intel_context *ce = eb->context;
 	struct intel_timeline *tl;
 	struct i915_request *rq = NULL;
-	int err;
-
-	GEM_BUG_ON(eb->args->flags & __EXEC_ENGINE_PINNED);
-
-	if (unlikely(intel_context_is_banned(ce)))
-		return ERR_PTR(-EIO);
-
-	/*
-	 * Pinning the contexts may generate requests in order to acquire
-	 * GGTT space, so do this first before we reserve a seqno for
-	 * ourselves.
-	 */
-	err = intel_context_pin_ww(ce, &eb->ww);
-	if (err)
-		return ERR_PTR(err);
 
 	/*
 	 * Take a local wakeref for preparing to dispatch the execbuf as
@@ -2310,33 +2356,108 @@ static struct i915_request *eb_pin_engine(struct i915_execbuffer *eb, bool throt
 	 * taken on the engine, and the parent device.
 	 */
 	tl = intel_context_timeline_lock(ce);
-	if (IS_ERR(tl)) {
-		intel_context_unpin(ce);
-		return ERR_CAST(tl);
-	}
+	if (IS_ERR(tl))
+		return PTR_ERR(tl);
 
 	intel_context_enter(ce);
 	if (throttle)
 		rq = eb_throttle(eb, ce);
 	intel_context_timeline_unlock(tl);
 
+	if (rq) {
+		bool nonblock = eb->file->filp->f_flags & O_NONBLOCK;
+		long timeout = nonblock ? 0 : MAX_SCHEDULE_TIMEOUT;
+
+		if (i915_request_wait(rq, I915_WAIT_INTERRUPTIBLE,
+				      timeout) < 0) {
+			i915_request_put(rq);
+
+			tl = intel_context_timeline_lock(ce);
+			intel_context_exit(ce);
+			intel_context_timeline_unlock(tl);
+
+			if (nonblock)
+				return -EWOULDBLOCK;
+			else
+				return -EINTR;
+		}
+		i915_request_put(rq);
+	}
+
+	return 0;
+}
+
+static int eb_pin_engine(struct i915_execbuffer *eb, bool throttle)
+{
+	struct intel_context *ce = eb->context, *child;
+	int err;
+	int i = 0, j = 0;
+
+	GEM_BUG_ON(eb->args->flags & __EXEC_ENGINE_PINNED);
+
+	if (unlikely(intel_context_is_banned(ce)))
+		return -EIO;
+
+	/*
+	 * Pinning the contexts may generate requests in order to acquire
+	 * GGTT space, so do this first before we reserve a seqno for
+	 * ourselves.
+	 */
+	err = intel_context_pin_ww(ce, &eb->ww);
+	if (err)
+		return err;
+	for_each_child(ce, child) {
+		err = intel_context_pin_ww(child, &eb->ww);
+		GEM_BUG_ON(err);	/* perma-pinned should incr a counter */
+	}
+
+	for_each_child(ce, child) {
+		err = eb_pin_timeline(eb, child, throttle);
+		if (err)
+			goto unwind;
+		++i;
+	}
+	err = eb_pin_timeline(eb, ce, throttle);
+	if (err)
+		goto unwind;
+
 	eb->args->flags |= __EXEC_ENGINE_PINNED;
-	return rq;
+	return 0;
+
+unwind:
+	for_each_child(ce, child) {
+		if (j++ < i) {
+			mutex_lock(&child->timeline->mutex);
+			intel_context_exit(child);
+			mutex_unlock(&child->timeline->mutex);
+		}
+	}
+	for_each_child(ce, child)
+		intel_context_unpin(child);
+	intel_context_unpin(ce);
+	return err;
 }
 
 static void eb_unpin_engine(struct i915_execbuffer *eb)
 {
-	struct intel_context *ce = eb->context;
-	struct intel_timeline *tl = ce->timeline;
+	struct intel_context *ce = eb->context, *child;
 
 	if (!(eb->args->flags & __EXEC_ENGINE_PINNED))
 		return;
 
 	eb->args->flags &= ~__EXEC_ENGINE_PINNED;
 
