drm/i915: Move GEM object domain management from struct_mutex to local

Use the per-object local lock to control the cache domain of the
individual GEM objects, not struct_mutex. This is a huge leap forward
for us in terms of object-level synchronisation; execbuffers are
coordinated using the ww_mutex and pread/pwrite is finally fully
serialised again.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190528092956.14910-10-chris@chris-wilson.co.uk

32 files changed, 445 insertions, 180 deletions

diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 2d5bf69dbc4b..44d6e0bcf6a6 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -93,6 +93,7 @@ gem-y += \
 	gem/i915_gem_dmabuf.o \
 	gem/i915_gem_domain.o \
 	gem/i915_gem_execbuffer.o \
+	gem/i915_gem_fence.o \
 	gem/i915_gem_internal.o \
 	gem/i915_gem_object.o \
 	gem/i915_gem_mman.o \
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_clflush.c b/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
index 45d238d784fc..537aa2337cc8 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
@@ -95,6 +95,8 @@ bool i915_gem_clflush_object(struct drm_i915_gem_object *obj,
 {
 	struct clflush *clflush;
 
+	assert_object_held(obj);
+
 	/*
 	 * Stolen memory is always coherent with the GPU as it is explicitly
 	 * marked as wc by the system, or the system is cache-coherent.
@@ -144,9 +146,7 @@ bool i915_gem_clflush_object(struct drm_i915_gem_object *obj,
 						true, I915_FENCE_TIMEOUT,
 						I915_FENCE_GFP);
 
-		reservation_object_lock(obj->resv, NULL);
 		reservation_object_add_excl_fence(obj->resv, &clflush->dma);
-		reservation_object_unlock(obj->resv);
 
 		i915_sw_fence_commit(&clflush->wait);
 	} else if (obj->mm.pages) {
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
index 625397deb701..a93e233cfaa9 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_dmabuf.c
@@ -151,7 +151,6 @@ static int i915_gem_dmabuf_mmap(struct dma_buf *dma_buf, struct vm_area_struct *
 static int i915_gem_begin_cpu_access(struct dma_buf *dma_buf, enum dma_data_direction direction)
 {
 	struct drm_i915_gem_object *obj = dma_buf_to_obj(dma_buf);
-	struct drm_device *dev = obj->base.dev;
 	bool write = (direction == DMA_BIDIRECTIONAL || direction == DMA_TO_DEVICE);
 	int err;
 
@@ -159,12 +158,12 @@ static int i915_gem_begin_cpu_access(struct dma_buf *dma_buf, enum dma_data_dire
 	if (err)
 		return err;
 
-	err = i915_mutex_lock_interruptible(dev);
+	err = i915_gem_object_lock_interruptible(obj);
 	if (err)
 		goto out;
 
 	err = i915_gem_object_set_to_cpu_domain(obj, write);
-	mutex_unlock(&dev->struct_mutex);
+	i915_gem_object_unlock(obj);
 
 out:
 	i915_gem_object_unpin_pages(obj);
@@ -174,19 +173,18 @@ out:
 static int i915_gem_end_cpu_access(struct dma_buf *dma_buf, enum dma_data_direction direction)
 {
 	struct drm_i915_gem_object *obj = dma_buf_to_obj(dma_buf);
-	struct drm_device *dev = obj->base.dev;
 	int err;
 
 	err = i915_gem_object_pin_pages(obj);
 	if (err)
 		return err;
 
-	err = i915_mutex_lock_interruptible(dev);
+	err = i915_gem_object_lock_interruptible(obj);
 	if (err)
 		goto out;
 
 	err = i915_gem_object_set_to_gtt_domain(obj, false);
-	mutex_unlock(&dev->struct_mutex);
+	i915_gem_object_unlock(obj);
 
 out:
 	i915_gem_object_unpin_pages(obj);
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_domain.c b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
index bbc7fb758186..cce96e6c6e52 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_domain.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
@@ -29,9 +29,9 @@ void i915_gem_object_flush_if_display(struct drm_i915_gem_object *obj)
 	if (!READ_ONCE(obj->pin_global))
 		return;
 
-	mutex_lock(&obj->base.dev->struct_mutex);
+	i915_gem_object_lock(obj);
 	__i915_gem_object_flush_for_display(obj);
-	mutex_unlock(&obj->base.dev->struct_mutex);
+	i915_gem_object_unlock(obj);
 }
 
 /**
@@ -47,11 +47,10 @@ i915_gem_object_set_to_wc_domain(struct drm_i915_gem_object *obj, bool write)
 {
 	int ret;
 
-	lockdep_assert_held(&obj->base.dev->struct_mutex);
+	assert_object_held(obj);
 
 	ret = i915_gem_object_wait(obj,
 				   I915_WAIT_INTERRUPTIBLE |
-				   I915_WAIT_LOCKED |
 				   (write ? I915_WAIT_ALL : 0),
 				   MAX_SCHEDULE_TIMEOUT);
 	if (ret)
@@ -109,11 +108,10 @@ i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write)
 {
 	int ret;
 
-	lockdep_assert_held(&obj->base.dev->struct_mutex);
+	assert_object_held(obj);
 
 	ret = i915_gem_object_wait(obj,
 				   I915_WAIT_INTERRUPTIBLE |
-				   I915_WAIT_LOCKED |
 				   (write ? I915_WAIT_ALL : 0),
 				   MAX_SCHEDULE_TIMEOUT);
 	if (ret)
@@ -179,7 +177,7 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
 	struct i915_vma *vma;
 	int ret;
 
-	lockdep_assert_held(&obj->base.dev->struct_mutex);
+	assert_object_held(obj);
 
 	if (obj->cache_level == cache_level)
 		return 0;
@@ -228,7 +226,6 @@ restart:
 		 */
 		ret = i915_gem_object_wait(obj,
 					   I915_WAIT_INTERRUPTIBLE |
-					   I915_WAIT_LOCKED |
 					   I915_WAIT_ALL,
 					   MAX_SCHEDULE_TIMEOUT);
 		if (ret)
@@ -372,12 +369,16 @@ int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data,
 	if (ret)
 		goto out;
 
-	ret = i915_mutex_lock_interruptible(dev);
+	ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
 	if (ret)
 		goto out;
 
-	ret = i915_gem_object_set_cache_level(obj, level);
-	mutex_unlock(&dev->struct_mutex);
+	ret = i915_gem_object_lock_interruptible(obj);
+	if (ret == 0) {
+		ret = i915_gem_object_set_cache_level(obj, level);
+		i915_gem_object_unlock(obj);
+	}
+	mutex_unlock(&i915->drm.struct_mutex);
 
 out:
 	i915_gem_object_put(obj);
@@ -399,7 +400,7 @@ i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
 	struct i915_vma *vma;
 	int ret;
 
-	lockdep_assert_held(&obj->base.dev->struct_mutex);
+	assert_object_held(obj);
 
 	/* Mark the global pin early so that we account for the
 	 * display coherency whilst setting up the cache domains.
@@ -484,16 +485,18 @@ static void i915_gem_object_bump_inactive_ggtt(struct drm_i915_gem_object *obj)
 void
 i915_gem_object_unpin_from_display_plane(struct i915_vma *vma)
 {
-	lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
+	struct drm_i915_gem_object *obj = vma->obj;
+
+	assert_object_held(obj);
 
-	if (WARN_ON(vma->obj->pin_global == 0))
+	if (WARN_ON(obj->pin_global == 0))
 		return;
 
-	if (--vma->obj->pin_global == 0)
+	if (--obj->pin_global == 0)
 		vma->display_alignment = I915_GTT_MIN_ALIGNMENT;
 