-	mutex_lock(&tl->mutex);
+	for_each_child(ce, child) {
+		mutex_lock(&child->timeline->mutex);
+		intel_context_exit(child);
+		mutex_unlock(&child->timeline->mutex);
+
+		intel_context_unpin(child);
+	}
+
+	mutex_lock(&ce->timeline->mutex);
 	intel_context_exit(ce);
-	mutex_unlock(&tl->mutex);
+	mutex_unlock(&ce->timeline->mutex);
 
 	intel_context_unpin(ce);
 }
@@ -2387,7 +2508,7 @@ eb_select_legacy_ring(struct i915_execbuffer *eb)
 static int
 eb_select_engine(struct i915_execbuffer *eb)
 {
-	struct intel_context *ce;
+	struct intel_context *ce, *child;
 	unsigned int idx;
 	int err;
 
@@ -2400,6 +2521,20 @@ eb_select_engine(struct i915_execbuffer *eb)
 	if (IS_ERR(ce))
 		return PTR_ERR(ce);
 
+	if (intel_context_is_parallel(ce)) {
+		if (eb->buffer_count < ce->parallel.number_children + 1) {
+			intel_context_put(ce);
+			return -EINVAL;
+		}
+		if (eb->batch_start_offset || eb->args->batch_len) {
+			intel_context_put(ce);
+			return -EINVAL;
+		}
+	}
+	eb->num_batches = ce->parallel.number_children + 1;
+
+	for_each_child(ce, child)
+		intel_context_get(child);
 	intel_gt_pm_get(ce->engine->gt);
 
 	if (!test_bit(CONTEXT_ALLOC_BIT, &ce->flags)) {
@@ -2407,6 +2542,13 @@ eb_select_engine(struct i915_execbuffer *eb)
 		if (err)
 			goto err;
 	}
+	for_each_child(ce, child) {
+		if (!test_bit(CONTEXT_ALLOC_BIT, &child->flags)) {
+			err = intel_context_alloc_state(child);
+			if (err)
+				goto err;
+		}
+	}
 
 	/*
 	 * ABI: Before userspace accesses the GPU (e.g. execbuffer), report
@@ -2417,7 +2559,7 @@ eb_select_engine(struct i915_execbuffer *eb)
 		goto err;
 
 	eb->context = ce;
-	eb->engine = ce->engine;
+	eb->gt = ce->engine->gt;
 
 	/*
 	 * Make sure engine pool stays alive even if we call intel_context_put
@@ -2428,6 +2570,8 @@ eb_select_engine(struct i915_execbuffer *eb)
 
 err:
 	intel_gt_pm_put(ce->engine->gt);
+	for_each_child(ce, child)
+		intel_context_put(child);
 	intel_context_put(ce);
 	return err;
 }
@@ -2435,7 +2579,11 @@ err:
 static void
 eb_put_engine(struct i915_execbuffer *eb)
 {
-	intel_gt_pm_put(eb->engine->gt);
+	struct intel_context *child;
+
+	intel_gt_pm_put(eb->gt);
+	for_each_child(eb->context, child)
+		intel_context_put(child);
 	intel_context_put(eb->context);
 }
 
@@ -2658,7 +2806,8 @@ static void put_fence_array(struct eb_fence *fences, int num_fences)
 }
 
 static int
-await_fence_array(struct i915_execbuffer *eb)
+await_fence_array(struct i915_execbuffer *eb,
+		  struct i915_request *rq)
 {
 	unsigned int n;
 	int err;
@@ -2672,8 +2821,7 @@ await_fence_array(struct i915_execbuffer *eb)
 		if (!eb->fences[n].dma_fence)
 			continue;
 
-		err = i915_request_await_dma_fence(eb->request,
-						   eb->fences[n].dma_fence);
+		err = i915_request_await_dma_fence(rq, eb->fences[n].dma_fence);
 		if (err < 0)
 			return err;
 	}
@@ -2681,9 +2829,9 @@ await_fence_array(struct i915_execbuffer *eb)
 	return 0;
 }
 
-static void signal_fence_array(const struct i915_execbuffer *eb)
+static void signal_fence_array(const struct i915_execbuffer *eb,
+			       struct dma_fence * const fence)
 {
-	struct dma_fence * const fence = &eb->request->fence;
 	unsigned int n;
 
 	for (n = 0; n < eb->num_fences; n++) {
@@ -2731,9 +2879,9 @@ static void retire_requests(struct intel_timeline *tl, struct i915_request *end)
 			break;
 }
 
-static int eb_request_add(struct i915_execbuffer *eb, int err)
+static int eb_request_add(struct i915_execbuffer *eb, struct i915_request *rq,
+			  int err, bool last_parallel)
 {
-	struct i915_request *rq = eb->request;
 	struct intel_timeline * const tl = i915_request_timeline(rq);
 	struct i915_sched_attr attr = {};
 	struct i915_request *prev;
@@ -2755,6 +2903,17 @@ static int eb_request_add(struct i915_execbuffer *eb, int err)
 		err = -ENOENT; /* override any transient errors */
 	}
 
+	if (intel_context_is_parallel(eb->context)) {
+		if (err) {
+			__i915_request_skip(rq);
+			set_bit(I915_FENCE_FLAG_SKIP_PARALLEL,
+				&rq->fence.flags);
+		}
+		if (last_parallel)
+			set_bit(I915_FENCE_FLAG_SUBMIT_PARALLEL,
+				&rq->fence.flags);
+	}
+
 	__i915_request_queue(rq, &attr);
 
 	/* Try to clean up the client's timeline after submitting the request */
@@ -2766,6 +2925,25 @@ static int eb_request_add(struct i915_execbuffer *eb, int err)
 	return err;
 }
 
+static int eb_requests_add(struct i915_execbuffer *eb, int err)
+{
+	int i;
+
+	/*
+	 * We iterate in reverse order of creation to release timeline mutexes in
+	 * same order.
+	 */
+	for_each_batch_add_order(eb, i) {
+		struct i915_request *rq = eb->requests[i];
+
+		if (!rq)
+			continue;
+		err |= eb_request_add(eb, rq, err, i == 0);
+	}
+
+	return err;
+}
+
 static const i915_user_extension_fn execbuf_extensions[] = {
 	[DRM_I915_GEM_EXECBUFFER_EXT_TIMELINE_FENCES] = parse_timeline_fences,
 };
@@ -2792,6 +2970,185 @@ parse_execbuf2_extensions(struct drm_i915_gem_execbuffer2 *args,
 				    eb);
 }
 