 	/* Bump the LRU to try and avoid premature eviction whilst flipping  */
-	i915_gem_object_bump_inactive_ggtt(vma->obj);
+	i915_gem_object_bump_inactive_ggtt(obj);
 
 	i915_vma_unpin(vma);
 }
@@ -511,11 +514,10 @@ i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write)
 {
 	int ret;
 
-	lockdep_assert_held(&obj->base.dev->struct_mutex);
+	assert_object_held(obj);
 
 	ret = i915_gem_object_wait(obj,
 				   I915_WAIT_INTERRUPTIBLE |
-				   I915_WAIT_LOCKED |
 				   (write ? I915_WAIT_ALL : 0),
 				   MAX_SCHEDULE_TIMEOUT);
 	if (ret)
@@ -637,7 +639,7 @@ i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
 	if (err)
 		goto out;
 
-	err = i915_mutex_lock_interruptible(dev);
+	err = i915_gem_object_lock_interruptible(obj);
 	if (err)
 		goto out_unpin;
 
@@ -651,7 +653,7 @@ i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
 	/* And bump the LRU for this access */
 	i915_gem_object_bump_inactive_ggtt(obj);
 
-	mutex_unlock(&dev->struct_mutex);
+	i915_gem_object_unlock(obj);
 
 	if (write_domain != 0)
 		intel_fb_obj_invalidate(obj,
@@ -674,22 +676,23 @@ int i915_gem_object_prepare_read(struct drm_i915_gem_object *obj,
 {
 	int ret;
 
-	lockdep_assert_held(&obj->base.dev->struct_mutex);
-
 	*needs_clflush = 0;
 	if (!i915_gem_object_has_struct_page(obj))
 		return -ENODEV;
 
+	ret = i915_gem_object_lock_interruptible(obj);
+	if (ret)
+		return ret;
+
 	ret = i915_gem_object_wait(obj,
-				   I915_WAIT_INTERRUPTIBLE |
-				   I915_WAIT_LOCKED,
+				   I915_WAIT_INTERRUPTIBLE,
 				   MAX_SCHEDULE_TIMEOUT);
 	if (ret)
-		return ret;
+		goto err_unlock;
 
 	ret = i915_gem_object_pin_pages(obj);
 	if (ret)
-		return ret;
+		goto err_unlock;
 
 	if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ ||
 	    !static_cpu_has(X86_FEATURE_CLFLUSH)) {
@@ -717,6 +720,8 @@ out:
 
 err_unpin:
 	i915_gem_object_unpin_pages(obj);
+err_unlock:
+	i915_gem_object_unlock(obj);
 	return ret;
 }
 
@@ -725,23 +730,24 @@ int i915_gem_object_prepare_write(struct drm_i915_gem_object *obj,
 {
 	int ret;
 
-	lockdep_assert_held(&obj->base.dev->struct_mutex);
-
 	*needs_clflush = 0;
 	if (!i915_gem_object_has_struct_page(obj))
 		return -ENODEV;
 
+	ret = i915_gem_object_lock_interruptible(obj);
+	if (ret)
+		return ret;
+
 	ret = i915_gem_object_wait(obj,
 				   I915_WAIT_INTERRUPTIBLE |
-				   I915_WAIT_LOCKED |
 				   I915_WAIT_ALL,
 				   MAX_SCHEDULE_TIMEOUT);
 	if (ret)
-		return ret;
+		goto err_unlock;
 
 	ret = i915_gem_object_pin_pages(obj);
 	if (ret)
-		return ret;
+		goto err_unlock;
 
 	if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE ||
 	    !static_cpu_has(X86_FEATURE_CLFLUSH)) {
@@ -778,5 +784,7 @@ out:
 
 err_unpin:
 	i915_gem_object_unpin_pages(obj);
+err_unlock:
+	i915_gem_object_unlock(obj);
 	return ret;
 }
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 09e64bf33842..ed522fdfbe7f 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -1075,7 +1075,9 @@ static void *reloc_iomap(struct drm_i915_gem_object *obj,
 		if (use_cpu_reloc(cache, obj))
 			return NULL;
 
+		i915_gem_object_lock(obj);
 		err = i915_gem_object_set_to_gtt_domain(obj, true);
+		i915_gem_object_unlock(obj);
 		if (err)
 			return ERR_PTR(err);
 
@@ -1164,6 +1166,26 @@ static void clflush_write32(u32 *addr, u32 value, unsigned int flushes)
 		*addr = value;
 }
 
+static int reloc_move_to_gpu(struct i915_request *rq, struct i915_vma *vma)
+{
+	struct drm_i915_gem_object *obj = vma->obj;
+	int err;
+
+	i915_vma_lock(vma);
+
+	if (obj->cache_dirty & ~obj->cache_coherent)
+		i915_gem_clflush_object(obj, 0);
+	obj->write_domain = 0;
+
+	err = i915_request_await_object(rq, vma->obj, true);
+	if (err == 0)
+		err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+
+	i915_vma_unlock(vma);
+
+	return err;
+}
+
 static int __reloc_gpu_alloc(struct i915_execbuffer *eb,
 			     struct i915_vma *vma,
 			     unsigned int len)
@@ -1175,15 +1197,6 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb,
 	u32 *cmd;
 	int err;
 
-	if (DBG_FORCE_RELOC == FORCE_GPU_RELOC) {
-		obj = vma->obj;
-		if (obj->cache_dirty & ~obj->cache_coherent)
-			i915_gem_clflush_object(obj, 0);
-		obj->write_domain = 0;
-	}
-
-	GEM_BUG_ON(vma->obj->write_domain & I915_GEM_DOMAIN_CPU);
-
 	obj = i915_gem_batch_pool_get(&eb->engine->batch_pool, PAGE_SIZE);
 	if (IS_ERR(obj))
 		return PTR_ERR(obj);
@@ -1212,7 +1225,7 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb,
 		goto err_unpin;
 	}
 
-	err = i915_request_await_object(rq, vma->obj, true);
+	err = reloc_move_to_gpu(rq, vma);
 	if (err)
 		goto err_request;
 
@@ -1220,14 +1233,12 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb,
 					batch->node.start, PAGE_SIZE,
 					cache->gen > 5 ? 0 : I915_DISPATCH_SECURE);
 	if (err)
-		goto err_request;
+		goto skip_request;
 
+	i915_vma_lock(batch);
 	GEM_BUG_ON(!reservation_object_test_signaled_rcu(batch->resv, true));
 	err = i915_vma_move_to_active(batch, rq, 0);
-	if (err)
-		goto skip_request;
-
-	err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+	i915_vma_unlock(batch);
 	if (err)
 		goto skip_request;
 
@@ -1837,24 +1848,59 @@ slow:
 static int eb_move_to_gpu(struct i915_execbuffer *eb)
 {
 	const unsigned int count = eb->buffer_count;
+	struct ww_acquire_ctx acquire;
 	unsigned int i;
-	int err;
+	int err = 0;
+
+	ww_acquire_init(&acquire, &reservation_ww_class);
 
 	for (i = 0; i < count; i++) {
+		struct i915_vma *vma = eb->vma[i];
+
+		err = ww_mutex_lock_interruptible(&vma->resv->lock, &acquire);
+		if (!err)
+			continue;
+
+		GEM_BUG_ON(err == -EALREADY); /* No duplicate vma */
+
+		if (err == -EDEADLK) {
+			GEM_BUG_ON(i == 0);
+			do {
+				int j = i - 1;
+
+				ww_mutex_unlock(&eb->vma[j]->resv->lock);
+
+				swap(eb->flags[i], eb->flags[j]);
+				swap(eb->vma[i],  eb->vma[j]);
+				eb->vma[i]->exec_flags = &eb->flags[i];
+			} while (--i);
+			GEM_BUG_ON(vma != eb->vma[0]);
+			vma->exec_flags = &eb->flags[0];
+
+			err = ww_mutex_lock_slow_interruptible(&vma->resv->lock,
+							       &acquire);
+		}
+		if (err)
+			break;
+	}
+	ww_acquire_done(&acquire);
+
+	while (i--) {
 		unsigned int flags = eb->flags[i];
 		struct i915_vma *vma = eb->vma[i];
 		struct drm_i915_gem_object *obj = vma->obj;
 