+static void eb_requests_get(struct i915_execbuffer *eb)
+{
+	unsigned int i;
+
+	for_each_batch_create_order(eb, i) {
+		if (!eb->requests[i])
+			break;
+
+		i915_request_get(eb->requests[i]);
+	}
+}
+
+static void eb_requests_put(struct i915_execbuffer *eb)
+{
+	unsigned int i;
+
+	for_each_batch_create_order(eb, i) {
+		if (!eb->requests[i])
+			break;
+
+		i915_request_put(eb->requests[i]);
+	}
+}
+
+static struct sync_file *
+eb_composite_fence_create(struct i915_execbuffer *eb, int out_fence_fd)
+{
+	struct sync_file *out_fence = NULL;
+	struct dma_fence_array *fence_array;
+	struct dma_fence **fences;
+	unsigned int i;
+
+	GEM_BUG_ON(!intel_context_is_parent(eb->context));
+
+	fences = kmalloc_array(eb->num_batches, sizeof(*fences), GFP_KERNEL);
+	if (!fences)
+		return ERR_PTR(-ENOMEM);
+
+	for_each_batch_create_order(eb, i) {
+		fences[i] = &eb->requests[i]->fence;
+		__set_bit(I915_FENCE_FLAG_COMPOSITE,
+			  &eb->requests[i]->fence.flags);
+	}
+
+	fence_array = dma_fence_array_create(eb->num_batches,
+					     fences,
+					     eb->context->parallel.fence_context,
+					     eb->context->parallel.seqno,
+					     false);
+	if (!fence_array) {
+		kfree(fences);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	/* Move ownership to the dma_fence_array created above */
+	for_each_batch_create_order(eb, i)
+		dma_fence_get(fences[i]);
+
+	if (out_fence_fd != -1) {
+		out_fence = sync_file_create(&fence_array->base);
+		/* sync_file now owns fence_arry, drop creation ref */
+		dma_fence_put(&fence_array->base);
+		if (!out_fence)
+			return ERR_PTR(-ENOMEM);
+	}
+
+	eb->composite_fence = &fence_array->base;
+
+	return out_fence;
+}
+
+static struct sync_file *
+eb_fences_add(struct i915_execbuffer *eb, struct i915_request *rq,
+	      struct dma_fence *in_fence, int out_fence_fd)
+{
+	struct sync_file *out_fence = NULL;
+	int err;
+
+	if (unlikely(eb->gem_context->syncobj)) {
+		struct dma_fence *fence;
+
+		fence = drm_syncobj_fence_get(eb->gem_context->syncobj);
+		err = i915_request_await_dma_fence(rq, fence);
+		dma_fence_put(fence);
+		if (err)
+			return ERR_PTR(err);
+	}
+
+	if (in_fence) {
+		if (eb->args->flags & I915_EXEC_FENCE_SUBMIT)
+			err = i915_request_await_execution(rq, in_fence);
+		else
+			err = i915_request_await_dma_fence(rq, in_fence);
+		if (err < 0)
+			return ERR_PTR(err);
+	}
+
+	if (eb->fences) {
+		err = await_fence_array(eb, rq);
+		if (err)
+			return ERR_PTR(err);
+	}
+
+	if (intel_context_is_parallel(eb->context)) {
+		out_fence = eb_composite_fence_create(eb, out_fence_fd);
+		if (IS_ERR(out_fence))
+			return ERR_PTR(-ENOMEM);
+	} else if (out_fence_fd != -1) {
+		out_fence = sync_file_create(&rq->fence);
+		if (!out_fence)
+			return ERR_PTR(-ENOMEM);
+	}
+
+	return out_fence;
+}
+
+static struct intel_context *
+eb_find_context(struct i915_execbuffer *eb, unsigned int context_number)
+{
+	struct intel_context *child;
+
+	if (likely(context_number == 0))
+		return eb->context;
+
+	for_each_child(eb->context, child)
+		if (!--context_number)
+			return child;
+
+	GEM_BUG_ON("Context not found");
+
+	return NULL;
+}
+
+static struct sync_file *
+eb_requests_create(struct i915_execbuffer *eb, struct dma_fence *in_fence,
+		   int out_fence_fd)
+{
+	struct sync_file *out_fence = NULL;
+	unsigned int i;
+
+	for_each_batch_create_order(eb, i) {
+		/* Allocate a request for this batch buffer nice and early. */
+		eb->requests[i] = i915_request_create(eb_find_context(eb, i));
+		if (IS_ERR(eb->requests[i])) {
+			out_fence = ERR_PTR(PTR_ERR(eb->requests[i]));
+			eb->requests[i] = NULL;
+			return out_fence;
+		}
+
+		/*
+		 * Only the first request added (committed to backend) has to
+		 * take the in fences into account as all subsequent requests
+		 * will have fences inserted inbetween them.
+		 */
+		if (i + 1 == eb->num_batches) {
+			out_fence = eb_fences_add(eb, eb->requests[i],
+						  in_fence, out_fence_fd);
+			if (IS_ERR(out_fence))
+				return out_fence;
+		}
+
+		/*
+		 * Whilst this request exists, batch_obj will be on the
+		 * active_list, and so will hold the active reference. Only when
+		 * this request is retired will the batch_obj be moved onto
+		 * the inactive_list and lose its active reference. Hence we do
+		 * not need to explicitly hold another reference here.
+		 */
+		eb->requests[i]->batch = eb->batches[i]->vma;
+		if (eb->batch_pool) {
+			GEM_BUG_ON(intel_context_is_parallel(eb->context));
+			intel_gt_buffer_pool_mark_active(eb->batch_pool,
+							 eb->requests[i]);
+		}
+	}
+
+	return out_fence;
+}
+
 static int
 i915_gem_do_execbuffer(struct drm_device *dev,
 		       struct drm_file *file,
@@ -2802,7 +3159,6 @@ i915_gem_do_execbuffer(struct drm_device *dev,
 	struct i915_execbuffer eb;
 	struct dma_fence *in_fence = NULL;
 	struct sync_file *out_fence = NULL;
-	struct i915_vma *batch;
 	int out_fence_fd = -1;
 	int err;
 
@@ -2826,12 +3182,15 @@ i915_gem_do_execbuffer(struct drm_device *dev,
 
 	eb.buffer_count = args->buffer_count;
 	eb.batch_start_offset = args->batch_start_offset;
-	eb.batch_len = args->batch_len;
 	eb.trampoline = NULL;
 
 	eb.fences = NULL;
 	eb.num_fences = 0;
 
+	memset(eb.requests, 0, sizeof(struct i915_request *) *
+	       ARRAY_SIZE(eb.requests));
+	eb.composite_fence = NULL;
+
 	eb.batch_flags = 0;
 	if (args->flags & I915_EXEC_SECURE) {
 		if (GRAPHICS_VER(i915) >= 11)
@@ -2915,70 +3274,25 @@ i915_gem_do_execbuffer(struct drm_device *dev,
 
 	ww_acquire_done(&eb.ww.ctx);
 
-	batch = eb.batch->vma;
-
-	/* Allocate a request for this batch buffer nice and early. */
-	eb.request = i915_request_create(eb.context);
-	if (IS_ERR(eb.request)) {
-		err = PTR_ERR(eb.request);
-		goto err_vma;
-	}
-
-	if (unlikely(eb.gem_context->syncobj)) {
-		struct dma_fence *fence;
-
-		fence = drm_syncobj_fence_get(eb.gem_context->syncobj);
-		err = i915_request_await_dma_fence(eb.request, fence);
-		dma_fence_put(fence);
-		if (err)
-			goto err_ext;
-	}
-
-	if (in_fence) {
-		if (args->flags & I915_EXEC_FENCE_SUBMIT)
-			err = i915_request_await_execution(eb.request,
-							   in_fence);
-		else
-			err = i915_request_await_dma_fence(eb.request,
-							   in_fence);
-		if (err < 0)
-			goto err_request;
-	}
-
-	if (eb.fences) {
-		err = await_fence_array(&eb);
-		if (err)
+	out_fence = eb_requests_create(&eb, in_fence, out_fence_fd);
+	if (IS_ERR(out_fence)) {
+		err = PTR_ERR(out_fence);
+		if (eb.requests[0])
 			goto err_request;
+		else
+			goto err_vma;
 	}
 
-	if (out_fence_fd != -1) {
-		out_fence = sync_file_create(&eb.request->fence);
-		if (!out_fence) {
-			err = -ENOMEM;
-			goto err_request;
-		}
-	}
-
-	/*
-	 * Whilst this request exists, batch_obj will be on the
-	 * active_list, and so will hold the active reference. Only when this
-	 * request is retired will the the batch_obj be moved onto the
-	 * inactive_list and lose its active reference. Hence we do not need
-	 * to explicitly hold another reference here.
-	 */
-	eb.request->batch = batch;
-	if (eb.batch_pool)
-		intel_gt_buffer_pool_mark_active(eb.batch_pool, eb.request);
-
-	trace_i915_request_queue(eb.request, eb.batch_flags);
-	err = eb_submit(&eb, batch);
+	err = eb_submit(&eb);
 
 err_request:
-	i915_request_get(eb.request);
-	err = eb_request_add(&eb, err);
+	eb_requests_get(&eb);
+	err = eb_requests_add(&eb, err);
 
 	if (eb.fences)
-		signal_fence_array(&eb);
+		signal_fence_array(&eb, eb.composite_fence ?
+				   eb.composite_fence :
+				   &eb.requests[0]->fence);
 
 	if (out_fence) {
 		if (err == 0) {
@@ -2993,10 +3307,15 @@ err_request:
 
 	if (unlikely(eb.gem_context->syncobj)) {
 		drm_syncobj_replace_fence(eb.gem_context->syncobj,
-					  &eb.request->fence);
+					  eb.composite_fence ?
+					  eb.composite_fence :
+					  &eb.requests[0]->fence);
 	}
 