+		assert_vma_held(vma);
+
 		if (flags & EXEC_OBJECT_CAPTURE) {
 			struct i915_capture_list *capture;
 
 			capture = kmalloc(sizeof(*capture), GFP_KERNEL);
-			if (unlikely(!capture))
-				return -ENOMEM;
-
-			capture->next = eb->request->capture_list;
-			capture->vma = eb->vma[i];
-			eb->request->capture_list = capture;
+			if (capture) {
+				capture->next = eb->request->capture_list;
+				capture->vma = vma;
+				eb->request->capture_list = capture;
+			}
 		}
 
 		/*
@@ -1874,24 +1920,15 @@ static int eb_move_to_gpu(struct i915_execbuffer *eb)
 				flags &= ~EXEC_OBJECT_ASYNC;
 		}
 
-		if (flags & EXEC_OBJECT_ASYNC)
-			continue;
-
-		err = i915_request_await_object
-			(eb->request, obj, flags & EXEC_OBJECT_WRITE);
-		if (err)
-			return err;
-	}
+		if (err == 0 && !(flags & EXEC_OBJECT_ASYNC)) {
+			err = i915_request_await_object
+				(eb->request, obj, flags & EXEC_OBJECT_WRITE);
+		}
 
-	for (i = 0; i < count; i++) {
-		unsigned int flags = eb->flags[i];
-		struct i915_vma *vma = eb->vma[i];
+		if (err == 0)
+			err = i915_vma_move_to_active(vma, eb->request, flags);
 
-		err = i915_vma_move_to_active(vma, eb->request, flags);
-		if (unlikely(err)) {
-			i915_request_skip(eb->request, err);
-			return err;
-		}
+		i915_vma_unlock(vma);
 
 		__eb_unreserve_vma(vma, flags);
 		vma->exec_flags = NULL;
@@ -1899,12 +1936,20 @@ static int eb_move_to_gpu(struct i915_execbuffer *eb)
 		if (unlikely(flags & __EXEC_OBJECT_HAS_REF))
 			i915_vma_put(vma);
 	}
+	ww_acquire_fini(&acquire);
+
+	if (unlikely(err))
+		goto err_skip;
+
 	eb->exec = NULL;
 
 	/* Unconditionally flush any chipset caches (for streaming writes). */
 	i915_gem_chipset_flush(eb->i915);
-
 	return 0;
+
+err_skip:
+	i915_request_skip(eb->request, err);
+	return err;
 }
 
 static bool i915_gem_check_execbuffer(struct drm_i915_gem_execbuffer2 *exec)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_fence.c b/drivers/gpu/drm/i915/gem/i915_gem_fence.c
new file mode 100644
index 000000000000..57dbc0862713
--- /dev/null
+++ b/drivers/gpu/drm/i915/gem/i915_gem_fence.c
@@ -0,0 +1,96 @@
+/*
+ * SPDX-License-Identifier: MIT
+ *
+ * Copyright © 2019 Intel Corporation
+ */
+
+#include "i915_drv.h"
+#include "i915_gem_object.h"
+
+struct stub_fence {
+	struct dma_fence dma;
+	struct i915_sw_fence chain;
+};
+
+static int __i915_sw_fence_call
+stub_notify(struct i915_sw_fence *fence, enum i915_sw_fence_notify state)
+{
+	struct stub_fence *stub = container_of(fence, typeof(*stub), chain);
+
+	switch (state) {
+	case FENCE_COMPLETE:
+		dma_fence_signal(&stub->dma);
+		break;
+
+	case FENCE_FREE:
+		dma_fence_put(&stub->dma);
+		break;
+	}
+
+	return NOTIFY_DONE;
+}
+
+static const char *stub_driver_name(struct dma_fence *fence)
+{
+	return DRIVER_NAME;
+}
+
+static const char *stub_timeline_name(struct dma_fence *fence)
+{
+	return "object";
+}
+
+static void stub_release(struct dma_fence *fence)
+{
+	struct stub_fence *stub = container_of(fence, typeof(*stub), dma);
+
+	i915_sw_fence_fini(&stub->chain);
+
+	BUILD_BUG_ON(offsetof(typeof(*stub), dma));
+	dma_fence_free(&stub->dma);
+}
+
+static const struct dma_fence_ops stub_fence_ops = {
+	.get_driver_name = stub_driver_name,
+	.get_timeline_name = stub_timeline_name,
+	.release = stub_release,
+};
+
+struct dma_fence *
+i915_gem_object_lock_fence(struct drm_i915_gem_object *obj)
+{
+	struct stub_fence *stub;
+
+	assert_object_held(obj);
+
+	stub = kmalloc(sizeof(*stub), GFP_KERNEL);
+	if (!stub)
+		return NULL;
+
+	i915_sw_fence_init(&stub->chain, stub_notify);
+	dma_fence_init(&stub->dma, &stub_fence_ops, &stub->chain.wait.lock,
+		       to_i915(obj->base.dev)->mm.unordered_timeline,
+		       0);
+
+	if (i915_sw_fence_await_reservation(&stub->chain,
+					    obj->resv, NULL,
+					    true, I915_FENCE_TIMEOUT,
+					    I915_FENCE_GFP) < 0)
+		goto err;
+
+	reservation_object_add_excl_fence(obj->resv, &stub->dma);
+
+	return &stub->dma;
+
+err:
+	stub_release(&stub->dma);
+	return NULL;
+}
+
+void i915_gem_object_unlock_fence(struct drm_i915_gem_object *obj,
+				  struct dma_fence *fence)
+{
+	struct stub_fence *stub = container_of(fence, typeof(*stub), dma);
+
+	i915_sw_fence_commit(&stub->chain);
+}
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c
index 457e694a5c3f..a6a3452d2b3e 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c
@@ -378,6 +378,8 @@ i915_gem_object_flush_write_domain(struct drm_i915_gem_object *obj,
 	struct drm_i915_private *dev_priv = to_i915(obj->base.dev);
 	struct i915_vma *vma;
 
+	assert_object_held(obj);
+
 	if (!(obj->write_domain & flush_domains))
 		return;
 
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index 8cf082abb0ab..b0488517e945 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -99,16 +99,29 @@ i915_gem_object_put(struct drm_i915_gem_object *obj)
 	__drm_gem_object_put(&obj->base);
 }
 
+#define assert_object_held(obj) reservation_object_assert_held((obj)->resv)
+
 static inline void i915_gem_object_lock(struct drm_i915_gem_object *obj)
 {
 	reservation_object_lock(obj->resv, NULL);
 }
 
+static inline int
+i915_gem_object_lock_interruptible(struct drm_i915_gem_object *obj)
+{
+	return reservation_object_lock_interruptible(obj->resv, NULL);
+}
+
 static inline void i915_gem_object_unlock(struct drm_i915_gem_object *obj)
 {
 	reservation_object_unlock(obj->resv);
 }
 
+struct dma_fence *
+i915_gem_object_lock_fence(struct drm_i915_gem_object *obj);
+void i915_gem_object_unlock_fence(struct drm_i915_gem_object *obj,
+				  struct dma_fence *fence);
+
 static inline void
 i915_gem_object_set_readonly(struct drm_i915_gem_object *obj)
 {
@@ -372,6 +385,7 @@ static inline void
 i915_gem_object_finish_access(struct drm_i915_gem_object *obj)
 {
 	i915_gem_object_unpin_pages(obj);
+	i915_gem_object_unlock(obj);
 }
 
 static inline struct intel_engine_cs *
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
index ad662e558dfb..11890e96ed65 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c
@@ -187,12 +187,13 @@ void i915_gem_suspend_late(struct drm_i915_private *i915)
 	 * machine in an unusable condition.
 	 */
 
-	mutex_lock(&i915->drm.struct_mutex);
 	for (phase = phases; *phase; phase++) {
-		list_for_each_entry(obj, *phase, mm.link)
+		list_for_each_entry(obj, *phase, mm.link) {
+			i915_gem_object_lock(obj);
 			WARN_ON(i915_gem_object_set_to_gtt_domain(obj, false));
+			i915_gem_object_unlock(obj);
+		}
 	}
-	mutex_unlock(&i915->drm.struct_mutex);
 
 	intel_uc_sanitize(i915);
 	i915_gem_sanitize(i915);
diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
index 7b437f06a9be..465e0e1d4aa3 100644
--- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
@@ -960,10 +960,6 @@ static int gpu_write(struct i915_vma *vma,
 
 	GEM_BUG_ON(!intel_engine_can_store_dword(engine));
 
-	err = i915_gem_object_set_to_gtt_domain(vma->obj, true);
-	if (err)
-		return err;
-
 	batch = gpu_write_dw(vma, dword * sizeof(u32), value);
 	if (IS_ERR(batch))
 		return PTR_ERR(batch);
@@ -974,13 +970,19 @@ static int gpu_write(struct i915_vma *vma,
 		goto err_batch;
 	}
 
+	i915_vma_lock(batch);
 	err = i915_vma_move_to_active(batch, rq, 0);
+	i915_vma_unlock(batch);
 	if (err)
 		goto err_request;
 
 	i915_gem_object_set_active_reference(batch->obj);
 
-	err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+	i915_vma_lock(vma);
+	err = i915_gem_object_set_to_gtt_domain(vma->obj, false);
+	if (err == 0)
+		err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+	i915_vma_unlock(vma);
 	if (err)
 		goto err_request;
 
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
index 5495875b48b3..b5c5dd034d5c 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
@@ -78,7 +78,9 @@ static int gtt_set(struct drm_i915_gem_object *obj,
 	u32 __iomem *map;
 	int err;
 
+	i915_gem_object_lock(obj);
 	err = i915_gem_object_set_to_gtt_domain(obj, true);
+	i915_gem_object_unlock(obj);
 	if (err)
 		return err;
 
@@ -105,7 +107,9 @@ static int gtt_get(struct drm_i915_gem_object *obj,
 	u32 __iomem *map;
 	int err;
 
+	i915_gem_object_lock(obj);
 	err = i915_gem_object_set_to_gtt_domain(obj, false);
+	i915_gem_object_unlock(obj);
 	if (err)
 		return err;
 
@@ -131,7 +135,9 @@ static int wc_set(struct drm_i915_gem_object *obj,
 	u32 *map;
 	int err;
 
+	i915_gem_object_lock(obj);
 	err = i915_gem_object_set_to_wc_domain(obj, true);
+	i915_gem_object_unlock(obj);
 	if (err)
 		return err;
 
@@ -152,7 +158,9 @@ static int wc_get(struct drm_i915_gem_object *obj,
 	u32 *map;
 	int err;
 
+	i915_gem_object_lock(obj);
 	err = i915_gem_object_set_to_wc_domain(obj, false);
+	i915_gem_object_unlock(obj);
 	if (err)
 		return err;
 
@@ -176,7 +184,9 @@ static int gpu_set(struct drm_i915_gem_object *obj,
 	u32 *cs;
 	int err;
 
+	i915_gem_object_lock(obj);
 	err = i915_gem_object_set_to_gtt_domain(obj, true);
+	i915_gem_object_unlock(obj);
 	if (err)
 		return err;
 
@@ -215,7 +225,9 @@ static int gpu_set(struct drm_i915_gem_object *obj,
 	}
 	intel_ring_advance(rq, cs);
 
+	i915_vma_lock(vma);
 	err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+	i915_vma_unlock(vma);
 	i915_vma_unpin(vma);
 
 	i915_request_add(rq);
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
index 653ae08a277f..72eedd6c2a0a 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
@@ -209,7 +209,9 @@ gpu_fill_dw(struct i915_vma *vma, u64 offset, unsigned long count, u32 value)
 	i915_gem_object_flush_map(obj);
 	i915_gem_object_unpin_map(obj);
 
+	i915_gem_object_lock(obj);
 	err = i915_gem_object_set_to_gtt_domain(obj, false);
+	i915_gem_object_unlock(obj);
 	if (err)
 		goto err;
 
@@ -261,7 +263,9 @@ static int gpu_fill(struct drm_i915_gem_object *obj,
 	if (IS_ERR(vma))
 		return PTR_ERR(vma);
 
+	i915_gem_object_lock(obj);
 	err = i915_gem_object_set_to_gtt_domain(obj, false);
+	i915_gem_object_unlock(obj);
 	if (err)
 		return err;
 
@@ -302,11 +306,15 @@ static int gpu_fill(struct drm_i915_gem_object *obj,
 	if (err)
 		goto err_request;
 
+	i915_vma_lock(batch);
 	err = i915_vma_move_to_active(batch, rq, 0);
+	i915_vma_unlock(batch);
 	if (err)
 		goto skip_request;
 
+	i915_vma_lock(vma);
 	err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+	i915_vma_unlock(vma);
 	if (err)
 		goto skip_request;
 
@@ -754,7 +762,9 @@ emit_rpcs_query(struct drm_i915_gem_object *obj,
 	if (IS_ERR(vma))
 		return PTR_ERR(vma);
 
+	i915_gem_object_lock(obj);
 	err = i915_gem_object_set_to_gtt_domain(obj, false);
+	i915_gem_object_unlock(obj);
 	if (err)
 		return err;
 
@@ -780,11 +790,15 @@ emit_rpcs_query(struct drm_i915_gem_object *obj,
 	if (err)
 		goto err_request;
 
+	i915_vma_lock(batch);
 	err = i915_vma_move_to_active(batch, rq, 0);
+	i915_vma_unlock(batch);
 	if (err)
 		goto skip_request;
 
+	i915_vma_lock(vma);
 	err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+	i915_vma_unlock(vma);
 	if (err)
 		goto skip_request;
 
@@ -1345,7 +1359,9 @@ static int write_to_scratch(struct i915_gem_context *ctx,
 	if (err)
 		goto err_request;
 
+	i915_vma_lock(vma);
 	err = i915_vma_move_to_active(vma, rq, 0);
+	i915_vma_unlock(vma);
 	if (err)
 		goto skip_request;
 
@@ -1440,7 +1456,9 @@ static int read_from_scratch(struct i915_gem_context *ctx,
 	if (err)
 		goto err_request;
 