-	i915_request_put(eb.request);
+	if (!out_fence && eb.composite_fence)
+		dma_fence_put(eb.composite_fence);
+
+	eb_requests_put(&eb);
 
 err_vma:
 	eb_release_vmas(&eb, true);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_internal.c b/drivers/gpu/drm/i915/gem/i915_gem_internal.c
index e5ae9c06510c..a57a6b7013c2 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_internal.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_internal.c
@@ -134,6 +134,8 @@ static void i915_gem_object_put_pages_internal(struct drm_i915_gem_object *obj,
 	internal_free_pages(pages);
 
 	obj->mm.dirty = false;
+
+	__start_cpu_write(obj);
 }
 
 static const struct drm_i915_gem_object_ops i915_gem_object_internal_ops = {
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
index 76ce6a1500bc..1e426a42a36c 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
@@ -128,6 +128,32 @@ void i915_gem_object_set_cache_coherency(struct drm_i915_gem_object *obj,
 		!(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE);
 }
 
+bool i915_gem_object_can_bypass_llc(struct drm_i915_gem_object *obj)
+{
+	struct drm_i915_private *i915 = to_i915(obj->base.dev);
+
+	/*
+	 * This is purely from a security perspective, so we simply don't care
+	 * about non-userspace objects being able to bypass the LLC.
+	 */
+	if (!(obj->flags & I915_BO_ALLOC_USER))
+		return false;
+
+	/*
+	 * EHL and JSL add the 'Bypass LLC' MOCS entry, which should make it
+	 * possible for userspace to bypass the GTT caching bits set by the
+	 * kernel, as per the given object cache_level. This is troublesome
+	 * since the heavy flush we apply when first gathering the pages is
+	 * skipped if the kernel thinks the object is coherent with the GPU. As
+	 * a result it might be possible to bypass the cache and read the
+	 * contents of the page directly, which could be stale data. If it's
+	 * just a case of userspace shooting themselves in the foot then so be
+	 * it, but since i915 takes the stance of always zeroing memory before
+	 * handing it to userspace, we need to prevent this.
+	 */
+	return IS_JSL_EHL(i915);
+}
+
 static void i915_gem_close_object(struct drm_gem_object *gem, struct drm_file *file)
 {
 	struct drm_i915_gem_object *obj = to_intel_bo(gem);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index 9df3ee60604e..59201801cec5 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -514,6 +514,7 @@ i915_gem_object_finish_access(struct drm_i915_gem_object *obj)
 
 void i915_gem_object_set_cache_coherency(struct drm_i915_gem_object *obj,
 					 unsigned int cache_level);
+bool i915_gem_object_can_bypass_llc(struct drm_i915_gem_object *obj);
 void i915_gem_object_flush_if_display(struct drm_i915_gem_object *obj);
 void i915_gem_object_flush_if_display_locked(struct drm_i915_gem_object *obj);
 
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
index 7c3da4e3e737..da85169006d4 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h
@@ -427,6 +427,33 @@ struct drm_i915_gem_object {
 	 * can freely bypass the CPU cache when touching the pages with the GPU,
 	 * where the kernel is completely unaware. On such platform we need
 	 * apply the sledgehammer-on-acquire regardless of the @cache_coherent.
+	 *
+	 * Special care is taken on non-LLC platforms, to prevent potential
+	 * information leak. The driver currently ensures:
+	 *
+	 *   1. All userspace objects, by default, have @cache_level set as
+	 *   I915_CACHE_NONE. The only exception is userptr objects, where we
+	 *   instead force I915_CACHE_LLC, but we also don't allow userspace to
+	 *   ever change the @cache_level for such objects. Another special case
+	 *   is dma-buf, which doesn't rely on @cache_dirty,  but there we
+	 *   always do a forced flush when acquiring the pages, if there is a
+	 *   chance that the pages can be read directly from main memory with
+	 *   the GPU.
+	 *
+	 *   2. All I915_CACHE_NONE objects have @cache_dirty initially true.
+	 *
+	 *   3. All swapped-out objects(i.e shmem) have @cache_dirty set to
+	 *   true.
+	 *
+	 *   4. The @cache_dirty is never freely reset before the initial
+	 *   flush, even if userspace adjusts the @cache_level through the
+	 *   i915_gem_set_caching_ioctl.
+	 *
+	 *   5. All @cache_dirty objects(including swapped-in) are initially
+	 *   flushed with a synchronous call to drm_clflush_sg in
+	 *   __i915_gem_object_set_pages. The @cache_dirty can be freely reset
+	 *   at this point. All further asynchronous clfushes are never security
+	 *   critical, i.e userspace is free to race against itself.
 	 */
 	unsigned int cache_dirty:1;
 