+	i915_vma_lock(vma);
 	err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+	i915_vma_unlock(vma);
 	if (err)
 		goto skip_request;
 
@@ -1449,7 +1467,9 @@ static int read_from_scratch(struct i915_gem_context *ctx,
 
 	i915_request_add(rq);
 
+	i915_gem_object_lock(obj);
 	err = i915_gem_object_set_to_cpu_domain(obj, false);
+	i915_gem_object_unlock(obj);
 	if (err)
 		goto err;
 
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
index 12c90d8fe0fb..297f8864d392 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c
@@ -110,7 +110,9 @@ static int check_partial_mapping(struct drm_i915_gem_object *obj,
 		GEM_BUG_ON(view.partial.size > nreal);
 		cond_resched();
 
+		i915_gem_object_lock(obj);
 		err = i915_gem_object_set_to_gtt_domain(obj, true);
+		i915_gem_object_unlock(obj);
 		if (err) {
 			pr_err("Failed to flush to GTT write domain; err=%d\n",
 			       err);
@@ -142,7 +144,9 @@ static int check_partial_mapping(struct drm_i915_gem_object *obj,
 		if (offset >= obj->base.size)
 			continue;
 
+		i915_gem_object_lock(obj);
 		i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
+		i915_gem_object_unlock(obj);
 
 		p = i915_gem_object_get_page(obj, offset >> PAGE_SHIFT);
 		cpu = kmap(p) + offset_in_page(offset);
@@ -344,7 +348,9 @@ static int make_obj_busy(struct drm_i915_gem_object *obj)
 		return PTR_ERR(rq);
 	}
 
+	i915_vma_lock(vma);
 	err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+	i915_vma_unlock(vma);
 
 	i915_request_add(rq);
 
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c
index ed64012e5d24..94a15e3f6db8 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_phys.c
@@ -46,9 +46,9 @@ static int mock_phys_object(void *arg)
 	}
 
 	/* Make the object dirty so that put_pages must do copy back the data */
-	mutex_lock(&i915->drm.struct_mutex);
+	i915_gem_object_lock(obj);
 	err = i915_gem_object_set_to_gtt_domain(obj, true);
-	mutex_unlock(&i915->drm.struct_mutex);
+	i915_gem_object_unlock(obj);
 	if (err) {
 		pr_err("i915_gem_object_set_to_gtt_domain failed with err=%d\n",
 		       err);
diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
index 690d77f5ecf6..c3fa10fd9383 100644
--- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
+++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
@@ -117,7 +117,9 @@ static int move_to_active(struct i915_vma *vma,
 {
 	int err;
 
+	i915_vma_lock(vma);
 	err = i915_vma_move_to_active(vma, rq, flags);
+	i915_vma_unlock(vma);
 	if (err)
 		return err;
 
@@ -1252,7 +1254,9 @@ static int __igt_reset_evict_vma(struct drm_i915_private *i915,
 		}
 	}
 
+	i915_vma_lock(arg.vma);
 	err = i915_vma_move_to_active(arg.vma, rq, flags);
+	i915_vma_unlock(arg.vma);
 
 	if (flags & EXEC_OBJECT_NEEDS_FENCE)
 		i915_vma_unpin_fence(arg.vma);
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index dfacc46ae7d3..d162b0d3ec8e 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -1108,11 +1108,13 @@ static int smoke_submit(struct preempt_smoke *smoke,
 	}
 
 	if (vma) {
+		i915_vma_lock(vma);
 		err = rq->engine->emit_bb_start(rq,
 						vma->node.start,
 						PAGE_SIZE, 0);
 		if (!err)
 			err = i915_vma_move_to_active(vma, rq, 0);
+		i915_vma_unlock(vma);
 	}
 
 	i915_request_add(rq);
diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
index 9040cae38fc5..38a69acf60ae 100644
--- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
@@ -118,7 +118,9 @@ read_nonprivs(struct i915_gem_context *ctx, struct intel_engine_cs *engine)
 		goto err_pin;
 	}
 
+	i915_vma_lock(vma);
 	err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
+	i915_vma_unlock(vma);
 	if (err)
 		goto err_req;
 
@@ -195,8 +197,10 @@ static int check_whitelist(struct i915_gem_context *ctx,
 		return PTR_ERR(results);
 
 	err = 0;
+	i915_gem_object_lock(results);
 	igt_wedge_on_timeout(&wedge, ctx->i915, HZ / 5) /* a safety net! */
 		err = i915_gem_object_set_to_cpu_domain(results, false);
+	i915_gem_object_unlock(results);
 	if (i915_terminally_wedged(ctx->i915))
 		err = -EIO;
 	if (err)
@@ -367,7 +371,9 @@ static struct i915_vma *create_batch(struct i915_gem_context *ctx)
 	if (err)
 		goto err_obj;
 
+	i915_gem_object_lock(obj);
 	err = i915_gem_object_set_to_wc_domain(obj, true);
+	i915_gem_object_unlock(obj);
 	if (err)
 		goto err_obj;
 
diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c
index e3608b170105..75cb98da6cc8 100644
--- a/drivers/gpu/drm/i915/gvt/cmd_parser.c
+++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c
@@ -2844,7 +2844,9 @@ static int shadow_indirect_ctx(struct intel_shadow_wa_ctx *wa_ctx)
 		goto put_obj;
 	}
 
+	i915_gem_object_lock(obj);
 	ret = i915_gem_object_set_to_cpu_domain(obj, false);
+	i915_gem_object_unlock(obj);
 	if (ret) {
 		gvt_vgpu_err("failed to set shadow indirect ctx to CPU\n");
 		goto unmap_src;
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index d66bf77f55fd..8a00e2d0c81c 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -509,18 +509,18 @@ static int prepare_shadow_batch_buffer(struct intel_vgpu_workload *workload)
 			}
 
 			ret = i915_gem_object_set_to_gtt_domain(bb->obj,
-					false);
+								false);
 			if (ret)
 				goto err;
 
-			i915_gem_object_finish_access(bb->obj);
-			bb->accessing = false;
-
 			ret = i915_vma_move_to_active(bb->vma,
 						      workload->req,
 						      0);
 			if (ret)
 				goto err;
+
+			i915_gem_object_finish_access(bb->obj);
+			bb->accessing = false;
 		}
 	}
 	return 0;
diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index c893bd4eb2c8..a28bcd2d7c09 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1058,19 +1058,20 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 	void *dst, *src;
 	int ret;
 
-	ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
+	ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
 	if (ret)
 		return ERR_PTR(ret);
 
-	ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
-	if (ret) {
-		dst = ERR_PTR(ret);
-		goto unpin_src;
-	}
-
 	dst = i915_gem_object_pin_map(dst_obj, I915_MAP_FORCE_WB);
+	i915_gem_object_finish_access(dst_obj);
 	if (IS_ERR(dst))
-		goto unpin_dst;
+		return dst;
+
+	ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
+	if (ret) {
+		i915_gem_object_unpin_map(dst_obj);
+		return ERR_PTR(ret);
+	}
 
 	src = ERR_PTR(-ENODEV);
 	if (src_needs_clflush &&
@@ -1116,13 +1117,11 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		}
 	}
 
+	i915_gem_object_finish_access(src_obj);
+
 	/* dst_obj is returned with vmap pinned */
 	*needs_clflush_after = dst_needs_clflush & CLFLUSH_AFTER;
 
-unpin_dst:
-	i915_gem_object_finish_access(dst_obj);
-unpin_src:
-	i915_gem_object_finish_access(src_obj);
 	return dst;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index cfe0d9ff4e5d..993489d0fb2e 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -104,19 +104,10 @@ int i915_gem_object_unbind(struct drm_i915_gem_object *obj)
 {
 	struct i915_vma *vma;
 	LIST_HEAD(still_in_list);
-	int ret;
+	int ret = 0;
 
 	lockdep_assert_held(&obj->base.dev->struct_mutex);
 
-	/* Closed vma are removed from the obj->vma_list - but they may
-	 * still have an active binding on the object. To remove those we
-	 * must wait for all rendering to complete to the object (as unbinding
-	 * must anyway), and retire the requests.
-	 */
-	ret = i915_gem_object_set_to_cpu_domain(obj, false);
-	if (ret)
-		return ret;
-
 	spin_lock(&obj->vma.lock);
 	while (!ret && (vma = list_first_entry_or_null(&obj->vma.list,
 						       struct i915_vma,
@@ -139,29 +130,17 @@ i915_gem_object_wait_fence(struct dma_fence *fence,
 			   unsigned int flags,
 			   long timeout)
 {
-	struct i915_request *rq;
-
 	BUILD_BUG_ON(I915_WAIT_INTERRUPTIBLE != 0x1);
 
 	if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
 		return timeout;
 
-	if (!dma_fence_is_i915(fence))
-		return dma_fence_wait_timeout(fence,
-					      flags & I915_WAIT_INTERRUPTIBLE,
-					      timeout);
+	if (dma_fence_is_i915(fence))
+		return i915_request_wait(to_request(fence), flags, timeout);
 
-	rq = to_request(fence);
-	if (i915_request_completed(rq))
-		goto out;
-
-	timeout = i915_request_wait(rq, flags, timeout);
-
-out:
-	if (flags & I915_WAIT_LOCKED && i915_request_completed(rq))
-		i915_request_retire_upto(rq);
-
-	return timeout;
+	return dma_fence_wait_timeout(fence,
+				      flags & I915_WAIT_INTERRUPTIBLE,
+				      timeout);
 }
 
 static long
@@ -487,21 +466,22 @@ static int
 i915_gem_shmem_pread(struct drm_i915_gem_object *obj,
 		     struct drm_i915_gem_pread *args)
 {
-	char __user *user_data;
-	u64 remain;
 	unsigned int needs_clflush;
 	unsigned int idx, offset;
+	struct dma_fence *fence;
+	char __user *user_data;
+	u64 remain;
 	int ret;
 
-	ret = mutex_lock_interruptible(&obj->base.dev->struct_mutex);
-	if (ret)
-		return ret;
-
 	ret = i915_gem_object_prepare_read(obj, &needs_clflush);
-	mutex_unlock(&obj->base.dev->struct_mutex);
 	if (ret)
 		return ret;
 
+	fence = i915_gem_object_lock_fence(obj);
+	i915_gem_object_finish_access(obj);
+	if (!fence)
+		return -ENOMEM;
+
 	remain = args->size;
 	user_data = u64_to_user_ptr(args->data_ptr);
 	offset = offset_in_page(args->offset);
@@ -519,7 +499,7 @@ i915_gem_shmem_pread(struct drm_i915_gem_object *obj,
 		offset = 0;
 	}
 
-	i915_gem_object_finish_access(obj);
+	i915_gem_object_unlock_fence(obj, fence);
 	return ret;
 }
 
@@ -555,8 +535,9 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
 	struct i915_ggtt *ggtt = &i915->ggtt;
 	intel_wakeref_t wakeref;
 	struct drm_mm_node node;
-	struct i915_vma *vma;
+	struct dma_fence *fence;
 	void __user *user_data;
+	struct i915_vma *vma;
 	u64 remain, offset;
 	int ret;
 
@@ -585,11 +566,24 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
 		GEM_BUG_ON(!node.allocated);
 	}
 
-	ret = i915_gem_object_set_to_gtt_domain(obj, false);
+	mutex_unlock(&i915->drm.struct_mutex);
+
+	ret = i915_gem_object_lock_interruptible(obj);
 	if (ret)
 		goto out_unpin;
 
-	mutex_unlock(&i915->drm.struct_mutex);
+	ret = i915_gem_object_set_to_gtt_domain(obj, false);
+	if (ret) {
+		i915_gem_object_unlock(obj);
+		goto out_unpin;
+	}
+
+	fence = i915_gem_object_lock_fence(obj);
+	i915_gem_object_unlock(obj);
+	if (!fence) {
+		ret = -ENOMEM;
+		goto out_unpin;
+	}
 
 	user_data = u64_to_user_ptr(args->data_ptr);
 	remain = args->size;
@@ -627,8 +621,9 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
 		offset += page_length;
 	}
 
-	mutex_lock(&i915->drm.struct_mutex);
+	i915_gem_object_unlock_fence(obj, fence);
 out_unpin:
+	mutex_lock(&i915->drm.struct_mutex);
 	if (node.allocated) {
 		wmb();
 		ggtt->vm.clear_range(&ggtt->vm, node.start, node.size);
@@ -739,6 +734,7 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
 	struct i915_ggtt *ggtt = &i915->ggtt;
 	intel_wakeref_t wakeref;
 	struct drm_mm_node node;
+	struct dma_fence *fence;
 	struct i915_vma *vma;
 	u64 remain, offset;
 	void __user *user_data;
@@ -786,11 +782,24 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
 		GEM_BUG_ON(!node.allocated);
 	}
 
-	ret = i915_gem_object_set_to_gtt_domain(obj, true);
+	mutex_unlock(&i915->drm.struct_mutex);
+
+	ret = i915_gem_object_lock_interruptible(obj);
 	if (ret)
 		goto out_unpin;
 
-	mutex_unlock(&i915->drm.struct_mutex);
+	ret = i915_gem_object_set_to_gtt_domain(obj, true);
+	if (ret) {
+		i915_gem_object_unlock(obj);
+		goto out_unpin;
+	}
+
+	fence = i915_gem_object_lock_fence(obj);
+	i915_gem_object_unlock(obj);
+	if (!fence) {
+		ret = -ENOMEM;
+		goto out_unpin;
+	}
 
 	intel_fb_obj_invalidate(obj, ORIGIN_CPU);
 
@@ -835,8 +844,9 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
 	}
 	intel_fb_obj_flush(obj, ORIGIN_CPU);
 
-	mutex_lock(&i915->drm.struct_mutex);
+	i915_gem_object_unlock_fence(obj, fence);
 out_unpin:
+	mutex_lock(&i915->drm.struct_mutex);
 	if (node.allocated) {
 		wmb();
 		ggtt->vm.clear_range(&ggtt->vm, node.start, node.size);
@@ -882,23 +892,23 @@ static int
 i915_gem_shmem_pwrite(struct drm_i915_gem_object *obj,
 		      const struct drm_i915_gem_pwrite *args)
 {
-	struct drm_i915_private *i915 = to_i915(obj->base.dev);
-	void __user *user_data;
-	u64 remain;
 	unsigned int partial_cacheline_write;
 	unsigned int needs_clflush;
 	unsigned int offset, idx;
+	struct dma_fence *fence;
+	void __user *user_data;
+	u64 remain;
 	int ret;
 
-	ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
-	if (ret)
-		return ret;
-
 	ret = i915_gem_object_prepare_write(obj, &needs_clflush);
-	mutex_unlock(&i915->drm.struct_mutex);
 	if (ret)
 		return ret;
 