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
index 11f072193f3b..d77da59fae04 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_shmem.c
@@ -182,22 +182,7 @@ rebuild_st:
 	if (i915_gem_object_needs_bit17_swizzle(obj))
 		i915_gem_object_do_bit_17_swizzle(obj, st);
 
-	/*
-	 * EHL and JSL add the 'Bypass LLC' MOCS entry, which should make it
-	 * possible for userspace to bypass the GTT caching bits set by the
-	 * kernel, as per the given object cache_level. This is troublesome
-	 * since the heavy flush we apply when first gathering the pages is
-	 * skipped if the kernel thinks the object is coherent with the GPU. As
-	 * a result it might be possible to bypass the cache and read the
-	 * contents of the page directly, which could be stale data. If it's
-	 * just a case of userspace shooting themselves in the foot then so be
-	 * it, but since i915 takes the stance of always zeroing memory before
-	 * handing it to userspace, we need to prevent this.
-	 *
-	 * By setting cache_dirty here we make the clflush in set_pages
-	 * unconditional on such platforms.
-	 */
-	if (IS_JSL_EHL(i915) && obj->flags & I915_BO_ALLOC_USER)
+	if (i915_gem_object_can_bypass_llc(obj))
 		obj->cache_dirty = true;
 
 	__i915_gem_object_set_pages(obj, st, sg_page_sizes);
@@ -301,6 +286,8 @@ __i915_gem_object_release_shmem(struct drm_i915_gem_object *obj,
 				struct sg_table *pages,
 				bool needs_clflush)
 {
+	struct drm_i915_private *i915 = to_i915(obj->base.dev);
+
 	GEM_BUG_ON(obj->mm.madv == __I915_MADV_PURGED);
 
 	if (obj->mm.madv == I915_MADV_DONTNEED)
@@ -312,6 +299,16 @@ __i915_gem_object_release_shmem(struct drm_i915_gem_object *obj,
 		drm_clflush_sg(pages);
 
 	__start_cpu_write(obj);
+	/*
+	 * On non-LLC platforms, force the flush-on-acquire if this is ever
+	 * swapped-in. Our async flush path is not trust worthy enough yet(and
+	 * happens in the wrong order), and with some tricks it's conceivable
+	 * for userspace to change the cache-level to I915_CACHE_NONE after the
+	 * pages are swapped-in, and since execbuf binds the object before doing
+	 * the async flush, we have a race window.
+	 */
+	if (!HAS_LLC(i915))
+		obj->cache_dirty = true;
 }
 
 void i915_gem_object_put_pages_shmem(struct drm_i915_gem_object *obj, struct sg_table *pages)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
index 8ea0fa665e53..3173c9f9a040 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
@@ -165,8 +165,11 @@ alloc_table:
 		goto err;
 	}
 
-	sg_page_sizes = i915_sg_dma_sizes(st->sgl);
+	WARN_ON_ONCE(!(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE));
+	if (i915_gem_object_can_bypass_llc(obj))
+		obj->cache_dirty = true;
 
+	sg_page_sizes = i915_sg_dma_sizes(st->sgl);
 	__i915_gem_object_set_pages(obj, st, sg_page_sizes);
 
 	return 0;
@@ -546,7 +549,8 @@ i915_gem_userptr_ioctl(struct drm_device *dev,
 		return -ENOMEM;
 
 	drm_gem_private_object_init(dev, &obj->base, args->user_size);
-	i915_gem_object_init(obj, &i915_gem_userptr_ops, &lock_class, 0);
+	i915_gem_object_init(obj, &i915_gem_userptr_ops, &lock_class,
+			     I915_BO_ALLOC_USER);
 	obj->mem_flags = I915_BO_FLAG_STRUCT_PAGE;
 	obj->read_domains = I915_GEM_DOMAIN_CPU;
 	obj->write_domain = I915_GEM_DOMAIN_CPU;
diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
index 41d0680f3bd7..b2003133deaf 100644
--- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
@@ -136,6 +136,8 @@ static void put_huge_pages(struct drm_i915_gem_object *obj,
 	huge_pages_free_pages(pages);
 
 	obj->mm.dirty = false;
+
+	__start_cpu_write(obj);
 }
 
 static const struct drm_i915_gem_object_ops huge_page_ops = {
@@ -152,6 +154,7 @@ huge_pages_object(struct drm_i915_private *i915,
 {
 	static struct lock_class_key lock_class;
 	struct drm_i915_gem_object *obj;
+	unsigned int cache_level;
 