+	fence = i915_gem_object_lock_fence(obj);
+	i915_gem_object_finish_access(obj);
+	if (!fence)
+		return -ENOMEM;
+
 	/* If we don't overwrite a cacheline completely we need to be
 	 * careful to have up-to-date data by first clflushing. Don't
 	 * overcomplicate things and flush the entire patch.
@@ -926,7 +936,8 @@ i915_gem_shmem_pwrite(struct drm_i915_gem_object *obj,
 	}
 
 	intel_fb_obj_flush(obj, ORIGIN_CPU);
-	i915_gem_object_finish_access(obj);
+	i915_gem_object_unlock_fence(obj, fence);
+
 	return ret;
 }
 
@@ -1805,7 +1816,9 @@ static int __intel_engines_record_defaults(struct drm_i915_private *i915)
 		if (err)
 			goto err_active;
 
+		i915_gem_object_lock(state->obj);
 		err = i915_gem_object_set_to_cpu_domain(state->obj, false);
+		i915_gem_object_unlock(state->obj);
 		if (err)
 			goto err_active;
 
@@ -2256,12 +2269,13 @@ int i915_gem_freeze_late(struct drm_i915_private *i915)
 	i915_gem_shrink(i915, -1UL, NULL, I915_SHRINK_UNBOUND);
 	i915_gem_drain_freed_objects(i915);
 
-	mutex_lock(&i915->drm.struct_mutex);
 	for (phase = phases; *phase; phase++) {
-		list_for_each_entry(obj, *phase, mm.link)
+		list_for_each_entry(obj, *phase, mm.link) {
+			i915_gem_object_lock(obj);
 			WARN_ON(i915_gem_object_set_to_cpu_domain(obj, true));
+			i915_gem_object_unlock(obj);
+		}
 	}
-	mutex_unlock(&i915->drm.struct_mutex);
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index ccc4bdf356c1..7496cce0d798 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -3578,8 +3578,11 @@ void i915_gem_restore_gtt_mappings(struct drm_i915_private *dev_priv)
 		WARN_ON(i915_vma_bind(vma,
 				      obj ? obj->cache_level : 0,
 				      PIN_UPDATE));
-		if (obj)
+		if (obj) {
+			i915_gem_object_lock(obj);
 			WARN_ON(i915_gem_object_set_to_gtt_domain(obj, false));
+			i915_gem_object_unlock(obj);
+		}
 
 lock:
 		mutex_lock(&ggtt->vm.mutex);
diff --git a/drivers/gpu/drm/i915/i915_gem_render_state.c b/drivers/gpu/drm/i915/i915_gem_render_state.c
index f3b42b026fff..706ed71468e8 100644
--- a/drivers/gpu/drm/i915/i915_gem_render_state.c
+++ b/drivers/gpu/drm/i915/i915_gem_render_state.c
@@ -222,7 +222,9 @@ int i915_gem_render_state_emit(struct i915_request *rq)
 			goto err_unpin;
 	}
 
+	i915_vma_lock(so.vma);
 	err = i915_vma_move_to_active(so.vma, rq, 0);
+	i915_vma_unlock(so.vma);
 err_unpin:
 	i915_vma_unpin(so.vma);
 err_vma:
diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
index cf405ffda045..db94d7b6c5a6 100644
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -840,13 +840,14 @@ void i915_vma_destroy(struct i915_vma *vma)
 {
 	lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
 
-	GEM_BUG_ON(i915_vma_is_active(vma));
 	GEM_BUG_ON(i915_vma_is_pinned(vma));
 
 	if (i915_vma_is_closed(vma))
 		list_del(&vma->closed_link);
 
 	WARN_ON(i915_vma_unbind(vma));
+	GEM_BUG_ON(i915_vma_is_active(vma));
+
 	__i915_vma_destroy(vma);
 }
 
@@ -908,12 +909,10 @@ static void export_fence(struct i915_vma *vma,
 	 * handle an error right now. Worst case should be missed
 	 * synchronisation leading to rendering corruption.
 	 */
-	reservation_object_lock(resv, NULL);
 	if (flags & EXEC_OBJECT_WRITE)
 		reservation_object_add_excl_fence(resv, &rq->fence);
 	else if (reservation_object_reserve_shared(resv, 1) == 0)
 		reservation_object_add_shared_fence(resv, &rq->fence);
-	reservation_object_unlock(resv);
 }
 
 int i915_vma_move_to_active(struct i915_vma *vma,
@@ -922,7 +921,8 @@ int i915_vma_move_to_active(struct i915_vma *vma,
 {
 	struct drm_i915_gem_object *obj = vma->obj;
 
-	lockdep_assert_held(&rq->i915->drm.struct_mutex);
+	assert_vma_held(vma);
+	assert_object_held(obj);
 	GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
 
 	/*
diff --git a/drivers/gpu/drm/i915/i915_vma.h b/drivers/gpu/drm/i915/i915_vma.h
index 754a762d90b4..2657c99fe187 100644
--- a/drivers/gpu/drm/i915/i915_vma.h
+++ b/drivers/gpu/drm/i915/i915_vma.h
@@ -298,6 +298,18 @@ void i915_vma_close(struct i915_vma *vma);
 void i915_vma_reopen(struct i915_vma *vma);
 void i915_vma_destroy(struct i915_vma *vma);
 
+#define assert_vma_held(vma) reservation_object_assert_held((vma)->resv)
+
+static inline void i915_vma_lock(struct i915_vma *vma)
+{
+	reservation_object_lock(vma->resv, NULL);
+}
+
+static inline void i915_vma_unlock(struct i915_vma *vma)
+{
+	reservation_object_unlock(vma->resv);
+}
+
 int __i915_vma_do_pin(struct i915_vma *vma,
 		      u64 size, u64 alignment, u64 flags);
 static inline int __must_check
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index f454cf2450b5..0e3abc7f65e3 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -2113,6 +2113,7 @@ intel_pin_and_fence_fb_obj(struct drm_framebuffer *fb,
 	 * pin/unpin/fence and not more.
 	 */
 	wakeref = intel_runtime_pm_get(dev_priv);
+	i915_gem_object_lock(obj);
 
 	atomic_inc(&dev_priv->gpu_error.pending_fb_pin);
 
@@ -2167,6 +2168,7 @@ intel_pin_and_fence_fb_obj(struct drm_framebuffer *fb,
 err:
 	atomic_dec(&dev_priv->gpu_error.pending_fb_pin);
 
+	i915_gem_object_unlock(obj);
 	intel_runtime_pm_put(dev_priv, wakeref);
 	return vma;
 }
@@ -2175,9 +2177,12 @@ void intel_unpin_fb_vma(struct i915_vma *vma, unsigned long flags)
 {
 	lockdep_assert_held(&vma->vm->i915->drm.struct_mutex);
 
+	i915_gem_object_lock(vma->obj);
 	if (flags & PLANE_HAS_FENCE)
 		i915_vma_unpin_fence(vma);
 	i915_gem_object_unpin_from_display_plane(vma);
+	i915_gem_object_unlock(vma->obj);
+
 	i915_vma_put(vma);
 }
 
diff --git a/drivers/gpu/drm/i915/intel_guc_log.c b/drivers/gpu/drm/i915/intel_guc_log.c
index 7146524264dd..67eadc82c396 100644
--- a/drivers/gpu/drm/i915/intel_guc_log.c
+++ b/drivers/gpu/drm/i915/intel_guc_log.c
@@ -343,8 +343,6 @@ static void capture_logs_work(struct work_struct *work)
 
 static int guc_log_map(struct intel_guc_log *log)
 {
-	struct intel_guc *guc = log_to_guc(log);
-	struct drm_i915_private *dev_priv = guc_to_i915(guc);
 	void *vaddr;
 	int ret;
 
@@ -353,9 +351,9 @@ static int guc_log_map(struct intel_guc_log *log)
 	if (!log->vma)
 		return -ENODEV;
 
-	mutex_lock(&dev_priv->drm.struct_mutex);
+	i915_gem_object_lock(log->vma->obj);
 	ret = i915_gem_object_set_to_wc_domain(log->vma->obj, true);
-	mutex_unlock(&dev_priv->drm.struct_mutex);
+	i915_gem_object_unlock(log->vma->obj);
 	if (ret)
 		return ret;
 
diff --git a/drivers/gpu/drm/i915/intel_overlay.c b/drivers/gpu/drm/i915/intel_overlay.c
index 80dcd879fc58..a2ac06a08715 100644
--- a/drivers/gpu/drm/i915/intel_overlay.c
+++ b/drivers/gpu/drm/i915/intel_overlay.c
@@ -765,8 +765,10 @@ static int intel_overlay_do_put_image(struct intel_overlay *overlay,
 
 	atomic_inc(&dev_priv->gpu_error.pending_fb_pin);
 
+	i915_gem_object_lock(new_bo);
 	vma = i915_gem_object_pin_to_display_plane(new_bo,
 						   0, NULL, PIN_MAPPABLE);
+	i915_gem_object_unlock(new_bo);
 	if (IS_ERR(vma)) {
 		ret = PTR_ERR(vma);
 		goto out_pin_section;
@@ -1305,15 +1307,20 @@ out_unlock:
 
 static int get_registers(struct intel_overlay *overlay, bool use_phys)
 {
+	struct drm_i915_private *i915 = overlay->i915;
 	struct drm_i915_gem_object *obj;
 	struct i915_vma *vma;
 	int err;
 
-	obj = i915_gem_object_create_stolen(overlay->i915, PAGE_SIZE);
+	mutex_lock(&i915->drm.struct_mutex);
+
+	obj = i915_gem_object_create_stolen(i915, PAGE_SIZE);
 	if (obj == NULL)
-		obj = i915_gem_object_create_internal(overlay->i915, PAGE_SIZE);
-	if (IS_ERR(obj))
-		return PTR_ERR(obj);
+		obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
+	if (IS_ERR(obj)) {
+		err = PTR_ERR(obj);
+		goto err_unlock;
+	}
 
 	vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0, PIN_MAPPABLE);
 	if (IS_ERR(vma)) {
@@ -1334,10 +1341,13 @@ static int get_registers(struct intel_overlay *overlay, bool use_phys)
 	}
 
 	overlay->reg_bo = obj;
+	mutex_unlock(&i915->drm.struct_mutex);
 	return 0;
 
 err_put_bo:
 	i915_gem_object_put(obj);
+err_unlock:
+	mutex_unlock(&i915->drm.struct_mutex);
 	return err;
 }
 
@@ -1363,18 +1373,16 @@ void intel_overlay_setup(struct drm_i915_private *dev_priv)
 
 	INIT_ACTIVE_REQUEST(&overlay->last_flip);
 
-	mutex_lock(&dev_priv->drm.struct_mutex);
-
 	ret = get_registers(overlay, OVERLAY_NEEDS_PHYSICAL(dev_priv));
 	if (ret)
 		goto out_free;
 
+	i915_gem_object_lock(overlay->reg_bo);
 	ret = i915_gem_object_set_to_gtt_domain(overlay->reg_bo, true);
+	i915_gem_object_unlock(overlay->reg_bo);
 	if (ret)
 		goto out_reg_bo;
 
-	mutex_unlock(&dev_priv->drm.struct_mutex);
-
 	memset_io(overlay->regs, 0, sizeof(struct overlay_registers));
 	update_polyphase_filter(overlay->regs);
 	update_reg_attrs(overlay, overlay->regs);
@@ -1386,7 +1394,6 @@ void intel_overlay_setup(struct drm_i915_private *dev_priv)
 out_reg_bo:
 	i915_gem_object_put(overlay->reg_bo);
 out_free:
-	mutex_unlock(&dev_priv->drm.struct_mutex);
 	kfree(overlay);
 }
 
diff --git a/drivers/gpu/drm/i915/intel_uc_fw.c b/drivers/gpu/drm/i915/intel_uc_fw.c
index ec1e8c4deb4d..f342ddd47df8 100644
--- a/drivers/gpu/drm/i915/intel_uc_fw.c
+++ b/drivers/gpu/drm/i915/intel_uc_fw.c
@@ -246,15 +246,13 @@ int intel_uc_fw_upload(struct intel_uc_fw *uc_fw,
 			 intel_uc_fw_type_repr(uc_fw->type),
 			 intel_uc_fw_status_repr(uc_fw->load_status));
 
-	intel_uc_fw_ggtt_bind(uc_fw);
-
 	/* Call custom loader */
+	intel_uc_fw_ggtt_bind(uc_fw);
 	err = xfer(uc_fw);
+	intel_uc_fw_ggtt_unbind(uc_fw);
 	if (err)
 		goto fail;
 
-	intel_uc_fw_ggtt_unbind(uc_fw);
-
 	uc_fw->load_status = INTEL_UC_FIRMWARE_SUCCESS;
 	DRM_DEBUG_DRIVER("%s fw load %s\n",
 			 intel_uc_fw_type_repr(uc_fw->type),
diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c b/drivers/gpu/drm/i915/selftests/i915_request.c
index 4fd5356c6577..2c5479ca1f69 100644
--- a/drivers/gpu/drm/i915/selftests/i915_request.c
+++ b/drivers/gpu/drm/i915/selftests/i915_request.c
@@ -874,7 +874,9 @@ static int live_all_engines(void *arg)
 			i915_gem_object_set_active_reference(batch->obj);
 		}
 
+		i915_vma_lock(batch);
 		err = i915_vma_move_to_active(batch, request[id], 0);
+		i915_vma_unlock(batch);
 		GEM_BUG_ON(err);
 
 		i915_request_get(request[id]);
@@ -989,7 +991,9 @@ static int live_sequential_engines(void *arg)
 		GEM_BUG_ON(err);
 		request[id]->batch = batch;
 
+		i915_vma_lock(batch);
 		err = i915_vma_move_to_active(batch, request[id], 0);
+		i915_vma_unlock(batch);
 		GEM_BUG_ON(err);
 
 		i915_gem_object_set_active_reference(batch->obj);
diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c
index 6919207883f6..6f3e41c0cb3f 100644
--- a/drivers/gpu/drm/i915/selftests/i915_vma.c
+++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
@@ -886,7 +886,9 @@ static int igt_vma_remapped_gtt(void *arg)
 			unsigned int x, y;
 			int err;
 
+			i915_gem_object_lock(obj);
 			err = i915_gem_object_set_to_gtt_domain(obj, true);
+			i915_gem_object_unlock(obj);
 			if (err)
 				goto out;
 
diff --git a/drivers/gpu/drm/i915/selftests/igt_spinner.c b/drivers/gpu/drm/i915/selftests/igt_spinner.c
index 38d6f1b10c54..15c0f0af9658 100644
--- a/drivers/gpu/drm/i915/selftests/igt_spinner.c
+++ b/drivers/gpu/drm/i915/selftests/igt_spinner.c
@@ -76,7 +76,9 @@ static int move_to_active(struct i915_vma *vma,
 {
 	int err;
 
+	i915_vma_lock(vma);
 	err = i915_vma_move_to_active(vma, rq, flags);
+	i915_vma_unlock(vma);
 	if (err)
 		return err;