 	GEM_BUG_ON(!size);
 	GEM_BUG_ON(!IS_ALIGNED(size, BIT(__ffs(page_mask))));
@@ -173,7 +176,9 @@ huge_pages_object(struct drm_i915_private *i915,
 
 	obj->write_domain = I915_GEM_DOMAIN_CPU;
 	obj->read_domains = I915_GEM_DOMAIN_CPU;
-	obj->cache_level = I915_CACHE_NONE;
+
+	cache_level = HAS_LLC(i915) ? I915_CACHE_LLC : I915_CACHE_NONE;
+	i915_gem_object_set_cache_coherency(obj, cache_level);
 
 	obj->mm.page_mask = page_mask;
 
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_client_blt.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_client_blt.c
index ecbcbb86ae1e..8402ed925a69 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_client_blt.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_client_blt.c
@@ -17,13 +17,20 @@
 #include "huge_gem_object.h"
 #include "mock_context.h"
 
+enum client_tiling {
+	CLIENT_TILING_LINEAR,
+	CLIENT_TILING_X,
+	CLIENT_TILING_Y,
+	CLIENT_NUM_TILING_TYPES
+};
+
 #define WIDTH 512
 #define HEIGHT 32
 
 struct blit_buffer {
 	struct i915_vma *vma;
 	u32 start_val;
-	u32 tiling;
+	enum client_tiling tiling;
 };
 
 struct tiled_blits {
@@ -53,9 +60,9 @@ static int prepare_blit(const struct tiled_blits *t,
 	*cs++ = MI_LOAD_REGISTER_IMM(1);
 	*cs++ = i915_mmio_reg_offset(BCS_SWCTRL);
 	cmd = (BCS_SRC_Y | BCS_DST_Y) << 16;
-	if (src->tiling == I915_TILING_Y)
+	if (src->tiling == CLIENT_TILING_Y)
 		cmd |= BCS_SRC_Y;
-	if (dst->tiling == I915_TILING_Y)
+	if (dst->tiling == CLIENT_TILING_Y)
 		cmd |= BCS_DST_Y;
 	*cs++ = cmd;
 
@@ -172,7 +179,7 @@ static int tiled_blits_create_buffers(struct tiled_blits *t,
 
 		t->buffers[i].vma = vma;
 		t->buffers[i].tiling =
-			i915_prandom_u32_max_state(I915_TILING_Y + 1, prng);
+			i915_prandom_u32_max_state(CLIENT_TILING_Y + 1, prng);
 	}
 
 	return 0;
@@ -197,17 +204,17 @@ static u64 swizzle_bit(unsigned int bit, u64 offset)
 static u64 tiled_offset(const struct intel_gt *gt,
 			u64 v,
 			unsigned int stride,
-			unsigned int tiling)
+			enum client_tiling tiling)
 {
 	unsigned int swizzle;
 	u64 x, y;
 
-	if (tiling == I915_TILING_NONE)
+	if (tiling == CLIENT_TILING_LINEAR)
 		return v;
 
 	y = div64_u64_rem(v, stride, &x);
 
-	if (tiling == I915_TILING_X) {
+	if (tiling == CLIENT_TILING_X) {
 		v = div64_u64_rem(y, 8, &y) * stride * 8;
 		v += y * 512;
 		v += div64_u64_rem(x, 512, &x) << 12;
@@ -244,12 +251,12 @@ static u64 tiled_offset(const struct intel_gt *gt,
 	return v;
 }
 
-static const char *repr_tiling(int tiling)
+static const char *repr_tiling(enum client_tiling tiling)
 {
 	switch (tiling) {
-	case I915_TILING_NONE: return "linear";
-	case I915_TILING_X: return "X";
-	case I915_TILING_Y: return "Y";
+	case CLIENT_TILING_LINEAR: return "linear";
+	case CLIENT_TILING_X: return "X";
+	case CLIENT_TILING_Y: return "Y";
 	default: return "unknown";
 	}
 }